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 <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.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 static int sctp_memory_pressure
;
106 static atomic_long_t sctp_memory_allocated
;
107 struct percpu_counter sctp_sockets_allocated
;
109 static void sctp_enter_memory_pressure(struct sock
*sk
)
111 sctp_memory_pressure
= 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association
*asoc
)
120 if (asoc
->ep
->sndbuf_policy
)
121 amt
= asoc
->sndbuf_used
;
123 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
125 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
126 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
129 amt
= sk_stream_wspace(asoc
->base
.sk
);
134 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
148 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
150 struct sctp_association
*asoc
= chunk
->asoc
;
151 struct sock
*sk
= asoc
->base
.sk
;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc
);
156 skb_set_owner_w(chunk
->skb
, sk
);
158 chunk
->skb
->destructor
= sctp_wfree
;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk
->skb
)->destructor_arg
= chunk
;
162 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
163 sizeof(struct sk_buff
) +
164 sizeof(struct sctp_chunk
);
166 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
167 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
168 sk_mem_charge(sk
, chunk
->skb
->truesize
);
171 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
177 /* Verify basic sockaddr. */
178 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
182 /* Is this a valid SCTP address? */
183 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
186 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
195 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
197 struct sctp_association
*asoc
= NULL
;
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk
, UDP
)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
205 if (!sctp_sstate(sk
, ESTABLISHED
))
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk
)->ep
->asocs
))
210 asoc
= list_entry(sctp_sk(sk
)->ep
->asocs
.next
,
211 struct sctp_association
, asocs
);
215 /* Otherwise this is a UDP-style socket. */
216 if (!id
|| (id
== (sctp_assoc_t
)-1))
219 spin_lock_bh(&sctp_assocs_id_lock
);
220 asoc
= (struct sctp_association
*)idr_find(&sctp_assocs_id
, (int)id
);
221 spin_unlock_bh(&sctp_assocs_id_lock
);
223 if (!asoc
|| (asoc
->base
.sk
!= sk
) || asoc
->base
.dead
)
229 /* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
233 static struct sctp_transport
*sctp_addr_id2transport(struct sock
*sk
,
234 struct sockaddr_storage
*addr
,
237 struct sctp_association
*addr_asoc
= NULL
, *id_asoc
= NULL
;
238 struct sctp_transport
*transport
;
239 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
241 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
248 id_asoc
= sctp_id2assoc(sk
, id
);
249 if (id_asoc
&& (id_asoc
!= addr_asoc
))
252 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
253 (union sctp_addr
*)addr
);
258 /* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
261 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
263 * sd - the socket descriptor returned by socket().
264 * addr - the address structure (struct sockaddr_in or struct
265 * sockaddr_in6 [RFC 2553]),
266 * addr_len - the size of the address structure.
268 static int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__
, sk
,
277 /* Disallow binding twice. */
278 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
279 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
289 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
291 /* Verify this is a valid sockaddr. */
292 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
293 union sctp_addr
*addr
, int len
)
297 /* Check minimum size. */
298 if (len
< sizeof (struct sockaddr
))
301 /* V4 mapped address are really of AF_INET family */
302 if (addr
->sa
.sa_family
== AF_INET6
&&
303 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
304 if (!opt
->pf
->af_supported(AF_INET
, opt
))
307 /* Does this PF support this AF? */
308 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
312 /* If we get this far, af is valid. */
313 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
315 if (len
< af
->sockaddr_len
)
321 /* Bind a local address either to an endpoint or to an association. */
322 static int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
324 struct net
*net
= sock_net(sk
);
325 struct sctp_sock
*sp
= sctp_sk(sk
);
326 struct sctp_endpoint
*ep
= sp
->ep
;
327 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
332 /* Common sockaddr verification. */
333 af
= sctp_sockaddr_af(sp
, addr
, len
);
335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336 __func__
, sk
, addr
, len
);
340 snum
= ntohs(addr
->v4
.sin_port
);
342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343 __func__
, sk
, &addr
->sa
, bp
->port
, snum
, len
);
345 /* PF specific bind() address verification. */
346 if (!sp
->pf
->bind_verify(sp
, addr
))
347 return -EADDRNOTAVAIL
;
349 /* We must either be unbound, or bind to the same port.
350 * It's OK to allow 0 ports if we are already bound.
351 * We'll just inhert an already bound port in this case
356 else if (snum
!= bp
->port
) {
357 pr_debug("%s: new port %d doesn't match existing port "
358 "%d\n", __func__
, snum
, bp
->port
);
363 if (snum
&& snum
< PROT_SOCK
&&
364 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
367 /* See if the address matches any of the addresses we may have
368 * already bound before checking against other endpoints.
370 if (sctp_bind_addr_match(bp
, addr
, sp
))
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
377 addr
->v4
.sin_port
= htons(snum
);
378 if ((ret
= sctp_get_port_local(sk
, addr
))) {
382 /* Refresh ephemeral port. */
384 bp
->port
= inet_sk(sk
)->inet_num
;
386 /* Add the address to the bind address list.
387 * Use GFP_ATOMIC since BHs will be disabled.
389 ret
= sctp_add_bind_addr(bp
, addr
, af
->sockaddr_len
,
390 SCTP_ADDR_SRC
, GFP_ATOMIC
);
392 /* Copy back into socket for getsockname() use. */
394 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
395 sp
->pf
->to_sk_saddr(addr
, sk
);
401 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
403 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
404 * at any one time. If a sender, after sending an ASCONF chunk, decides
405 * it needs to transfer another ASCONF Chunk, it MUST wait until the
406 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
407 * subsequent ASCONF. Note this restriction binds each side, so at any
408 * time two ASCONF may be in-transit on any given association (one sent
409 * from each endpoint).
411 static int sctp_send_asconf(struct sctp_association
*asoc
,
412 struct sctp_chunk
*chunk
)
414 struct net
*net
= sock_net(asoc
->base
.sk
);
417 /* If there is an outstanding ASCONF chunk, queue it for later
420 if (asoc
->addip_last_asconf
) {
421 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
425 /* Hold the chunk until an ASCONF_ACK is received. */
426 sctp_chunk_hold(chunk
);
427 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
429 sctp_chunk_free(chunk
);
431 asoc
->addip_last_asconf
= chunk
;
437 /* Add a list of addresses as bind addresses to local endpoint or
440 * Basically run through each address specified in the addrs/addrcnt
441 * array/length pair, determine if it is IPv6 or IPv4 and call
442 * sctp_do_bind() on it.
444 * If any of them fails, then the operation will be reversed and the
445 * ones that were added will be removed.
447 * Only sctp_setsockopt_bindx() is supposed to call this function.
449 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
454 struct sockaddr
*sa_addr
;
457 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__
, sk
,
461 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
462 /* The list may contain either IPv4 or IPv6 address;
463 * determine the address length for walking thru the list.
466 af
= sctp_get_af_specific(sa_addr
->sa_family
);
472 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
475 addr_buf
+= af
->sockaddr_len
;
479 /* Failed. Cleanup the ones that have been added */
481 sctp_bindx_rem(sk
, addrs
, cnt
);
489 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
490 * associations that are part of the endpoint indicating that a list of local
491 * addresses are added to the endpoint.
493 * If any of the addresses is already in the bind address list of the
494 * association, we do not send the chunk for that association. But it will not
495 * affect other associations.
497 * Only sctp_setsockopt_bindx() is supposed to call this function.
499 static int sctp_send_asconf_add_ip(struct sock
*sk
,
500 struct sockaddr
*addrs
,
503 struct net
*net
= sock_net(sk
);
504 struct sctp_sock
*sp
;
505 struct sctp_endpoint
*ep
;
506 struct sctp_association
*asoc
;
507 struct sctp_bind_addr
*bp
;
508 struct sctp_chunk
*chunk
;
509 struct sctp_sockaddr_entry
*laddr
;
510 union sctp_addr
*addr
;
511 union sctp_addr saveaddr
;
518 if (!net
->sctp
.addip_enable
)
524 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
525 __func__
, sk
, addrs
, addrcnt
);
527 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
528 if (!asoc
->peer
.asconf_capable
)
531 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
534 if (!sctp_state(asoc
, ESTABLISHED
))
537 /* Check if any address in the packed array of addresses is
538 * in the bind address list of the association. If so,
539 * do not send the asconf chunk to its peer, but continue with
540 * other associations.
543 for (i
= 0; i
< addrcnt
; i
++) {
545 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
551 if (sctp_assoc_lookup_laddr(asoc
, addr
))
554 addr_buf
+= af
->sockaddr_len
;
559 /* Use the first valid address in bind addr list of
560 * association as Address Parameter of ASCONF CHUNK.
562 bp
= &asoc
->base
.bind_addr
;
563 p
= bp
->address_list
.next
;
564 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
565 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
566 addrcnt
, SCTP_PARAM_ADD_IP
);
572 /* Add the new addresses to the bind address list with
573 * use_as_src set to 0.
576 for (i
= 0; i
< addrcnt
; i
++) {
578 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
579 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
580 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
582 SCTP_ADDR_NEW
, GFP_ATOMIC
);
583 addr_buf
+= af
->sockaddr_len
;
585 if (asoc
->src_out_of_asoc_ok
) {
586 struct sctp_transport
*trans
;
588 list_for_each_entry(trans
,
589 &asoc
->peer
.transport_addr_list
, transports
) {
590 /* Clear the source and route cache */
591 dst_release(trans
->dst
);
592 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
593 2*asoc
->pathmtu
, 4380));
594 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
595 trans
->rto
= asoc
->rto_initial
;
596 sctp_max_rto(asoc
, trans
);
597 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
598 sctp_transport_route(trans
, NULL
,
599 sctp_sk(asoc
->base
.sk
));
602 retval
= sctp_send_asconf(asoc
, chunk
);
609 /* Remove a list of addresses from bind addresses list. Do not remove the
612 * Basically run through each address specified in the addrs/addrcnt
613 * array/length pair, determine if it is IPv6 or IPv4 and call
614 * sctp_del_bind() on it.
616 * If any of them fails, then the operation will be reversed and the
617 * ones that were removed will be added back.
619 * At least one address has to be left; if only one address is
620 * available, the operation will return -EBUSY.
622 * Only sctp_setsockopt_bindx() is supposed to call this function.
624 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
626 struct sctp_sock
*sp
= sctp_sk(sk
);
627 struct sctp_endpoint
*ep
= sp
->ep
;
629 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
632 union sctp_addr
*sa_addr
;
635 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
636 __func__
, sk
, addrs
, addrcnt
);
639 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
640 /* If the bind address list is empty or if there is only one
641 * bind address, there is nothing more to be removed (we need
642 * at least one address here).
644 if (list_empty(&bp
->address_list
) ||
645 (sctp_list_single_entry(&bp
->address_list
))) {
651 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
657 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
658 retval
= -EADDRNOTAVAIL
;
662 if (sa_addr
->v4
.sin_port
&&
663 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
668 if (!sa_addr
->v4
.sin_port
)
669 sa_addr
->v4
.sin_port
= htons(bp
->port
);
671 /* FIXME - There is probably a need to check if sk->sk_saddr and
672 * sk->sk_rcv_addr are currently set to one of the addresses to
673 * be removed. This is something which needs to be looked into
674 * when we are fixing the outstanding issues with multi-homing
675 * socket routing and failover schemes. Refer to comments in
676 * sctp_do_bind(). -daisy
678 retval
= sctp_del_bind_addr(bp
, sa_addr
);
680 addr_buf
+= af
->sockaddr_len
;
683 /* Failed. Add the ones that has been removed back */
685 sctp_bindx_add(sk
, addrs
, cnt
);
693 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
694 * the associations that are part of the endpoint indicating that a list of
695 * local addresses are removed from the endpoint.
697 * If any of the addresses is already in the bind address list of the
698 * association, we do not send the chunk for that association. But it will not
699 * affect other associations.
701 * Only sctp_setsockopt_bindx() is supposed to call this function.
703 static int sctp_send_asconf_del_ip(struct sock
*sk
,
704 struct sockaddr
*addrs
,
707 struct net
*net
= sock_net(sk
);
708 struct sctp_sock
*sp
;
709 struct sctp_endpoint
*ep
;
710 struct sctp_association
*asoc
;
711 struct sctp_transport
*transport
;
712 struct sctp_bind_addr
*bp
;
713 struct sctp_chunk
*chunk
;
714 union sctp_addr
*laddr
;
717 struct sctp_sockaddr_entry
*saddr
;
723 if (!net
->sctp
.addip_enable
)
729 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
730 __func__
, sk
, addrs
, addrcnt
);
732 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
734 if (!asoc
->peer
.asconf_capable
)
737 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
740 if (!sctp_state(asoc
, ESTABLISHED
))
743 /* Check if any address in the packed array of addresses is
744 * not present in the bind address list of the association.
745 * If so, do not send the asconf chunk to its peer, but
746 * continue with other associations.
749 for (i
= 0; i
< addrcnt
; i
++) {
751 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
757 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
760 addr_buf
+= af
->sockaddr_len
;
765 /* Find one address in the association's bind address list
766 * that is not in the packed array of addresses. This is to
767 * make sure that we do not delete all the addresses in the
770 bp
= &asoc
->base
.bind_addr
;
771 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
773 if ((laddr
== NULL
) && (addrcnt
== 1)) {
774 if (asoc
->asconf_addr_del_pending
)
776 asoc
->asconf_addr_del_pending
=
777 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
778 if (asoc
->asconf_addr_del_pending
== NULL
) {
782 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
784 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
786 if (addrs
->sa_family
== AF_INET
) {
787 struct sockaddr_in
*sin
;
789 sin
= (struct sockaddr_in
*)addrs
;
790 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
791 } else if (addrs
->sa_family
== AF_INET6
) {
792 struct sockaddr_in6
*sin6
;
794 sin6
= (struct sockaddr_in6
*)addrs
;
795 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
798 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
799 __func__
, asoc
, &asoc
->asconf_addr_del_pending
->sa
,
800 asoc
->asconf_addr_del_pending
);
802 asoc
->src_out_of_asoc_ok
= 1;
810 /* We do not need RCU protection throughout this loop
811 * because this is done under a socket lock from the
814 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
822 /* Reset use_as_src flag for the addresses in the bind address
823 * list that are to be deleted.
826 for (i
= 0; i
< addrcnt
; i
++) {
828 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
829 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
830 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
831 saddr
->state
= SCTP_ADDR_DEL
;
833 addr_buf
+= af
->sockaddr_len
;
836 /* Update the route and saddr entries for all the transports
837 * as some of the addresses in the bind address list are
838 * about to be deleted and cannot be used as source addresses.
840 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
842 dst_release(transport
->dst
);
843 sctp_transport_route(transport
, NULL
,
844 sctp_sk(asoc
->base
.sk
));
848 /* We don't need to transmit ASCONF */
850 retval
= sctp_send_asconf(asoc
, chunk
);
856 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
857 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
859 struct sock
*sk
= sctp_opt2sk(sp
);
860 union sctp_addr
*addr
;
863 /* It is safe to write port space in caller. */
865 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
866 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
869 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
872 if (addrw
->state
== SCTP_ADDR_NEW
)
873 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
875 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
878 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
881 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
884 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
885 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
888 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
889 * Section 3.1.2 for this usage.
891 * addrs is a pointer to an array of one or more socket addresses. Each
892 * address is contained in its appropriate structure (i.e. struct
893 * sockaddr_in or struct sockaddr_in6) the family of the address type
894 * must be used to distinguish the address length (note that this
895 * representation is termed a "packed array" of addresses). The caller
896 * specifies the number of addresses in the array with addrcnt.
898 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
899 * -1, and sets errno to the appropriate error code.
901 * For SCTP, the port given in each socket address must be the same, or
902 * sctp_bindx() will fail, setting errno to EINVAL.
904 * The flags parameter is formed from the bitwise OR of zero or more of
905 * the following currently defined flags:
907 * SCTP_BINDX_ADD_ADDR
909 * SCTP_BINDX_REM_ADDR
911 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
912 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
913 * addresses from the association. The two flags are mutually exclusive;
914 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
915 * not remove all addresses from an association; sctp_bindx() will
916 * reject such an attempt with EINVAL.
918 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
919 * additional addresses with an endpoint after calling bind(). Or use
920 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
921 * socket is associated with so that no new association accepted will be
922 * associated with those addresses. If the endpoint supports dynamic
923 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
924 * endpoint to send the appropriate message to the peer to change the
925 * peers address lists.
927 * Adding and removing addresses from a connected association is
928 * optional functionality. Implementations that do not support this
929 * functionality should return EOPNOTSUPP.
931 * Basically do nothing but copying the addresses from user to kernel
932 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
933 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
936 * We don't use copy_from_user() for optimization: we first do the
937 * sanity checks (buffer size -fast- and access check-healthy
938 * pointer); if all of those succeed, then we can alloc the memory
939 * (expensive operation) needed to copy the data to kernel. Then we do
940 * the copying without checking the user space area
941 * (__copy_from_user()).
943 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
946 * sk The sk of the socket
947 * addrs The pointer to the addresses in user land
948 * addrssize Size of the addrs buffer
949 * op Operation to perform (add or remove, see the flags of
952 * Returns 0 if ok, <0 errno code on error.
954 static int sctp_setsockopt_bindx(struct sock
*sk
,
955 struct sockaddr __user
*addrs
,
956 int addrs_size
, int op
)
958 struct sockaddr
*kaddrs
;
962 struct sockaddr
*sa_addr
;
966 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
967 __func__
, sk
, addrs
, addrs_size
, op
);
969 if (unlikely(addrs_size
<= 0))
972 /* Check the user passed a healthy pointer. */
973 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
976 /* Alloc space for the address array in kernel memory. */
977 kaddrs
= kmalloc(addrs_size
, GFP_USER
| __GFP_NOWARN
);
978 if (unlikely(!kaddrs
))
981 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
986 /* Walk through the addrs buffer and count the number of addresses. */
988 while (walk_size
< addrs_size
) {
989 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
995 af
= sctp_get_af_specific(sa_addr
->sa_family
);
997 /* If the address family is not supported or if this address
998 * causes the address buffer to overflow return EINVAL.
1000 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1005 addr_buf
+= af
->sockaddr_len
;
1006 walk_size
+= af
->sockaddr_len
;
1011 case SCTP_BINDX_ADD_ADDR
:
1012 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1015 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1018 case SCTP_BINDX_REM_ADDR
:
1019 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1022 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1036 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1038 * Common routine for handling connect() and sctp_connectx().
1039 * Connect will come in with just a single address.
1041 static int __sctp_connect(struct sock
*sk
,
1042 struct sockaddr
*kaddrs
,
1044 sctp_assoc_t
*assoc_id
)
1046 struct net
*net
= sock_net(sk
);
1047 struct sctp_sock
*sp
;
1048 struct sctp_endpoint
*ep
;
1049 struct sctp_association
*asoc
= NULL
;
1050 struct sctp_association
*asoc2
;
1051 struct sctp_transport
*transport
;
1058 union sctp_addr
*sa_addr
= NULL
;
1060 unsigned short port
;
1061 unsigned int f_flags
= 0;
1066 /* connect() cannot be done on a socket that is already in ESTABLISHED
1067 * state - UDP-style peeled off socket or a TCP-style socket that
1068 * is already connected.
1069 * It cannot be done even on a TCP-style listening socket.
1071 if (sctp_sstate(sk
, ESTABLISHED
) ||
1072 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1077 /* Walk through the addrs buffer and count the number of addresses. */
1079 while (walk_size
< addrs_size
) {
1082 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1088 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1090 /* If the address family is not supported or if this address
1091 * causes the address buffer to overflow return EINVAL.
1093 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1098 port
= ntohs(sa_addr
->v4
.sin_port
);
1100 /* Save current address so we can work with it */
1101 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1103 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1107 /* Make sure the destination port is correctly set
1110 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1115 /* Check if there already is a matching association on the
1116 * endpoint (other than the one created here).
1118 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1119 if (asoc2
&& asoc2
!= asoc
) {
1120 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1127 /* If we could not find a matching association on the endpoint,
1128 * make sure that there is no peeled-off association matching
1129 * the peer address even on another socket.
1131 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1132 err
= -EADDRNOTAVAIL
;
1137 /* If a bind() or sctp_bindx() is not called prior to
1138 * an sctp_connectx() call, the system picks an
1139 * ephemeral port and will choose an address set
1140 * equivalent to binding with a wildcard address.
1142 if (!ep
->base
.bind_addr
.port
) {
1143 if (sctp_autobind(sk
)) {
1149 * If an unprivileged user inherits a 1-many
1150 * style socket with open associations on a
1151 * privileged port, it MAY be permitted to
1152 * accept new associations, but it SHOULD NOT
1153 * be permitted to open new associations.
1155 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1156 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1162 scope
= sctp_scope(&to
);
1163 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1169 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1177 /* Prime the peer's transport structures. */
1178 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1186 addr_buf
+= af
->sockaddr_len
;
1187 walk_size
+= af
->sockaddr_len
;
1190 /* In case the user of sctp_connectx() wants an association
1191 * id back, assign one now.
1194 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1199 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1204 /* Initialize sk's dport and daddr for getpeername() */
1205 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1206 sp
->pf
->to_sk_daddr(sa_addr
, sk
);
1209 /* in-kernel sockets don't generally have a file allocated to them
1210 * if all they do is call sock_create_kern().
1212 if (sk
->sk_socket
->file
)
1213 f_flags
= sk
->sk_socket
->file
->f_flags
;
1215 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1217 err
= sctp_wait_for_connect(asoc
, &timeo
);
1218 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1219 *assoc_id
= asoc
->assoc_id
;
1221 /* Don't free association on exit. */
1225 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1226 __func__
, asoc
, kaddrs
, err
);
1229 /* sctp_primitive_ASSOCIATE may have added this association
1230 * To the hash table, try to unhash it, just in case, its a noop
1231 * if it wasn't hashed so we're safe
1233 sctp_association_free(asoc
);
1238 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1298 * Returns >=0 if ok, <0 errno code on error.
1300 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1301 struct sockaddr __user
*addrs
,
1303 sctp_assoc_t
*assoc_id
)
1305 struct sockaddr
*kaddrs
;
1306 gfp_t gfp
= GFP_KERNEL
;
1309 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1310 __func__
, sk
, addrs
, addrs_size
);
1312 if (unlikely(addrs_size
<= 0))
1315 /* Check the user passed a healthy pointer. */
1316 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1319 /* Alloc space for the address array in kernel memory. */
1320 if (sk
->sk_socket
->file
)
1321 gfp
= GFP_USER
| __GFP_NOWARN
;
1322 kaddrs
= kmalloc(addrs_size
, gfp
);
1323 if (unlikely(!kaddrs
))
1326 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1329 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1338 * This is an older interface. It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1341 static int sctp_setsockopt_connectx_old(struct sock
*sk
,
1342 struct sockaddr __user
*addrs
,
1345 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1349 * New interface for the API. The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call. Error is always negative and association id is
1354 static int sctp_setsockopt_connectx(struct sock
*sk
,
1355 struct sockaddr __user
*addrs
,
1358 sctp_assoc_t assoc_id
= 0;
1361 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1377 #ifdef CONFIG_COMPAT
1378 struct compat_sctp_getaddrs_old
{
1379 sctp_assoc_t assoc_id
;
1381 compat_uptr_t addrs
; /* struct sockaddr * */
1385 static int sctp_getsockopt_connectx3(struct sock
*sk
, int len
,
1386 char __user
*optval
,
1389 struct sctp_getaddrs_old param
;
1390 sctp_assoc_t assoc_id
= 0;
1393 #ifdef CONFIG_COMPAT
1394 if (in_compat_syscall()) {
1395 struct compat_sctp_getaddrs_old param32
;
1397 if (len
< sizeof(param32
))
1399 if (copy_from_user(¶m32
, optval
, sizeof(param32
)))
1402 param
.assoc_id
= param32
.assoc_id
;
1403 param
.addr_num
= param32
.addr_num
;
1404 param
.addrs
= compat_ptr(param32
.addrs
);
1408 if (len
< sizeof(param
))
1410 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1414 err
= __sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)
1415 param
.addrs
, param
.addr_num
,
1417 if (err
== 0 || err
== -EINPROGRESS
) {
1418 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1420 if (put_user(sizeof(assoc_id
), optlen
))
1427 /* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1434 * ret = close(int sd);
1436 * sd - the socket descriptor of the associations to be closed.
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1446 * 4.1.6 close() - TCP Style Syntax
1448 * Applications use close() to gracefully close down an association.
1452 * int close(int sd);
1454 * sd - the socket descriptor of the association to be closed.
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1459 * API 7.1.4 SO_LINGER
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive. The linger option structure is:
1465 * int l_onoff; // option on/off
1466 * int l_linger; // linger time
1469 * To enable the option, set l_onoff to 1. If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive. If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error. If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms. If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1477 static void sctp_close(struct sock
*sk
, long timeout
)
1479 struct net
*net
= sock_net(sk
);
1480 struct sctp_endpoint
*ep
;
1481 struct sctp_association
*asoc
;
1482 struct list_head
*pos
, *temp
;
1483 unsigned int data_was_unread
;
1485 pr_debug("%s: sk:%p, timeout:%ld\n", __func__
, sk
, timeout
);
1488 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1489 sk
->sk_state
= SCTP_SS_CLOSING
;
1491 ep
= sctp_sk(sk
)->ep
;
1493 /* Clean up any skbs sitting on the receive queue. */
1494 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1495 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1497 /* Walk all associations on an endpoint. */
1498 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1499 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1501 if (sctp_style(sk
, TCP
)) {
1502 /* A closed association can still be in the list if
1503 * it belongs to a TCP-style listening socket that is
1504 * not yet accepted. If so, free it. If not, send an
1505 * ABORT or SHUTDOWN based on the linger options.
1507 if (sctp_state(asoc
, CLOSED
)) {
1508 sctp_association_free(asoc
);
1513 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1514 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1515 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1516 struct sctp_chunk
*chunk
;
1518 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1519 sctp_primitive_ABORT(net
, asoc
, chunk
);
1521 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1524 /* On a TCP-style socket, block for at most linger_time if set. */
1525 if (sctp_style(sk
, TCP
) && timeout
)
1526 sctp_wait_for_close(sk
, timeout
);
1528 /* This will run the backlog queue. */
1531 /* Supposedly, no process has access to the socket, but
1532 * the net layers still may.
1533 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1534 * held and that should be grabbed before socket lock.
1536 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
1539 /* Hold the sock, since sk_common_release() will put sock_put()
1540 * and we have just a little more cleanup.
1543 sk_common_release(sk
);
1546 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
1550 SCTP_DBG_OBJCNT_DEC(sock
);
1553 /* Handle EPIPE error. */
1554 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1557 err
= sock_error(sk
) ? : -EPIPE
;
1558 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1559 send_sig(SIGPIPE
, current
, 0);
1563 /* API 3.1.3 sendmsg() - UDP Style Syntax
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1568 * ssize_t sendmsg(int socket, const struct msghdr *message,
1571 * socket - the socket descriptor of the endpoint.
1572 * message - pointer to the msghdr structure which contains a single
1573 * user message and possibly some ancillary data.
1575 * See Section 5 for complete description of the data
1578 * flags - flags sent or received with the user message, see Section
1579 * 5 for complete description of the flags.
1581 * Note: This function could use a rewrite especially when explicit
1582 * connect support comes in.
1584 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1586 static int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1588 static int sctp_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t msg_len
)
1590 struct net
*net
= sock_net(sk
);
1591 struct sctp_sock
*sp
;
1592 struct sctp_endpoint
*ep
;
1593 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1594 struct sctp_transport
*transport
, *chunk_tp
;
1595 struct sctp_chunk
*chunk
;
1597 struct sockaddr
*msg_name
= NULL
;
1598 struct sctp_sndrcvinfo default_sinfo
;
1599 struct sctp_sndrcvinfo
*sinfo
;
1600 struct sctp_initmsg
*sinit
;
1601 sctp_assoc_t associd
= 0;
1602 sctp_cmsgs_t cmsgs
= { NULL
};
1604 bool fill_sinfo_ttl
= false, wait_connect
= false;
1605 struct sctp_datamsg
*datamsg
;
1606 int msg_flags
= msg
->msg_flags
;
1607 __u16 sinfo_flags
= 0;
1615 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__
, sk
,
1618 /* We cannot send a message over a TCP-style listening socket. */
1619 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1624 /* Parse out the SCTP CMSGs. */
1625 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1627 pr_debug("%s: msghdr parse err:%x\n", __func__
, err
);
1631 /* Fetch the destination address for this packet. This
1632 * address only selects the association--it is not necessarily
1633 * the address we will send to.
1634 * For a peeled-off socket, msg_name is ignored.
1636 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1637 int msg_namelen
= msg
->msg_namelen
;
1639 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1644 if (msg_namelen
> sizeof(to
))
1645 msg_namelen
= sizeof(to
);
1646 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1647 msg_name
= msg
->msg_name
;
1651 if (cmsgs
.sinfo
!= NULL
) {
1652 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1653 default_sinfo
.sinfo_stream
= cmsgs
.sinfo
->snd_sid
;
1654 default_sinfo
.sinfo_flags
= cmsgs
.sinfo
->snd_flags
;
1655 default_sinfo
.sinfo_ppid
= cmsgs
.sinfo
->snd_ppid
;
1656 default_sinfo
.sinfo_context
= cmsgs
.sinfo
->snd_context
;
1657 default_sinfo
.sinfo_assoc_id
= cmsgs
.sinfo
->snd_assoc_id
;
1659 sinfo
= &default_sinfo
;
1660 fill_sinfo_ttl
= true;
1662 sinfo
= cmsgs
.srinfo
;
1664 /* Did the user specify SNDINFO/SNDRCVINFO? */
1666 sinfo_flags
= sinfo
->sinfo_flags
;
1667 associd
= sinfo
->sinfo_assoc_id
;
1670 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1671 msg_len
, sinfo_flags
);
1673 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1674 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1679 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1680 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1681 * If SCTP_ABORT is set, the message length could be non zero with
1682 * the msg_iov set to the user abort reason.
1684 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1685 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1690 /* If SCTP_ADDR_OVER is set, there must be an address
1691 * specified in msg_name.
1693 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1700 pr_debug("%s: about to look up association\n", __func__
);
1704 /* If a msg_name has been specified, assume this is to be used. */
1706 /* Look for a matching association on the endpoint. */
1707 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1709 /* If we could not find a matching association on the
1710 * endpoint, make sure that it is not a TCP-style
1711 * socket that already has an association or there is
1712 * no peeled-off association on another socket.
1714 if ((sctp_style(sk
, TCP
) &&
1715 sctp_sstate(sk
, ESTABLISHED
)) ||
1716 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1717 err
= -EADDRNOTAVAIL
;
1722 asoc
= sctp_id2assoc(sk
, associd
);
1730 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1732 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1733 * socket that has an association in CLOSED state. This can
1734 * happen when an accepted socket has an association that is
1737 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1742 if (sinfo_flags
& SCTP_EOF
) {
1743 pr_debug("%s: shutting down association:%p\n",
1746 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1750 if (sinfo_flags
& SCTP_ABORT
) {
1752 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1758 pr_debug("%s: aborting association:%p\n",
1761 sctp_primitive_ABORT(net
, asoc
, chunk
);
1767 /* Do we need to create the association? */
1769 pr_debug("%s: there is no association yet\n", __func__
);
1771 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1776 /* Check for invalid stream against the stream counts,
1777 * either the default or the user specified stream counts.
1780 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1781 /* Check against the defaults. */
1782 if (sinfo
->sinfo_stream
>=
1783 sp
->initmsg
.sinit_num_ostreams
) {
1788 /* Check against the requested. */
1789 if (sinfo
->sinfo_stream
>=
1790 sinit
->sinit_num_ostreams
) {
1798 * API 3.1.2 bind() - UDP Style Syntax
1799 * If a bind() or sctp_bindx() is not called prior to a
1800 * sendmsg() call that initiates a new association, the
1801 * system picks an ephemeral port and will choose an address
1802 * set equivalent to binding with a wildcard address.
1804 if (!ep
->base
.bind_addr
.port
) {
1805 if (sctp_autobind(sk
)) {
1811 * If an unprivileged user inherits a one-to-many
1812 * style socket with open associations on a privileged
1813 * port, it MAY be permitted to accept new associations,
1814 * but it SHOULD NOT be permitted to open new
1817 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1818 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1824 scope
= sctp_scope(&to
);
1825 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1831 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1837 /* If the SCTP_INIT ancillary data is specified, set all
1838 * the association init values accordingly.
1841 if (sinit
->sinit_num_ostreams
) {
1842 asoc
->c
.sinit_num_ostreams
=
1843 sinit
->sinit_num_ostreams
;
1845 if (sinit
->sinit_max_instreams
) {
1846 asoc
->c
.sinit_max_instreams
=
1847 sinit
->sinit_max_instreams
;
1849 if (sinit
->sinit_max_attempts
) {
1850 asoc
->max_init_attempts
1851 = sinit
->sinit_max_attempts
;
1853 if (sinit
->sinit_max_init_timeo
) {
1854 asoc
->max_init_timeo
=
1855 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1859 /* Prime the peer's transport structures. */
1860 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1867 /* ASSERT: we have a valid association at this point. */
1868 pr_debug("%s: we have a valid association\n", __func__
);
1871 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1872 * one with some defaults.
1874 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1875 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1876 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1877 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1878 default_sinfo
.sinfo_context
= asoc
->default_context
;
1879 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1880 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1882 sinfo
= &default_sinfo
;
1883 } else if (fill_sinfo_ttl
) {
1884 /* In case SNDINFO was specified, we still need to fill
1885 * it with a default ttl from the assoc here.
1887 sinfo
->sinfo_timetolive
= asoc
->default_timetolive
;
1890 /* API 7.1.7, the sndbuf size per association bounds the
1891 * maximum size of data that can be sent in a single send call.
1893 if (msg_len
> sk
->sk_sndbuf
) {
1898 if (asoc
->pmtu_pending
)
1899 sctp_assoc_pending_pmtu(sk
, asoc
);
1901 /* If fragmentation is disabled and the message length exceeds the
1902 * association fragmentation point, return EMSGSIZE. The I-D
1903 * does not specify what this error is, but this looks like
1906 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1911 /* Check for invalid stream. */
1912 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1917 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1918 if (!sctp_wspace(asoc
)) {
1919 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1924 /* If an address is passed with the sendto/sendmsg call, it is used
1925 * to override the primary destination address in the TCP model, or
1926 * when SCTP_ADDR_OVER flag is set in the UDP model.
1928 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1929 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1930 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1938 /* Auto-connect, if we aren't connected already. */
1939 if (sctp_state(asoc
, CLOSED
)) {
1940 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1944 wait_connect
= true;
1945 pr_debug("%s: we associated primitively\n", __func__
);
1948 /* Break the message into multiple chunks of maximum size. */
1949 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, &msg
->msg_iter
);
1950 if (IS_ERR(datamsg
)) {
1951 err
= PTR_ERR(datamsg
);
1955 /* Now send the (possibly) fragmented message. */
1956 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1957 /* Do accounting for the write space. */
1958 sctp_set_owner_w(chunk
);
1960 chunk
->transport
= chunk_tp
;
1963 /* Send it to the lower layers. Note: all chunks
1964 * must either fail or succeed. The lower layer
1965 * works that way today. Keep it that way or this
1968 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1969 sctp_datamsg_put(datamsg
);
1970 /* Did the lower layer accept the chunk? */
1974 pr_debug("%s: we sent primitively\n", __func__
);
1978 if (unlikely(wait_connect
)) {
1979 timeo
= sock_sndtimeo(sk
, msg_flags
& MSG_DONTWAIT
);
1980 sctp_wait_for_connect(asoc
, &timeo
);
1983 /* If we are already past ASSOCIATE, the lower
1984 * layers are responsible for association cleanup.
1990 sctp_association_free(asoc
);
1995 return sctp_error(sk
, msg_flags
, err
);
2002 err
= sock_error(sk
);
2012 /* This is an extended version of skb_pull() that removes the data from the
2013 * start of a skb even when data is spread across the list of skb's in the
2014 * frag_list. len specifies the total amount of data that needs to be removed.
2015 * when 'len' bytes could be removed from the skb, it returns 0.
2016 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2017 * could not be removed.
2019 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2021 struct sk_buff
*list
;
2022 int skb_len
= skb_headlen(skb
);
2025 if (len
<= skb_len
) {
2026 __skb_pull(skb
, len
);
2030 __skb_pull(skb
, skb_len
);
2032 skb_walk_frags(skb
, list
) {
2033 rlen
= sctp_skb_pull(list
, len
);
2034 skb
->len
-= (len
-rlen
);
2035 skb
->data_len
-= (len
-rlen
);
2046 /* API 3.1.3 recvmsg() - UDP Style Syntax
2048 * ssize_t recvmsg(int socket, struct msghdr *message,
2051 * socket - the socket descriptor of the endpoint.
2052 * message - pointer to the msghdr structure which contains a single
2053 * user message and possibly some ancillary data.
2055 * See Section 5 for complete description of the data
2058 * flags - flags sent or received with the user message, see Section
2059 * 5 for complete description of the flags.
2061 static int sctp_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
2062 int noblock
, int flags
, int *addr_len
)
2064 struct sctp_ulpevent
*event
= NULL
;
2065 struct sctp_sock
*sp
= sctp_sk(sk
);
2066 struct sk_buff
*skb
;
2071 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2072 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2077 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2082 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2086 /* Get the total length of the skb including any skb's in the
2095 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
2097 event
= sctp_skb2event(skb
);
2102 sock_recv_ts_and_drops(msg
, sk
, skb
);
2103 if (sctp_ulpevent_is_notification(event
)) {
2104 msg
->msg_flags
|= MSG_NOTIFICATION
;
2105 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2107 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2110 /* Check if we allow SCTP_NXTINFO. */
2111 if (sp
->recvnxtinfo
)
2112 sctp_ulpevent_read_nxtinfo(event
, msg
, sk
);
2113 /* Check if we allow SCTP_RCVINFO. */
2114 if (sp
->recvrcvinfo
)
2115 sctp_ulpevent_read_rcvinfo(event
, msg
);
2116 /* Check if we allow SCTP_SNDRCVINFO. */
2117 if (sp
->subscribe
.sctp_data_io_event
)
2118 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2122 /* If skb's length exceeds the user's buffer, update the skb and
2123 * push it back to the receive_queue so that the next call to
2124 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2126 if (skb_len
> copied
) {
2127 msg
->msg_flags
&= ~MSG_EOR
;
2128 if (flags
& MSG_PEEK
)
2130 sctp_skb_pull(skb
, copied
);
2131 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2133 /* When only partial message is copied to the user, increase
2134 * rwnd by that amount. If all the data in the skb is read,
2135 * rwnd is updated when the event is freed.
2137 if (!sctp_ulpevent_is_notification(event
))
2138 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2140 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2141 (event
->msg_flags
& MSG_EOR
))
2142 msg
->msg_flags
|= MSG_EOR
;
2144 msg
->msg_flags
&= ~MSG_EOR
;
2147 if (flags
& MSG_PEEK
) {
2148 /* Release the skb reference acquired after peeking the skb in
2149 * sctp_skb_recv_datagram().
2153 /* Free the event which includes releasing the reference to
2154 * the owner of the skb, freeing the skb and updating the
2157 sctp_ulpevent_free(event
);
2164 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2166 * This option is a on/off flag. If enabled no SCTP message
2167 * fragmentation will be performed. Instead if a message being sent
2168 * exceeds the current PMTU size, the message will NOT be sent and
2169 * instead a error will be indicated to the user.
2171 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2172 char __user
*optval
,
2173 unsigned int optlen
)
2177 if (optlen
< sizeof(int))
2180 if (get_user(val
, (int __user
*)optval
))
2183 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2188 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2189 unsigned int optlen
)
2191 struct sctp_association
*asoc
;
2192 struct sctp_ulpevent
*event
;
2194 if (optlen
> sizeof(struct sctp_event_subscribe
))
2196 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2199 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2200 * if there is no data to be sent or retransmit, the stack will
2201 * immediately send up this notification.
2203 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2204 &sctp_sk(sk
)->subscribe
)) {
2205 asoc
= sctp_id2assoc(sk
, 0);
2207 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2208 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2213 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2220 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2222 * This socket option is applicable to the UDP-style socket only. When
2223 * set it will cause associations that are idle for more than the
2224 * specified number of seconds to automatically close. An association
2225 * being idle is defined an association that has NOT sent or received
2226 * user data. The special value of '0' indicates that no automatic
2227 * close of any associations should be performed. The option expects an
2228 * integer defining the number of seconds of idle time before an
2229 * association is closed.
2231 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2232 unsigned int optlen
)
2234 struct sctp_sock
*sp
= sctp_sk(sk
);
2235 struct net
*net
= sock_net(sk
);
2237 /* Applicable to UDP-style socket only */
2238 if (sctp_style(sk
, TCP
))
2240 if (optlen
!= sizeof(int))
2242 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2245 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2246 sp
->autoclose
= net
->sctp
.max_autoclose
;
2251 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2253 * Applications can enable or disable heartbeats for any peer address of
2254 * an association, modify an address's heartbeat interval, force a
2255 * heartbeat to be sent immediately, and adjust the address's maximum
2256 * number of retransmissions sent before an address is considered
2257 * unreachable. The following structure is used to access and modify an
2258 * address's parameters:
2260 * struct sctp_paddrparams {
2261 * sctp_assoc_t spp_assoc_id;
2262 * struct sockaddr_storage spp_address;
2263 * uint32_t spp_hbinterval;
2264 * uint16_t spp_pathmaxrxt;
2265 * uint32_t spp_pathmtu;
2266 * uint32_t spp_sackdelay;
2267 * uint32_t spp_flags;
2270 * spp_assoc_id - (one-to-many style socket) This is filled in the
2271 * application, and identifies the association for
2273 * spp_address - This specifies which address is of interest.
2274 * spp_hbinterval - This contains the value of the heartbeat interval,
2275 * in milliseconds. If a value of zero
2276 * is present in this field then no changes are to
2277 * be made to this parameter.
2278 * spp_pathmaxrxt - This contains the maximum number of
2279 * retransmissions before this address shall be
2280 * considered unreachable. If a value of zero
2281 * is present in this field then no changes are to
2282 * be made to this parameter.
2283 * spp_pathmtu - When Path MTU discovery is disabled the value
2284 * specified here will be the "fixed" path mtu.
2285 * Note that if the spp_address field is empty
2286 * then all associations on this address will
2287 * have this fixed path mtu set upon them.
2289 * spp_sackdelay - When delayed sack is enabled, this value specifies
2290 * the number of milliseconds that sacks will be delayed
2291 * for. This value will apply to all addresses of an
2292 * association if the spp_address field is empty. Note
2293 * also, that if delayed sack is enabled and this
2294 * value is set to 0, no change is made to the last
2295 * recorded delayed sack timer value.
2297 * spp_flags - These flags are used to control various features
2298 * on an association. The flag field may contain
2299 * zero or more of the following options.
2301 * SPP_HB_ENABLE - Enable heartbeats on the
2302 * specified address. Note that if the address
2303 * field is empty all addresses for the association
2304 * have heartbeats enabled upon them.
2306 * SPP_HB_DISABLE - Disable heartbeats on the
2307 * speicifed address. Note that if the address
2308 * field is empty all addresses for the association
2309 * will have their heartbeats disabled. Note also
2310 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2311 * mutually exclusive, only one of these two should
2312 * be specified. Enabling both fields will have
2313 * undetermined results.
2315 * SPP_HB_DEMAND - Request a user initiated heartbeat
2316 * to be made immediately.
2318 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2319 * heartbeat delayis to be set to the value of 0
2322 * SPP_PMTUD_ENABLE - This field will enable PMTU
2323 * discovery upon the specified address. Note that
2324 * if the address feild is empty then all addresses
2325 * on the association are effected.
2327 * SPP_PMTUD_DISABLE - This field will disable PMTU
2328 * discovery upon the specified address. Note that
2329 * if the address feild is empty then all addresses
2330 * on the association are effected. Not also that
2331 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2332 * exclusive. Enabling both will have undetermined
2335 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2336 * on delayed sack. The time specified in spp_sackdelay
2337 * is used to specify the sack delay for this address. Note
2338 * that if spp_address is empty then all addresses will
2339 * enable delayed sack and take on the sack delay
2340 * value specified in spp_sackdelay.
2341 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2342 * off delayed sack. If the spp_address field is blank then
2343 * delayed sack is disabled for the entire association. Note
2344 * also that this field is mutually exclusive to
2345 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2348 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2349 struct sctp_transport
*trans
,
2350 struct sctp_association
*asoc
,
2351 struct sctp_sock
*sp
,
2354 int sackdelay_change
)
2358 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2359 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2361 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2366 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2367 * this field is ignored. Note also that a value of zero indicates
2368 * the current setting should be left unchanged.
2370 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2372 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2373 * set. This lets us use 0 value when this flag
2376 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2377 params
->spp_hbinterval
= 0;
2379 if (params
->spp_hbinterval
||
2380 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2383 msecs_to_jiffies(params
->spp_hbinterval
);
2386 msecs_to_jiffies(params
->spp_hbinterval
);
2388 sp
->hbinterval
= params
->spp_hbinterval
;
2395 trans
->param_flags
=
2396 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2399 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2402 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2406 /* When Path MTU discovery is disabled the value specified here will
2407 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2408 * include the flag SPP_PMTUD_DISABLE for this field to have any
2411 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2413 trans
->pathmtu
= params
->spp_pathmtu
;
2414 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2416 asoc
->pathmtu
= params
->spp_pathmtu
;
2417 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2419 sp
->pathmtu
= params
->spp_pathmtu
;
2425 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2426 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2427 trans
->param_flags
=
2428 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2430 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2431 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2435 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2438 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2442 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2443 * value of this field is ignored. Note also that a value of zero
2444 * indicates the current setting should be left unchanged.
2446 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2449 msecs_to_jiffies(params
->spp_sackdelay
);
2452 msecs_to_jiffies(params
->spp_sackdelay
);
2454 sp
->sackdelay
= params
->spp_sackdelay
;
2458 if (sackdelay_change
) {
2460 trans
->param_flags
=
2461 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2465 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2469 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2474 /* Note that a value of zero indicates the current setting should be
2477 if (params
->spp_pathmaxrxt
) {
2479 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2481 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2483 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2490 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2491 char __user
*optval
,
2492 unsigned int optlen
)
2494 struct sctp_paddrparams params
;
2495 struct sctp_transport
*trans
= NULL
;
2496 struct sctp_association
*asoc
= NULL
;
2497 struct sctp_sock
*sp
= sctp_sk(sk
);
2499 int hb_change
, pmtud_change
, sackdelay_change
;
2501 if (optlen
!= sizeof(struct sctp_paddrparams
))
2504 if (copy_from_user(¶ms
, optval
, optlen
))
2507 /* Validate flags and value parameters. */
2508 hb_change
= params
.spp_flags
& SPP_HB
;
2509 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2510 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2512 if (hb_change
== SPP_HB
||
2513 pmtud_change
== SPP_PMTUD
||
2514 sackdelay_change
== SPP_SACKDELAY
||
2515 params
.spp_sackdelay
> 500 ||
2516 (params
.spp_pathmtu
&&
2517 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2520 /* If an address other than INADDR_ANY is specified, and
2521 * no transport is found, then the request is invalid.
2523 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2524 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2525 params
.spp_assoc_id
);
2530 /* Get association, if assoc_id != 0 and the socket is a one
2531 * to many style socket, and an association was not found, then
2532 * the id was invalid.
2534 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2535 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2538 /* Heartbeat demand can only be sent on a transport or
2539 * association, but not a socket.
2541 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2544 /* Process parameters. */
2545 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2546 hb_change
, pmtud_change
,
2552 /* If changes are for association, also apply parameters to each
2555 if (!trans
&& asoc
) {
2556 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2558 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2559 hb_change
, pmtud_change
,
2567 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2569 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2572 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2574 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2578 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2580 * This option will effect the way delayed acks are performed. This
2581 * option allows you to get or set the delayed ack time, in
2582 * milliseconds. It also allows changing the delayed ack frequency.
2583 * Changing the frequency to 1 disables the delayed sack algorithm. If
2584 * the assoc_id is 0, then this sets or gets the endpoints default
2585 * values. If the assoc_id field is non-zero, then the set or get
2586 * effects the specified association for the one to many model (the
2587 * assoc_id field is ignored by the one to one model). Note that if
2588 * sack_delay or sack_freq are 0 when setting this option, then the
2589 * current values will remain unchanged.
2591 * struct sctp_sack_info {
2592 * sctp_assoc_t sack_assoc_id;
2593 * uint32_t sack_delay;
2594 * uint32_t sack_freq;
2597 * sack_assoc_id - This parameter, indicates which association the user
2598 * is performing an action upon. Note that if this field's value is
2599 * zero then the endpoints default value is changed (effecting future
2600 * associations only).
2602 * sack_delay - This parameter contains the number of milliseconds that
2603 * the user is requesting the delayed ACK timer be set to. Note that
2604 * this value is defined in the standard to be between 200 and 500
2607 * sack_freq - This parameter contains the number of packets that must
2608 * be received before a sack is sent without waiting for the delay
2609 * timer to expire. The default value for this is 2, setting this
2610 * value to 1 will disable the delayed sack algorithm.
2613 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2614 char __user
*optval
, unsigned int optlen
)
2616 struct sctp_sack_info params
;
2617 struct sctp_transport
*trans
= NULL
;
2618 struct sctp_association
*asoc
= NULL
;
2619 struct sctp_sock
*sp
= sctp_sk(sk
);
2621 if (optlen
== sizeof(struct sctp_sack_info
)) {
2622 if (copy_from_user(¶ms
, optval
, optlen
))
2625 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2627 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2628 pr_warn_ratelimited(DEPRECATED
2630 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2631 "Use struct sctp_sack_info instead\n",
2632 current
->comm
, task_pid_nr(current
));
2633 if (copy_from_user(¶ms
, optval
, optlen
))
2636 if (params
.sack_delay
== 0)
2637 params
.sack_freq
= 1;
2639 params
.sack_freq
= 0;
2643 /* Validate value parameter. */
2644 if (params
.sack_delay
> 500)
2647 /* Get association, if sack_assoc_id != 0 and the socket is a one
2648 * to many style socket, and an association was not found, then
2649 * the id was invalid.
2651 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2652 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2655 if (params
.sack_delay
) {
2658 msecs_to_jiffies(params
.sack_delay
);
2660 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2662 sp
->sackdelay
= params
.sack_delay
;
2664 sctp_spp_sackdelay_enable(sp
->param_flags
);
2668 if (params
.sack_freq
== 1) {
2671 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2674 sctp_spp_sackdelay_disable(sp
->param_flags
);
2676 } else if (params
.sack_freq
> 1) {
2678 asoc
->sackfreq
= params
.sack_freq
;
2680 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2682 sp
->sackfreq
= params
.sack_freq
;
2684 sctp_spp_sackdelay_enable(sp
->param_flags
);
2688 /* If change is for association, also apply to each transport. */
2690 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2692 if (params
.sack_delay
) {
2694 msecs_to_jiffies(params
.sack_delay
);
2695 trans
->param_flags
=
2696 sctp_spp_sackdelay_enable(trans
->param_flags
);
2698 if (params
.sack_freq
== 1) {
2699 trans
->param_flags
=
2700 sctp_spp_sackdelay_disable(trans
->param_flags
);
2701 } else if (params
.sack_freq
> 1) {
2702 trans
->sackfreq
= params
.sack_freq
;
2703 trans
->param_flags
=
2704 sctp_spp_sackdelay_enable(trans
->param_flags
);
2712 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2714 * Applications can specify protocol parameters for the default association
2715 * initialization. The option name argument to setsockopt() and getsockopt()
2718 * Setting initialization parameters is effective only on an unconnected
2719 * socket (for UDP-style sockets only future associations are effected
2720 * by the change). With TCP-style sockets, this option is inherited by
2721 * sockets derived from a listener socket.
2723 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2725 struct sctp_initmsg sinit
;
2726 struct sctp_sock
*sp
= sctp_sk(sk
);
2728 if (optlen
!= sizeof(struct sctp_initmsg
))
2730 if (copy_from_user(&sinit
, optval
, optlen
))
2733 if (sinit
.sinit_num_ostreams
)
2734 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2735 if (sinit
.sinit_max_instreams
)
2736 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2737 if (sinit
.sinit_max_attempts
)
2738 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2739 if (sinit
.sinit_max_init_timeo
)
2740 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2746 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2748 * Applications that wish to use the sendto() system call may wish to
2749 * specify a default set of parameters that would normally be supplied
2750 * through the inclusion of ancillary data. This socket option allows
2751 * such an application to set the default sctp_sndrcvinfo structure.
2752 * The application that wishes to use this socket option simply passes
2753 * in to this call the sctp_sndrcvinfo structure defined in Section
2754 * 5.2.2) The input parameters accepted by this call include
2755 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2756 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2757 * to this call if the caller is using the UDP model.
2759 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2760 char __user
*optval
,
2761 unsigned int optlen
)
2763 struct sctp_sock
*sp
= sctp_sk(sk
);
2764 struct sctp_association
*asoc
;
2765 struct sctp_sndrcvinfo info
;
2767 if (optlen
!= sizeof(info
))
2769 if (copy_from_user(&info
, optval
, optlen
))
2771 if (info
.sinfo_flags
&
2772 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2773 SCTP_ABORT
| SCTP_EOF
))
2776 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2777 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2780 asoc
->default_stream
= info
.sinfo_stream
;
2781 asoc
->default_flags
= info
.sinfo_flags
;
2782 asoc
->default_ppid
= info
.sinfo_ppid
;
2783 asoc
->default_context
= info
.sinfo_context
;
2784 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2786 sp
->default_stream
= info
.sinfo_stream
;
2787 sp
->default_flags
= info
.sinfo_flags
;
2788 sp
->default_ppid
= info
.sinfo_ppid
;
2789 sp
->default_context
= info
.sinfo_context
;
2790 sp
->default_timetolive
= info
.sinfo_timetolive
;
2796 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2797 * (SCTP_DEFAULT_SNDINFO)
2799 static int sctp_setsockopt_default_sndinfo(struct sock
*sk
,
2800 char __user
*optval
,
2801 unsigned int optlen
)
2803 struct sctp_sock
*sp
= sctp_sk(sk
);
2804 struct sctp_association
*asoc
;
2805 struct sctp_sndinfo info
;
2807 if (optlen
!= sizeof(info
))
2809 if (copy_from_user(&info
, optval
, optlen
))
2811 if (info
.snd_flags
&
2812 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2813 SCTP_ABORT
| SCTP_EOF
))
2816 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
2817 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
2820 asoc
->default_stream
= info
.snd_sid
;
2821 asoc
->default_flags
= info
.snd_flags
;
2822 asoc
->default_ppid
= info
.snd_ppid
;
2823 asoc
->default_context
= info
.snd_context
;
2825 sp
->default_stream
= info
.snd_sid
;
2826 sp
->default_flags
= info
.snd_flags
;
2827 sp
->default_ppid
= info
.snd_ppid
;
2828 sp
->default_context
= info
.snd_context
;
2834 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2836 * Requests that the local SCTP stack use the enclosed peer address as
2837 * the association primary. The enclosed address must be one of the
2838 * association peer's addresses.
2840 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2841 unsigned int optlen
)
2843 struct sctp_prim prim
;
2844 struct sctp_transport
*trans
;
2846 if (optlen
!= sizeof(struct sctp_prim
))
2849 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2852 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2856 sctp_assoc_set_primary(trans
->asoc
, trans
);
2862 * 7.1.5 SCTP_NODELAY
2864 * Turn on/off any Nagle-like algorithm. This means that packets are
2865 * generally sent as soon as possible and no unnecessary delays are
2866 * introduced, at the cost of more packets in the network. Expects an
2867 * integer boolean flag.
2869 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2870 unsigned int optlen
)
2874 if (optlen
< sizeof(int))
2876 if (get_user(val
, (int __user
*)optval
))
2879 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2885 * 7.1.1 SCTP_RTOINFO
2887 * The protocol parameters used to initialize and bound retransmission
2888 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2889 * and modify these parameters.
2890 * All parameters are time values, in milliseconds. A value of 0, when
2891 * modifying the parameters, indicates that the current value should not
2895 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2897 struct sctp_rtoinfo rtoinfo
;
2898 struct sctp_association
*asoc
;
2899 unsigned long rto_min
, rto_max
;
2900 struct sctp_sock
*sp
= sctp_sk(sk
);
2902 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2905 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2908 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2910 /* Set the values to the specific association */
2911 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2914 rto_max
= rtoinfo
.srto_max
;
2915 rto_min
= rtoinfo
.srto_min
;
2918 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2920 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2923 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2925 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2927 if (rto_min
> rto_max
)
2931 if (rtoinfo
.srto_initial
!= 0)
2933 msecs_to_jiffies(rtoinfo
.srto_initial
);
2934 asoc
->rto_max
= rto_max
;
2935 asoc
->rto_min
= rto_min
;
2937 /* If there is no association or the association-id = 0
2938 * set the values to the endpoint.
2940 if (rtoinfo
.srto_initial
!= 0)
2941 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2942 sp
->rtoinfo
.srto_max
= rto_max
;
2943 sp
->rtoinfo
.srto_min
= rto_min
;
2951 * 7.1.2 SCTP_ASSOCINFO
2953 * This option is used to tune the maximum retransmission attempts
2954 * of the association.
2955 * Returns an error if the new association retransmission value is
2956 * greater than the sum of the retransmission value of the peer.
2957 * See [SCTP] for more information.
2960 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2963 struct sctp_assocparams assocparams
;
2964 struct sctp_association
*asoc
;
2966 if (optlen
!= sizeof(struct sctp_assocparams
))
2968 if (copy_from_user(&assocparams
, optval
, optlen
))
2971 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2973 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2976 /* Set the values to the specific association */
2978 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2981 struct sctp_transport
*peer_addr
;
2983 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2985 path_sum
+= peer_addr
->pathmaxrxt
;
2989 /* Only validate asocmaxrxt if we have more than
2990 * one path/transport. We do this because path
2991 * retransmissions are only counted when we have more
2995 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2998 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
3001 if (assocparams
.sasoc_cookie_life
!= 0)
3002 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
3004 /* Set the values to the endpoint */
3005 struct sctp_sock
*sp
= sctp_sk(sk
);
3007 if (assocparams
.sasoc_asocmaxrxt
!= 0)
3008 sp
->assocparams
.sasoc_asocmaxrxt
=
3009 assocparams
.sasoc_asocmaxrxt
;
3010 if (assocparams
.sasoc_cookie_life
!= 0)
3011 sp
->assocparams
.sasoc_cookie_life
=
3012 assocparams
.sasoc_cookie_life
;
3018 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3020 * This socket option is a boolean flag which turns on or off mapped V4
3021 * addresses. If this option is turned on and the socket is type
3022 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3023 * If this option is turned off, then no mapping will be done of V4
3024 * addresses and a user will receive both PF_INET6 and PF_INET type
3025 * addresses on the socket.
3027 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3030 struct sctp_sock
*sp
= sctp_sk(sk
);
3032 if (optlen
< sizeof(int))
3034 if (get_user(val
, (int __user
*)optval
))
3045 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3046 * This option will get or set the maximum size to put in any outgoing
3047 * SCTP DATA chunk. If a message is larger than this size it will be
3048 * fragmented by SCTP into the specified size. Note that the underlying
3049 * SCTP implementation may fragment into smaller sized chunks when the
3050 * PMTU of the underlying association is smaller than the value set by
3051 * the user. The default value for this option is '0' which indicates
3052 * the user is NOT limiting fragmentation and only the PMTU will effect
3053 * SCTP's choice of DATA chunk size. Note also that values set larger
3054 * than the maximum size of an IP datagram will effectively let SCTP
3055 * control fragmentation (i.e. the same as setting this option to 0).
3057 * The following structure is used to access and modify this parameter:
3059 * struct sctp_assoc_value {
3060 * sctp_assoc_t assoc_id;
3061 * uint32_t assoc_value;
3064 * assoc_id: This parameter is ignored for one-to-one style sockets.
3065 * For one-to-many style sockets this parameter indicates which
3066 * association the user is performing an action upon. Note that if
3067 * this field's value is zero then the endpoints default value is
3068 * changed (effecting future associations only).
3069 * assoc_value: This parameter specifies the maximum size in bytes.
3071 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3073 struct sctp_assoc_value params
;
3074 struct sctp_association
*asoc
;
3075 struct sctp_sock
*sp
= sctp_sk(sk
);
3078 if (optlen
== sizeof(int)) {
3079 pr_warn_ratelimited(DEPRECATED
3081 "Use of int in maxseg socket option.\n"
3082 "Use struct sctp_assoc_value instead\n",
3083 current
->comm
, task_pid_nr(current
));
3084 if (copy_from_user(&val
, optval
, optlen
))
3086 params
.assoc_id
= 0;
3087 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3088 if (copy_from_user(¶ms
, optval
, optlen
))
3090 val
= params
.assoc_value
;
3094 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3097 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3098 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3103 val
= asoc
->pathmtu
;
3104 val
-= sp
->pf
->af
->net_header_len
;
3105 val
-= sizeof(struct sctphdr
) +
3106 sizeof(struct sctp_data_chunk
);
3108 asoc
->user_frag
= val
;
3109 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3111 sp
->user_frag
= val
;
3119 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3121 * Requests that the peer mark the enclosed address as the association
3122 * primary. The enclosed address must be one of the association's
3123 * locally bound addresses. The following structure is used to make a
3124 * set primary request:
3126 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3127 unsigned int optlen
)
3129 struct net
*net
= sock_net(sk
);
3130 struct sctp_sock
*sp
;
3131 struct sctp_association
*asoc
= NULL
;
3132 struct sctp_setpeerprim prim
;
3133 struct sctp_chunk
*chunk
;
3139 if (!net
->sctp
.addip_enable
)
3142 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3145 if (copy_from_user(&prim
, optval
, optlen
))
3148 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3152 if (!asoc
->peer
.asconf_capable
)
3155 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3158 if (!sctp_state(asoc
, ESTABLISHED
))
3161 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3165 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3166 return -EADDRNOTAVAIL
;
3168 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3169 return -EADDRNOTAVAIL
;
3171 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3172 chunk
= sctp_make_asconf_set_prim(asoc
,
3173 (union sctp_addr
*)&prim
.sspp_addr
);
3177 err
= sctp_send_asconf(asoc
, chunk
);
3179 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3184 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3185 unsigned int optlen
)
3187 struct sctp_setadaptation adaptation
;
3189 if (optlen
!= sizeof(struct sctp_setadaptation
))
3191 if (copy_from_user(&adaptation
, optval
, optlen
))
3194 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3200 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3202 * The context field in the sctp_sndrcvinfo structure is normally only
3203 * used when a failed message is retrieved holding the value that was
3204 * sent down on the actual send call. This option allows the setting of
3205 * a default context on an association basis that will be received on
3206 * reading messages from the peer. This is especially helpful in the
3207 * one-2-many model for an application to keep some reference to an
3208 * internal state machine that is processing messages on the
3209 * association. Note that the setting of this value only effects
3210 * received messages from the peer and does not effect the value that is
3211 * saved with outbound messages.
3213 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3214 unsigned int optlen
)
3216 struct sctp_assoc_value params
;
3217 struct sctp_sock
*sp
;
3218 struct sctp_association
*asoc
;
3220 if (optlen
!= sizeof(struct sctp_assoc_value
))
3222 if (copy_from_user(¶ms
, optval
, optlen
))
3227 if (params
.assoc_id
!= 0) {
3228 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3231 asoc
->default_rcv_context
= params
.assoc_value
;
3233 sp
->default_rcv_context
= params
.assoc_value
;
3240 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3242 * This options will at a minimum specify if the implementation is doing
3243 * fragmented interleave. Fragmented interleave, for a one to many
3244 * socket, is when subsequent calls to receive a message may return
3245 * parts of messages from different associations. Some implementations
3246 * may allow you to turn this value on or off. If so, when turned off,
3247 * no fragment interleave will occur (which will cause a head of line
3248 * blocking amongst multiple associations sharing the same one to many
3249 * socket). When this option is turned on, then each receive call may
3250 * come from a different association (thus the user must receive data
3251 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3252 * association each receive belongs to.
3254 * This option takes a boolean value. A non-zero value indicates that
3255 * fragmented interleave is on. A value of zero indicates that
3256 * fragmented interleave is off.
3258 * Note that it is important that an implementation that allows this
3259 * option to be turned on, have it off by default. Otherwise an unaware
3260 * application using the one to many model may become confused and act
3263 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3264 char __user
*optval
,
3265 unsigned int optlen
)
3269 if (optlen
!= sizeof(int))
3271 if (get_user(val
, (int __user
*)optval
))
3274 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3280 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3281 * (SCTP_PARTIAL_DELIVERY_POINT)
3283 * This option will set or get the SCTP partial delivery point. This
3284 * point is the size of a message where the partial delivery API will be
3285 * invoked to help free up rwnd space for the peer. Setting this to a
3286 * lower value will cause partial deliveries to happen more often. The
3287 * calls argument is an integer that sets or gets the partial delivery
3288 * point. Note also that the call will fail if the user attempts to set
3289 * this value larger than the socket receive buffer size.
3291 * Note that any single message having a length smaller than or equal to
3292 * the SCTP partial delivery point will be delivered in one single read
3293 * call as long as the user provided buffer is large enough to hold the
3296 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3297 char __user
*optval
,
3298 unsigned int optlen
)
3302 if (optlen
!= sizeof(u32
))
3304 if (get_user(val
, (int __user
*)optval
))
3307 /* Note: We double the receive buffer from what the user sets
3308 * it to be, also initial rwnd is based on rcvbuf/2.
3310 if (val
> (sk
->sk_rcvbuf
>> 1))
3313 sctp_sk(sk
)->pd_point
= val
;
3315 return 0; /* is this the right error code? */
3319 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3321 * This option will allow a user to change the maximum burst of packets
3322 * that can be emitted by this association. Note that the default value
3323 * is 4, and some implementations may restrict this setting so that it
3324 * can only be lowered.
3326 * NOTE: This text doesn't seem right. Do this on a socket basis with
3327 * future associations inheriting the socket value.
3329 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3330 char __user
*optval
,
3331 unsigned int optlen
)
3333 struct sctp_assoc_value params
;
3334 struct sctp_sock
*sp
;
3335 struct sctp_association
*asoc
;
3339 if (optlen
== sizeof(int)) {
3340 pr_warn_ratelimited(DEPRECATED
3342 "Use of int in max_burst socket option deprecated.\n"
3343 "Use struct sctp_assoc_value instead\n",
3344 current
->comm
, task_pid_nr(current
));
3345 if (copy_from_user(&val
, optval
, optlen
))
3347 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3348 if (copy_from_user(¶ms
, optval
, optlen
))
3350 val
= params
.assoc_value
;
3351 assoc_id
= params
.assoc_id
;
3357 if (assoc_id
!= 0) {
3358 asoc
= sctp_id2assoc(sk
, assoc_id
);
3361 asoc
->max_burst
= val
;
3363 sp
->max_burst
= val
;
3369 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3371 * This set option adds a chunk type that the user is requesting to be
3372 * received only in an authenticated way. Changes to the list of chunks
3373 * will only effect future associations on the socket.
3375 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3376 char __user
*optval
,
3377 unsigned int optlen
)
3379 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3380 struct sctp_authchunk val
;
3382 if (!ep
->auth_enable
)
3385 if (optlen
!= sizeof(struct sctp_authchunk
))
3387 if (copy_from_user(&val
, optval
, optlen
))
3390 switch (val
.sauth_chunk
) {
3392 case SCTP_CID_INIT_ACK
:
3393 case SCTP_CID_SHUTDOWN_COMPLETE
:
3398 /* add this chunk id to the endpoint */
3399 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3403 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3405 * This option gets or sets the list of HMAC algorithms that the local
3406 * endpoint requires the peer to use.
3408 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3409 char __user
*optval
,
3410 unsigned int optlen
)
3412 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3413 struct sctp_hmacalgo
*hmacs
;
3417 if (!ep
->auth_enable
)
3420 if (optlen
< sizeof(struct sctp_hmacalgo
))
3423 hmacs
= memdup_user(optval
, optlen
);
3425 return PTR_ERR(hmacs
);
3427 idents
= hmacs
->shmac_num_idents
;
3428 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3429 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3434 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3441 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3443 * This option will set a shared secret key which is used to build an
3444 * association shared key.
3446 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3447 char __user
*optval
,
3448 unsigned int optlen
)
3450 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3451 struct sctp_authkey
*authkey
;
3452 struct sctp_association
*asoc
;
3455 if (!ep
->auth_enable
)
3458 if (optlen
<= sizeof(struct sctp_authkey
))
3461 authkey
= memdup_user(optval
, optlen
);
3462 if (IS_ERR(authkey
))
3463 return PTR_ERR(authkey
);
3465 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3470 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3471 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3476 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3483 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3485 * This option will get or set the active shared key to be used to build
3486 * the association shared key.
3488 static int sctp_setsockopt_active_key(struct sock
*sk
,
3489 char __user
*optval
,
3490 unsigned int optlen
)
3492 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3493 struct sctp_authkeyid val
;
3494 struct sctp_association
*asoc
;
3496 if (!ep
->auth_enable
)
3499 if (optlen
!= sizeof(struct sctp_authkeyid
))
3501 if (copy_from_user(&val
, optval
, optlen
))
3504 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3505 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3508 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3512 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3514 * This set option will delete a shared secret key from use.
3516 static int sctp_setsockopt_del_key(struct sock
*sk
,
3517 char __user
*optval
,
3518 unsigned int optlen
)
3520 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3521 struct sctp_authkeyid val
;
3522 struct sctp_association
*asoc
;
3524 if (!ep
->auth_enable
)
3527 if (optlen
!= sizeof(struct sctp_authkeyid
))
3529 if (copy_from_user(&val
, optval
, optlen
))
3532 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3533 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3536 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3541 * 8.1.23 SCTP_AUTO_ASCONF
3543 * This option will enable or disable the use of the automatic generation of
3544 * ASCONF chunks to add and delete addresses to an existing association. Note
3545 * that this option has two caveats namely: a) it only affects sockets that
3546 * are bound to all addresses available to the SCTP stack, and b) the system
3547 * administrator may have an overriding control that turns the ASCONF feature
3548 * off no matter what setting the socket option may have.
3549 * This option expects an integer boolean flag, where a non-zero value turns on
3550 * the option, and a zero value turns off the option.
3551 * Note. In this implementation, socket operation overrides default parameter
3552 * being set by sysctl as well as FreeBSD implementation
3554 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3555 unsigned int optlen
)
3558 struct sctp_sock
*sp
= sctp_sk(sk
);
3560 if (optlen
< sizeof(int))
3562 if (get_user(val
, (int __user
*)optval
))
3564 if (!sctp_is_ep_boundall(sk
) && val
)
3566 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3569 spin_lock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3570 if (val
== 0 && sp
->do_auto_asconf
) {
3571 list_del(&sp
->auto_asconf_list
);
3572 sp
->do_auto_asconf
= 0;
3573 } else if (val
&& !sp
->do_auto_asconf
) {
3574 list_add_tail(&sp
->auto_asconf_list
,
3575 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3576 sp
->do_auto_asconf
= 1;
3578 spin_unlock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3583 * SCTP_PEER_ADDR_THLDS
3585 * This option allows us to alter the partially failed threshold for one or all
3586 * transports in an association. See Section 6.1 of:
3587 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3589 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3590 char __user
*optval
,
3591 unsigned int optlen
)
3593 struct sctp_paddrthlds val
;
3594 struct sctp_transport
*trans
;
3595 struct sctp_association
*asoc
;
3597 if (optlen
< sizeof(struct sctp_paddrthlds
))
3599 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3600 sizeof(struct sctp_paddrthlds
)))
3604 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3605 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3608 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3610 if (val
.spt_pathmaxrxt
)
3611 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3612 trans
->pf_retrans
= val
.spt_pathpfthld
;
3615 if (val
.spt_pathmaxrxt
)
3616 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3617 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3619 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3624 if (val
.spt_pathmaxrxt
)
3625 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3626 trans
->pf_retrans
= val
.spt_pathpfthld
;
3632 static int sctp_setsockopt_recvrcvinfo(struct sock
*sk
,
3633 char __user
*optval
,
3634 unsigned int optlen
)
3638 if (optlen
< sizeof(int))
3640 if (get_user(val
, (int __user
*) optval
))
3643 sctp_sk(sk
)->recvrcvinfo
= (val
== 0) ? 0 : 1;
3648 static int sctp_setsockopt_recvnxtinfo(struct sock
*sk
,
3649 char __user
*optval
,
3650 unsigned int optlen
)
3654 if (optlen
< sizeof(int))
3656 if (get_user(val
, (int __user
*) optval
))
3659 sctp_sk(sk
)->recvnxtinfo
= (val
== 0) ? 0 : 1;
3664 static int sctp_setsockopt_pr_supported(struct sock
*sk
,
3665 char __user
*optval
,
3666 unsigned int optlen
)
3668 struct sctp_assoc_value params
;
3669 struct sctp_association
*asoc
;
3670 int retval
= -EINVAL
;
3672 if (optlen
!= sizeof(params
))
3675 if (copy_from_user(¶ms
, optval
, optlen
)) {
3680 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3682 asoc
->prsctp_enable
= !!params
.assoc_value
;
3683 } else if (!params
.assoc_id
) {
3684 struct sctp_sock
*sp
= sctp_sk(sk
);
3686 sp
->ep
->prsctp_enable
= !!params
.assoc_value
;
3697 static int sctp_setsockopt_default_prinfo(struct sock
*sk
,
3698 char __user
*optval
,
3699 unsigned int optlen
)
3701 struct sctp_default_prinfo info
;
3702 struct sctp_association
*asoc
;
3703 int retval
= -EINVAL
;
3705 if (optlen
!= sizeof(info
))
3708 if (copy_from_user(&info
, optval
, sizeof(info
))) {
3713 if (info
.pr_policy
& ~SCTP_PR_SCTP_MASK
)
3716 if (info
.pr_policy
== SCTP_PR_SCTP_NONE
)
3719 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
3721 SCTP_PR_SET_POLICY(asoc
->default_flags
, info
.pr_policy
);
3722 asoc
->default_timetolive
= info
.pr_value
;
3723 } else if (!info
.pr_assoc_id
) {
3724 struct sctp_sock
*sp
= sctp_sk(sk
);
3726 SCTP_PR_SET_POLICY(sp
->default_flags
, info
.pr_policy
);
3727 sp
->default_timetolive
= info
.pr_value
;
3738 /* API 6.2 setsockopt(), getsockopt()
3740 * Applications use setsockopt() and getsockopt() to set or retrieve
3741 * socket options. Socket options are used to change the default
3742 * behavior of sockets calls. They are described in Section 7.
3746 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3747 * int __user *optlen);
3748 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3751 * sd - the socket descript.
3752 * level - set to IPPROTO_SCTP for all SCTP options.
3753 * optname - the option name.
3754 * optval - the buffer to store the value of the option.
3755 * optlen - the size of the buffer.
3757 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3758 char __user
*optval
, unsigned int optlen
)
3762 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3764 /* I can hardly begin to describe how wrong this is. This is
3765 * so broken as to be worse than useless. The API draft
3766 * REALLY is NOT helpful here... I am not convinced that the
3767 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3768 * are at all well-founded.
3770 if (level
!= SOL_SCTP
) {
3771 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3772 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3779 case SCTP_SOCKOPT_BINDX_ADD
:
3780 /* 'optlen' is the size of the addresses buffer. */
3781 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3782 optlen
, SCTP_BINDX_ADD_ADDR
);
3785 case SCTP_SOCKOPT_BINDX_REM
:
3786 /* 'optlen' is the size of the addresses buffer. */
3787 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3788 optlen
, SCTP_BINDX_REM_ADDR
);
3791 case SCTP_SOCKOPT_CONNECTX_OLD
:
3792 /* 'optlen' is the size of the addresses buffer. */
3793 retval
= sctp_setsockopt_connectx_old(sk
,
3794 (struct sockaddr __user
*)optval
,
3798 case SCTP_SOCKOPT_CONNECTX
:
3799 /* 'optlen' is the size of the addresses buffer. */
3800 retval
= sctp_setsockopt_connectx(sk
,
3801 (struct sockaddr __user
*)optval
,
3805 case SCTP_DISABLE_FRAGMENTS
:
3806 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3810 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3813 case SCTP_AUTOCLOSE
:
3814 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3817 case SCTP_PEER_ADDR_PARAMS
:
3818 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3821 case SCTP_DELAYED_SACK
:
3822 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3824 case SCTP_PARTIAL_DELIVERY_POINT
:
3825 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3829 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3831 case SCTP_DEFAULT_SEND_PARAM
:
3832 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3835 case SCTP_DEFAULT_SNDINFO
:
3836 retval
= sctp_setsockopt_default_sndinfo(sk
, optval
, optlen
);
3838 case SCTP_PRIMARY_ADDR
:
3839 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3841 case SCTP_SET_PEER_PRIMARY_ADDR
:
3842 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3845 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3848 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3850 case SCTP_ASSOCINFO
:
3851 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3853 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3854 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3857 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3859 case SCTP_ADAPTATION_LAYER
:
3860 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3863 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3865 case SCTP_FRAGMENT_INTERLEAVE
:
3866 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3868 case SCTP_MAX_BURST
:
3869 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3871 case SCTP_AUTH_CHUNK
:
3872 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3874 case SCTP_HMAC_IDENT
:
3875 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3878 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3880 case SCTP_AUTH_ACTIVE_KEY
:
3881 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3883 case SCTP_AUTH_DELETE_KEY
:
3884 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3886 case SCTP_AUTO_ASCONF
:
3887 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3889 case SCTP_PEER_ADDR_THLDS
:
3890 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3892 case SCTP_RECVRCVINFO
:
3893 retval
= sctp_setsockopt_recvrcvinfo(sk
, optval
, optlen
);
3895 case SCTP_RECVNXTINFO
:
3896 retval
= sctp_setsockopt_recvnxtinfo(sk
, optval
, optlen
);
3898 case SCTP_PR_SUPPORTED
:
3899 retval
= sctp_setsockopt_pr_supported(sk
, optval
, optlen
);
3901 case SCTP_DEFAULT_PRINFO
:
3902 retval
= sctp_setsockopt_default_prinfo(sk
, optval
, optlen
);
3905 retval
= -ENOPROTOOPT
;
3915 /* API 3.1.6 connect() - UDP Style Syntax
3917 * An application may use the connect() call in the UDP model to initiate an
3918 * association without sending data.
3922 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3924 * sd: the socket descriptor to have a new association added to.
3926 * nam: the address structure (either struct sockaddr_in or struct
3927 * sockaddr_in6 defined in RFC2553 [7]).
3929 * len: the size of the address.
3931 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3939 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
3942 /* Validate addr_len before calling common connect/connectx routine. */
3943 af
= sctp_get_af_specific(addr
->sa_family
);
3944 if (!af
|| addr_len
< af
->sockaddr_len
) {
3947 /* Pass correct addr len to common routine (so it knows there
3948 * is only one address being passed.
3950 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3957 /* FIXME: Write comments. */
3958 static int sctp_disconnect(struct sock
*sk
, int flags
)
3960 return -EOPNOTSUPP
; /* STUB */
3963 /* 4.1.4 accept() - TCP Style Syntax
3965 * Applications use accept() call to remove an established SCTP
3966 * association from the accept queue of the endpoint. A new socket
3967 * descriptor will be returned from accept() to represent the newly
3968 * formed association.
3970 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3972 struct sctp_sock
*sp
;
3973 struct sctp_endpoint
*ep
;
3974 struct sock
*newsk
= NULL
;
3975 struct sctp_association
*asoc
;
3984 if (!sctp_style(sk
, TCP
)) {
3985 error
= -EOPNOTSUPP
;
3989 if (!sctp_sstate(sk
, LISTENING
)) {
3994 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3996 error
= sctp_wait_for_accept(sk
, timeo
);
4000 /* We treat the list of associations on the endpoint as the accept
4001 * queue and pick the first association on the list.
4003 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
4005 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
4011 /* Populate the fields of the newsk from the oldsk and migrate the
4012 * asoc to the newsk.
4014 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
4022 /* The SCTP ioctl handler. */
4023 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
4030 * SEQPACKET-style sockets in LISTENING state are valid, for
4031 * SCTP, so only discard TCP-style sockets in LISTENING state.
4033 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4038 struct sk_buff
*skb
;
4039 unsigned int amount
= 0;
4041 skb
= skb_peek(&sk
->sk_receive_queue
);
4044 * We will only return the amount of this packet since
4045 * that is all that will be read.
4049 rc
= put_user(amount
, (int __user
*)arg
);
4061 /* This is the function which gets called during socket creation to
4062 * initialized the SCTP-specific portion of the sock.
4063 * The sock structure should already be zero-filled memory.
4065 static int sctp_init_sock(struct sock
*sk
)
4067 struct net
*net
= sock_net(sk
);
4068 struct sctp_sock
*sp
;
4070 pr_debug("%s: sk:%p\n", __func__
, sk
);
4074 /* Initialize the SCTP per socket area. */
4075 switch (sk
->sk_type
) {
4076 case SOCK_SEQPACKET
:
4077 sp
->type
= SCTP_SOCKET_UDP
;
4080 sp
->type
= SCTP_SOCKET_TCP
;
4083 return -ESOCKTNOSUPPORT
;
4086 sk
->sk_gso_type
= SKB_GSO_SCTP
;
4088 /* Initialize default send parameters. These parameters can be
4089 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4091 sp
->default_stream
= 0;
4092 sp
->default_ppid
= 0;
4093 sp
->default_flags
= 0;
4094 sp
->default_context
= 0;
4095 sp
->default_timetolive
= 0;
4097 sp
->default_rcv_context
= 0;
4098 sp
->max_burst
= net
->sctp
.max_burst
;
4100 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
4102 /* Initialize default setup parameters. These parameters
4103 * can be modified with the SCTP_INITMSG socket option or
4104 * overridden by the SCTP_INIT CMSG.
4106 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
4107 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
4108 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
4109 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
4111 /* Initialize default RTO related parameters. These parameters can
4112 * be modified for with the SCTP_RTOINFO socket option.
4114 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
4115 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
4116 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
4118 /* Initialize default association related parameters. These parameters
4119 * can be modified with the SCTP_ASSOCINFO socket option.
4121 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
4122 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
4123 sp
->assocparams
.sasoc_peer_rwnd
= 0;
4124 sp
->assocparams
.sasoc_local_rwnd
= 0;
4125 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
4127 /* Initialize default event subscriptions. By default, all the
4130 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
4132 /* Default Peer Address Parameters. These defaults can
4133 * be modified via SCTP_PEER_ADDR_PARAMS
4135 sp
->hbinterval
= net
->sctp
.hb_interval
;
4136 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
4137 sp
->pathmtu
= 0; /* allow default discovery */
4138 sp
->sackdelay
= net
->sctp
.sack_timeout
;
4140 sp
->param_flags
= SPP_HB_ENABLE
|
4142 SPP_SACKDELAY_ENABLE
;
4144 /* If enabled no SCTP message fragmentation will be performed.
4145 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4147 sp
->disable_fragments
= 0;
4149 /* Enable Nagle algorithm by default. */
4152 sp
->recvrcvinfo
= 0;
4153 sp
->recvnxtinfo
= 0;
4155 /* Enable by default. */
4158 /* Auto-close idle associations after the configured
4159 * number of seconds. A value of 0 disables this
4160 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4161 * for UDP-style sockets only.
4165 /* User specified fragmentation limit. */
4168 sp
->adaptation_ind
= 0;
4170 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
4172 /* Control variables for partial data delivery. */
4173 atomic_set(&sp
->pd_mode
, 0);
4174 skb_queue_head_init(&sp
->pd_lobby
);
4175 sp
->frag_interleave
= 0;
4177 /* Create a per socket endpoint structure. Even if we
4178 * change the data structure relationships, this may still
4179 * be useful for storing pre-connect address information.
4181 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4187 sk
->sk_destruct
= sctp_destruct_sock
;
4189 SCTP_DBG_OBJCNT_INC(sock
);
4192 percpu_counter_inc(&sctp_sockets_allocated
);
4193 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4195 /* Nothing can fail after this block, otherwise
4196 * sctp_destroy_sock() will be called without addr_wq_lock held
4198 if (net
->sctp
.default_auto_asconf
) {
4199 spin_lock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4200 list_add_tail(&sp
->auto_asconf_list
,
4201 &net
->sctp
.auto_asconf_splist
);
4202 sp
->do_auto_asconf
= 1;
4203 spin_unlock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4205 sp
->do_auto_asconf
= 0;
4213 /* Cleanup any SCTP per socket resources. Must be called with
4214 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4216 static void sctp_destroy_sock(struct sock
*sk
)
4218 struct sctp_sock
*sp
;
4220 pr_debug("%s: sk:%p\n", __func__
, sk
);
4222 /* Release our hold on the endpoint. */
4224 /* This could happen during socket init, thus we bail out
4225 * early, since the rest of the below is not setup either.
4230 if (sp
->do_auto_asconf
) {
4231 sp
->do_auto_asconf
= 0;
4232 list_del(&sp
->auto_asconf_list
);
4234 sctp_endpoint_free(sp
->ep
);
4236 percpu_counter_dec(&sctp_sockets_allocated
);
4237 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4241 /* Triggered when there are no references on the socket anymore */
4242 static void sctp_destruct_sock(struct sock
*sk
)
4244 struct sctp_sock
*sp
= sctp_sk(sk
);
4246 /* Free up the HMAC transform. */
4247 crypto_free_shash(sp
->hmac
);
4249 inet_sock_destruct(sk
);
4252 /* API 4.1.7 shutdown() - TCP Style Syntax
4253 * int shutdown(int socket, int how);
4255 * sd - the socket descriptor of the association to be closed.
4256 * how - Specifies the type of shutdown. The values are
4259 * Disables further receive operations. No SCTP
4260 * protocol action is taken.
4262 * Disables further send operations, and initiates
4263 * the SCTP shutdown sequence.
4265 * Disables further send and receive operations
4266 * and initiates the SCTP shutdown sequence.
4268 static void sctp_shutdown(struct sock
*sk
, int how
)
4270 struct net
*net
= sock_net(sk
);
4271 struct sctp_endpoint
*ep
;
4272 struct sctp_association
*asoc
;
4274 if (!sctp_style(sk
, TCP
))
4277 if (how
& SEND_SHUTDOWN
) {
4278 sk
->sk_state
= SCTP_SS_CLOSING
;
4279 ep
= sctp_sk(sk
)->ep
;
4280 if (!list_empty(&ep
->asocs
)) {
4281 asoc
= list_entry(ep
->asocs
.next
,
4282 struct sctp_association
, asocs
);
4283 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4288 int sctp_get_sctp_info(struct sock
*sk
, struct sctp_association
*asoc
,
4289 struct sctp_info
*info
)
4291 struct sctp_transport
*prim
;
4292 struct list_head
*pos
;
4295 memset(info
, 0, sizeof(*info
));
4297 struct sctp_sock
*sp
= sctp_sk(sk
);
4299 info
->sctpi_s_autoclose
= sp
->autoclose
;
4300 info
->sctpi_s_adaptation_ind
= sp
->adaptation_ind
;
4301 info
->sctpi_s_pd_point
= sp
->pd_point
;
4302 info
->sctpi_s_nodelay
= sp
->nodelay
;
4303 info
->sctpi_s_disable_fragments
= sp
->disable_fragments
;
4304 info
->sctpi_s_v4mapped
= sp
->v4mapped
;
4305 info
->sctpi_s_frag_interleave
= sp
->frag_interleave
;
4306 info
->sctpi_s_type
= sp
->type
;
4311 info
->sctpi_tag
= asoc
->c
.my_vtag
;
4312 info
->sctpi_state
= asoc
->state
;
4313 info
->sctpi_rwnd
= asoc
->a_rwnd
;
4314 info
->sctpi_unackdata
= asoc
->unack_data
;
4315 info
->sctpi_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4316 info
->sctpi_instrms
= asoc
->c
.sinit_max_instreams
;
4317 info
->sctpi_outstrms
= asoc
->c
.sinit_num_ostreams
;
4318 list_for_each(pos
, &asoc
->base
.inqueue
.in_chunk_list
)
4319 info
->sctpi_inqueue
++;
4320 list_for_each(pos
, &asoc
->outqueue
.out_chunk_list
)
4321 info
->sctpi_outqueue
++;
4322 info
->sctpi_overall_error
= asoc
->overall_error_count
;
4323 info
->sctpi_max_burst
= asoc
->max_burst
;
4324 info
->sctpi_maxseg
= asoc
->frag_point
;
4325 info
->sctpi_peer_rwnd
= asoc
->peer
.rwnd
;
4326 info
->sctpi_peer_tag
= asoc
->c
.peer_vtag
;
4328 mask
= asoc
->peer
.ecn_capable
<< 1;
4329 mask
= (mask
| asoc
->peer
.ipv4_address
) << 1;
4330 mask
= (mask
| asoc
->peer
.ipv6_address
) << 1;
4331 mask
= (mask
| asoc
->peer
.hostname_address
) << 1;
4332 mask
= (mask
| asoc
->peer
.asconf_capable
) << 1;
4333 mask
= (mask
| asoc
->peer
.prsctp_capable
) << 1;
4334 mask
= (mask
| asoc
->peer
.auth_capable
);
4335 info
->sctpi_peer_capable
= mask
;
4336 mask
= asoc
->peer
.sack_needed
<< 1;
4337 mask
= (mask
| asoc
->peer
.sack_generation
) << 1;
4338 mask
= (mask
| asoc
->peer
.zero_window_announced
);
4339 info
->sctpi_peer_sack
= mask
;
4341 info
->sctpi_isacks
= asoc
->stats
.isacks
;
4342 info
->sctpi_osacks
= asoc
->stats
.osacks
;
4343 info
->sctpi_opackets
= asoc
->stats
.opackets
;
4344 info
->sctpi_ipackets
= asoc
->stats
.ipackets
;
4345 info
->sctpi_rtxchunks
= asoc
->stats
.rtxchunks
;
4346 info
->sctpi_outofseqtsns
= asoc
->stats
.outofseqtsns
;
4347 info
->sctpi_idupchunks
= asoc
->stats
.idupchunks
;
4348 info
->sctpi_gapcnt
= asoc
->stats
.gapcnt
;
4349 info
->sctpi_ouodchunks
= asoc
->stats
.ouodchunks
;
4350 info
->sctpi_iuodchunks
= asoc
->stats
.iuodchunks
;
4351 info
->sctpi_oodchunks
= asoc
->stats
.oodchunks
;
4352 info
->sctpi_iodchunks
= asoc
->stats
.iodchunks
;
4353 info
->sctpi_octrlchunks
= asoc
->stats
.octrlchunks
;
4354 info
->sctpi_ictrlchunks
= asoc
->stats
.ictrlchunks
;
4356 prim
= asoc
->peer
.primary_path
;
4357 memcpy(&info
->sctpi_p_address
, &prim
->ipaddr
,
4358 sizeof(struct sockaddr_storage
));
4359 info
->sctpi_p_state
= prim
->state
;
4360 info
->sctpi_p_cwnd
= prim
->cwnd
;
4361 info
->sctpi_p_srtt
= prim
->srtt
;
4362 info
->sctpi_p_rto
= jiffies_to_msecs(prim
->rto
);
4363 info
->sctpi_p_hbinterval
= prim
->hbinterval
;
4364 info
->sctpi_p_pathmaxrxt
= prim
->pathmaxrxt
;
4365 info
->sctpi_p_sackdelay
= jiffies_to_msecs(prim
->sackdelay
);
4366 info
->sctpi_p_ssthresh
= prim
->ssthresh
;
4367 info
->sctpi_p_partial_bytes_acked
= prim
->partial_bytes_acked
;
4368 info
->sctpi_p_flight_size
= prim
->flight_size
;
4369 info
->sctpi_p_error
= prim
->error_count
;
4373 EXPORT_SYMBOL_GPL(sctp_get_sctp_info
);
4375 /* use callback to avoid exporting the core structure */
4376 int sctp_transport_walk_start(struct rhashtable_iter
*iter
)
4380 err
= rhashtable_walk_init(&sctp_transport_hashtable
, iter
,
4385 err
= rhashtable_walk_start(iter
);
4386 if (err
&& err
!= -EAGAIN
) {
4387 rhashtable_walk_exit(iter
);
4394 void sctp_transport_walk_stop(struct rhashtable_iter
*iter
)
4396 rhashtable_walk_stop(iter
);
4397 rhashtable_walk_exit(iter
);
4400 struct sctp_transport
*sctp_transport_get_next(struct net
*net
,
4401 struct rhashtable_iter
*iter
)
4403 struct sctp_transport
*t
;
4405 t
= rhashtable_walk_next(iter
);
4406 for (; t
; t
= rhashtable_walk_next(iter
)) {
4408 if (PTR_ERR(t
) == -EAGAIN
)
4413 if (net_eq(sock_net(t
->asoc
->base
.sk
), net
) &&
4414 t
->asoc
->peer
.primary_path
== t
)
4421 struct sctp_transport
*sctp_transport_get_idx(struct net
*net
,
4422 struct rhashtable_iter
*iter
,
4425 void *obj
= SEQ_START_TOKEN
;
4427 while (pos
&& (obj
= sctp_transport_get_next(net
, iter
)) &&
4434 int sctp_for_each_endpoint(int (*cb
)(struct sctp_endpoint
*, void *),
4438 struct sctp_ep_common
*epb
;
4439 struct sctp_hashbucket
*head
;
4441 for (head
= sctp_ep_hashtable
; hash
< sctp_ep_hashsize
;
4443 read_lock(&head
->lock
);
4444 sctp_for_each_hentry(epb
, &head
->chain
) {
4445 err
= cb(sctp_ep(epb
), p
);
4449 read_unlock(&head
->lock
);
4454 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint
);
4456 int sctp_transport_lookup_process(int (*cb
)(struct sctp_transport
*, void *),
4458 const union sctp_addr
*laddr
,
4459 const union sctp_addr
*paddr
, void *p
)
4461 struct sctp_transport
*transport
;
4465 transport
= sctp_addrs_lookup_transport(net
, laddr
, paddr
);
4466 if (!transport
|| !sctp_transport_hold(transport
))
4468 err
= cb(transport
, p
);
4469 sctp_transport_put(transport
);
4475 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process
);
4477 int sctp_for_each_transport(int (*cb
)(struct sctp_transport
*, void *),
4478 struct net
*net
, int pos
, void *p
) {
4479 struct rhashtable_iter hti
;
4483 err
= sctp_transport_walk_start(&hti
);
4487 sctp_transport_get_idx(net
, &hti
, pos
);
4488 obj
= sctp_transport_get_next(net
, &hti
);
4489 for (; obj
&& !IS_ERR(obj
); obj
= sctp_transport_get_next(net
, &hti
)) {
4490 struct sctp_transport
*transport
= obj
;
4492 if (!sctp_transport_hold(transport
))
4494 err
= cb(transport
, p
);
4495 sctp_transport_put(transport
);
4499 sctp_transport_walk_stop(&hti
);
4503 EXPORT_SYMBOL_GPL(sctp_for_each_transport
);
4505 /* 7.2.1 Association Status (SCTP_STATUS)
4507 * Applications can retrieve current status information about an
4508 * association, including association state, peer receiver window size,
4509 * number of unacked data chunks, and number of data chunks pending
4510 * receipt. This information is read-only.
4512 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4513 char __user
*optval
,
4516 struct sctp_status status
;
4517 struct sctp_association
*asoc
= NULL
;
4518 struct sctp_transport
*transport
;
4519 sctp_assoc_t associd
;
4522 if (len
< sizeof(status
)) {
4527 len
= sizeof(status
);
4528 if (copy_from_user(&status
, optval
, len
)) {
4533 associd
= status
.sstat_assoc_id
;
4534 asoc
= sctp_id2assoc(sk
, associd
);
4540 transport
= asoc
->peer
.primary_path
;
4542 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4543 status
.sstat_state
= sctp_assoc_to_state(asoc
);
4544 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4545 status
.sstat_unackdata
= asoc
->unack_data
;
4547 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4548 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4549 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4550 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4551 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4552 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4553 transport
->af_specific
->sockaddr_len
);
4554 /* Map ipv4 address into v4-mapped-on-v6 address. */
4555 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
4556 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4557 status
.sstat_primary
.spinfo_state
= transport
->state
;
4558 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4559 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4560 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4561 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4563 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4564 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4566 if (put_user(len
, optlen
)) {
4571 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4572 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4573 status
.sstat_assoc_id
);
4575 if (copy_to_user(optval
, &status
, len
)) {
4585 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4587 * Applications can retrieve information about a specific peer address
4588 * of an association, including its reachability state, congestion
4589 * window, and retransmission timer values. This information is
4592 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4593 char __user
*optval
,
4596 struct sctp_paddrinfo pinfo
;
4597 struct sctp_transport
*transport
;
4600 if (len
< sizeof(pinfo
)) {
4605 len
= sizeof(pinfo
);
4606 if (copy_from_user(&pinfo
, optval
, len
)) {
4611 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4612 pinfo
.spinfo_assoc_id
);
4616 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4617 pinfo
.spinfo_state
= transport
->state
;
4618 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4619 pinfo
.spinfo_srtt
= transport
->srtt
;
4620 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4621 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4623 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4624 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4626 if (put_user(len
, optlen
)) {
4631 if (copy_to_user(optval
, &pinfo
, len
)) {
4640 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4642 * This option is a on/off flag. If enabled no SCTP message
4643 * fragmentation will be performed. Instead if a message being sent
4644 * exceeds the current PMTU size, the message will NOT be sent and
4645 * instead a error will be indicated to the user.
4647 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4648 char __user
*optval
, int __user
*optlen
)
4652 if (len
< sizeof(int))
4656 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4657 if (put_user(len
, optlen
))
4659 if (copy_to_user(optval
, &val
, len
))
4664 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4666 * This socket option is used to specify various notifications and
4667 * ancillary data the user wishes to receive.
4669 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4674 if (len
> sizeof(struct sctp_event_subscribe
))
4675 len
= sizeof(struct sctp_event_subscribe
);
4676 if (put_user(len
, optlen
))
4678 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4683 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4685 * This socket option is applicable to the UDP-style socket only. When
4686 * set it will cause associations that are idle for more than the
4687 * specified number of seconds to automatically close. An association
4688 * being idle is defined an association that has NOT sent or received
4689 * user data. The special value of '0' indicates that no automatic
4690 * close of any associations should be performed. The option expects an
4691 * integer defining the number of seconds of idle time before an
4692 * association is closed.
4694 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4696 /* Applicable to UDP-style socket only */
4697 if (sctp_style(sk
, TCP
))
4699 if (len
< sizeof(int))
4702 if (put_user(len
, optlen
))
4704 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4709 /* Helper routine to branch off an association to a new socket. */
4710 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4712 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4713 struct sctp_sock
*sp
= sctp_sk(sk
);
4714 struct socket
*sock
;
4720 /* An association cannot be branched off from an already peeled-off
4721 * socket, nor is this supported for tcp style sockets.
4723 if (!sctp_style(sk
, UDP
))
4726 /* Create a new socket. */
4727 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4731 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4733 /* Make peeled-off sockets more like 1-1 accepted sockets.
4734 * Set the daddr and initialize id to something more random
4736 sp
->pf
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4738 /* Populate the fields of the newsk from the oldsk and migrate the
4739 * asoc to the newsk.
4741 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4747 EXPORT_SYMBOL(sctp_do_peeloff
);
4749 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4751 sctp_peeloff_arg_t peeloff
;
4752 struct socket
*newsock
;
4753 struct file
*newfile
;
4756 if (len
< sizeof(sctp_peeloff_arg_t
))
4758 len
= sizeof(sctp_peeloff_arg_t
);
4759 if (copy_from_user(&peeloff
, optval
, len
))
4762 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4766 /* Map the socket to an unused fd that can be returned to the user. */
4767 retval
= get_unused_fd_flags(0);
4769 sock_release(newsock
);
4773 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4774 if (IS_ERR(newfile
)) {
4775 put_unused_fd(retval
);
4776 sock_release(newsock
);
4777 return PTR_ERR(newfile
);
4780 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4783 /* Return the fd mapped to the new socket. */
4784 if (put_user(len
, optlen
)) {
4786 put_unused_fd(retval
);
4789 peeloff
.sd
= retval
;
4790 if (copy_to_user(optval
, &peeloff
, len
)) {
4792 put_unused_fd(retval
);
4795 fd_install(retval
, newfile
);
4800 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4802 * Applications can enable or disable heartbeats for any peer address of
4803 * an association, modify an address's heartbeat interval, force a
4804 * heartbeat to be sent immediately, and adjust the address's maximum
4805 * number of retransmissions sent before an address is considered
4806 * unreachable. The following structure is used to access and modify an
4807 * address's parameters:
4809 * struct sctp_paddrparams {
4810 * sctp_assoc_t spp_assoc_id;
4811 * struct sockaddr_storage spp_address;
4812 * uint32_t spp_hbinterval;
4813 * uint16_t spp_pathmaxrxt;
4814 * uint32_t spp_pathmtu;
4815 * uint32_t spp_sackdelay;
4816 * uint32_t spp_flags;
4819 * spp_assoc_id - (one-to-many style socket) This is filled in the
4820 * application, and identifies the association for
4822 * spp_address - This specifies which address is of interest.
4823 * spp_hbinterval - This contains the value of the heartbeat interval,
4824 * in milliseconds. If a value of zero
4825 * is present in this field then no changes are to
4826 * be made to this parameter.
4827 * spp_pathmaxrxt - This contains the maximum number of
4828 * retransmissions before this address shall be
4829 * considered unreachable. If a value of zero
4830 * is present in this field then no changes are to
4831 * be made to this parameter.
4832 * spp_pathmtu - When Path MTU discovery is disabled the value
4833 * specified here will be the "fixed" path mtu.
4834 * Note that if the spp_address field is empty
4835 * then all associations on this address will
4836 * have this fixed path mtu set upon them.
4838 * spp_sackdelay - When delayed sack is enabled, this value specifies
4839 * the number of milliseconds that sacks will be delayed
4840 * for. This value will apply to all addresses of an
4841 * association if the spp_address field is empty. Note
4842 * also, that if delayed sack is enabled and this
4843 * value is set to 0, no change is made to the last
4844 * recorded delayed sack timer value.
4846 * spp_flags - These flags are used to control various features
4847 * on an association. The flag field may contain
4848 * zero or more of the following options.
4850 * SPP_HB_ENABLE - Enable heartbeats on the
4851 * specified address. Note that if the address
4852 * field is empty all addresses for the association
4853 * have heartbeats enabled upon them.
4855 * SPP_HB_DISABLE - Disable heartbeats on the
4856 * speicifed address. Note that if the address
4857 * field is empty all addresses for the association
4858 * will have their heartbeats disabled. Note also
4859 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4860 * mutually exclusive, only one of these two should
4861 * be specified. Enabling both fields will have
4862 * undetermined results.
4864 * SPP_HB_DEMAND - Request a user initiated heartbeat
4865 * to be made immediately.
4867 * SPP_PMTUD_ENABLE - This field will enable PMTU
4868 * discovery upon the specified address. Note that
4869 * if the address feild is empty then all addresses
4870 * on the association are effected.
4872 * SPP_PMTUD_DISABLE - This field will disable PMTU
4873 * discovery upon the specified address. Note that
4874 * if the address feild is empty then all addresses
4875 * on the association are effected. Not also that
4876 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4877 * exclusive. Enabling both will have undetermined
4880 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4881 * on delayed sack. The time specified in spp_sackdelay
4882 * is used to specify the sack delay for this address. Note
4883 * that if spp_address is empty then all addresses will
4884 * enable delayed sack and take on the sack delay
4885 * value specified in spp_sackdelay.
4886 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4887 * off delayed sack. If the spp_address field is blank then
4888 * delayed sack is disabled for the entire association. Note
4889 * also that this field is mutually exclusive to
4890 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4893 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4894 char __user
*optval
, int __user
*optlen
)
4896 struct sctp_paddrparams params
;
4897 struct sctp_transport
*trans
= NULL
;
4898 struct sctp_association
*asoc
= NULL
;
4899 struct sctp_sock
*sp
= sctp_sk(sk
);
4901 if (len
< sizeof(struct sctp_paddrparams
))
4903 len
= sizeof(struct sctp_paddrparams
);
4904 if (copy_from_user(¶ms
, optval
, len
))
4907 /* If an address other than INADDR_ANY is specified, and
4908 * no transport is found, then the request is invalid.
4910 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4911 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4912 params
.spp_assoc_id
);
4914 pr_debug("%s: failed no transport\n", __func__
);
4919 /* Get association, if assoc_id != 0 and the socket is a one
4920 * to many style socket, and an association was not found, then
4921 * the id was invalid.
4923 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4924 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4925 pr_debug("%s: failed no association\n", __func__
);
4930 /* Fetch transport values. */
4931 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4932 params
.spp_pathmtu
= trans
->pathmtu
;
4933 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4934 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4936 /*draft-11 doesn't say what to return in spp_flags*/
4937 params
.spp_flags
= trans
->param_flags
;
4939 /* Fetch association values. */
4940 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4941 params
.spp_pathmtu
= asoc
->pathmtu
;
4942 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4943 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4945 /*draft-11 doesn't say what to return in spp_flags*/
4946 params
.spp_flags
= asoc
->param_flags
;
4948 /* Fetch socket values. */
4949 params
.spp_hbinterval
= sp
->hbinterval
;
4950 params
.spp_pathmtu
= sp
->pathmtu
;
4951 params
.spp_sackdelay
= sp
->sackdelay
;
4952 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4954 /*draft-11 doesn't say what to return in spp_flags*/
4955 params
.spp_flags
= sp
->param_flags
;
4958 if (copy_to_user(optval
, ¶ms
, len
))
4961 if (put_user(len
, optlen
))
4968 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4970 * This option will effect the way delayed acks are performed. This
4971 * option allows you to get or set the delayed ack time, in
4972 * milliseconds. It also allows changing the delayed ack frequency.
4973 * Changing the frequency to 1 disables the delayed sack algorithm. If
4974 * the assoc_id is 0, then this sets or gets the endpoints default
4975 * values. If the assoc_id field is non-zero, then the set or get
4976 * effects the specified association for the one to many model (the
4977 * assoc_id field is ignored by the one to one model). Note that if
4978 * sack_delay or sack_freq are 0 when setting this option, then the
4979 * current values will remain unchanged.
4981 * struct sctp_sack_info {
4982 * sctp_assoc_t sack_assoc_id;
4983 * uint32_t sack_delay;
4984 * uint32_t sack_freq;
4987 * sack_assoc_id - This parameter, indicates which association the user
4988 * is performing an action upon. Note that if this field's value is
4989 * zero then the endpoints default value is changed (effecting future
4990 * associations only).
4992 * sack_delay - This parameter contains the number of milliseconds that
4993 * the user is requesting the delayed ACK timer be set to. Note that
4994 * this value is defined in the standard to be between 200 and 500
4997 * sack_freq - This parameter contains the number of packets that must
4998 * be received before a sack is sent without waiting for the delay
4999 * timer to expire. The default value for this is 2, setting this
5000 * value to 1 will disable the delayed sack algorithm.
5002 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
5003 char __user
*optval
,
5006 struct sctp_sack_info params
;
5007 struct sctp_association
*asoc
= NULL
;
5008 struct sctp_sock
*sp
= sctp_sk(sk
);
5010 if (len
>= sizeof(struct sctp_sack_info
)) {
5011 len
= sizeof(struct sctp_sack_info
);
5013 if (copy_from_user(¶ms
, optval
, len
))
5015 } else if (len
== sizeof(struct sctp_assoc_value
)) {
5016 pr_warn_ratelimited(DEPRECATED
5018 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5019 "Use struct sctp_sack_info instead\n",
5020 current
->comm
, task_pid_nr(current
));
5021 if (copy_from_user(¶ms
, optval
, len
))
5026 /* Get association, if sack_assoc_id != 0 and the socket is a one
5027 * to many style socket, and an association was not found, then
5028 * the id was invalid.
5030 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
5031 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
5035 /* Fetch association values. */
5036 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5037 params
.sack_delay
= jiffies_to_msecs(
5039 params
.sack_freq
= asoc
->sackfreq
;
5042 params
.sack_delay
= 0;
5043 params
.sack_freq
= 1;
5046 /* Fetch socket values. */
5047 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5048 params
.sack_delay
= sp
->sackdelay
;
5049 params
.sack_freq
= sp
->sackfreq
;
5051 params
.sack_delay
= 0;
5052 params
.sack_freq
= 1;
5056 if (copy_to_user(optval
, ¶ms
, len
))
5059 if (put_user(len
, optlen
))
5065 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5067 * Applications can specify protocol parameters for the default association
5068 * initialization. The option name argument to setsockopt() and getsockopt()
5071 * Setting initialization parameters is effective only on an unconnected
5072 * socket (for UDP-style sockets only future associations are effected
5073 * by the change). With TCP-style sockets, this option is inherited by
5074 * sockets derived from a listener socket.
5076 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
5078 if (len
< sizeof(struct sctp_initmsg
))
5080 len
= sizeof(struct sctp_initmsg
);
5081 if (put_user(len
, optlen
))
5083 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
5089 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
5090 char __user
*optval
, int __user
*optlen
)
5092 struct sctp_association
*asoc
;
5094 struct sctp_getaddrs getaddrs
;
5095 struct sctp_transport
*from
;
5097 union sctp_addr temp
;
5098 struct sctp_sock
*sp
= sctp_sk(sk
);
5103 if (len
< sizeof(struct sctp_getaddrs
))
5106 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5109 /* For UDP-style sockets, id specifies the association to query. */
5110 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5114 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5115 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5117 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
5119 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
5120 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5121 ->addr_to_user(sp
, &temp
);
5122 if (space_left
< addrlen
)
5124 if (copy_to_user(to
, &temp
, addrlen
))
5128 space_left
-= addrlen
;
5131 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
5133 bytes_copied
= ((char __user
*)to
) - optval
;
5134 if (put_user(bytes_copied
, optlen
))
5140 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
5141 size_t space_left
, int *bytes_copied
)
5143 struct sctp_sockaddr_entry
*addr
;
5144 union sctp_addr temp
;
5147 struct net
*net
= sock_net(sk
);
5150 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
5154 if ((PF_INET
== sk
->sk_family
) &&
5155 (AF_INET6
== addr
->a
.sa
.sa_family
))
5157 if ((PF_INET6
== sk
->sk_family
) &&
5158 inet_v6_ipv6only(sk
) &&
5159 (AF_INET
== addr
->a
.sa
.sa_family
))
5161 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5162 if (!temp
.v4
.sin_port
)
5163 temp
.v4
.sin_port
= htons(port
);
5165 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5166 ->addr_to_user(sctp_sk(sk
), &temp
);
5168 if (space_left
< addrlen
) {
5172 memcpy(to
, &temp
, addrlen
);
5176 space_left
-= addrlen
;
5177 *bytes_copied
+= addrlen
;
5185 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
5186 char __user
*optval
, int __user
*optlen
)
5188 struct sctp_bind_addr
*bp
;
5189 struct sctp_association
*asoc
;
5191 struct sctp_getaddrs getaddrs
;
5192 struct sctp_sockaddr_entry
*addr
;
5194 union sctp_addr temp
;
5195 struct sctp_sock
*sp
= sctp_sk(sk
);
5199 int bytes_copied
= 0;
5203 if (len
< sizeof(struct sctp_getaddrs
))
5206 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5210 * For UDP-style sockets, id specifies the association to query.
5211 * If the id field is set to the value '0' then the locally bound
5212 * addresses are returned without regard to any particular
5215 if (0 == getaddrs
.assoc_id
) {
5216 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
5218 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5221 bp
= &asoc
->base
.bind_addr
;
5224 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5225 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5227 addrs
= kmalloc(space_left
, GFP_USER
| __GFP_NOWARN
);
5231 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5232 * addresses from the global local address list.
5234 if (sctp_list_single_entry(&bp
->address_list
)) {
5235 addr
= list_entry(bp
->address_list
.next
,
5236 struct sctp_sockaddr_entry
, list
);
5237 if (sctp_is_any(sk
, &addr
->a
)) {
5238 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
5239 space_left
, &bytes_copied
);
5249 /* Protection on the bound address list is not needed since
5250 * in the socket option context we hold a socket lock and
5251 * thus the bound address list can't change.
5253 list_for_each_entry(addr
, &bp
->address_list
, list
) {
5254 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5255 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5256 ->addr_to_user(sp
, &temp
);
5257 if (space_left
< addrlen
) {
5258 err
= -ENOMEM
; /*fixme: right error?*/
5261 memcpy(buf
, &temp
, addrlen
);
5263 bytes_copied
+= addrlen
;
5265 space_left
-= addrlen
;
5269 if (copy_to_user(to
, addrs
, bytes_copied
)) {
5273 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
5277 if (put_user(bytes_copied
, optlen
))
5284 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5286 * Requests that the local SCTP stack use the enclosed peer address as
5287 * the association primary. The enclosed address must be one of the
5288 * association peer's addresses.
5290 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
5291 char __user
*optval
, int __user
*optlen
)
5293 struct sctp_prim prim
;
5294 struct sctp_association
*asoc
;
5295 struct sctp_sock
*sp
= sctp_sk(sk
);
5297 if (len
< sizeof(struct sctp_prim
))
5300 len
= sizeof(struct sctp_prim
);
5302 if (copy_from_user(&prim
, optval
, len
))
5305 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
5309 if (!asoc
->peer
.primary_path
)
5312 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
5313 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
5315 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sp
,
5316 (union sctp_addr
*)&prim
.ssp_addr
);
5318 if (put_user(len
, optlen
))
5320 if (copy_to_user(optval
, &prim
, len
))
5327 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5329 * Requests that the local endpoint set the specified Adaptation Layer
5330 * Indication parameter for all future INIT and INIT-ACK exchanges.
5332 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
5333 char __user
*optval
, int __user
*optlen
)
5335 struct sctp_setadaptation adaptation
;
5337 if (len
< sizeof(struct sctp_setadaptation
))
5340 len
= sizeof(struct sctp_setadaptation
);
5342 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
5344 if (put_user(len
, optlen
))
5346 if (copy_to_user(optval
, &adaptation
, len
))
5354 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5356 * Applications that wish to use the sendto() system call may wish to
5357 * specify a default set of parameters that would normally be supplied
5358 * through the inclusion of ancillary data. This socket option allows
5359 * such an application to set the default sctp_sndrcvinfo structure.
5362 * The application that wishes to use this socket option simply passes
5363 * in to this call the sctp_sndrcvinfo structure defined in Section
5364 * 5.2.2) The input parameters accepted by this call include
5365 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5366 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5367 * to this call if the caller is using the UDP model.
5369 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5371 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
5372 int len
, char __user
*optval
,
5375 struct sctp_sock
*sp
= sctp_sk(sk
);
5376 struct sctp_association
*asoc
;
5377 struct sctp_sndrcvinfo info
;
5379 if (len
< sizeof(info
))
5384 if (copy_from_user(&info
, optval
, len
))
5387 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
5388 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
5391 info
.sinfo_stream
= asoc
->default_stream
;
5392 info
.sinfo_flags
= asoc
->default_flags
;
5393 info
.sinfo_ppid
= asoc
->default_ppid
;
5394 info
.sinfo_context
= asoc
->default_context
;
5395 info
.sinfo_timetolive
= asoc
->default_timetolive
;
5397 info
.sinfo_stream
= sp
->default_stream
;
5398 info
.sinfo_flags
= sp
->default_flags
;
5399 info
.sinfo_ppid
= sp
->default_ppid
;
5400 info
.sinfo_context
= sp
->default_context
;
5401 info
.sinfo_timetolive
= sp
->default_timetolive
;
5404 if (put_user(len
, optlen
))
5406 if (copy_to_user(optval
, &info
, len
))
5412 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5413 * (SCTP_DEFAULT_SNDINFO)
5415 static int sctp_getsockopt_default_sndinfo(struct sock
*sk
, int len
,
5416 char __user
*optval
,
5419 struct sctp_sock
*sp
= sctp_sk(sk
);
5420 struct sctp_association
*asoc
;
5421 struct sctp_sndinfo info
;
5423 if (len
< sizeof(info
))
5428 if (copy_from_user(&info
, optval
, len
))
5431 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
5432 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
5435 info
.snd_sid
= asoc
->default_stream
;
5436 info
.snd_flags
= asoc
->default_flags
;
5437 info
.snd_ppid
= asoc
->default_ppid
;
5438 info
.snd_context
= asoc
->default_context
;
5440 info
.snd_sid
= sp
->default_stream
;
5441 info
.snd_flags
= sp
->default_flags
;
5442 info
.snd_ppid
= sp
->default_ppid
;
5443 info
.snd_context
= sp
->default_context
;
5446 if (put_user(len
, optlen
))
5448 if (copy_to_user(optval
, &info
, len
))
5456 * 7.1.5 SCTP_NODELAY
5458 * Turn on/off any Nagle-like algorithm. This means that packets are
5459 * generally sent as soon as possible and no unnecessary delays are
5460 * introduced, at the cost of more packets in the network. Expects an
5461 * integer boolean flag.
5464 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5465 char __user
*optval
, int __user
*optlen
)
5469 if (len
< sizeof(int))
5473 val
= (sctp_sk(sk
)->nodelay
== 1);
5474 if (put_user(len
, optlen
))
5476 if (copy_to_user(optval
, &val
, len
))
5483 * 7.1.1 SCTP_RTOINFO
5485 * The protocol parameters used to initialize and bound retransmission
5486 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5487 * and modify these parameters.
5488 * All parameters are time values, in milliseconds. A value of 0, when
5489 * modifying the parameters, indicates that the current value should not
5493 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5494 char __user
*optval
,
5495 int __user
*optlen
) {
5496 struct sctp_rtoinfo rtoinfo
;
5497 struct sctp_association
*asoc
;
5499 if (len
< sizeof (struct sctp_rtoinfo
))
5502 len
= sizeof(struct sctp_rtoinfo
);
5504 if (copy_from_user(&rtoinfo
, optval
, len
))
5507 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5509 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5512 /* Values corresponding to the specific association. */
5514 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5515 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5516 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5518 /* Values corresponding to the endpoint. */
5519 struct sctp_sock
*sp
= sctp_sk(sk
);
5521 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5522 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5523 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5526 if (put_user(len
, optlen
))
5529 if (copy_to_user(optval
, &rtoinfo
, len
))
5537 * 7.1.2 SCTP_ASSOCINFO
5539 * This option is used to tune the maximum retransmission attempts
5540 * of the association.
5541 * Returns an error if the new association retransmission value is
5542 * greater than the sum of the retransmission value of the peer.
5543 * See [SCTP] for more information.
5546 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5547 char __user
*optval
,
5551 struct sctp_assocparams assocparams
;
5552 struct sctp_association
*asoc
;
5553 struct list_head
*pos
;
5556 if (len
< sizeof (struct sctp_assocparams
))
5559 len
= sizeof(struct sctp_assocparams
);
5561 if (copy_from_user(&assocparams
, optval
, len
))
5564 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5566 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5569 /* Values correspoinding to the specific association */
5571 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5572 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5573 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5574 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5576 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5580 assocparams
.sasoc_number_peer_destinations
= cnt
;
5582 /* Values corresponding to the endpoint */
5583 struct sctp_sock
*sp
= sctp_sk(sk
);
5585 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5586 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5587 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5588 assocparams
.sasoc_cookie_life
=
5589 sp
->assocparams
.sasoc_cookie_life
;
5590 assocparams
.sasoc_number_peer_destinations
=
5592 sasoc_number_peer_destinations
;
5595 if (put_user(len
, optlen
))
5598 if (copy_to_user(optval
, &assocparams
, len
))
5605 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5607 * This socket option is a boolean flag which turns on or off mapped V4
5608 * addresses. If this option is turned on and the socket is type
5609 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5610 * If this option is turned off, then no mapping will be done of V4
5611 * addresses and a user will receive both PF_INET6 and PF_INET type
5612 * addresses on the socket.
5614 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5615 char __user
*optval
, int __user
*optlen
)
5618 struct sctp_sock
*sp
= sctp_sk(sk
);
5620 if (len
< sizeof(int))
5625 if (put_user(len
, optlen
))
5627 if (copy_to_user(optval
, &val
, len
))
5634 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5635 * (chapter and verse is quoted at sctp_setsockopt_context())
5637 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5638 char __user
*optval
, int __user
*optlen
)
5640 struct sctp_assoc_value params
;
5641 struct sctp_sock
*sp
;
5642 struct sctp_association
*asoc
;
5644 if (len
< sizeof(struct sctp_assoc_value
))
5647 len
= sizeof(struct sctp_assoc_value
);
5649 if (copy_from_user(¶ms
, optval
, len
))
5654 if (params
.assoc_id
!= 0) {
5655 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5658 params
.assoc_value
= asoc
->default_rcv_context
;
5660 params
.assoc_value
= sp
->default_rcv_context
;
5663 if (put_user(len
, optlen
))
5665 if (copy_to_user(optval
, ¶ms
, len
))
5672 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5673 * This option will get or set the maximum size to put in any outgoing
5674 * SCTP DATA chunk. If a message is larger than this size it will be
5675 * fragmented by SCTP into the specified size. Note that the underlying
5676 * SCTP implementation may fragment into smaller sized chunks when the
5677 * PMTU of the underlying association is smaller than the value set by
5678 * the user. The default value for this option is '0' which indicates
5679 * the user is NOT limiting fragmentation and only the PMTU will effect
5680 * SCTP's choice of DATA chunk size. Note also that values set larger
5681 * than the maximum size of an IP datagram will effectively let SCTP
5682 * control fragmentation (i.e. the same as setting this option to 0).
5684 * The following structure is used to access and modify this parameter:
5686 * struct sctp_assoc_value {
5687 * sctp_assoc_t assoc_id;
5688 * uint32_t assoc_value;
5691 * assoc_id: This parameter is ignored for one-to-one style sockets.
5692 * For one-to-many style sockets this parameter indicates which
5693 * association the user is performing an action upon. Note that if
5694 * this field's value is zero then the endpoints default value is
5695 * changed (effecting future associations only).
5696 * assoc_value: This parameter specifies the maximum size in bytes.
5698 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5699 char __user
*optval
, int __user
*optlen
)
5701 struct sctp_assoc_value params
;
5702 struct sctp_association
*asoc
;
5704 if (len
== sizeof(int)) {
5705 pr_warn_ratelimited(DEPRECATED
5707 "Use of int in maxseg socket option.\n"
5708 "Use struct sctp_assoc_value instead\n",
5709 current
->comm
, task_pid_nr(current
));
5710 params
.assoc_id
= 0;
5711 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5712 len
= sizeof(struct sctp_assoc_value
);
5713 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5718 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5719 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5723 params
.assoc_value
= asoc
->frag_point
;
5725 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5727 if (put_user(len
, optlen
))
5729 if (len
== sizeof(int)) {
5730 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5733 if (copy_to_user(optval
, ¶ms
, len
))
5741 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5742 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5744 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5745 char __user
*optval
, int __user
*optlen
)
5749 if (len
< sizeof(int))
5754 val
= sctp_sk(sk
)->frag_interleave
;
5755 if (put_user(len
, optlen
))
5757 if (copy_to_user(optval
, &val
, len
))
5764 * 7.1.25. Set or Get the sctp partial delivery point
5765 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5767 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5768 char __user
*optval
,
5773 if (len
< sizeof(u32
))
5778 val
= sctp_sk(sk
)->pd_point
;
5779 if (put_user(len
, optlen
))
5781 if (copy_to_user(optval
, &val
, len
))
5788 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5789 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5791 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5792 char __user
*optval
,
5795 struct sctp_assoc_value params
;
5796 struct sctp_sock
*sp
;
5797 struct sctp_association
*asoc
;
5799 if (len
== sizeof(int)) {
5800 pr_warn_ratelimited(DEPRECATED
5802 "Use of int in max_burst socket option.\n"
5803 "Use struct sctp_assoc_value instead\n",
5804 current
->comm
, task_pid_nr(current
));
5805 params
.assoc_id
= 0;
5806 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5807 len
= sizeof(struct sctp_assoc_value
);
5808 if (copy_from_user(¶ms
, optval
, len
))
5815 if (params
.assoc_id
!= 0) {
5816 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5819 params
.assoc_value
= asoc
->max_burst
;
5821 params
.assoc_value
= sp
->max_burst
;
5823 if (len
== sizeof(int)) {
5824 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5827 if (copy_to_user(optval
, ¶ms
, len
))
5835 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5836 char __user
*optval
, int __user
*optlen
)
5838 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5839 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5840 struct sctp_hmac_algo_param
*hmacs
;
5845 if (!ep
->auth_enable
)
5848 hmacs
= ep
->auth_hmacs_list
;
5849 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5851 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5854 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5855 num_idents
= data_len
/ sizeof(u16
);
5857 if (put_user(len
, optlen
))
5859 if (put_user(num_idents
, &p
->shmac_num_idents
))
5861 for (i
= 0; i
< num_idents
; i
++) {
5862 __u16 hmacid
= ntohs(hmacs
->hmac_ids
[i
]);
5864 if (copy_to_user(&p
->shmac_idents
[i
], &hmacid
, sizeof(__u16
)))
5870 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5871 char __user
*optval
, int __user
*optlen
)
5873 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5874 struct sctp_authkeyid val
;
5875 struct sctp_association
*asoc
;
5877 if (!ep
->auth_enable
)
5880 if (len
< sizeof(struct sctp_authkeyid
))
5882 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5885 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5886 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5890 val
.scact_keynumber
= asoc
->active_key_id
;
5892 val
.scact_keynumber
= ep
->active_key_id
;
5894 len
= sizeof(struct sctp_authkeyid
);
5895 if (put_user(len
, optlen
))
5897 if (copy_to_user(optval
, &val
, len
))
5903 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5904 char __user
*optval
, int __user
*optlen
)
5906 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5907 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5908 struct sctp_authchunks val
;
5909 struct sctp_association
*asoc
;
5910 struct sctp_chunks_param
*ch
;
5914 if (!ep
->auth_enable
)
5917 if (len
< sizeof(struct sctp_authchunks
))
5920 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5923 to
= p
->gauth_chunks
;
5924 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5928 ch
= asoc
->peer
.peer_chunks
;
5932 /* See if the user provided enough room for all the data */
5933 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5934 if (len
< num_chunks
)
5937 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5940 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5941 if (put_user(len
, optlen
))
5943 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5948 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5949 char __user
*optval
, int __user
*optlen
)
5951 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5952 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5953 struct sctp_authchunks val
;
5954 struct sctp_association
*asoc
;
5955 struct sctp_chunks_param
*ch
;
5959 if (!ep
->auth_enable
)
5962 if (len
< sizeof(struct sctp_authchunks
))
5965 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5968 to
= p
->gauth_chunks
;
5969 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5970 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5974 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5976 ch
= ep
->auth_chunk_list
;
5981 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5982 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5985 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5988 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5989 if (put_user(len
, optlen
))
5991 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5998 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5999 * This option gets the current number of associations that are attached
6000 * to a one-to-many style socket. The option value is an uint32_t.
6002 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
6003 char __user
*optval
, int __user
*optlen
)
6005 struct sctp_sock
*sp
= sctp_sk(sk
);
6006 struct sctp_association
*asoc
;
6009 if (sctp_style(sk
, TCP
))
6012 if (len
< sizeof(u32
))
6017 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6021 if (put_user(len
, optlen
))
6023 if (copy_to_user(optval
, &val
, len
))
6030 * 8.1.23 SCTP_AUTO_ASCONF
6031 * See the corresponding setsockopt entry as description
6033 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
6034 char __user
*optval
, int __user
*optlen
)
6038 if (len
< sizeof(int))
6042 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
6044 if (put_user(len
, optlen
))
6046 if (copy_to_user(optval
, &val
, len
))
6052 * 8.2.6. Get the Current Identifiers of Associations
6053 * (SCTP_GET_ASSOC_ID_LIST)
6055 * This option gets the current list of SCTP association identifiers of
6056 * the SCTP associations handled by a one-to-many style socket.
6058 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
6059 char __user
*optval
, int __user
*optlen
)
6061 struct sctp_sock
*sp
= sctp_sk(sk
);
6062 struct sctp_association
*asoc
;
6063 struct sctp_assoc_ids
*ids
;
6066 if (sctp_style(sk
, TCP
))
6069 if (len
< sizeof(struct sctp_assoc_ids
))
6072 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6076 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
6079 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
6081 ids
= kmalloc(len
, GFP_USER
| __GFP_NOWARN
);
6085 ids
->gaids_number_of_ids
= num
;
6087 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6088 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
6091 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
6101 * SCTP_PEER_ADDR_THLDS
6103 * This option allows us to fetch the partially failed threshold for one or all
6104 * transports in an association. See Section 6.1 of:
6105 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6107 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
6108 char __user
*optval
,
6112 struct sctp_paddrthlds val
;
6113 struct sctp_transport
*trans
;
6114 struct sctp_association
*asoc
;
6116 if (len
< sizeof(struct sctp_paddrthlds
))
6118 len
= sizeof(struct sctp_paddrthlds
);
6119 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
6122 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
6123 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
6127 val
.spt_pathpfthld
= asoc
->pf_retrans
;
6128 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
6130 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
6135 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
6136 val
.spt_pathpfthld
= trans
->pf_retrans
;
6139 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
6146 * SCTP_GET_ASSOC_STATS
6148 * This option retrieves local per endpoint statistics. It is modeled
6149 * after OpenSolaris' implementation
6151 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
6152 char __user
*optval
,
6155 struct sctp_assoc_stats sas
;
6156 struct sctp_association
*asoc
= NULL
;
6158 /* User must provide at least the assoc id */
6159 if (len
< sizeof(sctp_assoc_t
))
6162 /* Allow the struct to grow and fill in as much as possible */
6163 len
= min_t(size_t, len
, sizeof(sas
));
6165 if (copy_from_user(&sas
, optval
, len
))
6168 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
6172 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
6173 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
6174 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
6175 sas
.sas_osacks
= asoc
->stats
.osacks
;
6176 sas
.sas_isacks
= asoc
->stats
.isacks
;
6177 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
6178 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
6179 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
6180 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
6181 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
6182 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
6183 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
6184 sas
.sas_opackets
= asoc
->stats
.opackets
;
6185 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
6187 /* New high max rto observed, will return 0 if not a single
6188 * RTO update took place. obs_rto_ipaddr will be bogus
6191 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
6192 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
6193 sizeof(struct sockaddr_storage
));
6195 /* Mark beginning of a new observation period */
6196 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
6198 if (put_user(len
, optlen
))
6201 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
6203 if (copy_to_user(optval
, &sas
, len
))
6209 static int sctp_getsockopt_recvrcvinfo(struct sock
*sk
, int len
,
6210 char __user
*optval
,
6215 if (len
< sizeof(int))
6219 if (sctp_sk(sk
)->recvrcvinfo
)
6221 if (put_user(len
, optlen
))
6223 if (copy_to_user(optval
, &val
, len
))
6229 static int sctp_getsockopt_recvnxtinfo(struct sock
*sk
, int len
,
6230 char __user
*optval
,
6235 if (len
< sizeof(int))
6239 if (sctp_sk(sk
)->recvnxtinfo
)
6241 if (put_user(len
, optlen
))
6243 if (copy_to_user(optval
, &val
, len
))
6249 static int sctp_getsockopt_pr_supported(struct sock
*sk
, int len
,
6250 char __user
*optval
,
6253 struct sctp_assoc_value params
;
6254 struct sctp_association
*asoc
;
6255 int retval
= -EFAULT
;
6257 if (len
< sizeof(params
)) {
6262 len
= sizeof(params
);
6263 if (copy_from_user(¶ms
, optval
, len
))
6266 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
6268 params
.assoc_value
= asoc
->prsctp_enable
;
6269 } else if (!params
.assoc_id
) {
6270 struct sctp_sock
*sp
= sctp_sk(sk
);
6272 params
.assoc_value
= sp
->ep
->prsctp_enable
;
6278 if (put_user(len
, optlen
))
6281 if (copy_to_user(optval
, ¶ms
, len
))
6290 static int sctp_getsockopt_default_prinfo(struct sock
*sk
, int len
,
6291 char __user
*optval
,
6294 struct sctp_default_prinfo info
;
6295 struct sctp_association
*asoc
;
6296 int retval
= -EFAULT
;
6298 if (len
< sizeof(info
)) {
6304 if (copy_from_user(&info
, optval
, len
))
6307 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
6309 info
.pr_policy
= SCTP_PR_POLICY(asoc
->default_flags
);
6310 info
.pr_value
= asoc
->default_timetolive
;
6311 } else if (!info
.pr_assoc_id
) {
6312 struct sctp_sock
*sp
= sctp_sk(sk
);
6314 info
.pr_policy
= SCTP_PR_POLICY(sp
->default_flags
);
6315 info
.pr_value
= sp
->default_timetolive
;
6321 if (put_user(len
, optlen
))
6324 if (copy_to_user(optval
, &info
, len
))
6333 static int sctp_getsockopt_pr_assocstatus(struct sock
*sk
, int len
,
6334 char __user
*optval
,
6337 struct sctp_prstatus params
;
6338 struct sctp_association
*asoc
;
6340 int retval
= -EINVAL
;
6342 if (len
< sizeof(params
))
6345 len
= sizeof(params
);
6346 if (copy_from_user(¶ms
, optval
, len
)) {
6351 policy
= params
.sprstat_policy
;
6352 if (policy
& ~SCTP_PR_SCTP_MASK
)
6355 asoc
= sctp_id2assoc(sk
, params
.sprstat_assoc_id
);
6359 if (policy
== SCTP_PR_SCTP_NONE
) {
6360 params
.sprstat_abandoned_unsent
= 0;
6361 params
.sprstat_abandoned_sent
= 0;
6362 for (policy
= 0; policy
<= SCTP_PR_INDEX(MAX
); policy
++) {
6363 params
.sprstat_abandoned_unsent
+=
6364 asoc
->abandoned_unsent
[policy
];
6365 params
.sprstat_abandoned_sent
+=
6366 asoc
->abandoned_sent
[policy
];
6369 params
.sprstat_abandoned_unsent
=
6370 asoc
->abandoned_unsent
[__SCTP_PR_INDEX(policy
)];
6371 params
.sprstat_abandoned_sent
=
6372 asoc
->abandoned_sent
[__SCTP_PR_INDEX(policy
)];
6375 if (put_user(len
, optlen
)) {
6380 if (copy_to_user(optval
, ¶ms
, len
)) {
6391 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
6392 char __user
*optval
, int __user
*optlen
)
6397 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
6399 /* I can hardly begin to describe how wrong this is. This is
6400 * so broken as to be worse than useless. The API draft
6401 * REALLY is NOT helpful here... I am not convinced that the
6402 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6403 * are at all well-founded.
6405 if (level
!= SOL_SCTP
) {
6406 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6408 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
6412 if (get_user(len
, optlen
))
6419 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
6421 case SCTP_DISABLE_FRAGMENTS
:
6422 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
6426 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
6428 case SCTP_AUTOCLOSE
:
6429 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
6431 case SCTP_SOCKOPT_PEELOFF
:
6432 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
6434 case SCTP_PEER_ADDR_PARAMS
:
6435 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
6438 case SCTP_DELAYED_SACK
:
6439 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
6443 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
6445 case SCTP_GET_PEER_ADDRS
:
6446 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
6449 case SCTP_GET_LOCAL_ADDRS
:
6450 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
6453 case SCTP_SOCKOPT_CONNECTX3
:
6454 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
6456 case SCTP_DEFAULT_SEND_PARAM
:
6457 retval
= sctp_getsockopt_default_send_param(sk
, len
,
6460 case SCTP_DEFAULT_SNDINFO
:
6461 retval
= sctp_getsockopt_default_sndinfo(sk
, len
,
6464 case SCTP_PRIMARY_ADDR
:
6465 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
6468 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
6471 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
6473 case SCTP_ASSOCINFO
:
6474 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
6476 case SCTP_I_WANT_MAPPED_V4_ADDR
:
6477 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
6480 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
6482 case SCTP_GET_PEER_ADDR_INFO
:
6483 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
6486 case SCTP_ADAPTATION_LAYER
:
6487 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
6491 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
6493 case SCTP_FRAGMENT_INTERLEAVE
:
6494 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
6497 case SCTP_PARTIAL_DELIVERY_POINT
:
6498 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
6501 case SCTP_MAX_BURST
:
6502 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
6505 case SCTP_AUTH_CHUNK
:
6506 case SCTP_AUTH_DELETE_KEY
:
6507 retval
= -EOPNOTSUPP
;
6509 case SCTP_HMAC_IDENT
:
6510 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
6512 case SCTP_AUTH_ACTIVE_KEY
:
6513 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
6515 case SCTP_PEER_AUTH_CHUNKS
:
6516 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
6519 case SCTP_LOCAL_AUTH_CHUNKS
:
6520 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
6523 case SCTP_GET_ASSOC_NUMBER
:
6524 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
6526 case SCTP_GET_ASSOC_ID_LIST
:
6527 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
6529 case SCTP_AUTO_ASCONF
:
6530 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
6532 case SCTP_PEER_ADDR_THLDS
:
6533 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
6535 case SCTP_GET_ASSOC_STATS
:
6536 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
6538 case SCTP_RECVRCVINFO
:
6539 retval
= sctp_getsockopt_recvrcvinfo(sk
, len
, optval
, optlen
);
6541 case SCTP_RECVNXTINFO
:
6542 retval
= sctp_getsockopt_recvnxtinfo(sk
, len
, optval
, optlen
);
6544 case SCTP_PR_SUPPORTED
:
6545 retval
= sctp_getsockopt_pr_supported(sk
, len
, optval
, optlen
);
6547 case SCTP_DEFAULT_PRINFO
:
6548 retval
= sctp_getsockopt_default_prinfo(sk
, len
, optval
,
6551 case SCTP_PR_ASSOC_STATUS
:
6552 retval
= sctp_getsockopt_pr_assocstatus(sk
, len
, optval
,
6556 retval
= -ENOPROTOOPT
;
6564 static int sctp_hash(struct sock
*sk
)
6570 static void sctp_unhash(struct sock
*sk
)
6575 /* Check if port is acceptable. Possibly find first available port.
6577 * The port hash table (contained in the 'global' SCTP protocol storage
6578 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6579 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6580 * list (the list number is the port number hashed out, so as you
6581 * would expect from a hash function, all the ports in a given list have
6582 * such a number that hashes out to the same list number; you were
6583 * expecting that, right?); so each list has a set of ports, with a
6584 * link to the socket (struct sock) that uses it, the port number and
6585 * a fastreuse flag (FIXME: NPI ipg).
6587 static struct sctp_bind_bucket
*sctp_bucket_create(
6588 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
6590 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
6592 struct sctp_bind_hashbucket
*head
; /* hash list */
6593 struct sctp_bind_bucket
*pp
;
6594 unsigned short snum
;
6597 snum
= ntohs(addr
->v4
.sin_port
);
6599 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
6604 /* Search for an available port. */
6605 int low
, high
, remaining
, index
;
6607 struct net
*net
= sock_net(sk
);
6609 inet_get_local_port_range(net
, &low
, &high
);
6610 remaining
= (high
- low
) + 1;
6611 rover
= prandom_u32() % remaining
+ low
;
6615 if ((rover
< low
) || (rover
> high
))
6617 if (inet_is_local_reserved_port(net
, rover
))
6619 index
= sctp_phashfn(sock_net(sk
), rover
);
6620 head
= &sctp_port_hashtable
[index
];
6621 spin_lock(&head
->lock
);
6622 sctp_for_each_hentry(pp
, &head
->chain
)
6623 if ((pp
->port
== rover
) &&
6624 net_eq(sock_net(sk
), pp
->net
))
6628 spin_unlock(&head
->lock
);
6629 } while (--remaining
> 0);
6631 /* Exhausted local port range during search? */
6636 /* OK, here is the one we will use. HEAD (the port
6637 * hash table list entry) is non-NULL and we hold it's
6642 /* We are given an specific port number; we verify
6643 * that it is not being used. If it is used, we will
6644 * exahust the search in the hash list corresponding
6645 * to the port number (snum) - we detect that with the
6646 * port iterator, pp being NULL.
6648 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
6649 spin_lock(&head
->lock
);
6650 sctp_for_each_hentry(pp
, &head
->chain
) {
6651 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
6658 if (!hlist_empty(&pp
->owner
)) {
6659 /* We had a port hash table hit - there is an
6660 * available port (pp != NULL) and it is being
6661 * used by other socket (pp->owner not empty); that other
6662 * socket is going to be sk2.
6664 int reuse
= sk
->sk_reuse
;
6667 pr_debug("%s: found a possible match\n", __func__
);
6669 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6670 sk
->sk_state
!= SCTP_SS_LISTENING
)
6673 /* Run through the list of sockets bound to the port
6674 * (pp->port) [via the pointers bind_next and
6675 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6676 * we get the endpoint they describe and run through
6677 * the endpoint's list of IP (v4 or v6) addresses,
6678 * comparing each of the addresses with the address of
6679 * the socket sk. If we find a match, then that means
6680 * that this port/socket (sk) combination are already
6683 sk_for_each_bound(sk2
, &pp
->owner
) {
6684 struct sctp_endpoint
*ep2
;
6685 ep2
= sctp_sk(sk2
)->ep
;
6688 (reuse
&& sk2
->sk_reuse
&&
6689 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6692 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6693 sctp_sk(sk2
), sctp_sk(sk
))) {
6699 pr_debug("%s: found a match\n", __func__
);
6702 /* If there was a hash table miss, create a new port. */
6704 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6707 /* In either case (hit or miss), make sure fastreuse is 1 only
6708 * if sk->sk_reuse is too (that is, if the caller requested
6709 * SO_REUSEADDR on this socket -sk-).
6711 if (hlist_empty(&pp
->owner
)) {
6712 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6716 } else if (pp
->fastreuse
&&
6717 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6720 /* We are set, so fill up all the data in the hash table
6721 * entry, tie the socket list information with the rest of the
6722 * sockets FIXME: Blurry, NPI (ipg).
6725 if (!sctp_sk(sk
)->bind_hash
) {
6726 inet_sk(sk
)->inet_num
= snum
;
6727 sk_add_bind_node(sk
, &pp
->owner
);
6728 sctp_sk(sk
)->bind_hash
= pp
;
6733 spin_unlock(&head
->lock
);
6740 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6741 * port is requested.
6743 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6745 union sctp_addr addr
;
6746 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6748 /* Set up a dummy address struct from the sk. */
6749 af
->from_sk(&addr
, sk
);
6750 addr
.v4
.sin_port
= htons(snum
);
6752 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6753 return !!sctp_get_port_local(sk
, &addr
);
6757 * Move a socket to LISTENING state.
6759 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6761 struct sctp_sock
*sp
= sctp_sk(sk
);
6762 struct sctp_endpoint
*ep
= sp
->ep
;
6763 struct crypto_shash
*tfm
= NULL
;
6766 /* Allocate HMAC for generating cookie. */
6767 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6768 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6769 tfm
= crypto_alloc_shash(alg
, 0, 0);
6771 net_info_ratelimited("failed to load transform for %s: %ld\n",
6772 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6775 sctp_sk(sk
)->hmac
= tfm
;
6779 * If a bind() or sctp_bindx() is not called prior to a listen()
6780 * call that allows new associations to be accepted, the system
6781 * picks an ephemeral port and will choose an address set equivalent
6782 * to binding with a wildcard address.
6784 * This is not currently spelled out in the SCTP sockets
6785 * extensions draft, but follows the practice as seen in TCP
6789 sk
->sk_state
= SCTP_SS_LISTENING
;
6790 if (!ep
->base
.bind_addr
.port
) {
6791 if (sctp_autobind(sk
))
6794 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6795 sk
->sk_state
= SCTP_SS_CLOSED
;
6800 sk
->sk_max_ack_backlog
= backlog
;
6801 sctp_hash_endpoint(ep
);
6806 * 4.1.3 / 5.1.3 listen()
6808 * By default, new associations are not accepted for UDP style sockets.
6809 * An application uses listen() to mark a socket as being able to
6810 * accept new associations.
6812 * On TCP style sockets, applications use listen() to ready the SCTP
6813 * endpoint for accepting inbound associations.
6815 * On both types of endpoints a backlog of '0' disables listening.
6817 * Move a socket to LISTENING state.
6819 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6821 struct sock
*sk
= sock
->sk
;
6822 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6825 if (unlikely(backlog
< 0))
6830 /* Peeled-off sockets are not allowed to listen(). */
6831 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6834 if (sock
->state
!= SS_UNCONNECTED
)
6837 /* If backlog is zero, disable listening. */
6839 if (sctp_sstate(sk
, CLOSED
))
6843 sctp_unhash_endpoint(ep
);
6844 sk
->sk_state
= SCTP_SS_CLOSED
;
6846 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6850 /* If we are already listening, just update the backlog */
6851 if (sctp_sstate(sk
, LISTENING
))
6852 sk
->sk_max_ack_backlog
= backlog
;
6854 err
= sctp_listen_start(sk
, backlog
);
6866 * This function is done by modeling the current datagram_poll() and the
6867 * tcp_poll(). Note that, based on these implementations, we don't
6868 * lock the socket in this function, even though it seems that,
6869 * ideally, locking or some other mechanisms can be used to ensure
6870 * the integrity of the counters (sndbuf and wmem_alloc) used
6871 * in this place. We assume that we don't need locks either until proven
6874 * Another thing to note is that we include the Async I/O support
6875 * here, again, by modeling the current TCP/UDP code. We don't have
6876 * a good way to test with it yet.
6878 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6880 struct sock
*sk
= sock
->sk
;
6881 struct sctp_sock
*sp
= sctp_sk(sk
);
6884 poll_wait(file
, sk_sleep(sk
), wait
);
6886 sock_rps_record_flow(sk
);
6888 /* A TCP-style listening socket becomes readable when the accept queue
6891 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6892 return (!list_empty(&sp
->ep
->asocs
)) ?
6893 (POLLIN
| POLLRDNORM
) : 0;
6897 /* Is there any exceptional events? */
6898 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6900 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6901 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6902 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6903 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6906 /* Is it readable? Reconsider this code with TCP-style support. */
6907 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6908 mask
|= POLLIN
| POLLRDNORM
;
6910 /* The association is either gone or not ready. */
6911 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6914 /* Is it writable? */
6915 if (sctp_writeable(sk
)) {
6916 mask
|= POLLOUT
| POLLWRNORM
;
6918 sk_set_bit(SOCKWQ_ASYNC_NOSPACE
, sk
);
6920 * Since the socket is not locked, the buffer
6921 * might be made available after the writeable check and
6922 * before the bit is set. This could cause a lost I/O
6923 * signal. tcp_poll() has a race breaker for this race
6924 * condition. Based on their implementation, we put
6925 * in the following code to cover it as well.
6927 if (sctp_writeable(sk
))
6928 mask
|= POLLOUT
| POLLWRNORM
;
6933 /********************************************************************
6934 * 2nd Level Abstractions
6935 ********************************************************************/
6937 static struct sctp_bind_bucket
*sctp_bucket_create(
6938 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6940 struct sctp_bind_bucket
*pp
;
6942 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6944 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6947 INIT_HLIST_HEAD(&pp
->owner
);
6949 hlist_add_head(&pp
->node
, &head
->chain
);
6954 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6955 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6957 if (pp
&& hlist_empty(&pp
->owner
)) {
6958 __hlist_del(&pp
->node
);
6959 kmem_cache_free(sctp_bucket_cachep
, pp
);
6960 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6964 /* Release this socket's reference to a local port. */
6965 static inline void __sctp_put_port(struct sock
*sk
)
6967 struct sctp_bind_hashbucket
*head
=
6968 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6969 inet_sk(sk
)->inet_num
)];
6970 struct sctp_bind_bucket
*pp
;
6972 spin_lock(&head
->lock
);
6973 pp
= sctp_sk(sk
)->bind_hash
;
6974 __sk_del_bind_node(sk
);
6975 sctp_sk(sk
)->bind_hash
= NULL
;
6976 inet_sk(sk
)->inet_num
= 0;
6977 sctp_bucket_destroy(pp
);
6978 spin_unlock(&head
->lock
);
6981 void sctp_put_port(struct sock
*sk
)
6984 __sctp_put_port(sk
);
6989 * The system picks an ephemeral port and choose an address set equivalent
6990 * to binding with a wildcard address.
6991 * One of those addresses will be the primary address for the association.
6992 * This automatically enables the multihoming capability of SCTP.
6994 static int sctp_autobind(struct sock
*sk
)
6996 union sctp_addr autoaddr
;
7000 /* Initialize a local sockaddr structure to INADDR_ANY. */
7001 af
= sctp_sk(sk
)->pf
->af
;
7003 port
= htons(inet_sk(sk
)->inet_num
);
7004 af
->inaddr_any(&autoaddr
, port
);
7006 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
7009 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7012 * 4.2 The cmsghdr Structure *
7014 * When ancillary data is sent or received, any number of ancillary data
7015 * objects can be specified by the msg_control and msg_controllen members of
7016 * the msghdr structure, because each object is preceded by
7017 * a cmsghdr structure defining the object's length (the cmsg_len member).
7018 * Historically Berkeley-derived implementations have passed only one object
7019 * at a time, but this API allows multiple objects to be
7020 * passed in a single call to sendmsg() or recvmsg(). The following example
7021 * shows two ancillary data objects in a control buffer.
7023 * |<--------------------------- msg_controllen -------------------------->|
7026 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7028 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7031 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7033 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7036 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7037 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7039 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7041 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7048 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
7050 struct cmsghdr
*cmsg
;
7051 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
7053 for_each_cmsghdr(cmsg
, my_msg
) {
7054 if (!CMSG_OK(my_msg
, cmsg
))
7057 /* Should we parse this header or ignore? */
7058 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
7061 /* Strictly check lengths following example in SCM code. */
7062 switch (cmsg
->cmsg_type
) {
7064 /* SCTP Socket API Extension
7065 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7067 * This cmsghdr structure provides information for
7068 * initializing new SCTP associations with sendmsg().
7069 * The SCTP_INITMSG socket option uses this same data
7070 * structure. This structure is not used for
7073 * cmsg_level cmsg_type cmsg_data[]
7074 * ------------ ------------ ----------------------
7075 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7077 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_initmsg
)))
7080 cmsgs
->init
= CMSG_DATA(cmsg
);
7084 /* SCTP Socket API Extension
7085 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7087 * This cmsghdr structure specifies SCTP options for
7088 * sendmsg() and describes SCTP header information
7089 * about a received message through recvmsg().
7091 * cmsg_level cmsg_type cmsg_data[]
7092 * ------------ ------------ ----------------------
7093 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7095 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
7098 cmsgs
->srinfo
= CMSG_DATA(cmsg
);
7100 if (cmsgs
->srinfo
->sinfo_flags
&
7101 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7102 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7103 SCTP_ABORT
| SCTP_EOF
))
7108 /* SCTP Socket API Extension
7109 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7111 * This cmsghdr structure specifies SCTP options for
7112 * sendmsg(). This structure and SCTP_RCVINFO replaces
7113 * SCTP_SNDRCV which has been deprecated.
7115 * cmsg_level cmsg_type cmsg_data[]
7116 * ------------ ------------ ---------------------
7117 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7119 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndinfo
)))
7122 cmsgs
->sinfo
= CMSG_DATA(cmsg
);
7124 if (cmsgs
->sinfo
->snd_flags
&
7125 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7126 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7127 SCTP_ABORT
| SCTP_EOF
))
7139 * Wait for a packet..
7140 * Note: This function is the same function as in core/datagram.c
7141 * with a few modifications to make lksctp work.
7143 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
7148 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7150 /* Socket errors? */
7151 error
= sock_error(sk
);
7155 if (!skb_queue_empty(&sk
->sk_receive_queue
))
7158 /* Socket shut down? */
7159 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7162 /* Sequenced packets can come disconnected. If so we report the
7167 /* Is there a good reason to think that we may receive some data? */
7168 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
7171 /* Handle signals. */
7172 if (signal_pending(current
))
7175 /* Let another process have a go. Since we are going to sleep
7176 * anyway. Note: This may cause odd behaviors if the message
7177 * does not fit in the user's buffer, but this seems to be the
7178 * only way to honor MSG_DONTWAIT realistically.
7181 *timeo_p
= schedule_timeout(*timeo_p
);
7185 finish_wait(sk_sleep(sk
), &wait
);
7189 error
= sock_intr_errno(*timeo_p
);
7192 finish_wait(sk_sleep(sk
), &wait
);
7197 /* Receive a datagram.
7198 * Note: This is pretty much the same routine as in core/datagram.c
7199 * with a few changes to make lksctp work.
7201 struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
7202 int noblock
, int *err
)
7205 struct sk_buff
*skb
;
7208 timeo
= sock_rcvtimeo(sk
, noblock
);
7210 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
7211 MAX_SCHEDULE_TIMEOUT
);
7214 /* Again only user level code calls this function,
7215 * so nothing interrupt level
7216 * will suddenly eat the receive_queue.
7218 * Look at current nfs client by the way...
7219 * However, this function was correct in any case. 8)
7221 if (flags
& MSG_PEEK
) {
7222 skb
= skb_peek(&sk
->sk_receive_queue
);
7224 atomic_inc(&skb
->users
);
7226 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
7232 /* Caller is allowed not to check sk->sk_err before calling. */
7233 error
= sock_error(sk
);
7237 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7240 if (sk_can_busy_loop(sk
) &&
7241 sk_busy_loop(sk
, noblock
))
7244 /* User doesn't want to wait. */
7248 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
7257 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7258 static void __sctp_write_space(struct sctp_association
*asoc
)
7260 struct sock
*sk
= asoc
->base
.sk
;
7262 if (sctp_wspace(asoc
) <= 0)
7265 if (waitqueue_active(&asoc
->wait
))
7266 wake_up_interruptible(&asoc
->wait
);
7268 if (sctp_writeable(sk
)) {
7269 struct socket_wq
*wq
;
7272 wq
= rcu_dereference(sk
->sk_wq
);
7274 if (waitqueue_active(&wq
->wait
))
7275 wake_up_interruptible(&wq
->wait
);
7277 /* Note that we try to include the Async I/O support
7278 * here by modeling from the current TCP/UDP code.
7279 * We have not tested with it yet.
7281 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
7282 sock_wake_async(wq
, SOCK_WAKE_SPACE
, POLL_OUT
);
7288 static void sctp_wake_up_waiters(struct sock
*sk
,
7289 struct sctp_association
*asoc
)
7291 struct sctp_association
*tmp
= asoc
;
7293 /* We do accounting for the sndbuf space per association,
7294 * so we only need to wake our own association.
7296 if (asoc
->ep
->sndbuf_policy
)
7297 return __sctp_write_space(asoc
);
7299 /* If association goes down and is just flushing its
7300 * outq, then just normally notify others.
7302 if (asoc
->base
.dead
)
7303 return sctp_write_space(sk
);
7305 /* Accounting for the sndbuf space is per socket, so we
7306 * need to wake up others, try to be fair and in case of
7307 * other associations, let them have a go first instead
7308 * of just doing a sctp_write_space() call.
7310 * Note that we reach sctp_wake_up_waiters() only when
7311 * associations free up queued chunks, thus we are under
7312 * lock and the list of associations on a socket is
7313 * guaranteed not to change.
7315 for (tmp
= list_next_entry(tmp
, asocs
); 1;
7316 tmp
= list_next_entry(tmp
, asocs
)) {
7317 /* Manually skip the head element. */
7318 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
7320 /* Wake up association. */
7321 __sctp_write_space(tmp
);
7322 /* We've reached the end. */
7328 /* Do accounting for the sndbuf space.
7329 * Decrement the used sndbuf space of the corresponding association by the
7330 * data size which was just transmitted(freed).
7332 static void sctp_wfree(struct sk_buff
*skb
)
7334 struct sctp_chunk
*chunk
= skb_shinfo(skb
)->destructor_arg
;
7335 struct sctp_association
*asoc
= chunk
->asoc
;
7336 struct sock
*sk
= asoc
->base
.sk
;
7338 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
7339 sizeof(struct sk_buff
) +
7340 sizeof(struct sctp_chunk
);
7342 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
7345 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7347 sk
->sk_wmem_queued
-= skb
->truesize
;
7348 sk_mem_uncharge(sk
, skb
->truesize
);
7351 sctp_wake_up_waiters(sk
, asoc
);
7353 sctp_association_put(asoc
);
7356 /* Do accounting for the receive space on the socket.
7357 * Accounting for the association is done in ulpevent.c
7358 * We set this as a destructor for the cloned data skbs so that
7359 * accounting is done at the correct time.
7361 void sctp_sock_rfree(struct sk_buff
*skb
)
7363 struct sock
*sk
= skb
->sk
;
7364 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
7366 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
7369 * Mimic the behavior of sock_rfree
7371 sk_mem_uncharge(sk
, event
->rmem_len
);
7375 /* Helper function to wait for space in the sndbuf. */
7376 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
7379 struct sock
*sk
= asoc
->base
.sk
;
7381 long current_timeo
= *timeo_p
;
7384 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
7387 /* Increment the association's refcnt. */
7388 sctp_association_hold(asoc
);
7390 /* Wait on the association specific sndbuf space. */
7392 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7393 TASK_INTERRUPTIBLE
);
7396 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7399 if (signal_pending(current
))
7400 goto do_interrupted
;
7401 if (msg_len
<= sctp_wspace(asoc
))
7404 /* Let another process have a go. Since we are going
7408 current_timeo
= schedule_timeout(current_timeo
);
7409 BUG_ON(sk
!= asoc
->base
.sk
);
7412 *timeo_p
= current_timeo
;
7416 finish_wait(&asoc
->wait
, &wait
);
7418 /* Release the association's refcnt. */
7419 sctp_association_put(asoc
);
7428 err
= sock_intr_errno(*timeo_p
);
7436 void sctp_data_ready(struct sock
*sk
)
7438 struct socket_wq
*wq
;
7441 wq
= rcu_dereference(sk
->sk_wq
);
7442 if (skwq_has_sleeper(wq
))
7443 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
7444 POLLRDNORM
| POLLRDBAND
);
7445 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
7449 /* If socket sndbuf has changed, wake up all per association waiters. */
7450 void sctp_write_space(struct sock
*sk
)
7452 struct sctp_association
*asoc
;
7454 /* Wake up the tasks in each wait queue. */
7455 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
7456 __sctp_write_space(asoc
);
7460 /* Is there any sndbuf space available on the socket?
7462 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7463 * associations on the same socket. For a UDP-style socket with
7464 * multiple associations, it is possible for it to be "unwriteable"
7465 * prematurely. I assume that this is acceptable because
7466 * a premature "unwriteable" is better than an accidental "writeable" which
7467 * would cause an unwanted block under certain circumstances. For the 1-1
7468 * UDP-style sockets or TCP-style sockets, this code should work.
7471 static int sctp_writeable(struct sock
*sk
)
7475 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
7481 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7482 * returns immediately with EINPROGRESS.
7484 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
7486 struct sock
*sk
= asoc
->base
.sk
;
7488 long current_timeo
= *timeo_p
;
7491 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
7493 /* Increment the association's refcnt. */
7494 sctp_association_hold(asoc
);
7497 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7498 TASK_INTERRUPTIBLE
);
7501 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7503 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7506 if (signal_pending(current
))
7507 goto do_interrupted
;
7509 if (sctp_state(asoc
, ESTABLISHED
))
7512 /* Let another process have a go. Since we are going
7516 current_timeo
= schedule_timeout(current_timeo
);
7519 *timeo_p
= current_timeo
;
7523 finish_wait(&asoc
->wait
, &wait
);
7525 /* Release the association's refcnt. */
7526 sctp_association_put(asoc
);
7531 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
7534 err
= -ECONNREFUSED
;
7538 err
= sock_intr_errno(*timeo_p
);
7546 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
7548 struct sctp_endpoint
*ep
;
7552 ep
= sctp_sk(sk
)->ep
;
7556 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
7557 TASK_INTERRUPTIBLE
);
7559 if (list_empty(&ep
->asocs
)) {
7561 timeo
= schedule_timeout(timeo
);
7566 if (!sctp_sstate(sk
, LISTENING
))
7570 if (!list_empty(&ep
->asocs
))
7573 err
= sock_intr_errno(timeo
);
7574 if (signal_pending(current
))
7582 finish_wait(sk_sleep(sk
), &wait
);
7587 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
7592 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7593 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
7596 timeout
= schedule_timeout(timeout
);
7598 } while (!signal_pending(current
) && timeout
);
7600 finish_wait(sk_sleep(sk
), &wait
);
7603 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
7605 struct sk_buff
*frag
;
7610 /* Don't forget the fragments. */
7611 skb_walk_frags(skb
, frag
)
7612 sctp_skb_set_owner_r_frag(frag
, sk
);
7615 sctp_skb_set_owner_r(skb
, sk
);
7618 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
7619 struct sctp_association
*asoc
)
7621 struct inet_sock
*inet
= inet_sk(sk
);
7622 struct inet_sock
*newinet
;
7624 newsk
->sk_type
= sk
->sk_type
;
7625 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
7626 newsk
->sk_flags
= sk
->sk_flags
;
7627 newsk
->sk_tsflags
= sk
->sk_tsflags
;
7628 newsk
->sk_no_check_tx
= sk
->sk_no_check_tx
;
7629 newsk
->sk_no_check_rx
= sk
->sk_no_check_rx
;
7630 newsk
->sk_reuse
= sk
->sk_reuse
;
7632 newsk
->sk_shutdown
= sk
->sk_shutdown
;
7633 newsk
->sk_destruct
= sctp_destruct_sock
;
7634 newsk
->sk_family
= sk
->sk_family
;
7635 newsk
->sk_protocol
= IPPROTO_SCTP
;
7636 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
7637 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
7638 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
7639 newsk
->sk_lingertime
= sk
->sk_lingertime
;
7640 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
7641 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
7642 newsk
->sk_rxhash
= sk
->sk_rxhash
;
7644 newinet
= inet_sk(newsk
);
7646 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7647 * getsockname() and getpeername()
7649 newinet
->inet_sport
= inet
->inet_sport
;
7650 newinet
->inet_saddr
= inet
->inet_saddr
;
7651 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
7652 newinet
->inet_dport
= htons(asoc
->peer
.port
);
7653 newinet
->pmtudisc
= inet
->pmtudisc
;
7654 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
7656 newinet
->uc_ttl
= inet
->uc_ttl
;
7657 newinet
->mc_loop
= 1;
7658 newinet
->mc_ttl
= 1;
7659 newinet
->mc_index
= 0;
7660 newinet
->mc_list
= NULL
;
7662 if (newsk
->sk_flags
& SK_FLAGS_TIMESTAMP
)
7663 net_enable_timestamp();
7665 security_sk_clone(sk
, newsk
);
7668 static inline void sctp_copy_descendant(struct sock
*sk_to
,
7669 const struct sock
*sk_from
)
7671 int ancestor_size
= sizeof(struct inet_sock
) +
7672 sizeof(struct sctp_sock
) -
7673 offsetof(struct sctp_sock
, auto_asconf_list
);
7675 if (sk_from
->sk_family
== PF_INET6
)
7676 ancestor_size
+= sizeof(struct ipv6_pinfo
);
7678 __inet_sk_copy_descendant(sk_to
, sk_from
, ancestor_size
);
7681 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7682 * and its messages to the newsk.
7684 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
7685 struct sctp_association
*assoc
,
7686 sctp_socket_type_t type
)
7688 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
7689 struct sctp_sock
*newsp
= sctp_sk(newsk
);
7690 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
7691 struct sctp_endpoint
*newep
= newsp
->ep
;
7692 struct sk_buff
*skb
, *tmp
;
7693 struct sctp_ulpevent
*event
;
7694 struct sctp_bind_hashbucket
*head
;
7696 /* Migrate socket buffer sizes and all the socket level options to the
7699 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
7700 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
7701 /* Brute force copy old sctp opt. */
7702 sctp_copy_descendant(newsk
, oldsk
);
7704 /* Restore the ep value that was overwritten with the above structure
7710 /* Hook this new socket in to the bind_hash list. */
7711 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7712 inet_sk(oldsk
)->inet_num
)];
7713 spin_lock_bh(&head
->lock
);
7714 pp
= sctp_sk(oldsk
)->bind_hash
;
7715 sk_add_bind_node(newsk
, &pp
->owner
);
7716 sctp_sk(newsk
)->bind_hash
= pp
;
7717 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7718 spin_unlock_bh(&head
->lock
);
7720 /* Copy the bind_addr list from the original endpoint to the new
7721 * endpoint so that we can handle restarts properly
7723 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7724 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7726 /* Move any messages in the old socket's receive queue that are for the
7727 * peeled off association to the new socket's receive queue.
7729 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7730 event
= sctp_skb2event(skb
);
7731 if (event
->asoc
== assoc
) {
7732 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7733 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7734 sctp_skb_set_owner_r_frag(skb
, newsk
);
7738 /* Clean up any messages pending delivery due to partial
7739 * delivery. Three cases:
7740 * 1) No partial deliver; no work.
7741 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7742 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7744 skb_queue_head_init(&newsp
->pd_lobby
);
7745 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7747 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7748 struct sk_buff_head
*queue
;
7750 /* Decide which queue to move pd_lobby skbs to. */
7751 if (assoc
->ulpq
.pd_mode
) {
7752 queue
= &newsp
->pd_lobby
;
7754 queue
= &newsk
->sk_receive_queue
;
7756 /* Walk through the pd_lobby, looking for skbs that
7757 * need moved to the new socket.
7759 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7760 event
= sctp_skb2event(skb
);
7761 if (event
->asoc
== assoc
) {
7762 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7763 __skb_queue_tail(queue
, skb
);
7764 sctp_skb_set_owner_r_frag(skb
, newsk
);
7768 /* Clear up any skbs waiting for the partial
7769 * delivery to finish.
7771 if (assoc
->ulpq
.pd_mode
)
7772 sctp_clear_pd(oldsk
, NULL
);
7776 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7777 sctp_skb_set_owner_r_frag(skb
, newsk
);
7779 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7780 sctp_skb_set_owner_r_frag(skb
, newsk
);
7782 /* Set the type of socket to indicate that it is peeled off from the
7783 * original UDP-style socket or created with the accept() call on a
7784 * TCP-style socket..
7788 /* Mark the new socket "in-use" by the user so that any packets
7789 * that may arrive on the association after we've moved it are
7790 * queued to the backlog. This prevents a potential race between
7791 * backlog processing on the old socket and new-packet processing
7792 * on the new socket.
7794 * The caller has just allocated newsk so we can guarantee that other
7795 * paths won't try to lock it and then oldsk.
7797 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7798 sctp_assoc_migrate(assoc
, newsk
);
7800 /* If the association on the newsk is already closed before accept()
7801 * is called, set RCV_SHUTDOWN flag.
7803 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
)) {
7804 newsk
->sk_state
= SCTP_SS_CLOSED
;
7805 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7807 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7810 release_sock(newsk
);
7814 /* This proto struct describes the ULP interface for SCTP. */
7815 struct proto sctp_prot
= {
7817 .owner
= THIS_MODULE
,
7818 .close
= sctp_close
,
7819 .connect
= sctp_connect
,
7820 .disconnect
= sctp_disconnect
,
7821 .accept
= sctp_accept
,
7822 .ioctl
= sctp_ioctl
,
7823 .init
= sctp_init_sock
,
7824 .destroy
= sctp_destroy_sock
,
7825 .shutdown
= sctp_shutdown
,
7826 .setsockopt
= sctp_setsockopt
,
7827 .getsockopt
= sctp_getsockopt
,
7828 .sendmsg
= sctp_sendmsg
,
7829 .recvmsg
= sctp_recvmsg
,
7831 .backlog_rcv
= sctp_backlog_rcv
,
7833 .unhash
= sctp_unhash
,
7834 .get_port
= sctp_get_port
,
7835 .obj_size
= sizeof(struct sctp_sock
),
7836 .sysctl_mem
= sysctl_sctp_mem
,
7837 .sysctl_rmem
= sysctl_sctp_rmem
,
7838 .sysctl_wmem
= sysctl_sctp_wmem
,
7839 .memory_pressure
= &sctp_memory_pressure
,
7840 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7841 .memory_allocated
= &sctp_memory_allocated
,
7842 .sockets_allocated
= &sctp_sockets_allocated
,
7845 #if IS_ENABLED(CONFIG_IPV6)
7847 #include <net/transp_v6.h>
7848 static void sctp_v6_destroy_sock(struct sock
*sk
)
7850 sctp_destroy_sock(sk
);
7851 inet6_destroy_sock(sk
);
7854 struct proto sctpv6_prot
= {
7856 .owner
= THIS_MODULE
,
7857 .close
= sctp_close
,
7858 .connect
= sctp_connect
,
7859 .disconnect
= sctp_disconnect
,
7860 .accept
= sctp_accept
,
7861 .ioctl
= sctp_ioctl
,
7862 .init
= sctp_init_sock
,
7863 .destroy
= sctp_v6_destroy_sock
,
7864 .shutdown
= sctp_shutdown
,
7865 .setsockopt
= sctp_setsockopt
,
7866 .getsockopt
= sctp_getsockopt
,
7867 .sendmsg
= sctp_sendmsg
,
7868 .recvmsg
= sctp_recvmsg
,
7870 .backlog_rcv
= sctp_backlog_rcv
,
7872 .unhash
= sctp_unhash
,
7873 .get_port
= sctp_get_port
,
7874 .obj_size
= sizeof(struct sctp6_sock
),
7875 .sysctl_mem
= sysctl_sctp_mem
,
7876 .sysctl_rmem
= sysctl_sctp_rmem
,
7877 .sysctl_wmem
= sysctl_sctp_wmem
,
7878 .memory_pressure
= &sctp_memory_pressure
,
7879 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7880 .memory_allocated
= &sctp_memory_allocated
,
7881 .sockets_allocated
= &sctp_sockets_allocated
,
7883 #endif /* IS_ENABLED(CONFIG_IPV6) */