1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock
*sk
);
84 static void sctp_wfree(struct sk_buff
*skb
);
85 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
87 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
);
88 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
89 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
90 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
91 static void sctp_destruct_sock(struct sock
*sk
);
92 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
93 union sctp_addr
*addr
, int len
);
94 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
95 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
96 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
97 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
98 static int sctp_send_asconf(struct sctp_association
*asoc
,
99 struct sctp_chunk
*chunk
);
100 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
101 static int sctp_autobind(struct sock
*sk
);
102 static void sctp_sock_migrate(struct sock
*, struct sock
*,
103 struct sctp_association
*, sctp_socket_type_t
);
105 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
, SCTP_ADDR_SRC
, GFP_ATOMIC
);
391 /* Copy back into socket for getsockname() use. */
393 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
394 sp
->pf
->to_sk_saddr(addr
, sk
);
400 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
402 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
403 * at any one time. If a sender, after sending an ASCONF chunk, decides
404 * it needs to transfer another ASCONF Chunk, it MUST wait until the
405 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
406 * subsequent ASCONF. Note this restriction binds each side, so at any
407 * time two ASCONF may be in-transit on any given association (one sent
408 * from each endpoint).
410 static int sctp_send_asconf(struct sctp_association
*asoc
,
411 struct sctp_chunk
*chunk
)
413 struct net
*net
= sock_net(asoc
->base
.sk
);
416 /* If there is an outstanding ASCONF chunk, queue it for later
419 if (asoc
->addip_last_asconf
) {
420 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
424 /* Hold the chunk until an ASCONF_ACK is received. */
425 sctp_chunk_hold(chunk
);
426 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
428 sctp_chunk_free(chunk
);
430 asoc
->addip_last_asconf
= chunk
;
436 /* Add a list of addresses as bind addresses to local endpoint or
439 * Basically run through each address specified in the addrs/addrcnt
440 * array/length pair, determine if it is IPv6 or IPv4 and call
441 * sctp_do_bind() on it.
443 * If any of them fails, then the operation will be reversed and the
444 * ones that were added will be removed.
446 * Only sctp_setsockopt_bindx() is supposed to call this function.
448 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
453 struct sockaddr
*sa_addr
;
456 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__
, sk
,
460 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
461 /* The list may contain either IPv4 or IPv6 address;
462 * determine the address length for walking thru the list.
465 af
= sctp_get_af_specific(sa_addr
->sa_family
);
471 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
474 addr_buf
+= af
->sockaddr_len
;
478 /* Failed. Cleanup the ones that have been added */
480 sctp_bindx_rem(sk
, addrs
, cnt
);
488 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
489 * associations that are part of the endpoint indicating that a list of local
490 * addresses are added to the endpoint.
492 * If any of the addresses is already in the bind address list of the
493 * association, we do not send the chunk for that association. But it will not
494 * affect other associations.
496 * Only sctp_setsockopt_bindx() is supposed to call this function.
498 static int sctp_send_asconf_add_ip(struct sock
*sk
,
499 struct sockaddr
*addrs
,
502 struct net
*net
= sock_net(sk
);
503 struct sctp_sock
*sp
;
504 struct sctp_endpoint
*ep
;
505 struct sctp_association
*asoc
;
506 struct sctp_bind_addr
*bp
;
507 struct sctp_chunk
*chunk
;
508 struct sctp_sockaddr_entry
*laddr
;
509 union sctp_addr
*addr
;
510 union sctp_addr saveaddr
;
517 if (!net
->sctp
.addip_enable
)
523 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
524 __func__
, sk
, addrs
, addrcnt
);
526 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
527 if (!asoc
->peer
.asconf_capable
)
530 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
533 if (!sctp_state(asoc
, ESTABLISHED
))
536 /* Check if any address in the packed array of addresses is
537 * in the bind address list of the association. If so,
538 * do not send the asconf chunk to its peer, but continue with
539 * other associations.
542 for (i
= 0; i
< addrcnt
; i
++) {
544 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
550 if (sctp_assoc_lookup_laddr(asoc
, addr
))
553 addr_buf
+= af
->sockaddr_len
;
558 /* Use the first valid address in bind addr list of
559 * association as Address Parameter of ASCONF CHUNK.
561 bp
= &asoc
->base
.bind_addr
;
562 p
= bp
->address_list
.next
;
563 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
564 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
565 addrcnt
, SCTP_PARAM_ADD_IP
);
571 /* Add the new addresses to the bind address list with
572 * use_as_src set to 0.
575 for (i
= 0; i
< addrcnt
; i
++) {
577 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
578 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
579 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
580 SCTP_ADDR_NEW
, GFP_ATOMIC
);
581 addr_buf
+= af
->sockaddr_len
;
583 if (asoc
->src_out_of_asoc_ok
) {
584 struct sctp_transport
*trans
;
586 list_for_each_entry(trans
,
587 &asoc
->peer
.transport_addr_list
, transports
) {
588 /* Clear the source and route cache */
589 dst_release(trans
->dst
);
590 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
591 2*asoc
->pathmtu
, 4380));
592 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
593 trans
->rto
= asoc
->rto_initial
;
594 sctp_max_rto(asoc
, trans
);
595 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
596 sctp_transport_route(trans
, NULL
,
597 sctp_sk(asoc
->base
.sk
));
600 retval
= sctp_send_asconf(asoc
, chunk
);
607 /* Remove a list of addresses from bind addresses list. Do not remove the
610 * Basically run through each address specified in the addrs/addrcnt
611 * array/length pair, determine if it is IPv6 or IPv4 and call
612 * sctp_del_bind() on it.
614 * If any of them fails, then the operation will be reversed and the
615 * ones that were removed will be added back.
617 * At least one address has to be left; if only one address is
618 * available, the operation will return -EBUSY.
620 * Only sctp_setsockopt_bindx() is supposed to call this function.
622 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
624 struct sctp_sock
*sp
= sctp_sk(sk
);
625 struct sctp_endpoint
*ep
= sp
->ep
;
627 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
630 union sctp_addr
*sa_addr
;
633 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
634 __func__
, sk
, addrs
, addrcnt
);
637 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
638 /* If the bind address list is empty or if there is only one
639 * bind address, there is nothing more to be removed (we need
640 * at least one address here).
642 if (list_empty(&bp
->address_list
) ||
643 (sctp_list_single_entry(&bp
->address_list
))) {
649 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
655 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
656 retval
= -EADDRNOTAVAIL
;
660 if (sa_addr
->v4
.sin_port
&&
661 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
666 if (!sa_addr
->v4
.sin_port
)
667 sa_addr
->v4
.sin_port
= htons(bp
->port
);
669 /* FIXME - There is probably a need to check if sk->sk_saddr and
670 * sk->sk_rcv_addr are currently set to one of the addresses to
671 * be removed. This is something which needs to be looked into
672 * when we are fixing the outstanding issues with multi-homing
673 * socket routing and failover schemes. Refer to comments in
674 * sctp_do_bind(). -daisy
676 retval
= sctp_del_bind_addr(bp
, sa_addr
);
678 addr_buf
+= af
->sockaddr_len
;
681 /* Failed. Add the ones that has been removed back */
683 sctp_bindx_add(sk
, addrs
, cnt
);
691 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
692 * the associations that are part of the endpoint indicating that a list of
693 * local addresses are removed from the endpoint.
695 * If any of the addresses is already in the bind address list of the
696 * association, we do not send the chunk for that association. But it will not
697 * affect other associations.
699 * Only sctp_setsockopt_bindx() is supposed to call this function.
701 static int sctp_send_asconf_del_ip(struct sock
*sk
,
702 struct sockaddr
*addrs
,
705 struct net
*net
= sock_net(sk
);
706 struct sctp_sock
*sp
;
707 struct sctp_endpoint
*ep
;
708 struct sctp_association
*asoc
;
709 struct sctp_transport
*transport
;
710 struct sctp_bind_addr
*bp
;
711 struct sctp_chunk
*chunk
;
712 union sctp_addr
*laddr
;
715 struct sctp_sockaddr_entry
*saddr
;
721 if (!net
->sctp
.addip_enable
)
727 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
728 __func__
, sk
, addrs
, addrcnt
);
730 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
732 if (!asoc
->peer
.asconf_capable
)
735 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
738 if (!sctp_state(asoc
, ESTABLISHED
))
741 /* Check if any address in the packed array of addresses is
742 * not present in the bind address list of the association.
743 * If so, do not send the asconf chunk to its peer, but
744 * continue with other associations.
747 for (i
= 0; i
< addrcnt
; i
++) {
749 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
755 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
758 addr_buf
+= af
->sockaddr_len
;
763 /* Find one address in the association's bind address list
764 * that is not in the packed array of addresses. This is to
765 * make sure that we do not delete all the addresses in the
768 bp
= &asoc
->base
.bind_addr
;
769 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
771 if ((laddr
== NULL
) && (addrcnt
== 1)) {
772 if (asoc
->asconf_addr_del_pending
)
774 asoc
->asconf_addr_del_pending
=
775 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
776 if (asoc
->asconf_addr_del_pending
== NULL
) {
780 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
782 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
784 if (addrs
->sa_family
== AF_INET
) {
785 struct sockaddr_in
*sin
;
787 sin
= (struct sockaddr_in
*)addrs
;
788 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
789 } else if (addrs
->sa_family
== AF_INET6
) {
790 struct sockaddr_in6
*sin6
;
792 sin6
= (struct sockaddr_in6
*)addrs
;
793 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
796 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
797 __func__
, asoc
, &asoc
->asconf_addr_del_pending
->sa
,
798 asoc
->asconf_addr_del_pending
);
800 asoc
->src_out_of_asoc_ok
= 1;
808 /* We do not need RCU protection throughout this loop
809 * because this is done under a socket lock from the
812 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
820 /* Reset use_as_src flag for the addresses in the bind address
821 * list that are to be deleted.
824 for (i
= 0; i
< addrcnt
; i
++) {
826 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
827 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
828 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
829 saddr
->state
= SCTP_ADDR_DEL
;
831 addr_buf
+= af
->sockaddr_len
;
834 /* Update the route and saddr entries for all the transports
835 * as some of the addresses in the bind address list are
836 * about to be deleted and cannot be used as source addresses.
838 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
840 dst_release(transport
->dst
);
841 sctp_transport_route(transport
, NULL
,
842 sctp_sk(asoc
->base
.sk
));
846 /* We don't need to transmit ASCONF */
848 retval
= sctp_send_asconf(asoc
, chunk
);
854 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
855 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
857 struct sock
*sk
= sctp_opt2sk(sp
);
858 union sctp_addr
*addr
;
861 /* It is safe to write port space in caller. */
863 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
864 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
867 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
870 if (addrw
->state
== SCTP_ADDR_NEW
)
871 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
873 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
876 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
879 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
882 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
883 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
886 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
887 * Section 3.1.2 for this usage.
889 * addrs is a pointer to an array of one or more socket addresses. Each
890 * address is contained in its appropriate structure (i.e. struct
891 * sockaddr_in or struct sockaddr_in6) the family of the address type
892 * must be used to distinguish the address length (note that this
893 * representation is termed a "packed array" of addresses). The caller
894 * specifies the number of addresses in the array with addrcnt.
896 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
897 * -1, and sets errno to the appropriate error code.
899 * For SCTP, the port given in each socket address must be the same, or
900 * sctp_bindx() will fail, setting errno to EINVAL.
902 * The flags parameter is formed from the bitwise OR of zero or more of
903 * the following currently defined flags:
905 * SCTP_BINDX_ADD_ADDR
907 * SCTP_BINDX_REM_ADDR
909 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
910 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
911 * addresses from the association. The two flags are mutually exclusive;
912 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
913 * not remove all addresses from an association; sctp_bindx() will
914 * reject such an attempt with EINVAL.
916 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
917 * additional addresses with an endpoint after calling bind(). Or use
918 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
919 * socket is associated with so that no new association accepted will be
920 * associated with those addresses. If the endpoint supports dynamic
921 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
922 * endpoint to send the appropriate message to the peer to change the
923 * peers address lists.
925 * Adding and removing addresses from a connected association is
926 * optional functionality. Implementations that do not support this
927 * functionality should return EOPNOTSUPP.
929 * Basically do nothing but copying the addresses from user to kernel
930 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
931 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
934 * We don't use copy_from_user() for optimization: we first do the
935 * sanity checks (buffer size -fast- and access check-healthy
936 * pointer); if all of those succeed, then we can alloc the memory
937 * (expensive operation) needed to copy the data to kernel. Then we do
938 * the copying without checking the user space area
939 * (__copy_from_user()).
941 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
944 * sk The sk of the socket
945 * addrs The pointer to the addresses in user land
946 * addrssize Size of the addrs buffer
947 * op Operation to perform (add or remove, see the flags of
950 * Returns 0 if ok, <0 errno code on error.
952 static int sctp_setsockopt_bindx(struct sock
*sk
,
953 struct sockaddr __user
*addrs
,
954 int addrs_size
, int op
)
956 struct sockaddr
*kaddrs
;
960 struct sockaddr
*sa_addr
;
964 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
965 __func__
, sk
, addrs
, addrs_size
, op
);
967 if (unlikely(addrs_size
<= 0))
970 /* Check the user passed a healthy pointer. */
971 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
974 /* Alloc space for the address array in kernel memory. */
975 kaddrs
= kmalloc(addrs_size
, GFP_USER
| __GFP_NOWARN
);
976 if (unlikely(!kaddrs
))
979 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
984 /* Walk through the addrs buffer and count the number of addresses. */
986 while (walk_size
< addrs_size
) {
987 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
993 af
= sctp_get_af_specific(sa_addr
->sa_family
);
995 /* If the address family is not supported or if this address
996 * causes the address buffer to overflow return EINVAL.
998 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1003 addr_buf
+= af
->sockaddr_len
;
1004 walk_size
+= af
->sockaddr_len
;
1009 case SCTP_BINDX_ADD_ADDR
:
1010 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1013 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1016 case SCTP_BINDX_REM_ADDR
:
1017 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1020 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1034 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1036 * Common routine for handling connect() and sctp_connectx().
1037 * Connect will come in with just a single address.
1039 static int __sctp_connect(struct sock
*sk
,
1040 struct sockaddr
*kaddrs
,
1042 sctp_assoc_t
*assoc_id
)
1044 struct net
*net
= sock_net(sk
);
1045 struct sctp_sock
*sp
;
1046 struct sctp_endpoint
*ep
;
1047 struct sctp_association
*asoc
= NULL
;
1048 struct sctp_association
*asoc2
;
1049 struct sctp_transport
*transport
;
1056 union sctp_addr
*sa_addr
= NULL
;
1058 unsigned short port
;
1059 unsigned int f_flags
= 0;
1064 /* connect() cannot be done on a socket that is already in ESTABLISHED
1065 * state - UDP-style peeled off socket or a TCP-style socket that
1066 * is already connected.
1067 * It cannot be done even on a TCP-style listening socket.
1069 if (sctp_sstate(sk
, ESTABLISHED
) ||
1070 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1075 /* Walk through the addrs buffer and count the number of addresses. */
1077 while (walk_size
< addrs_size
) {
1080 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1086 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1088 /* If the address family is not supported or if this address
1089 * causes the address buffer to overflow return EINVAL.
1091 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1096 port
= ntohs(sa_addr
->v4
.sin_port
);
1098 /* Save current address so we can work with it */
1099 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1101 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1105 /* Make sure the destination port is correctly set
1108 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1113 /* Check if there already is a matching association on the
1114 * endpoint (other than the one created here).
1116 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1117 if (asoc2
&& asoc2
!= asoc
) {
1118 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1125 /* If we could not find a matching association on the endpoint,
1126 * make sure that there is no peeled-off association matching
1127 * the peer address even on another socket.
1129 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1130 err
= -EADDRNOTAVAIL
;
1135 /* If a bind() or sctp_bindx() is not called prior to
1136 * an sctp_connectx() call, the system picks an
1137 * ephemeral port and will choose an address set
1138 * equivalent to binding with a wildcard address.
1140 if (!ep
->base
.bind_addr
.port
) {
1141 if (sctp_autobind(sk
)) {
1147 * If an unprivileged user inherits a 1-many
1148 * style socket with open associations on a
1149 * privileged port, it MAY be permitted to
1150 * accept new associations, but it SHOULD NOT
1151 * be permitted to open new associations.
1153 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1154 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1160 scope
= sctp_scope(&to
);
1161 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1167 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1175 /* Prime the peer's transport structures. */
1176 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1184 addr_buf
+= af
->sockaddr_len
;
1185 walk_size
+= af
->sockaddr_len
;
1188 /* In case the user of sctp_connectx() wants an association
1189 * id back, assign one now.
1192 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1197 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1202 /* Initialize sk's dport and daddr for getpeername() */
1203 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1204 sp
->pf
->to_sk_daddr(sa_addr
, sk
);
1207 /* in-kernel sockets don't generally have a file allocated to them
1208 * if all they do is call sock_create_kern().
1210 if (sk
->sk_socket
->file
)
1211 f_flags
= sk
->sk_socket
->file
->f_flags
;
1213 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1215 err
= sctp_wait_for_connect(asoc
, &timeo
);
1216 if ((err
== 0 || err
== -EINPROGRESS
) && assoc_id
)
1217 *assoc_id
= asoc
->assoc_id
;
1219 /* Don't free association on exit. */
1223 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1224 __func__
, asoc
, kaddrs
, err
);
1227 /* sctp_primitive_ASSOCIATE may have added this association
1228 * To the hash table, try to unhash it, just in case, its a noop
1229 * if it wasn't hashed so we're safe
1231 sctp_association_free(asoc
);
1236 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1239 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1240 * sctp_assoc_t *asoc);
1242 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1243 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1244 * or IPv6 addresses.
1246 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1247 * Section 3.1.2 for this usage.
1249 * addrs is a pointer to an array of one or more socket addresses. Each
1250 * address is contained in its appropriate structure (i.e. struct
1251 * sockaddr_in or struct sockaddr_in6) the family of the address type
1252 * must be used to distengish the address length (note that this
1253 * representation is termed a "packed array" of addresses). The caller
1254 * specifies the number of addresses in the array with addrcnt.
1256 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1257 * the association id of the new association. On failure, sctp_connectx()
1258 * returns -1, and sets errno to the appropriate error code. The assoc_id
1259 * is not touched by the kernel.
1261 * For SCTP, the port given in each socket address must be the same, or
1262 * sctp_connectx() will fail, setting errno to EINVAL.
1264 * An application can use sctp_connectx to initiate an association with
1265 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1266 * allows a caller to specify multiple addresses at which a peer can be
1267 * reached. The way the SCTP stack uses the list of addresses to set up
1268 * the association is implementation dependent. This function only
1269 * specifies that the stack will try to make use of all the addresses in
1270 * the list when needed.
1272 * Note that the list of addresses passed in is only used for setting up
1273 * the association. It does not necessarily equal the set of addresses
1274 * the peer uses for the resulting association. If the caller wants to
1275 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1276 * retrieve them after the association has been set up.
1278 * Basically do nothing but copying the addresses from user to kernel
1279 * land and invoking either sctp_connectx(). This is used for tunneling
1280 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1282 * We don't use copy_from_user() for optimization: we first do the
1283 * sanity checks (buffer size -fast- and access check-healthy
1284 * pointer); if all of those succeed, then we can alloc the memory
1285 * (expensive operation) needed to copy the data to kernel. Then we do
1286 * the copying without checking the user space area
1287 * (__copy_from_user()).
1289 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1292 * sk The sk of the socket
1293 * addrs The pointer to the addresses in user land
1294 * addrssize Size of the addrs buffer
1296 * Returns >=0 if ok, <0 errno code on error.
1298 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1299 struct sockaddr __user
*addrs
,
1301 sctp_assoc_t
*assoc_id
)
1303 struct sockaddr
*kaddrs
;
1304 gfp_t gfp
= GFP_KERNEL
;
1307 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1308 __func__
, sk
, addrs
, addrs_size
);
1310 if (unlikely(addrs_size
<= 0))
1313 /* Check the user passed a healthy pointer. */
1314 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1317 /* Alloc space for the address array in kernel memory. */
1318 if (sk
->sk_socket
->file
)
1319 gfp
= GFP_USER
| __GFP_NOWARN
;
1320 kaddrs
= kmalloc(addrs_size
, gfp
);
1321 if (unlikely(!kaddrs
))
1324 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1327 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1336 * This is an older interface. It's kept for backward compatibility
1337 * to the option that doesn't provide association id.
1339 static int sctp_setsockopt_connectx_old(struct sock
*sk
,
1340 struct sockaddr __user
*addrs
,
1343 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1347 * New interface for the API. The since the API is done with a socket
1348 * option, to make it simple we feed back the association id is as a return
1349 * indication to the call. Error is always negative and association id is
1352 static int sctp_setsockopt_connectx(struct sock
*sk
,
1353 struct sockaddr __user
*addrs
,
1356 sctp_assoc_t assoc_id
= 0;
1359 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1368 * New (hopefully final) interface for the API.
1369 * We use the sctp_getaddrs_old structure so that use-space library
1370 * can avoid any unnecessary allocations. The only different part
1371 * is that we store the actual length of the address buffer into the
1372 * addrs_num structure member. That way we can re-use the existing
1375 #ifdef CONFIG_COMPAT
1376 struct compat_sctp_getaddrs_old
{
1377 sctp_assoc_t assoc_id
;
1379 compat_uptr_t addrs
; /* struct sockaddr * */
1383 static int sctp_getsockopt_connectx3(struct sock
*sk
, int len
,
1384 char __user
*optval
,
1387 struct sctp_getaddrs_old param
;
1388 sctp_assoc_t assoc_id
= 0;
1391 #ifdef CONFIG_COMPAT
1392 if (is_compat_task()) {
1393 struct compat_sctp_getaddrs_old param32
;
1395 if (len
< sizeof(param32
))
1397 if (copy_from_user(¶m32
, optval
, sizeof(param32
)))
1400 param
.assoc_id
= param32
.assoc_id
;
1401 param
.addr_num
= param32
.addr_num
;
1402 param
.addrs
= compat_ptr(param32
.addrs
);
1406 if (len
< sizeof(param
))
1408 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1412 err
= __sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)
1413 param
.addrs
, param
.addr_num
,
1415 if (err
== 0 || err
== -EINPROGRESS
) {
1416 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1418 if (put_user(sizeof(assoc_id
), optlen
))
1425 /* API 3.1.4 close() - UDP Style Syntax
1426 * Applications use close() to perform graceful shutdown (as described in
1427 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1428 * by a UDP-style socket.
1432 * ret = close(int sd);
1434 * sd - the socket descriptor of the associations to be closed.
1436 * To gracefully shutdown a specific association represented by the
1437 * UDP-style socket, an application should use the sendmsg() call,
1438 * passing no user data, but including the appropriate flag in the
1439 * ancillary data (see Section xxxx).
1441 * If sd in the close() call is a branched-off socket representing only
1442 * one association, the shutdown is performed on that association only.
1444 * 4.1.6 close() - TCP Style Syntax
1446 * Applications use close() to gracefully close down an association.
1450 * int close(int sd);
1452 * sd - the socket descriptor of the association to be closed.
1454 * After an application calls close() on a socket descriptor, no further
1455 * socket operations will succeed on that descriptor.
1457 * API 7.1.4 SO_LINGER
1459 * An application using the TCP-style socket can use this option to
1460 * perform the SCTP ABORT primitive. The linger option structure is:
1463 * int l_onoff; // option on/off
1464 * int l_linger; // linger time
1467 * To enable the option, set l_onoff to 1. If the l_linger value is set
1468 * to 0, calling close() is the same as the ABORT primitive. If the
1469 * value is set to a negative value, the setsockopt() call will return
1470 * an error. If the value is set to a positive value linger_time, the
1471 * close() can be blocked for at most linger_time ms. If the graceful
1472 * shutdown phase does not finish during this period, close() will
1473 * return but the graceful shutdown phase continues in the system.
1475 static void sctp_close(struct sock
*sk
, long timeout
)
1477 struct net
*net
= sock_net(sk
);
1478 struct sctp_endpoint
*ep
;
1479 struct sctp_association
*asoc
;
1480 struct list_head
*pos
, *temp
;
1481 unsigned int data_was_unread
;
1483 pr_debug("%s: sk:%p, timeout:%ld\n", __func__
, sk
, timeout
);
1486 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1487 sk
->sk_state
= SCTP_SS_CLOSING
;
1489 ep
= sctp_sk(sk
)->ep
;
1491 /* Clean up any skbs sitting on the receive queue. */
1492 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1493 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1495 /* Walk all associations on an endpoint. */
1496 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1497 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1499 if (sctp_style(sk
, TCP
)) {
1500 /* A closed association can still be in the list if
1501 * it belongs to a TCP-style listening socket that is
1502 * not yet accepted. If so, free it. If not, send an
1503 * ABORT or SHUTDOWN based on the linger options.
1505 if (sctp_state(asoc
, CLOSED
)) {
1506 sctp_association_free(asoc
);
1511 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1512 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1513 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1514 struct sctp_chunk
*chunk
;
1516 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1517 sctp_primitive_ABORT(net
, asoc
, chunk
);
1519 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1522 /* On a TCP-style socket, block for at most linger_time if set. */
1523 if (sctp_style(sk
, TCP
) && timeout
)
1524 sctp_wait_for_close(sk
, timeout
);
1526 /* This will run the backlog queue. */
1529 /* Supposedly, no process has access to the socket, but
1530 * the net layers still may.
1531 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1532 * held and that should be grabbed before socket lock.
1534 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
1537 /* Hold the sock, since sk_common_release() will put sock_put()
1538 * and we have just a little more cleanup.
1541 sk_common_release(sk
);
1544 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
1548 SCTP_DBG_OBJCNT_DEC(sock
);
1551 /* Handle EPIPE error. */
1552 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1555 err
= sock_error(sk
) ? : -EPIPE
;
1556 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1557 send_sig(SIGPIPE
, current
, 0);
1561 /* API 3.1.3 sendmsg() - UDP Style Syntax
1563 * An application uses sendmsg() and recvmsg() calls to transmit data to
1564 * and receive data from its peer.
1566 * ssize_t sendmsg(int socket, const struct msghdr *message,
1569 * socket - the socket descriptor of the endpoint.
1570 * message - pointer to the msghdr structure which contains a single
1571 * user message and possibly some ancillary data.
1573 * See Section 5 for complete description of the data
1576 * flags - flags sent or received with the user message, see Section
1577 * 5 for complete description of the flags.
1579 * Note: This function could use a rewrite especially when explicit
1580 * connect support comes in.
1582 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1584 static int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1586 static int sctp_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t msg_len
)
1588 struct net
*net
= sock_net(sk
);
1589 struct sctp_sock
*sp
;
1590 struct sctp_endpoint
*ep
;
1591 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1592 struct sctp_transport
*transport
, *chunk_tp
;
1593 struct sctp_chunk
*chunk
;
1595 struct sockaddr
*msg_name
= NULL
;
1596 struct sctp_sndrcvinfo default_sinfo
;
1597 struct sctp_sndrcvinfo
*sinfo
;
1598 struct sctp_initmsg
*sinit
;
1599 sctp_assoc_t associd
= 0;
1600 sctp_cmsgs_t cmsgs
= { NULL
};
1602 bool fill_sinfo_ttl
= false, wait_connect
= false;
1603 struct sctp_datamsg
*datamsg
;
1604 int msg_flags
= msg
->msg_flags
;
1605 __u16 sinfo_flags
= 0;
1613 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__
, sk
,
1616 /* We cannot send a message over a TCP-style listening socket. */
1617 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1622 /* Parse out the SCTP CMSGs. */
1623 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1625 pr_debug("%s: msghdr parse err:%x\n", __func__
, err
);
1629 /* Fetch the destination address for this packet. This
1630 * address only selects the association--it is not necessarily
1631 * the address we will send to.
1632 * For a peeled-off socket, msg_name is ignored.
1634 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1635 int msg_namelen
= msg
->msg_namelen
;
1637 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1642 if (msg_namelen
> sizeof(to
))
1643 msg_namelen
= sizeof(to
);
1644 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1645 msg_name
= msg
->msg_name
;
1649 if (cmsgs
.sinfo
!= NULL
) {
1650 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1651 default_sinfo
.sinfo_stream
= cmsgs
.sinfo
->snd_sid
;
1652 default_sinfo
.sinfo_flags
= cmsgs
.sinfo
->snd_flags
;
1653 default_sinfo
.sinfo_ppid
= cmsgs
.sinfo
->snd_ppid
;
1654 default_sinfo
.sinfo_context
= cmsgs
.sinfo
->snd_context
;
1655 default_sinfo
.sinfo_assoc_id
= cmsgs
.sinfo
->snd_assoc_id
;
1657 sinfo
= &default_sinfo
;
1658 fill_sinfo_ttl
= true;
1660 sinfo
= cmsgs
.srinfo
;
1662 /* Did the user specify SNDINFO/SNDRCVINFO? */
1664 sinfo_flags
= sinfo
->sinfo_flags
;
1665 associd
= sinfo
->sinfo_assoc_id
;
1668 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1669 msg_len
, sinfo_flags
);
1671 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1672 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1677 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1678 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1679 * If SCTP_ABORT is set, the message length could be non zero with
1680 * the msg_iov set to the user abort reason.
1682 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1683 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1688 /* If SCTP_ADDR_OVER is set, there must be an address
1689 * specified in msg_name.
1691 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1698 pr_debug("%s: about to look up association\n", __func__
);
1702 /* If a msg_name has been specified, assume this is to be used. */
1704 /* Look for a matching association on the endpoint. */
1705 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1707 /* If we could not find a matching association on the
1708 * endpoint, make sure that it is not a TCP-style
1709 * socket that already has an association or there is
1710 * no peeled-off association on another socket.
1712 if ((sctp_style(sk
, TCP
) &&
1713 sctp_sstate(sk
, ESTABLISHED
)) ||
1714 sctp_endpoint_is_peeled_off(ep
, &to
)) {
1715 err
= -EADDRNOTAVAIL
;
1720 asoc
= sctp_id2assoc(sk
, associd
);
1728 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1730 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1731 * socket that has an association in CLOSED state. This can
1732 * happen when an accepted socket has an association that is
1735 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1740 if (sinfo_flags
& SCTP_EOF
) {
1741 pr_debug("%s: shutting down association:%p\n",
1744 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1748 if (sinfo_flags
& SCTP_ABORT
) {
1750 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1756 pr_debug("%s: aborting association:%p\n",
1759 sctp_primitive_ABORT(net
, asoc
, chunk
);
1765 /* Do we need to create the association? */
1767 pr_debug("%s: there is no association yet\n", __func__
);
1769 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1774 /* Check for invalid stream against the stream counts,
1775 * either the default or the user specified stream counts.
1778 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1779 /* Check against the defaults. */
1780 if (sinfo
->sinfo_stream
>=
1781 sp
->initmsg
.sinit_num_ostreams
) {
1786 /* Check against the requested. */
1787 if (sinfo
->sinfo_stream
>=
1788 sinit
->sinit_num_ostreams
) {
1796 * API 3.1.2 bind() - UDP Style Syntax
1797 * If a bind() or sctp_bindx() is not called prior to a
1798 * sendmsg() call that initiates a new association, the
1799 * system picks an ephemeral port and will choose an address
1800 * set equivalent to binding with a wildcard address.
1802 if (!ep
->base
.bind_addr
.port
) {
1803 if (sctp_autobind(sk
)) {
1809 * If an unprivileged user inherits a one-to-many
1810 * style socket with open associations on a privileged
1811 * port, it MAY be permitted to accept new associations,
1812 * but it SHOULD NOT be permitted to open new
1815 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1816 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1822 scope
= sctp_scope(&to
);
1823 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1829 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1835 /* If the SCTP_INIT ancillary data is specified, set all
1836 * the association init values accordingly.
1839 if (sinit
->sinit_num_ostreams
) {
1840 asoc
->c
.sinit_num_ostreams
=
1841 sinit
->sinit_num_ostreams
;
1843 if (sinit
->sinit_max_instreams
) {
1844 asoc
->c
.sinit_max_instreams
=
1845 sinit
->sinit_max_instreams
;
1847 if (sinit
->sinit_max_attempts
) {
1848 asoc
->max_init_attempts
1849 = sinit
->sinit_max_attempts
;
1851 if (sinit
->sinit_max_init_timeo
) {
1852 asoc
->max_init_timeo
=
1853 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1857 /* Prime the peer's transport structures. */
1858 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1865 /* ASSERT: we have a valid association at this point. */
1866 pr_debug("%s: we have a valid association\n", __func__
);
1869 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1870 * one with some defaults.
1872 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1873 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1874 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1875 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1876 default_sinfo
.sinfo_context
= asoc
->default_context
;
1877 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1878 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1880 sinfo
= &default_sinfo
;
1881 } else if (fill_sinfo_ttl
) {
1882 /* In case SNDINFO was specified, we still need to fill
1883 * it with a default ttl from the assoc here.
1885 sinfo
->sinfo_timetolive
= asoc
->default_timetolive
;
1888 /* API 7.1.7, the sndbuf size per association bounds the
1889 * maximum size of data that can be sent in a single send call.
1891 if (msg_len
> sk
->sk_sndbuf
) {
1896 if (asoc
->pmtu_pending
)
1897 sctp_assoc_pending_pmtu(sk
, asoc
);
1899 /* If fragmentation is disabled and the message length exceeds the
1900 * association fragmentation point, return EMSGSIZE. The I-D
1901 * does not specify what this error is, but this looks like
1904 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1909 /* Check for invalid stream. */
1910 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1915 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1916 if (!sctp_wspace(asoc
)) {
1917 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1922 /* If an address is passed with the sendto/sendmsg call, it is used
1923 * to override the primary destination address in the TCP model, or
1924 * when SCTP_ADDR_OVER flag is set in the UDP model.
1926 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1927 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1928 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1936 /* Auto-connect, if we aren't connected already. */
1937 if (sctp_state(asoc
, CLOSED
)) {
1938 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1942 wait_connect
= true;
1943 pr_debug("%s: we associated primitively\n", __func__
);
1946 /* Break the message into multiple chunks of maximum size. */
1947 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, &msg
->msg_iter
);
1948 if (IS_ERR(datamsg
)) {
1949 err
= PTR_ERR(datamsg
);
1953 /* Now send the (possibly) fragmented message. */
1954 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1955 /* Do accounting for the write space. */
1956 sctp_set_owner_w(chunk
);
1958 chunk
->transport
= chunk_tp
;
1961 /* Send it to the lower layers. Note: all chunks
1962 * must either fail or succeed. The lower layer
1963 * works that way today. Keep it that way or this
1966 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1967 sctp_datamsg_put(datamsg
);
1968 /* Did the lower layer accept the chunk? */
1972 pr_debug("%s: we sent primitively\n", __func__
);
1976 if (unlikely(wait_connect
)) {
1977 timeo
= sock_sndtimeo(sk
, msg_flags
& MSG_DONTWAIT
);
1978 sctp_wait_for_connect(asoc
, &timeo
);
1981 /* If we are already past ASSOCIATE, the lower
1982 * layers are responsible for association cleanup.
1988 sctp_association_free(asoc
);
1993 return sctp_error(sk
, msg_flags
, err
);
2000 err
= sock_error(sk
);
2010 /* This is an extended version of skb_pull() that removes the data from the
2011 * start of a skb even when data is spread across the list of skb's in the
2012 * frag_list. len specifies the total amount of data that needs to be removed.
2013 * when 'len' bytes could be removed from the skb, it returns 0.
2014 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2015 * could not be removed.
2017 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2019 struct sk_buff
*list
;
2020 int skb_len
= skb_headlen(skb
);
2023 if (len
<= skb_len
) {
2024 __skb_pull(skb
, len
);
2028 __skb_pull(skb
, skb_len
);
2030 skb_walk_frags(skb
, list
) {
2031 rlen
= sctp_skb_pull(list
, len
);
2032 skb
->len
-= (len
-rlen
);
2033 skb
->data_len
-= (len
-rlen
);
2044 /* API 3.1.3 recvmsg() - UDP Style Syntax
2046 * ssize_t recvmsg(int socket, struct msghdr *message,
2049 * socket - the socket descriptor of the endpoint.
2050 * message - pointer to the msghdr structure which contains a single
2051 * user message and possibly some ancillary data.
2053 * See Section 5 for complete description of the data
2056 * flags - flags sent or received with the user message, see Section
2057 * 5 for complete description of the flags.
2059 static int sctp_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
2060 int noblock
, int flags
, int *addr_len
)
2062 struct sctp_ulpevent
*event
= NULL
;
2063 struct sctp_sock
*sp
= sctp_sk(sk
);
2064 struct sk_buff
*skb
;
2069 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2070 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2075 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
)) {
2080 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2084 /* Get the total length of the skb including any skb's in the
2093 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
2095 event
= sctp_skb2event(skb
);
2100 sock_recv_ts_and_drops(msg
, sk
, skb
);
2101 if (sctp_ulpevent_is_notification(event
)) {
2102 msg
->msg_flags
|= MSG_NOTIFICATION
;
2103 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2105 sp
->pf
->skb_msgname(skb
, msg
->msg_name
, addr_len
);
2108 /* Check if we allow SCTP_NXTINFO. */
2109 if (sp
->recvnxtinfo
)
2110 sctp_ulpevent_read_nxtinfo(event
, msg
, sk
);
2111 /* Check if we allow SCTP_RCVINFO. */
2112 if (sp
->recvrcvinfo
)
2113 sctp_ulpevent_read_rcvinfo(event
, msg
);
2114 /* Check if we allow SCTP_SNDRCVINFO. */
2115 if (sp
->subscribe
.sctp_data_io_event
)
2116 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2120 /* If skb's length exceeds the user's buffer, update the skb and
2121 * push it back to the receive_queue so that the next call to
2122 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2124 if (skb_len
> copied
) {
2125 msg
->msg_flags
&= ~MSG_EOR
;
2126 if (flags
& MSG_PEEK
)
2128 sctp_skb_pull(skb
, copied
);
2129 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2131 /* When only partial message is copied to the user, increase
2132 * rwnd by that amount. If all the data in the skb is read,
2133 * rwnd is updated when the event is freed.
2135 if (!sctp_ulpevent_is_notification(event
))
2136 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2138 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2139 (event
->msg_flags
& MSG_EOR
))
2140 msg
->msg_flags
|= MSG_EOR
;
2142 msg
->msg_flags
&= ~MSG_EOR
;
2145 if (flags
& MSG_PEEK
) {
2146 /* Release the skb reference acquired after peeking the skb in
2147 * sctp_skb_recv_datagram().
2151 /* Free the event which includes releasing the reference to
2152 * the owner of the skb, freeing the skb and updating the
2155 sctp_ulpevent_free(event
);
2162 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2164 * This option is a on/off flag. If enabled no SCTP message
2165 * fragmentation will be performed. Instead if a message being sent
2166 * exceeds the current PMTU size, the message will NOT be sent and
2167 * instead a error will be indicated to the user.
2169 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2170 char __user
*optval
,
2171 unsigned int optlen
)
2175 if (optlen
< sizeof(int))
2178 if (get_user(val
, (int __user
*)optval
))
2181 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2186 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2187 unsigned int optlen
)
2189 struct sctp_association
*asoc
;
2190 struct sctp_ulpevent
*event
;
2192 if (optlen
> sizeof(struct sctp_event_subscribe
))
2194 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2197 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2198 * if there is no data to be sent or retransmit, the stack will
2199 * immediately send up this notification.
2201 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2202 &sctp_sk(sk
)->subscribe
)) {
2203 asoc
= sctp_id2assoc(sk
, 0);
2205 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2206 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2211 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2218 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2220 * This socket option is applicable to the UDP-style socket only. When
2221 * set it will cause associations that are idle for more than the
2222 * specified number of seconds to automatically close. An association
2223 * being idle is defined an association that has NOT sent or received
2224 * user data. The special value of '0' indicates that no automatic
2225 * close of any associations should be performed. The option expects an
2226 * integer defining the number of seconds of idle time before an
2227 * association is closed.
2229 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2230 unsigned int optlen
)
2232 struct sctp_sock
*sp
= sctp_sk(sk
);
2233 struct net
*net
= sock_net(sk
);
2235 /* Applicable to UDP-style socket only */
2236 if (sctp_style(sk
, TCP
))
2238 if (optlen
!= sizeof(int))
2240 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2243 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2244 sp
->autoclose
= net
->sctp
.max_autoclose
;
2249 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2251 * Applications can enable or disable heartbeats for any peer address of
2252 * an association, modify an address's heartbeat interval, force a
2253 * heartbeat to be sent immediately, and adjust the address's maximum
2254 * number of retransmissions sent before an address is considered
2255 * unreachable. The following structure is used to access and modify an
2256 * address's parameters:
2258 * struct sctp_paddrparams {
2259 * sctp_assoc_t spp_assoc_id;
2260 * struct sockaddr_storage spp_address;
2261 * uint32_t spp_hbinterval;
2262 * uint16_t spp_pathmaxrxt;
2263 * uint32_t spp_pathmtu;
2264 * uint32_t spp_sackdelay;
2265 * uint32_t spp_flags;
2268 * spp_assoc_id - (one-to-many style socket) This is filled in the
2269 * application, and identifies the association for
2271 * spp_address - This specifies which address is of interest.
2272 * spp_hbinterval - This contains the value of the heartbeat interval,
2273 * in milliseconds. If a value of zero
2274 * is present in this field then no changes are to
2275 * be made to this parameter.
2276 * spp_pathmaxrxt - This contains the maximum number of
2277 * retransmissions before this address shall be
2278 * considered unreachable. If a value of zero
2279 * is present in this field then no changes are to
2280 * be made to this parameter.
2281 * spp_pathmtu - When Path MTU discovery is disabled the value
2282 * specified here will be the "fixed" path mtu.
2283 * Note that if the spp_address field is empty
2284 * then all associations on this address will
2285 * have this fixed path mtu set upon them.
2287 * spp_sackdelay - When delayed sack is enabled, this value specifies
2288 * the number of milliseconds that sacks will be delayed
2289 * for. This value will apply to all addresses of an
2290 * association if the spp_address field is empty. Note
2291 * also, that if delayed sack is enabled and this
2292 * value is set to 0, no change is made to the last
2293 * recorded delayed sack timer value.
2295 * spp_flags - These flags are used to control various features
2296 * on an association. The flag field may contain
2297 * zero or more of the following options.
2299 * SPP_HB_ENABLE - Enable heartbeats on the
2300 * specified address. Note that if the address
2301 * field is empty all addresses for the association
2302 * have heartbeats enabled upon them.
2304 * SPP_HB_DISABLE - Disable heartbeats on the
2305 * speicifed address. Note that if the address
2306 * field is empty all addresses for the association
2307 * will have their heartbeats disabled. Note also
2308 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2309 * mutually exclusive, only one of these two should
2310 * be specified. Enabling both fields will have
2311 * undetermined results.
2313 * SPP_HB_DEMAND - Request a user initiated heartbeat
2314 * to be made immediately.
2316 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2317 * heartbeat delayis to be set to the value of 0
2320 * SPP_PMTUD_ENABLE - This field will enable PMTU
2321 * discovery upon the specified address. Note that
2322 * if the address feild is empty then all addresses
2323 * on the association are effected.
2325 * SPP_PMTUD_DISABLE - This field will disable PMTU
2326 * discovery upon the specified address. Note that
2327 * if the address feild is empty then all addresses
2328 * on the association are effected. Not also that
2329 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2330 * exclusive. Enabling both will have undetermined
2333 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2334 * on delayed sack. The time specified in spp_sackdelay
2335 * is used to specify the sack delay for this address. Note
2336 * that if spp_address is empty then all addresses will
2337 * enable delayed sack and take on the sack delay
2338 * value specified in spp_sackdelay.
2339 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2340 * off delayed sack. If the spp_address field is blank then
2341 * delayed sack is disabled for the entire association. Note
2342 * also that this field is mutually exclusive to
2343 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2346 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2347 struct sctp_transport
*trans
,
2348 struct sctp_association
*asoc
,
2349 struct sctp_sock
*sp
,
2352 int sackdelay_change
)
2356 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2357 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2359 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2364 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2365 * this field is ignored. Note also that a value of zero indicates
2366 * the current setting should be left unchanged.
2368 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2370 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2371 * set. This lets us use 0 value when this flag
2374 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2375 params
->spp_hbinterval
= 0;
2377 if (params
->spp_hbinterval
||
2378 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2381 msecs_to_jiffies(params
->spp_hbinterval
);
2384 msecs_to_jiffies(params
->spp_hbinterval
);
2386 sp
->hbinterval
= params
->spp_hbinterval
;
2393 trans
->param_flags
=
2394 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2397 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2400 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2404 /* When Path MTU discovery is disabled the value specified here will
2405 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2406 * include the flag SPP_PMTUD_DISABLE for this field to have any
2409 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2411 trans
->pathmtu
= params
->spp_pathmtu
;
2412 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2414 asoc
->pathmtu
= params
->spp_pathmtu
;
2415 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2417 sp
->pathmtu
= params
->spp_pathmtu
;
2423 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2424 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2425 trans
->param_flags
=
2426 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2428 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2429 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2433 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2436 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2440 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2441 * value of this field is ignored. Note also that a value of zero
2442 * indicates the current setting should be left unchanged.
2444 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2447 msecs_to_jiffies(params
->spp_sackdelay
);
2450 msecs_to_jiffies(params
->spp_sackdelay
);
2452 sp
->sackdelay
= params
->spp_sackdelay
;
2456 if (sackdelay_change
) {
2458 trans
->param_flags
=
2459 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2463 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2467 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2472 /* Note that a value of zero indicates the current setting should be
2475 if (params
->spp_pathmaxrxt
) {
2477 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2479 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2481 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2488 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2489 char __user
*optval
,
2490 unsigned int optlen
)
2492 struct sctp_paddrparams params
;
2493 struct sctp_transport
*trans
= NULL
;
2494 struct sctp_association
*asoc
= NULL
;
2495 struct sctp_sock
*sp
= sctp_sk(sk
);
2497 int hb_change
, pmtud_change
, sackdelay_change
;
2499 if (optlen
!= sizeof(struct sctp_paddrparams
))
2502 if (copy_from_user(¶ms
, optval
, optlen
))
2505 /* Validate flags and value parameters. */
2506 hb_change
= params
.spp_flags
& SPP_HB
;
2507 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2508 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2510 if (hb_change
== SPP_HB
||
2511 pmtud_change
== SPP_PMTUD
||
2512 sackdelay_change
== SPP_SACKDELAY
||
2513 params
.spp_sackdelay
> 500 ||
2514 (params
.spp_pathmtu
&&
2515 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2518 /* If an address other than INADDR_ANY is specified, and
2519 * no transport is found, then the request is invalid.
2521 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2522 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2523 params
.spp_assoc_id
);
2528 /* Get association, if assoc_id != 0 and the socket is a one
2529 * to many style socket, and an association was not found, then
2530 * the id was invalid.
2532 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2533 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2536 /* Heartbeat demand can only be sent on a transport or
2537 * association, but not a socket.
2539 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2542 /* Process parameters. */
2543 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2544 hb_change
, pmtud_change
,
2550 /* If changes are for association, also apply parameters to each
2553 if (!trans
&& asoc
) {
2554 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2556 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2557 hb_change
, pmtud_change
,
2565 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2567 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2570 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2572 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2576 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2578 * This option will effect the way delayed acks are performed. This
2579 * option allows you to get or set the delayed ack time, in
2580 * milliseconds. It also allows changing the delayed ack frequency.
2581 * Changing the frequency to 1 disables the delayed sack algorithm. If
2582 * the assoc_id is 0, then this sets or gets the endpoints default
2583 * values. If the assoc_id field is non-zero, then the set or get
2584 * effects the specified association for the one to many model (the
2585 * assoc_id field is ignored by the one to one model). Note that if
2586 * sack_delay or sack_freq are 0 when setting this option, then the
2587 * current values will remain unchanged.
2589 * struct sctp_sack_info {
2590 * sctp_assoc_t sack_assoc_id;
2591 * uint32_t sack_delay;
2592 * uint32_t sack_freq;
2595 * sack_assoc_id - This parameter, indicates which association the user
2596 * is performing an action upon. Note that if this field's value is
2597 * zero then the endpoints default value is changed (effecting future
2598 * associations only).
2600 * sack_delay - This parameter contains the number of milliseconds that
2601 * the user is requesting the delayed ACK timer be set to. Note that
2602 * this value is defined in the standard to be between 200 and 500
2605 * sack_freq - This parameter contains the number of packets that must
2606 * be received before a sack is sent without waiting for the delay
2607 * timer to expire. The default value for this is 2, setting this
2608 * value to 1 will disable the delayed sack algorithm.
2611 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2612 char __user
*optval
, unsigned int optlen
)
2614 struct sctp_sack_info params
;
2615 struct sctp_transport
*trans
= NULL
;
2616 struct sctp_association
*asoc
= NULL
;
2617 struct sctp_sock
*sp
= sctp_sk(sk
);
2619 if (optlen
== sizeof(struct sctp_sack_info
)) {
2620 if (copy_from_user(¶ms
, optval
, optlen
))
2623 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2625 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2626 pr_warn_ratelimited(DEPRECATED
2628 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2629 "Use struct sctp_sack_info instead\n",
2630 current
->comm
, task_pid_nr(current
));
2631 if (copy_from_user(¶ms
, optval
, optlen
))
2634 if (params
.sack_delay
== 0)
2635 params
.sack_freq
= 1;
2637 params
.sack_freq
= 0;
2641 /* Validate value parameter. */
2642 if (params
.sack_delay
> 500)
2645 /* Get association, if sack_assoc_id != 0 and the socket is a one
2646 * to many style socket, and an association was not found, then
2647 * the id was invalid.
2649 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2650 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2653 if (params
.sack_delay
) {
2656 msecs_to_jiffies(params
.sack_delay
);
2658 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2660 sp
->sackdelay
= params
.sack_delay
;
2662 sctp_spp_sackdelay_enable(sp
->param_flags
);
2666 if (params
.sack_freq
== 1) {
2669 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2672 sctp_spp_sackdelay_disable(sp
->param_flags
);
2674 } else if (params
.sack_freq
> 1) {
2676 asoc
->sackfreq
= params
.sack_freq
;
2678 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2680 sp
->sackfreq
= params
.sack_freq
;
2682 sctp_spp_sackdelay_enable(sp
->param_flags
);
2686 /* If change is for association, also apply to each transport. */
2688 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2690 if (params
.sack_delay
) {
2692 msecs_to_jiffies(params
.sack_delay
);
2693 trans
->param_flags
=
2694 sctp_spp_sackdelay_enable(trans
->param_flags
);
2696 if (params
.sack_freq
== 1) {
2697 trans
->param_flags
=
2698 sctp_spp_sackdelay_disable(trans
->param_flags
);
2699 } else if (params
.sack_freq
> 1) {
2700 trans
->sackfreq
= params
.sack_freq
;
2701 trans
->param_flags
=
2702 sctp_spp_sackdelay_enable(trans
->param_flags
);
2710 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2712 * Applications can specify protocol parameters for the default association
2713 * initialization. The option name argument to setsockopt() and getsockopt()
2716 * Setting initialization parameters is effective only on an unconnected
2717 * socket (for UDP-style sockets only future associations are effected
2718 * by the change). With TCP-style sockets, this option is inherited by
2719 * sockets derived from a listener socket.
2721 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2723 struct sctp_initmsg sinit
;
2724 struct sctp_sock
*sp
= sctp_sk(sk
);
2726 if (optlen
!= sizeof(struct sctp_initmsg
))
2728 if (copy_from_user(&sinit
, optval
, optlen
))
2731 if (sinit
.sinit_num_ostreams
)
2732 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2733 if (sinit
.sinit_max_instreams
)
2734 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2735 if (sinit
.sinit_max_attempts
)
2736 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2737 if (sinit
.sinit_max_init_timeo
)
2738 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2744 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2746 * Applications that wish to use the sendto() system call may wish to
2747 * specify a default set of parameters that would normally be supplied
2748 * through the inclusion of ancillary data. This socket option allows
2749 * such an application to set the default sctp_sndrcvinfo structure.
2750 * The application that wishes to use this socket option simply passes
2751 * in to this call the sctp_sndrcvinfo structure defined in Section
2752 * 5.2.2) The input parameters accepted by this call include
2753 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2754 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2755 * to this call if the caller is using the UDP model.
2757 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2758 char __user
*optval
,
2759 unsigned int optlen
)
2761 struct sctp_sock
*sp
= sctp_sk(sk
);
2762 struct sctp_association
*asoc
;
2763 struct sctp_sndrcvinfo info
;
2765 if (optlen
!= sizeof(info
))
2767 if (copy_from_user(&info
, optval
, optlen
))
2769 if (info
.sinfo_flags
&
2770 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2771 SCTP_ABORT
| SCTP_EOF
))
2774 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2775 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2778 asoc
->default_stream
= info
.sinfo_stream
;
2779 asoc
->default_flags
= info
.sinfo_flags
;
2780 asoc
->default_ppid
= info
.sinfo_ppid
;
2781 asoc
->default_context
= info
.sinfo_context
;
2782 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2784 sp
->default_stream
= info
.sinfo_stream
;
2785 sp
->default_flags
= info
.sinfo_flags
;
2786 sp
->default_ppid
= info
.sinfo_ppid
;
2787 sp
->default_context
= info
.sinfo_context
;
2788 sp
->default_timetolive
= info
.sinfo_timetolive
;
2794 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2795 * (SCTP_DEFAULT_SNDINFO)
2797 static int sctp_setsockopt_default_sndinfo(struct sock
*sk
,
2798 char __user
*optval
,
2799 unsigned int optlen
)
2801 struct sctp_sock
*sp
= sctp_sk(sk
);
2802 struct sctp_association
*asoc
;
2803 struct sctp_sndinfo info
;
2805 if (optlen
!= sizeof(info
))
2807 if (copy_from_user(&info
, optval
, optlen
))
2809 if (info
.snd_flags
&
2810 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2811 SCTP_ABORT
| SCTP_EOF
))
2814 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
2815 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
2818 asoc
->default_stream
= info
.snd_sid
;
2819 asoc
->default_flags
= info
.snd_flags
;
2820 asoc
->default_ppid
= info
.snd_ppid
;
2821 asoc
->default_context
= info
.snd_context
;
2823 sp
->default_stream
= info
.snd_sid
;
2824 sp
->default_flags
= info
.snd_flags
;
2825 sp
->default_ppid
= info
.snd_ppid
;
2826 sp
->default_context
= info
.snd_context
;
2832 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2834 * Requests that the local SCTP stack use the enclosed peer address as
2835 * the association primary. The enclosed address must be one of the
2836 * association peer's addresses.
2838 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2839 unsigned int optlen
)
2841 struct sctp_prim prim
;
2842 struct sctp_transport
*trans
;
2844 if (optlen
!= sizeof(struct sctp_prim
))
2847 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2850 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2854 sctp_assoc_set_primary(trans
->asoc
, trans
);
2860 * 7.1.5 SCTP_NODELAY
2862 * Turn on/off any Nagle-like algorithm. This means that packets are
2863 * generally sent as soon as possible and no unnecessary delays are
2864 * introduced, at the cost of more packets in the network. Expects an
2865 * integer boolean flag.
2867 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2868 unsigned int optlen
)
2872 if (optlen
< sizeof(int))
2874 if (get_user(val
, (int __user
*)optval
))
2877 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2883 * 7.1.1 SCTP_RTOINFO
2885 * The protocol parameters used to initialize and bound retransmission
2886 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2887 * and modify these parameters.
2888 * All parameters are time values, in milliseconds. A value of 0, when
2889 * modifying the parameters, indicates that the current value should not
2893 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2895 struct sctp_rtoinfo rtoinfo
;
2896 struct sctp_association
*asoc
;
2897 unsigned long rto_min
, rto_max
;
2898 struct sctp_sock
*sp
= sctp_sk(sk
);
2900 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2903 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2906 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2908 /* Set the values to the specific association */
2909 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2912 rto_max
= rtoinfo
.srto_max
;
2913 rto_min
= rtoinfo
.srto_min
;
2916 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2918 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2921 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2923 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2925 if (rto_min
> rto_max
)
2929 if (rtoinfo
.srto_initial
!= 0)
2931 msecs_to_jiffies(rtoinfo
.srto_initial
);
2932 asoc
->rto_max
= rto_max
;
2933 asoc
->rto_min
= rto_min
;
2935 /* If there is no association or the association-id = 0
2936 * set the values to the endpoint.
2938 if (rtoinfo
.srto_initial
!= 0)
2939 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2940 sp
->rtoinfo
.srto_max
= rto_max
;
2941 sp
->rtoinfo
.srto_min
= rto_min
;
2949 * 7.1.2 SCTP_ASSOCINFO
2951 * This option is used to tune the maximum retransmission attempts
2952 * of the association.
2953 * Returns an error if the new association retransmission value is
2954 * greater than the sum of the retransmission value of the peer.
2955 * See [SCTP] for more information.
2958 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2961 struct sctp_assocparams assocparams
;
2962 struct sctp_association
*asoc
;
2964 if (optlen
!= sizeof(struct sctp_assocparams
))
2966 if (copy_from_user(&assocparams
, optval
, optlen
))
2969 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2971 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2974 /* Set the values to the specific association */
2976 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
2979 struct sctp_transport
*peer_addr
;
2981 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
2983 path_sum
+= peer_addr
->pathmaxrxt
;
2987 /* Only validate asocmaxrxt if we have more than
2988 * one path/transport. We do this because path
2989 * retransmissions are only counted when we have more
2993 assocparams
.sasoc_asocmaxrxt
> path_sum
)
2996 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
2999 if (assocparams
.sasoc_cookie_life
!= 0)
3000 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
3002 /* Set the values to the endpoint */
3003 struct sctp_sock
*sp
= sctp_sk(sk
);
3005 if (assocparams
.sasoc_asocmaxrxt
!= 0)
3006 sp
->assocparams
.sasoc_asocmaxrxt
=
3007 assocparams
.sasoc_asocmaxrxt
;
3008 if (assocparams
.sasoc_cookie_life
!= 0)
3009 sp
->assocparams
.sasoc_cookie_life
=
3010 assocparams
.sasoc_cookie_life
;
3016 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3018 * This socket option is a boolean flag which turns on or off mapped V4
3019 * addresses. If this option is turned on and the socket is type
3020 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3021 * If this option is turned off, then no mapping will be done of V4
3022 * addresses and a user will receive both PF_INET6 and PF_INET type
3023 * addresses on the socket.
3025 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3028 struct sctp_sock
*sp
= sctp_sk(sk
);
3030 if (optlen
< sizeof(int))
3032 if (get_user(val
, (int __user
*)optval
))
3043 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3044 * This option will get or set the maximum size to put in any outgoing
3045 * SCTP DATA chunk. If a message is larger than this size it will be
3046 * fragmented by SCTP into the specified size. Note that the underlying
3047 * SCTP implementation may fragment into smaller sized chunks when the
3048 * PMTU of the underlying association is smaller than the value set by
3049 * the user. The default value for this option is '0' which indicates
3050 * the user is NOT limiting fragmentation and only the PMTU will effect
3051 * SCTP's choice of DATA chunk size. Note also that values set larger
3052 * than the maximum size of an IP datagram will effectively let SCTP
3053 * control fragmentation (i.e. the same as setting this option to 0).
3055 * The following structure is used to access and modify this parameter:
3057 * struct sctp_assoc_value {
3058 * sctp_assoc_t assoc_id;
3059 * uint32_t assoc_value;
3062 * assoc_id: This parameter is ignored for one-to-one style sockets.
3063 * For one-to-many style sockets this parameter indicates which
3064 * association the user is performing an action upon. Note that if
3065 * this field's value is zero then the endpoints default value is
3066 * changed (effecting future associations only).
3067 * assoc_value: This parameter specifies the maximum size in bytes.
3069 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3071 struct sctp_assoc_value params
;
3072 struct sctp_association
*asoc
;
3073 struct sctp_sock
*sp
= sctp_sk(sk
);
3076 if (optlen
== sizeof(int)) {
3077 pr_warn_ratelimited(DEPRECATED
3079 "Use of int in maxseg socket option.\n"
3080 "Use struct sctp_assoc_value instead\n",
3081 current
->comm
, task_pid_nr(current
));
3082 if (copy_from_user(&val
, optval
, optlen
))
3084 params
.assoc_id
= 0;
3085 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3086 if (copy_from_user(¶ms
, optval
, optlen
))
3088 val
= params
.assoc_value
;
3092 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3095 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3096 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3101 val
= asoc
->pathmtu
;
3102 val
-= sp
->pf
->af
->net_header_len
;
3103 val
-= sizeof(struct sctphdr
) +
3104 sizeof(struct sctp_data_chunk
);
3106 asoc
->user_frag
= val
;
3107 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3109 sp
->user_frag
= val
;
3117 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3119 * Requests that the peer mark the enclosed address as the association
3120 * primary. The enclosed address must be one of the association's
3121 * locally bound addresses. The following structure is used to make a
3122 * set primary request:
3124 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3125 unsigned int optlen
)
3127 struct net
*net
= sock_net(sk
);
3128 struct sctp_sock
*sp
;
3129 struct sctp_association
*asoc
= NULL
;
3130 struct sctp_setpeerprim prim
;
3131 struct sctp_chunk
*chunk
;
3137 if (!net
->sctp
.addip_enable
)
3140 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3143 if (copy_from_user(&prim
, optval
, optlen
))
3146 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3150 if (!asoc
->peer
.asconf_capable
)
3153 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3156 if (!sctp_state(asoc
, ESTABLISHED
))
3159 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3163 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3164 return -EADDRNOTAVAIL
;
3166 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3167 return -EADDRNOTAVAIL
;
3169 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3170 chunk
= sctp_make_asconf_set_prim(asoc
,
3171 (union sctp_addr
*)&prim
.sspp_addr
);
3175 err
= sctp_send_asconf(asoc
, chunk
);
3177 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3182 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3183 unsigned int optlen
)
3185 struct sctp_setadaptation adaptation
;
3187 if (optlen
!= sizeof(struct sctp_setadaptation
))
3189 if (copy_from_user(&adaptation
, optval
, optlen
))
3192 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3198 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3200 * The context field in the sctp_sndrcvinfo structure is normally only
3201 * used when a failed message is retrieved holding the value that was
3202 * sent down on the actual send call. This option allows the setting of
3203 * a default context on an association basis that will be received on
3204 * reading messages from the peer. This is especially helpful in the
3205 * one-2-many model for an application to keep some reference to an
3206 * internal state machine that is processing messages on the
3207 * association. Note that the setting of this value only effects
3208 * received messages from the peer and does not effect the value that is
3209 * saved with outbound messages.
3211 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3212 unsigned int optlen
)
3214 struct sctp_assoc_value params
;
3215 struct sctp_sock
*sp
;
3216 struct sctp_association
*asoc
;
3218 if (optlen
!= sizeof(struct sctp_assoc_value
))
3220 if (copy_from_user(¶ms
, optval
, optlen
))
3225 if (params
.assoc_id
!= 0) {
3226 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3229 asoc
->default_rcv_context
= params
.assoc_value
;
3231 sp
->default_rcv_context
= params
.assoc_value
;
3238 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3240 * This options will at a minimum specify if the implementation is doing
3241 * fragmented interleave. Fragmented interleave, for a one to many
3242 * socket, is when subsequent calls to receive a message may return
3243 * parts of messages from different associations. Some implementations
3244 * may allow you to turn this value on or off. If so, when turned off,
3245 * no fragment interleave will occur (which will cause a head of line
3246 * blocking amongst multiple associations sharing the same one to many
3247 * socket). When this option is turned on, then each receive call may
3248 * come from a different association (thus the user must receive data
3249 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3250 * association each receive belongs to.
3252 * This option takes a boolean value. A non-zero value indicates that
3253 * fragmented interleave is on. A value of zero indicates that
3254 * fragmented interleave is off.
3256 * Note that it is important that an implementation that allows this
3257 * option to be turned on, have it off by default. Otherwise an unaware
3258 * application using the one to many model may become confused and act
3261 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3262 char __user
*optval
,
3263 unsigned int optlen
)
3267 if (optlen
!= sizeof(int))
3269 if (get_user(val
, (int __user
*)optval
))
3272 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3278 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3279 * (SCTP_PARTIAL_DELIVERY_POINT)
3281 * This option will set or get the SCTP partial delivery point. This
3282 * point is the size of a message where the partial delivery API will be
3283 * invoked to help free up rwnd space for the peer. Setting this to a
3284 * lower value will cause partial deliveries to happen more often. The
3285 * calls argument is an integer that sets or gets the partial delivery
3286 * point. Note also that the call will fail if the user attempts to set
3287 * this value larger than the socket receive buffer size.
3289 * Note that any single message having a length smaller than or equal to
3290 * the SCTP partial delivery point will be delivered in one single read
3291 * call as long as the user provided buffer is large enough to hold the
3294 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3295 char __user
*optval
,
3296 unsigned int optlen
)
3300 if (optlen
!= sizeof(u32
))
3302 if (get_user(val
, (int __user
*)optval
))
3305 /* Note: We double the receive buffer from what the user sets
3306 * it to be, also initial rwnd is based on rcvbuf/2.
3308 if (val
> (sk
->sk_rcvbuf
>> 1))
3311 sctp_sk(sk
)->pd_point
= val
;
3313 return 0; /* is this the right error code? */
3317 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3319 * This option will allow a user to change the maximum burst of packets
3320 * that can be emitted by this association. Note that the default value
3321 * is 4, and some implementations may restrict this setting so that it
3322 * can only be lowered.
3324 * NOTE: This text doesn't seem right. Do this on a socket basis with
3325 * future associations inheriting the socket value.
3327 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3328 char __user
*optval
,
3329 unsigned int optlen
)
3331 struct sctp_assoc_value params
;
3332 struct sctp_sock
*sp
;
3333 struct sctp_association
*asoc
;
3337 if (optlen
== sizeof(int)) {
3338 pr_warn_ratelimited(DEPRECATED
3340 "Use of int in max_burst socket option deprecated.\n"
3341 "Use struct sctp_assoc_value instead\n",
3342 current
->comm
, task_pid_nr(current
));
3343 if (copy_from_user(&val
, optval
, optlen
))
3345 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3346 if (copy_from_user(¶ms
, optval
, optlen
))
3348 val
= params
.assoc_value
;
3349 assoc_id
= params
.assoc_id
;
3355 if (assoc_id
!= 0) {
3356 asoc
= sctp_id2assoc(sk
, assoc_id
);
3359 asoc
->max_burst
= val
;
3361 sp
->max_burst
= val
;
3367 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3369 * This set option adds a chunk type that the user is requesting to be
3370 * received only in an authenticated way. Changes to the list of chunks
3371 * will only effect future associations on the socket.
3373 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3374 char __user
*optval
,
3375 unsigned int optlen
)
3377 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3378 struct sctp_authchunk val
;
3380 if (!ep
->auth_enable
)
3383 if (optlen
!= sizeof(struct sctp_authchunk
))
3385 if (copy_from_user(&val
, optval
, optlen
))
3388 switch (val
.sauth_chunk
) {
3390 case SCTP_CID_INIT_ACK
:
3391 case SCTP_CID_SHUTDOWN_COMPLETE
:
3396 /* add this chunk id to the endpoint */
3397 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3401 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3403 * This option gets or sets the list of HMAC algorithms that the local
3404 * endpoint requires the peer to use.
3406 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3407 char __user
*optval
,
3408 unsigned int optlen
)
3410 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3411 struct sctp_hmacalgo
*hmacs
;
3415 if (!ep
->auth_enable
)
3418 if (optlen
< sizeof(struct sctp_hmacalgo
))
3421 hmacs
= memdup_user(optval
, optlen
);
3423 return PTR_ERR(hmacs
);
3425 idents
= hmacs
->shmac_num_idents
;
3426 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3427 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3432 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3439 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3441 * This option will set a shared secret key which is used to build an
3442 * association shared key.
3444 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3445 char __user
*optval
,
3446 unsigned int optlen
)
3448 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3449 struct sctp_authkey
*authkey
;
3450 struct sctp_association
*asoc
;
3453 if (!ep
->auth_enable
)
3456 if (optlen
<= sizeof(struct sctp_authkey
))
3459 authkey
= memdup_user(optval
, optlen
);
3460 if (IS_ERR(authkey
))
3461 return PTR_ERR(authkey
);
3463 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3468 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3469 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3474 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3481 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3483 * This option will get or set the active shared key to be used to build
3484 * the association shared key.
3486 static int sctp_setsockopt_active_key(struct sock
*sk
,
3487 char __user
*optval
,
3488 unsigned int optlen
)
3490 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3491 struct sctp_authkeyid val
;
3492 struct sctp_association
*asoc
;
3494 if (!ep
->auth_enable
)
3497 if (optlen
!= sizeof(struct sctp_authkeyid
))
3499 if (copy_from_user(&val
, optval
, optlen
))
3502 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3503 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3506 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3510 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3512 * This set option will delete a shared secret key from use.
3514 static int sctp_setsockopt_del_key(struct sock
*sk
,
3515 char __user
*optval
,
3516 unsigned int optlen
)
3518 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3519 struct sctp_authkeyid val
;
3520 struct sctp_association
*asoc
;
3522 if (!ep
->auth_enable
)
3525 if (optlen
!= sizeof(struct sctp_authkeyid
))
3527 if (copy_from_user(&val
, optval
, optlen
))
3530 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3531 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3534 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3539 * 8.1.23 SCTP_AUTO_ASCONF
3541 * This option will enable or disable the use of the automatic generation of
3542 * ASCONF chunks to add and delete addresses to an existing association. Note
3543 * that this option has two caveats namely: a) it only affects sockets that
3544 * are bound to all addresses available to the SCTP stack, and b) the system
3545 * administrator may have an overriding control that turns the ASCONF feature
3546 * off no matter what setting the socket option may have.
3547 * This option expects an integer boolean flag, where a non-zero value turns on
3548 * the option, and a zero value turns off the option.
3549 * Note. In this implementation, socket operation overrides default parameter
3550 * being set by sysctl as well as FreeBSD implementation
3552 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3553 unsigned int optlen
)
3556 struct sctp_sock
*sp
= sctp_sk(sk
);
3558 if (optlen
< sizeof(int))
3560 if (get_user(val
, (int __user
*)optval
))
3562 if (!sctp_is_ep_boundall(sk
) && val
)
3564 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3567 spin_lock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3568 if (val
== 0 && sp
->do_auto_asconf
) {
3569 list_del(&sp
->auto_asconf_list
);
3570 sp
->do_auto_asconf
= 0;
3571 } else if (val
&& !sp
->do_auto_asconf
) {
3572 list_add_tail(&sp
->auto_asconf_list
,
3573 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3574 sp
->do_auto_asconf
= 1;
3576 spin_unlock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3581 * SCTP_PEER_ADDR_THLDS
3583 * This option allows us to alter the partially failed threshold for one or all
3584 * transports in an association. See Section 6.1 of:
3585 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3587 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3588 char __user
*optval
,
3589 unsigned int optlen
)
3591 struct sctp_paddrthlds val
;
3592 struct sctp_transport
*trans
;
3593 struct sctp_association
*asoc
;
3595 if (optlen
< sizeof(struct sctp_paddrthlds
))
3597 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3598 sizeof(struct sctp_paddrthlds
)))
3602 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3603 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3606 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3608 if (val
.spt_pathmaxrxt
)
3609 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3610 trans
->pf_retrans
= val
.spt_pathpfthld
;
3613 if (val
.spt_pathmaxrxt
)
3614 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3615 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3617 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3622 if (val
.spt_pathmaxrxt
)
3623 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3624 trans
->pf_retrans
= val
.spt_pathpfthld
;
3630 static int sctp_setsockopt_recvrcvinfo(struct sock
*sk
,
3631 char __user
*optval
,
3632 unsigned int optlen
)
3636 if (optlen
< sizeof(int))
3638 if (get_user(val
, (int __user
*) optval
))
3641 sctp_sk(sk
)->recvrcvinfo
= (val
== 0) ? 0 : 1;
3646 static int sctp_setsockopt_recvnxtinfo(struct sock
*sk
,
3647 char __user
*optval
,
3648 unsigned int optlen
)
3652 if (optlen
< sizeof(int))
3654 if (get_user(val
, (int __user
*) optval
))
3657 sctp_sk(sk
)->recvnxtinfo
= (val
== 0) ? 0 : 1;
3662 /* API 6.2 setsockopt(), getsockopt()
3664 * Applications use setsockopt() and getsockopt() to set or retrieve
3665 * socket options. Socket options are used to change the default
3666 * behavior of sockets calls. They are described in Section 7.
3670 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3671 * int __user *optlen);
3672 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3675 * sd - the socket descript.
3676 * level - set to IPPROTO_SCTP for all SCTP options.
3677 * optname - the option name.
3678 * optval - the buffer to store the value of the option.
3679 * optlen - the size of the buffer.
3681 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3682 char __user
*optval
, unsigned int optlen
)
3686 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3688 /* I can hardly begin to describe how wrong this is. This is
3689 * so broken as to be worse than useless. The API draft
3690 * REALLY is NOT helpful here... I am not convinced that the
3691 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3692 * are at all well-founded.
3694 if (level
!= SOL_SCTP
) {
3695 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3696 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3703 case SCTP_SOCKOPT_BINDX_ADD
:
3704 /* 'optlen' is the size of the addresses buffer. */
3705 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3706 optlen
, SCTP_BINDX_ADD_ADDR
);
3709 case SCTP_SOCKOPT_BINDX_REM
:
3710 /* 'optlen' is the size of the addresses buffer. */
3711 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3712 optlen
, SCTP_BINDX_REM_ADDR
);
3715 case SCTP_SOCKOPT_CONNECTX_OLD
:
3716 /* 'optlen' is the size of the addresses buffer. */
3717 retval
= sctp_setsockopt_connectx_old(sk
,
3718 (struct sockaddr __user
*)optval
,
3722 case SCTP_SOCKOPT_CONNECTX
:
3723 /* 'optlen' is the size of the addresses buffer. */
3724 retval
= sctp_setsockopt_connectx(sk
,
3725 (struct sockaddr __user
*)optval
,
3729 case SCTP_DISABLE_FRAGMENTS
:
3730 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3734 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3737 case SCTP_AUTOCLOSE
:
3738 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3741 case SCTP_PEER_ADDR_PARAMS
:
3742 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3745 case SCTP_DELAYED_SACK
:
3746 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3748 case SCTP_PARTIAL_DELIVERY_POINT
:
3749 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3753 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3755 case SCTP_DEFAULT_SEND_PARAM
:
3756 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3759 case SCTP_DEFAULT_SNDINFO
:
3760 retval
= sctp_setsockopt_default_sndinfo(sk
, optval
, optlen
);
3762 case SCTP_PRIMARY_ADDR
:
3763 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3765 case SCTP_SET_PEER_PRIMARY_ADDR
:
3766 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3769 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3772 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3774 case SCTP_ASSOCINFO
:
3775 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3777 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3778 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3781 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3783 case SCTP_ADAPTATION_LAYER
:
3784 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3787 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3789 case SCTP_FRAGMENT_INTERLEAVE
:
3790 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3792 case SCTP_MAX_BURST
:
3793 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3795 case SCTP_AUTH_CHUNK
:
3796 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3798 case SCTP_HMAC_IDENT
:
3799 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3802 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3804 case SCTP_AUTH_ACTIVE_KEY
:
3805 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3807 case SCTP_AUTH_DELETE_KEY
:
3808 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3810 case SCTP_AUTO_ASCONF
:
3811 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3813 case SCTP_PEER_ADDR_THLDS
:
3814 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3816 case SCTP_RECVRCVINFO
:
3817 retval
= sctp_setsockopt_recvrcvinfo(sk
, optval
, optlen
);
3819 case SCTP_RECVNXTINFO
:
3820 retval
= sctp_setsockopt_recvnxtinfo(sk
, optval
, optlen
);
3823 retval
= -ENOPROTOOPT
;
3833 /* API 3.1.6 connect() - UDP Style Syntax
3835 * An application may use the connect() call in the UDP model to initiate an
3836 * association without sending data.
3840 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3842 * sd: the socket descriptor to have a new association added to.
3844 * nam: the address structure (either struct sockaddr_in or struct
3845 * sockaddr_in6 defined in RFC2553 [7]).
3847 * len: the size of the address.
3849 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3857 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
3860 /* Validate addr_len before calling common connect/connectx routine. */
3861 af
= sctp_get_af_specific(addr
->sa_family
);
3862 if (!af
|| addr_len
< af
->sockaddr_len
) {
3865 /* Pass correct addr len to common routine (so it knows there
3866 * is only one address being passed.
3868 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3875 /* FIXME: Write comments. */
3876 static int sctp_disconnect(struct sock
*sk
, int flags
)
3878 return -EOPNOTSUPP
; /* STUB */
3881 /* 4.1.4 accept() - TCP Style Syntax
3883 * Applications use accept() call to remove an established SCTP
3884 * association from the accept queue of the endpoint. A new socket
3885 * descriptor will be returned from accept() to represent the newly
3886 * formed association.
3888 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3890 struct sctp_sock
*sp
;
3891 struct sctp_endpoint
*ep
;
3892 struct sock
*newsk
= NULL
;
3893 struct sctp_association
*asoc
;
3902 if (!sctp_style(sk
, TCP
)) {
3903 error
= -EOPNOTSUPP
;
3907 if (!sctp_sstate(sk
, LISTENING
)) {
3912 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
3914 error
= sctp_wait_for_accept(sk
, timeo
);
3918 /* We treat the list of associations on the endpoint as the accept
3919 * queue and pick the first association on the list.
3921 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
3923 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
3929 /* Populate the fields of the newsk from the oldsk and migrate the
3930 * asoc to the newsk.
3932 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
3940 /* The SCTP ioctl handler. */
3941 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
3948 * SEQPACKET-style sockets in LISTENING state are valid, for
3949 * SCTP, so only discard TCP-style sockets in LISTENING state.
3951 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
3956 struct sk_buff
*skb
;
3957 unsigned int amount
= 0;
3959 skb
= skb_peek(&sk
->sk_receive_queue
);
3962 * We will only return the amount of this packet since
3963 * that is all that will be read.
3967 rc
= put_user(amount
, (int __user
*)arg
);
3979 /* This is the function which gets called during socket creation to
3980 * initialized the SCTP-specific portion of the sock.
3981 * The sock structure should already be zero-filled memory.
3983 static int sctp_init_sock(struct sock
*sk
)
3985 struct net
*net
= sock_net(sk
);
3986 struct sctp_sock
*sp
;
3988 pr_debug("%s: sk:%p\n", __func__
, sk
);
3992 /* Initialize the SCTP per socket area. */
3993 switch (sk
->sk_type
) {
3994 case SOCK_SEQPACKET
:
3995 sp
->type
= SCTP_SOCKET_UDP
;
3998 sp
->type
= SCTP_SOCKET_TCP
;
4001 return -ESOCKTNOSUPPORT
;
4004 /* Initialize default send parameters. These parameters can be
4005 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4007 sp
->default_stream
= 0;
4008 sp
->default_ppid
= 0;
4009 sp
->default_flags
= 0;
4010 sp
->default_context
= 0;
4011 sp
->default_timetolive
= 0;
4013 sp
->default_rcv_context
= 0;
4014 sp
->max_burst
= net
->sctp
.max_burst
;
4016 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
4018 /* Initialize default setup parameters. These parameters
4019 * can be modified with the SCTP_INITMSG socket option or
4020 * overridden by the SCTP_INIT CMSG.
4022 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
4023 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
4024 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
4025 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
4027 /* Initialize default RTO related parameters. These parameters can
4028 * be modified for with the SCTP_RTOINFO socket option.
4030 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
4031 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
4032 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
4034 /* Initialize default association related parameters. These parameters
4035 * can be modified with the SCTP_ASSOCINFO socket option.
4037 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
4038 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
4039 sp
->assocparams
.sasoc_peer_rwnd
= 0;
4040 sp
->assocparams
.sasoc_local_rwnd
= 0;
4041 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
4043 /* Initialize default event subscriptions. By default, all the
4046 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
4048 /* Default Peer Address Parameters. These defaults can
4049 * be modified via SCTP_PEER_ADDR_PARAMS
4051 sp
->hbinterval
= net
->sctp
.hb_interval
;
4052 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
4053 sp
->pathmtu
= 0; /* allow default discovery */
4054 sp
->sackdelay
= net
->sctp
.sack_timeout
;
4056 sp
->param_flags
= SPP_HB_ENABLE
|
4058 SPP_SACKDELAY_ENABLE
;
4060 /* If enabled no SCTP message fragmentation will be performed.
4061 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4063 sp
->disable_fragments
= 0;
4065 /* Enable Nagle algorithm by default. */
4068 sp
->recvrcvinfo
= 0;
4069 sp
->recvnxtinfo
= 0;
4071 /* Enable by default. */
4074 /* Auto-close idle associations after the configured
4075 * number of seconds. A value of 0 disables this
4076 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4077 * for UDP-style sockets only.
4081 /* User specified fragmentation limit. */
4084 sp
->adaptation_ind
= 0;
4086 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
4088 /* Control variables for partial data delivery. */
4089 atomic_set(&sp
->pd_mode
, 0);
4090 skb_queue_head_init(&sp
->pd_lobby
);
4091 sp
->frag_interleave
= 0;
4093 /* Create a per socket endpoint structure. Even if we
4094 * change the data structure relationships, this may still
4095 * be useful for storing pre-connect address information.
4097 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4103 sk
->sk_destruct
= sctp_destruct_sock
;
4105 SCTP_DBG_OBJCNT_INC(sock
);
4108 percpu_counter_inc(&sctp_sockets_allocated
);
4109 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4111 /* Nothing can fail after this block, otherwise
4112 * sctp_destroy_sock() will be called without addr_wq_lock held
4114 if (net
->sctp
.default_auto_asconf
) {
4115 spin_lock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4116 list_add_tail(&sp
->auto_asconf_list
,
4117 &net
->sctp
.auto_asconf_splist
);
4118 sp
->do_auto_asconf
= 1;
4119 spin_unlock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4121 sp
->do_auto_asconf
= 0;
4129 /* Cleanup any SCTP per socket resources. Must be called with
4130 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4132 static void sctp_destroy_sock(struct sock
*sk
)
4134 struct sctp_sock
*sp
;
4136 pr_debug("%s: sk:%p\n", __func__
, sk
);
4138 /* Release our hold on the endpoint. */
4140 /* This could happen during socket init, thus we bail out
4141 * early, since the rest of the below is not setup either.
4146 if (sp
->do_auto_asconf
) {
4147 sp
->do_auto_asconf
= 0;
4148 list_del(&sp
->auto_asconf_list
);
4150 sctp_endpoint_free(sp
->ep
);
4152 percpu_counter_dec(&sctp_sockets_allocated
);
4153 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4157 /* Triggered when there are no references on the socket anymore */
4158 static void sctp_destruct_sock(struct sock
*sk
)
4160 struct sctp_sock
*sp
= sctp_sk(sk
);
4162 /* Free up the HMAC transform. */
4163 crypto_free_hash(sp
->hmac
);
4165 inet_sock_destruct(sk
);
4168 /* API 4.1.7 shutdown() - TCP Style Syntax
4169 * int shutdown(int socket, int how);
4171 * sd - the socket descriptor of the association to be closed.
4172 * how - Specifies the type of shutdown. The values are
4175 * Disables further receive operations. No SCTP
4176 * protocol action is taken.
4178 * Disables further send operations, and initiates
4179 * the SCTP shutdown sequence.
4181 * Disables further send and receive operations
4182 * and initiates the SCTP shutdown sequence.
4184 static void sctp_shutdown(struct sock
*sk
, int how
)
4186 struct net
*net
= sock_net(sk
);
4187 struct sctp_endpoint
*ep
;
4188 struct sctp_association
*asoc
;
4190 if (!sctp_style(sk
, TCP
))
4193 if (how
& SEND_SHUTDOWN
) {
4194 ep
= sctp_sk(sk
)->ep
;
4195 if (!list_empty(&ep
->asocs
)) {
4196 asoc
= list_entry(ep
->asocs
.next
,
4197 struct sctp_association
, asocs
);
4198 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4203 /* 7.2.1 Association Status (SCTP_STATUS)
4205 * Applications can retrieve current status information about an
4206 * association, including association state, peer receiver window size,
4207 * number of unacked data chunks, and number of data chunks pending
4208 * receipt. This information is read-only.
4210 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4211 char __user
*optval
,
4214 struct sctp_status status
;
4215 struct sctp_association
*asoc
= NULL
;
4216 struct sctp_transport
*transport
;
4217 sctp_assoc_t associd
;
4220 if (len
< sizeof(status
)) {
4225 len
= sizeof(status
);
4226 if (copy_from_user(&status
, optval
, len
)) {
4231 associd
= status
.sstat_assoc_id
;
4232 asoc
= sctp_id2assoc(sk
, associd
);
4238 transport
= asoc
->peer
.primary_path
;
4240 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4241 status
.sstat_state
= sctp_assoc_to_state(asoc
);
4242 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4243 status
.sstat_unackdata
= asoc
->unack_data
;
4245 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4246 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4247 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4248 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4249 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4250 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4251 transport
->af_specific
->sockaddr_len
);
4252 /* Map ipv4 address into v4-mapped-on-v6 address. */
4253 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
4254 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4255 status
.sstat_primary
.spinfo_state
= transport
->state
;
4256 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4257 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4258 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4259 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4261 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4262 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4264 if (put_user(len
, optlen
)) {
4269 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4270 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4271 status
.sstat_assoc_id
);
4273 if (copy_to_user(optval
, &status
, len
)) {
4283 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4285 * Applications can retrieve information about a specific peer address
4286 * of an association, including its reachability state, congestion
4287 * window, and retransmission timer values. This information is
4290 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4291 char __user
*optval
,
4294 struct sctp_paddrinfo pinfo
;
4295 struct sctp_transport
*transport
;
4298 if (len
< sizeof(pinfo
)) {
4303 len
= sizeof(pinfo
);
4304 if (copy_from_user(&pinfo
, optval
, len
)) {
4309 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4310 pinfo
.spinfo_assoc_id
);
4314 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4315 pinfo
.spinfo_state
= transport
->state
;
4316 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4317 pinfo
.spinfo_srtt
= transport
->srtt
;
4318 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4319 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4321 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4322 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4324 if (put_user(len
, optlen
)) {
4329 if (copy_to_user(optval
, &pinfo
, len
)) {
4338 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4340 * This option is a on/off flag. If enabled no SCTP message
4341 * fragmentation will be performed. Instead if a message being sent
4342 * exceeds the current PMTU size, the message will NOT be sent and
4343 * instead a error will be indicated to the user.
4345 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4346 char __user
*optval
, int __user
*optlen
)
4350 if (len
< sizeof(int))
4354 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4355 if (put_user(len
, optlen
))
4357 if (copy_to_user(optval
, &val
, len
))
4362 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4364 * This socket option is used to specify various notifications and
4365 * ancillary data the user wishes to receive.
4367 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4372 if (len
> sizeof(struct sctp_event_subscribe
))
4373 len
= sizeof(struct sctp_event_subscribe
);
4374 if (put_user(len
, optlen
))
4376 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4381 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4383 * This socket option is applicable to the UDP-style socket only. When
4384 * set it will cause associations that are idle for more than the
4385 * specified number of seconds to automatically close. An association
4386 * being idle is defined an association that has NOT sent or received
4387 * user data. The special value of '0' indicates that no automatic
4388 * close of any associations should be performed. The option expects an
4389 * integer defining the number of seconds of idle time before an
4390 * association is closed.
4392 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4394 /* Applicable to UDP-style socket only */
4395 if (sctp_style(sk
, TCP
))
4397 if (len
< sizeof(int))
4400 if (put_user(len
, optlen
))
4402 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4407 /* Helper routine to branch off an association to a new socket. */
4408 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4410 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4411 struct sctp_sock
*sp
= sctp_sk(sk
);
4412 struct socket
*sock
;
4418 /* An association cannot be branched off from an already peeled-off
4419 * socket, nor is this supported for tcp style sockets.
4421 if (!sctp_style(sk
, UDP
))
4424 /* Create a new socket. */
4425 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4429 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4431 /* Make peeled-off sockets more like 1-1 accepted sockets.
4432 * Set the daddr and initialize id to something more random
4434 sp
->pf
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4436 /* Populate the fields of the newsk from the oldsk and migrate the
4437 * asoc to the newsk.
4439 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4445 EXPORT_SYMBOL(sctp_do_peeloff
);
4447 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4449 sctp_peeloff_arg_t peeloff
;
4450 struct socket
*newsock
;
4451 struct file
*newfile
;
4454 if (len
< sizeof(sctp_peeloff_arg_t
))
4456 len
= sizeof(sctp_peeloff_arg_t
);
4457 if (copy_from_user(&peeloff
, optval
, len
))
4460 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4464 /* Map the socket to an unused fd that can be returned to the user. */
4465 retval
= get_unused_fd_flags(0);
4467 sock_release(newsock
);
4471 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4472 if (IS_ERR(newfile
)) {
4473 put_unused_fd(retval
);
4474 sock_release(newsock
);
4475 return PTR_ERR(newfile
);
4478 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4481 /* Return the fd mapped to the new socket. */
4482 if (put_user(len
, optlen
)) {
4484 put_unused_fd(retval
);
4487 peeloff
.sd
= retval
;
4488 if (copy_to_user(optval
, &peeloff
, len
)) {
4490 put_unused_fd(retval
);
4493 fd_install(retval
, newfile
);
4498 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4500 * Applications can enable or disable heartbeats for any peer address of
4501 * an association, modify an address's heartbeat interval, force a
4502 * heartbeat to be sent immediately, and adjust the address's maximum
4503 * number of retransmissions sent before an address is considered
4504 * unreachable. The following structure is used to access and modify an
4505 * address's parameters:
4507 * struct sctp_paddrparams {
4508 * sctp_assoc_t spp_assoc_id;
4509 * struct sockaddr_storage spp_address;
4510 * uint32_t spp_hbinterval;
4511 * uint16_t spp_pathmaxrxt;
4512 * uint32_t spp_pathmtu;
4513 * uint32_t spp_sackdelay;
4514 * uint32_t spp_flags;
4517 * spp_assoc_id - (one-to-many style socket) This is filled in the
4518 * application, and identifies the association for
4520 * spp_address - This specifies which address is of interest.
4521 * spp_hbinterval - This contains the value of the heartbeat interval,
4522 * in milliseconds. If a value of zero
4523 * is present in this field then no changes are to
4524 * be made to this parameter.
4525 * spp_pathmaxrxt - This contains the maximum number of
4526 * retransmissions before this address shall be
4527 * considered unreachable. If a value of zero
4528 * is present in this field then no changes are to
4529 * be made to this parameter.
4530 * spp_pathmtu - When Path MTU discovery is disabled the value
4531 * specified here will be the "fixed" path mtu.
4532 * Note that if the spp_address field is empty
4533 * then all associations on this address will
4534 * have this fixed path mtu set upon them.
4536 * spp_sackdelay - When delayed sack is enabled, this value specifies
4537 * the number of milliseconds that sacks will be delayed
4538 * for. This value will apply to all addresses of an
4539 * association if the spp_address field is empty. Note
4540 * also, that if delayed sack is enabled and this
4541 * value is set to 0, no change is made to the last
4542 * recorded delayed sack timer value.
4544 * spp_flags - These flags are used to control various features
4545 * on an association. The flag field may contain
4546 * zero or more of the following options.
4548 * SPP_HB_ENABLE - Enable heartbeats on the
4549 * specified address. Note that if the address
4550 * field is empty all addresses for the association
4551 * have heartbeats enabled upon them.
4553 * SPP_HB_DISABLE - Disable heartbeats on the
4554 * speicifed address. Note that if the address
4555 * field is empty all addresses for the association
4556 * will have their heartbeats disabled. Note also
4557 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4558 * mutually exclusive, only one of these two should
4559 * be specified. Enabling both fields will have
4560 * undetermined results.
4562 * SPP_HB_DEMAND - Request a user initiated heartbeat
4563 * to be made immediately.
4565 * SPP_PMTUD_ENABLE - This field will enable PMTU
4566 * discovery upon the specified address. Note that
4567 * if the address feild is empty then all addresses
4568 * on the association are effected.
4570 * SPP_PMTUD_DISABLE - This field will disable PMTU
4571 * discovery upon the specified address. Note that
4572 * if the address feild is empty then all addresses
4573 * on the association are effected. Not also that
4574 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4575 * exclusive. Enabling both will have undetermined
4578 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4579 * on delayed sack. The time specified in spp_sackdelay
4580 * is used to specify the sack delay for this address. Note
4581 * that if spp_address is empty then all addresses will
4582 * enable delayed sack and take on the sack delay
4583 * value specified in spp_sackdelay.
4584 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4585 * off delayed sack. If the spp_address field is blank then
4586 * delayed sack is disabled for the entire association. Note
4587 * also that this field is mutually exclusive to
4588 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4591 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4592 char __user
*optval
, int __user
*optlen
)
4594 struct sctp_paddrparams params
;
4595 struct sctp_transport
*trans
= NULL
;
4596 struct sctp_association
*asoc
= NULL
;
4597 struct sctp_sock
*sp
= sctp_sk(sk
);
4599 if (len
< sizeof(struct sctp_paddrparams
))
4601 len
= sizeof(struct sctp_paddrparams
);
4602 if (copy_from_user(¶ms
, optval
, len
))
4605 /* If an address other than INADDR_ANY is specified, and
4606 * no transport is found, then the request is invalid.
4608 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4609 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4610 params
.spp_assoc_id
);
4612 pr_debug("%s: failed no transport\n", __func__
);
4617 /* Get association, if assoc_id != 0 and the socket is a one
4618 * to many style socket, and an association was not found, then
4619 * the id was invalid.
4621 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4622 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4623 pr_debug("%s: failed no association\n", __func__
);
4628 /* Fetch transport values. */
4629 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4630 params
.spp_pathmtu
= trans
->pathmtu
;
4631 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4632 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4634 /*draft-11 doesn't say what to return in spp_flags*/
4635 params
.spp_flags
= trans
->param_flags
;
4637 /* Fetch association values. */
4638 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4639 params
.spp_pathmtu
= asoc
->pathmtu
;
4640 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4641 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4643 /*draft-11 doesn't say what to return in spp_flags*/
4644 params
.spp_flags
= asoc
->param_flags
;
4646 /* Fetch socket values. */
4647 params
.spp_hbinterval
= sp
->hbinterval
;
4648 params
.spp_pathmtu
= sp
->pathmtu
;
4649 params
.spp_sackdelay
= sp
->sackdelay
;
4650 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4652 /*draft-11 doesn't say what to return in spp_flags*/
4653 params
.spp_flags
= sp
->param_flags
;
4656 if (copy_to_user(optval
, ¶ms
, len
))
4659 if (put_user(len
, optlen
))
4666 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4668 * This option will effect the way delayed acks are performed. This
4669 * option allows you to get or set the delayed ack time, in
4670 * milliseconds. It also allows changing the delayed ack frequency.
4671 * Changing the frequency to 1 disables the delayed sack algorithm. If
4672 * the assoc_id is 0, then this sets or gets the endpoints default
4673 * values. If the assoc_id field is non-zero, then the set or get
4674 * effects the specified association for the one to many model (the
4675 * assoc_id field is ignored by the one to one model). Note that if
4676 * sack_delay or sack_freq are 0 when setting this option, then the
4677 * current values will remain unchanged.
4679 * struct sctp_sack_info {
4680 * sctp_assoc_t sack_assoc_id;
4681 * uint32_t sack_delay;
4682 * uint32_t sack_freq;
4685 * sack_assoc_id - This parameter, indicates which association the user
4686 * is performing an action upon. Note that if this field's value is
4687 * zero then the endpoints default value is changed (effecting future
4688 * associations only).
4690 * sack_delay - This parameter contains the number of milliseconds that
4691 * the user is requesting the delayed ACK timer be set to. Note that
4692 * this value is defined in the standard to be between 200 and 500
4695 * sack_freq - This parameter contains the number of packets that must
4696 * be received before a sack is sent without waiting for the delay
4697 * timer to expire. The default value for this is 2, setting this
4698 * value to 1 will disable the delayed sack algorithm.
4700 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
4701 char __user
*optval
,
4704 struct sctp_sack_info params
;
4705 struct sctp_association
*asoc
= NULL
;
4706 struct sctp_sock
*sp
= sctp_sk(sk
);
4708 if (len
>= sizeof(struct sctp_sack_info
)) {
4709 len
= sizeof(struct sctp_sack_info
);
4711 if (copy_from_user(¶ms
, optval
, len
))
4713 } else if (len
== sizeof(struct sctp_assoc_value
)) {
4714 pr_warn_ratelimited(DEPRECATED
4716 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4717 "Use struct sctp_sack_info instead\n",
4718 current
->comm
, task_pid_nr(current
));
4719 if (copy_from_user(¶ms
, optval
, len
))
4724 /* Get association, if sack_assoc_id != 0 and the socket is a one
4725 * to many style socket, and an association was not found, then
4726 * the id was invalid.
4728 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
4729 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
4733 /* Fetch association values. */
4734 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4735 params
.sack_delay
= jiffies_to_msecs(
4737 params
.sack_freq
= asoc
->sackfreq
;
4740 params
.sack_delay
= 0;
4741 params
.sack_freq
= 1;
4744 /* Fetch socket values. */
4745 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
4746 params
.sack_delay
= sp
->sackdelay
;
4747 params
.sack_freq
= sp
->sackfreq
;
4749 params
.sack_delay
= 0;
4750 params
.sack_freq
= 1;
4754 if (copy_to_user(optval
, ¶ms
, len
))
4757 if (put_user(len
, optlen
))
4763 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4765 * Applications can specify protocol parameters for the default association
4766 * initialization. The option name argument to setsockopt() and getsockopt()
4769 * Setting initialization parameters is effective only on an unconnected
4770 * socket (for UDP-style sockets only future associations are effected
4771 * by the change). With TCP-style sockets, this option is inherited by
4772 * sockets derived from a listener socket.
4774 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4776 if (len
< sizeof(struct sctp_initmsg
))
4778 len
= sizeof(struct sctp_initmsg
);
4779 if (put_user(len
, optlen
))
4781 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
4787 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
4788 char __user
*optval
, int __user
*optlen
)
4790 struct sctp_association
*asoc
;
4792 struct sctp_getaddrs getaddrs
;
4793 struct sctp_transport
*from
;
4795 union sctp_addr temp
;
4796 struct sctp_sock
*sp
= sctp_sk(sk
);
4801 if (len
< sizeof(struct sctp_getaddrs
))
4804 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4807 /* For UDP-style sockets, id specifies the association to query. */
4808 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4812 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4813 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4815 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
4817 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
4818 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
4819 ->addr_to_user(sp
, &temp
);
4820 if (space_left
< addrlen
)
4822 if (copy_to_user(to
, &temp
, addrlen
))
4826 space_left
-= addrlen
;
4829 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
4831 bytes_copied
= ((char __user
*)to
) - optval
;
4832 if (put_user(bytes_copied
, optlen
))
4838 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
4839 size_t space_left
, int *bytes_copied
)
4841 struct sctp_sockaddr_entry
*addr
;
4842 union sctp_addr temp
;
4845 struct net
*net
= sock_net(sk
);
4848 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
4852 if ((PF_INET
== sk
->sk_family
) &&
4853 (AF_INET6
== addr
->a
.sa
.sa_family
))
4855 if ((PF_INET6
== sk
->sk_family
) &&
4856 inet_v6_ipv6only(sk
) &&
4857 (AF_INET
== addr
->a
.sa
.sa_family
))
4859 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4860 if (!temp
.v4
.sin_port
)
4861 temp
.v4
.sin_port
= htons(port
);
4863 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
4864 ->addr_to_user(sctp_sk(sk
), &temp
);
4866 if (space_left
< addrlen
) {
4870 memcpy(to
, &temp
, addrlen
);
4874 space_left
-= addrlen
;
4875 *bytes_copied
+= addrlen
;
4883 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
4884 char __user
*optval
, int __user
*optlen
)
4886 struct sctp_bind_addr
*bp
;
4887 struct sctp_association
*asoc
;
4889 struct sctp_getaddrs getaddrs
;
4890 struct sctp_sockaddr_entry
*addr
;
4892 union sctp_addr temp
;
4893 struct sctp_sock
*sp
= sctp_sk(sk
);
4897 int bytes_copied
= 0;
4901 if (len
< sizeof(struct sctp_getaddrs
))
4904 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
4908 * For UDP-style sockets, id specifies the association to query.
4909 * If the id field is set to the value '0' then the locally bound
4910 * addresses are returned without regard to any particular
4913 if (0 == getaddrs
.assoc_id
) {
4914 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
4916 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
4919 bp
= &asoc
->base
.bind_addr
;
4922 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
4923 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
4925 addrs
= kmalloc(space_left
, GFP_USER
| __GFP_NOWARN
);
4929 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4930 * addresses from the global local address list.
4932 if (sctp_list_single_entry(&bp
->address_list
)) {
4933 addr
= list_entry(bp
->address_list
.next
,
4934 struct sctp_sockaddr_entry
, list
);
4935 if (sctp_is_any(sk
, &addr
->a
)) {
4936 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
4937 space_left
, &bytes_copied
);
4947 /* Protection on the bound address list is not needed since
4948 * in the socket option context we hold a socket lock and
4949 * thus the bound address list can't change.
4951 list_for_each_entry(addr
, &bp
->address_list
, list
) {
4952 memcpy(&temp
, &addr
->a
, sizeof(temp
));
4953 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
4954 ->addr_to_user(sp
, &temp
);
4955 if (space_left
< addrlen
) {
4956 err
= -ENOMEM
; /*fixme: right error?*/
4959 memcpy(buf
, &temp
, addrlen
);
4961 bytes_copied
+= addrlen
;
4963 space_left
-= addrlen
;
4967 if (copy_to_user(to
, addrs
, bytes_copied
)) {
4971 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
4975 if (put_user(bytes_copied
, optlen
))
4982 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4984 * Requests that the local SCTP stack use the enclosed peer address as
4985 * the association primary. The enclosed address must be one of the
4986 * association peer's addresses.
4988 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
4989 char __user
*optval
, int __user
*optlen
)
4991 struct sctp_prim prim
;
4992 struct sctp_association
*asoc
;
4993 struct sctp_sock
*sp
= sctp_sk(sk
);
4995 if (len
< sizeof(struct sctp_prim
))
4998 len
= sizeof(struct sctp_prim
);
5000 if (copy_from_user(&prim
, optval
, len
))
5003 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
5007 if (!asoc
->peer
.primary_path
)
5010 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
5011 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
5013 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sp
,
5014 (union sctp_addr
*)&prim
.ssp_addr
);
5016 if (put_user(len
, optlen
))
5018 if (copy_to_user(optval
, &prim
, len
))
5025 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5027 * Requests that the local endpoint set the specified Adaptation Layer
5028 * Indication parameter for all future INIT and INIT-ACK exchanges.
5030 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
5031 char __user
*optval
, int __user
*optlen
)
5033 struct sctp_setadaptation adaptation
;
5035 if (len
< sizeof(struct sctp_setadaptation
))
5038 len
= sizeof(struct sctp_setadaptation
);
5040 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
5042 if (put_user(len
, optlen
))
5044 if (copy_to_user(optval
, &adaptation
, len
))
5052 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5054 * Applications that wish to use the sendto() system call may wish to
5055 * specify a default set of parameters that would normally be supplied
5056 * through the inclusion of ancillary data. This socket option allows
5057 * such an application to set the default sctp_sndrcvinfo structure.
5060 * The application that wishes to use this socket option simply passes
5061 * in to this call the sctp_sndrcvinfo structure defined in Section
5062 * 5.2.2) The input parameters accepted by this call include
5063 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5064 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5065 * to this call if the caller is using the UDP model.
5067 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5069 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
5070 int len
, char __user
*optval
,
5073 struct sctp_sock
*sp
= sctp_sk(sk
);
5074 struct sctp_association
*asoc
;
5075 struct sctp_sndrcvinfo info
;
5077 if (len
< sizeof(info
))
5082 if (copy_from_user(&info
, optval
, len
))
5085 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
5086 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
5089 info
.sinfo_stream
= asoc
->default_stream
;
5090 info
.sinfo_flags
= asoc
->default_flags
;
5091 info
.sinfo_ppid
= asoc
->default_ppid
;
5092 info
.sinfo_context
= asoc
->default_context
;
5093 info
.sinfo_timetolive
= asoc
->default_timetolive
;
5095 info
.sinfo_stream
= sp
->default_stream
;
5096 info
.sinfo_flags
= sp
->default_flags
;
5097 info
.sinfo_ppid
= sp
->default_ppid
;
5098 info
.sinfo_context
= sp
->default_context
;
5099 info
.sinfo_timetolive
= sp
->default_timetolive
;
5102 if (put_user(len
, optlen
))
5104 if (copy_to_user(optval
, &info
, len
))
5110 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5111 * (SCTP_DEFAULT_SNDINFO)
5113 static int sctp_getsockopt_default_sndinfo(struct sock
*sk
, int len
,
5114 char __user
*optval
,
5117 struct sctp_sock
*sp
= sctp_sk(sk
);
5118 struct sctp_association
*asoc
;
5119 struct sctp_sndinfo info
;
5121 if (len
< sizeof(info
))
5126 if (copy_from_user(&info
, optval
, len
))
5129 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
5130 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
5133 info
.snd_sid
= asoc
->default_stream
;
5134 info
.snd_flags
= asoc
->default_flags
;
5135 info
.snd_ppid
= asoc
->default_ppid
;
5136 info
.snd_context
= asoc
->default_context
;
5138 info
.snd_sid
= sp
->default_stream
;
5139 info
.snd_flags
= sp
->default_flags
;
5140 info
.snd_ppid
= sp
->default_ppid
;
5141 info
.snd_context
= sp
->default_context
;
5144 if (put_user(len
, optlen
))
5146 if (copy_to_user(optval
, &info
, len
))
5154 * 7.1.5 SCTP_NODELAY
5156 * Turn on/off any Nagle-like algorithm. This means that packets are
5157 * generally sent as soon as possible and no unnecessary delays are
5158 * introduced, at the cost of more packets in the network. Expects an
5159 * integer boolean flag.
5162 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5163 char __user
*optval
, int __user
*optlen
)
5167 if (len
< sizeof(int))
5171 val
= (sctp_sk(sk
)->nodelay
== 1);
5172 if (put_user(len
, optlen
))
5174 if (copy_to_user(optval
, &val
, len
))
5181 * 7.1.1 SCTP_RTOINFO
5183 * The protocol parameters used to initialize and bound retransmission
5184 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5185 * and modify these parameters.
5186 * All parameters are time values, in milliseconds. A value of 0, when
5187 * modifying the parameters, indicates that the current value should not
5191 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5192 char __user
*optval
,
5193 int __user
*optlen
) {
5194 struct sctp_rtoinfo rtoinfo
;
5195 struct sctp_association
*asoc
;
5197 if (len
< sizeof (struct sctp_rtoinfo
))
5200 len
= sizeof(struct sctp_rtoinfo
);
5202 if (copy_from_user(&rtoinfo
, optval
, len
))
5205 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5207 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5210 /* Values corresponding to the specific association. */
5212 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5213 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5214 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5216 /* Values corresponding to the endpoint. */
5217 struct sctp_sock
*sp
= sctp_sk(sk
);
5219 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5220 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5221 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5224 if (put_user(len
, optlen
))
5227 if (copy_to_user(optval
, &rtoinfo
, len
))
5235 * 7.1.2 SCTP_ASSOCINFO
5237 * This option is used to tune the maximum retransmission attempts
5238 * of the association.
5239 * Returns an error if the new association retransmission value is
5240 * greater than the sum of the retransmission value of the peer.
5241 * See [SCTP] for more information.
5244 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5245 char __user
*optval
,
5249 struct sctp_assocparams assocparams
;
5250 struct sctp_association
*asoc
;
5251 struct list_head
*pos
;
5254 if (len
< sizeof (struct sctp_assocparams
))
5257 len
= sizeof(struct sctp_assocparams
);
5259 if (copy_from_user(&assocparams
, optval
, len
))
5262 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5264 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5267 /* Values correspoinding to the specific association */
5269 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5270 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5271 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5272 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5274 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5278 assocparams
.sasoc_number_peer_destinations
= cnt
;
5280 /* Values corresponding to the endpoint */
5281 struct sctp_sock
*sp
= sctp_sk(sk
);
5283 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5284 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5285 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5286 assocparams
.sasoc_cookie_life
=
5287 sp
->assocparams
.sasoc_cookie_life
;
5288 assocparams
.sasoc_number_peer_destinations
=
5290 sasoc_number_peer_destinations
;
5293 if (put_user(len
, optlen
))
5296 if (copy_to_user(optval
, &assocparams
, len
))
5303 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5305 * This socket option is a boolean flag which turns on or off mapped V4
5306 * addresses. If this option is turned on and the socket is type
5307 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5308 * If this option is turned off, then no mapping will be done of V4
5309 * addresses and a user will receive both PF_INET6 and PF_INET type
5310 * addresses on the socket.
5312 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5313 char __user
*optval
, int __user
*optlen
)
5316 struct sctp_sock
*sp
= sctp_sk(sk
);
5318 if (len
< sizeof(int))
5323 if (put_user(len
, optlen
))
5325 if (copy_to_user(optval
, &val
, len
))
5332 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5333 * (chapter and verse is quoted at sctp_setsockopt_context())
5335 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5336 char __user
*optval
, int __user
*optlen
)
5338 struct sctp_assoc_value params
;
5339 struct sctp_sock
*sp
;
5340 struct sctp_association
*asoc
;
5342 if (len
< sizeof(struct sctp_assoc_value
))
5345 len
= sizeof(struct sctp_assoc_value
);
5347 if (copy_from_user(¶ms
, optval
, len
))
5352 if (params
.assoc_id
!= 0) {
5353 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5356 params
.assoc_value
= asoc
->default_rcv_context
;
5358 params
.assoc_value
= sp
->default_rcv_context
;
5361 if (put_user(len
, optlen
))
5363 if (copy_to_user(optval
, ¶ms
, len
))
5370 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5371 * This option will get or set the maximum size to put in any outgoing
5372 * SCTP DATA chunk. If a message is larger than this size it will be
5373 * fragmented by SCTP into the specified size. Note that the underlying
5374 * SCTP implementation may fragment into smaller sized chunks when the
5375 * PMTU of the underlying association is smaller than the value set by
5376 * the user. The default value for this option is '0' which indicates
5377 * the user is NOT limiting fragmentation and only the PMTU will effect
5378 * SCTP's choice of DATA chunk size. Note also that values set larger
5379 * than the maximum size of an IP datagram will effectively let SCTP
5380 * control fragmentation (i.e. the same as setting this option to 0).
5382 * The following structure is used to access and modify this parameter:
5384 * struct sctp_assoc_value {
5385 * sctp_assoc_t assoc_id;
5386 * uint32_t assoc_value;
5389 * assoc_id: This parameter is ignored for one-to-one style sockets.
5390 * For one-to-many style sockets this parameter indicates which
5391 * association the user is performing an action upon. Note that if
5392 * this field's value is zero then the endpoints default value is
5393 * changed (effecting future associations only).
5394 * assoc_value: This parameter specifies the maximum size in bytes.
5396 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5397 char __user
*optval
, int __user
*optlen
)
5399 struct sctp_assoc_value params
;
5400 struct sctp_association
*asoc
;
5402 if (len
== sizeof(int)) {
5403 pr_warn_ratelimited(DEPRECATED
5405 "Use of int in maxseg socket option.\n"
5406 "Use struct sctp_assoc_value instead\n",
5407 current
->comm
, task_pid_nr(current
));
5408 params
.assoc_id
= 0;
5409 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5410 len
= sizeof(struct sctp_assoc_value
);
5411 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5416 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5417 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5421 params
.assoc_value
= asoc
->frag_point
;
5423 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5425 if (put_user(len
, optlen
))
5427 if (len
== sizeof(int)) {
5428 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5431 if (copy_to_user(optval
, ¶ms
, len
))
5439 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5440 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5442 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5443 char __user
*optval
, int __user
*optlen
)
5447 if (len
< sizeof(int))
5452 val
= sctp_sk(sk
)->frag_interleave
;
5453 if (put_user(len
, optlen
))
5455 if (copy_to_user(optval
, &val
, len
))
5462 * 7.1.25. Set or Get the sctp partial delivery point
5463 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5465 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5466 char __user
*optval
,
5471 if (len
< sizeof(u32
))
5476 val
= sctp_sk(sk
)->pd_point
;
5477 if (put_user(len
, optlen
))
5479 if (copy_to_user(optval
, &val
, len
))
5486 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5487 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5489 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5490 char __user
*optval
,
5493 struct sctp_assoc_value params
;
5494 struct sctp_sock
*sp
;
5495 struct sctp_association
*asoc
;
5497 if (len
== sizeof(int)) {
5498 pr_warn_ratelimited(DEPRECATED
5500 "Use of int in max_burst socket option.\n"
5501 "Use struct sctp_assoc_value instead\n",
5502 current
->comm
, task_pid_nr(current
));
5503 params
.assoc_id
= 0;
5504 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5505 len
= sizeof(struct sctp_assoc_value
);
5506 if (copy_from_user(¶ms
, optval
, len
))
5513 if (params
.assoc_id
!= 0) {
5514 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5517 params
.assoc_value
= asoc
->max_burst
;
5519 params
.assoc_value
= sp
->max_burst
;
5521 if (len
== sizeof(int)) {
5522 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5525 if (copy_to_user(optval
, ¶ms
, len
))
5533 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5534 char __user
*optval
, int __user
*optlen
)
5536 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5537 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5538 struct sctp_hmac_algo_param
*hmacs
;
5542 if (!ep
->auth_enable
)
5545 hmacs
= ep
->auth_hmacs_list
;
5546 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5548 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5551 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5552 num_idents
= data_len
/ sizeof(u16
);
5554 if (put_user(len
, optlen
))
5556 if (put_user(num_idents
, &p
->shmac_num_idents
))
5558 if (copy_to_user(p
->shmac_idents
, hmacs
->hmac_ids
, data_len
))
5563 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5564 char __user
*optval
, int __user
*optlen
)
5566 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5567 struct sctp_authkeyid val
;
5568 struct sctp_association
*asoc
;
5570 if (!ep
->auth_enable
)
5573 if (len
< sizeof(struct sctp_authkeyid
))
5575 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5578 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5579 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5583 val
.scact_keynumber
= asoc
->active_key_id
;
5585 val
.scact_keynumber
= ep
->active_key_id
;
5587 len
= sizeof(struct sctp_authkeyid
);
5588 if (put_user(len
, optlen
))
5590 if (copy_to_user(optval
, &val
, len
))
5596 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5597 char __user
*optval
, int __user
*optlen
)
5599 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5600 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5601 struct sctp_authchunks val
;
5602 struct sctp_association
*asoc
;
5603 struct sctp_chunks_param
*ch
;
5607 if (!ep
->auth_enable
)
5610 if (len
< sizeof(struct sctp_authchunks
))
5613 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5616 to
= p
->gauth_chunks
;
5617 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5621 ch
= asoc
->peer
.peer_chunks
;
5625 /* See if the user provided enough room for all the data */
5626 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5627 if (len
< num_chunks
)
5630 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5633 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5634 if (put_user(len
, optlen
))
5636 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5641 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5642 char __user
*optval
, int __user
*optlen
)
5644 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5645 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5646 struct sctp_authchunks val
;
5647 struct sctp_association
*asoc
;
5648 struct sctp_chunks_param
*ch
;
5652 if (!ep
->auth_enable
)
5655 if (len
< sizeof(struct sctp_authchunks
))
5658 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5661 to
= p
->gauth_chunks
;
5662 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5663 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5667 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5669 ch
= ep
->auth_chunk_list
;
5674 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5675 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
5678 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5681 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5682 if (put_user(len
, optlen
))
5684 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5691 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5692 * This option gets the current number of associations that are attached
5693 * to a one-to-many style socket. The option value is an uint32_t.
5695 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
5696 char __user
*optval
, int __user
*optlen
)
5698 struct sctp_sock
*sp
= sctp_sk(sk
);
5699 struct sctp_association
*asoc
;
5702 if (sctp_style(sk
, TCP
))
5705 if (len
< sizeof(u32
))
5710 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5714 if (put_user(len
, optlen
))
5716 if (copy_to_user(optval
, &val
, len
))
5723 * 8.1.23 SCTP_AUTO_ASCONF
5724 * See the corresponding setsockopt entry as description
5726 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
5727 char __user
*optval
, int __user
*optlen
)
5731 if (len
< sizeof(int))
5735 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
5737 if (put_user(len
, optlen
))
5739 if (copy_to_user(optval
, &val
, len
))
5745 * 8.2.6. Get the Current Identifiers of Associations
5746 * (SCTP_GET_ASSOC_ID_LIST)
5748 * This option gets the current list of SCTP association identifiers of
5749 * the SCTP associations handled by a one-to-many style socket.
5751 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
5752 char __user
*optval
, int __user
*optlen
)
5754 struct sctp_sock
*sp
= sctp_sk(sk
);
5755 struct sctp_association
*asoc
;
5756 struct sctp_assoc_ids
*ids
;
5759 if (sctp_style(sk
, TCP
))
5762 if (len
< sizeof(struct sctp_assoc_ids
))
5765 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5769 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
5772 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
5774 ids
= kmalloc(len
, GFP_USER
| __GFP_NOWARN
);
5778 ids
->gaids_number_of_ids
= num
;
5780 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
5781 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
5784 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
5794 * SCTP_PEER_ADDR_THLDS
5796 * This option allows us to fetch the partially failed threshold for one or all
5797 * transports in an association. See Section 6.1 of:
5798 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5800 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
5801 char __user
*optval
,
5805 struct sctp_paddrthlds val
;
5806 struct sctp_transport
*trans
;
5807 struct sctp_association
*asoc
;
5809 if (len
< sizeof(struct sctp_paddrthlds
))
5811 len
= sizeof(struct sctp_paddrthlds
);
5812 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
5815 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
5816 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
5820 val
.spt_pathpfthld
= asoc
->pf_retrans
;
5821 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
5823 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
5828 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
5829 val
.spt_pathpfthld
= trans
->pf_retrans
;
5832 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
5839 * SCTP_GET_ASSOC_STATS
5841 * This option retrieves local per endpoint statistics. It is modeled
5842 * after OpenSolaris' implementation
5844 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
5845 char __user
*optval
,
5848 struct sctp_assoc_stats sas
;
5849 struct sctp_association
*asoc
= NULL
;
5851 /* User must provide at least the assoc id */
5852 if (len
< sizeof(sctp_assoc_t
))
5855 /* Allow the struct to grow and fill in as much as possible */
5856 len
= min_t(size_t, len
, sizeof(sas
));
5858 if (copy_from_user(&sas
, optval
, len
))
5861 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
5865 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
5866 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
5867 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
5868 sas
.sas_osacks
= asoc
->stats
.osacks
;
5869 sas
.sas_isacks
= asoc
->stats
.isacks
;
5870 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
5871 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
5872 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
5873 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
5874 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
5875 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
5876 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
5877 sas
.sas_opackets
= asoc
->stats
.opackets
;
5878 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
5880 /* New high max rto observed, will return 0 if not a single
5881 * RTO update took place. obs_rto_ipaddr will be bogus
5884 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
5885 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
5886 sizeof(struct sockaddr_storage
));
5888 /* Mark beginning of a new observation period */
5889 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
5891 if (put_user(len
, optlen
))
5894 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
5896 if (copy_to_user(optval
, &sas
, len
))
5902 static int sctp_getsockopt_recvrcvinfo(struct sock
*sk
, int len
,
5903 char __user
*optval
,
5908 if (len
< sizeof(int))
5912 if (sctp_sk(sk
)->recvrcvinfo
)
5914 if (put_user(len
, optlen
))
5916 if (copy_to_user(optval
, &val
, len
))
5922 static int sctp_getsockopt_recvnxtinfo(struct sock
*sk
, int len
,
5923 char __user
*optval
,
5928 if (len
< sizeof(int))
5932 if (sctp_sk(sk
)->recvnxtinfo
)
5934 if (put_user(len
, optlen
))
5936 if (copy_to_user(optval
, &val
, len
))
5942 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
5943 char __user
*optval
, int __user
*optlen
)
5948 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
5950 /* I can hardly begin to describe how wrong this is. This is
5951 * so broken as to be worse than useless. The API draft
5952 * REALLY is NOT helpful here... I am not convinced that the
5953 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5954 * are at all well-founded.
5956 if (level
!= SOL_SCTP
) {
5957 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
5959 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
5963 if (get_user(len
, optlen
))
5970 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
5972 case SCTP_DISABLE_FRAGMENTS
:
5973 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
5977 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
5979 case SCTP_AUTOCLOSE
:
5980 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
5982 case SCTP_SOCKOPT_PEELOFF
:
5983 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
5985 case SCTP_PEER_ADDR_PARAMS
:
5986 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
5989 case SCTP_DELAYED_SACK
:
5990 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
5994 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
5996 case SCTP_GET_PEER_ADDRS
:
5997 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
6000 case SCTP_GET_LOCAL_ADDRS
:
6001 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
6004 case SCTP_SOCKOPT_CONNECTX3
:
6005 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
6007 case SCTP_DEFAULT_SEND_PARAM
:
6008 retval
= sctp_getsockopt_default_send_param(sk
, len
,
6011 case SCTP_DEFAULT_SNDINFO
:
6012 retval
= sctp_getsockopt_default_sndinfo(sk
, len
,
6015 case SCTP_PRIMARY_ADDR
:
6016 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
6019 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
6022 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
6024 case SCTP_ASSOCINFO
:
6025 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
6027 case SCTP_I_WANT_MAPPED_V4_ADDR
:
6028 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
6031 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
6033 case SCTP_GET_PEER_ADDR_INFO
:
6034 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
6037 case SCTP_ADAPTATION_LAYER
:
6038 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
6042 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
6044 case SCTP_FRAGMENT_INTERLEAVE
:
6045 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
6048 case SCTP_PARTIAL_DELIVERY_POINT
:
6049 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
6052 case SCTP_MAX_BURST
:
6053 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
6056 case SCTP_AUTH_CHUNK
:
6057 case SCTP_AUTH_DELETE_KEY
:
6058 retval
= -EOPNOTSUPP
;
6060 case SCTP_HMAC_IDENT
:
6061 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
6063 case SCTP_AUTH_ACTIVE_KEY
:
6064 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
6066 case SCTP_PEER_AUTH_CHUNKS
:
6067 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
6070 case SCTP_LOCAL_AUTH_CHUNKS
:
6071 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
6074 case SCTP_GET_ASSOC_NUMBER
:
6075 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
6077 case SCTP_GET_ASSOC_ID_LIST
:
6078 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
6080 case SCTP_AUTO_ASCONF
:
6081 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
6083 case SCTP_PEER_ADDR_THLDS
:
6084 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
6086 case SCTP_GET_ASSOC_STATS
:
6087 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
6089 case SCTP_RECVRCVINFO
:
6090 retval
= sctp_getsockopt_recvrcvinfo(sk
, len
, optval
, optlen
);
6092 case SCTP_RECVNXTINFO
:
6093 retval
= sctp_getsockopt_recvnxtinfo(sk
, len
, optval
, optlen
);
6096 retval
= -ENOPROTOOPT
;
6104 static void sctp_hash(struct sock
*sk
)
6109 static void sctp_unhash(struct sock
*sk
)
6114 /* Check if port is acceptable. Possibly find first available port.
6116 * The port hash table (contained in the 'global' SCTP protocol storage
6117 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6118 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6119 * list (the list number is the port number hashed out, so as you
6120 * would expect from a hash function, all the ports in a given list have
6121 * such a number that hashes out to the same list number; you were
6122 * expecting that, right?); so each list has a set of ports, with a
6123 * link to the socket (struct sock) that uses it, the port number and
6124 * a fastreuse flag (FIXME: NPI ipg).
6126 static struct sctp_bind_bucket
*sctp_bucket_create(
6127 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
6129 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
6131 struct sctp_bind_hashbucket
*head
; /* hash list */
6132 struct sctp_bind_bucket
*pp
;
6133 unsigned short snum
;
6136 snum
= ntohs(addr
->v4
.sin_port
);
6138 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
6143 /* Search for an available port. */
6144 int low
, high
, remaining
, index
;
6146 struct net
*net
= sock_net(sk
);
6148 inet_get_local_port_range(net
, &low
, &high
);
6149 remaining
= (high
- low
) + 1;
6150 rover
= prandom_u32() % remaining
+ low
;
6154 if ((rover
< low
) || (rover
> high
))
6156 if (inet_is_local_reserved_port(net
, rover
))
6158 index
= sctp_phashfn(sock_net(sk
), rover
);
6159 head
= &sctp_port_hashtable
[index
];
6160 spin_lock(&head
->lock
);
6161 sctp_for_each_hentry(pp
, &head
->chain
)
6162 if ((pp
->port
== rover
) &&
6163 net_eq(sock_net(sk
), pp
->net
))
6167 spin_unlock(&head
->lock
);
6168 } while (--remaining
> 0);
6170 /* Exhausted local port range during search? */
6175 /* OK, here is the one we will use. HEAD (the port
6176 * hash table list entry) is non-NULL and we hold it's
6181 /* We are given an specific port number; we verify
6182 * that it is not being used. If it is used, we will
6183 * exahust the search in the hash list corresponding
6184 * to the port number (snum) - we detect that with the
6185 * port iterator, pp being NULL.
6187 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
6188 spin_lock(&head
->lock
);
6189 sctp_for_each_hentry(pp
, &head
->chain
) {
6190 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
6197 if (!hlist_empty(&pp
->owner
)) {
6198 /* We had a port hash table hit - there is an
6199 * available port (pp != NULL) and it is being
6200 * used by other socket (pp->owner not empty); that other
6201 * socket is going to be sk2.
6203 int reuse
= sk
->sk_reuse
;
6206 pr_debug("%s: found a possible match\n", __func__
);
6208 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6209 sk
->sk_state
!= SCTP_SS_LISTENING
)
6212 /* Run through the list of sockets bound to the port
6213 * (pp->port) [via the pointers bind_next and
6214 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6215 * we get the endpoint they describe and run through
6216 * the endpoint's list of IP (v4 or v6) addresses,
6217 * comparing each of the addresses with the address of
6218 * the socket sk. If we find a match, then that means
6219 * that this port/socket (sk) combination are already
6222 sk_for_each_bound(sk2
, &pp
->owner
) {
6223 struct sctp_endpoint
*ep2
;
6224 ep2
= sctp_sk(sk2
)->ep
;
6227 (reuse
&& sk2
->sk_reuse
&&
6228 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6231 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6232 sctp_sk(sk2
), sctp_sk(sk
))) {
6238 pr_debug("%s: found a match\n", __func__
);
6241 /* If there was a hash table miss, create a new port. */
6243 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6246 /* In either case (hit or miss), make sure fastreuse is 1 only
6247 * if sk->sk_reuse is too (that is, if the caller requested
6248 * SO_REUSEADDR on this socket -sk-).
6250 if (hlist_empty(&pp
->owner
)) {
6251 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6255 } else if (pp
->fastreuse
&&
6256 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6259 /* We are set, so fill up all the data in the hash table
6260 * entry, tie the socket list information with the rest of the
6261 * sockets FIXME: Blurry, NPI (ipg).
6264 if (!sctp_sk(sk
)->bind_hash
) {
6265 inet_sk(sk
)->inet_num
= snum
;
6266 sk_add_bind_node(sk
, &pp
->owner
);
6267 sctp_sk(sk
)->bind_hash
= pp
;
6272 spin_unlock(&head
->lock
);
6279 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6280 * port is requested.
6282 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6284 union sctp_addr addr
;
6285 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6287 /* Set up a dummy address struct from the sk. */
6288 af
->from_sk(&addr
, sk
);
6289 addr
.v4
.sin_port
= htons(snum
);
6291 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6292 return !!sctp_get_port_local(sk
, &addr
);
6296 * Move a socket to LISTENING state.
6298 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6300 struct sctp_sock
*sp
= sctp_sk(sk
);
6301 struct sctp_endpoint
*ep
= sp
->ep
;
6302 struct crypto_hash
*tfm
= NULL
;
6305 /* Allocate HMAC for generating cookie. */
6306 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6307 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6308 tfm
= crypto_alloc_hash(alg
, 0, CRYPTO_ALG_ASYNC
);
6310 net_info_ratelimited("failed to load transform for %s: %ld\n",
6311 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6314 sctp_sk(sk
)->hmac
= tfm
;
6318 * If a bind() or sctp_bindx() is not called prior to a listen()
6319 * call that allows new associations to be accepted, the system
6320 * picks an ephemeral port and will choose an address set equivalent
6321 * to binding with a wildcard address.
6323 * This is not currently spelled out in the SCTP sockets
6324 * extensions draft, but follows the practice as seen in TCP
6328 sk
->sk_state
= SCTP_SS_LISTENING
;
6329 if (!ep
->base
.bind_addr
.port
) {
6330 if (sctp_autobind(sk
))
6333 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6334 sk
->sk_state
= SCTP_SS_CLOSED
;
6339 sk
->sk_max_ack_backlog
= backlog
;
6340 sctp_hash_endpoint(ep
);
6345 * 4.1.3 / 5.1.3 listen()
6347 * By default, new associations are not accepted for UDP style sockets.
6348 * An application uses listen() to mark a socket as being able to
6349 * accept new associations.
6351 * On TCP style sockets, applications use listen() to ready the SCTP
6352 * endpoint for accepting inbound associations.
6354 * On both types of endpoints a backlog of '0' disables listening.
6356 * Move a socket to LISTENING state.
6358 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6360 struct sock
*sk
= sock
->sk
;
6361 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6364 if (unlikely(backlog
< 0))
6369 /* Peeled-off sockets are not allowed to listen(). */
6370 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6373 if (sock
->state
!= SS_UNCONNECTED
)
6376 /* If backlog is zero, disable listening. */
6378 if (sctp_sstate(sk
, CLOSED
))
6382 sctp_unhash_endpoint(ep
);
6383 sk
->sk_state
= SCTP_SS_CLOSED
;
6385 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6389 /* If we are already listening, just update the backlog */
6390 if (sctp_sstate(sk
, LISTENING
))
6391 sk
->sk_max_ack_backlog
= backlog
;
6393 err
= sctp_listen_start(sk
, backlog
);
6405 * This function is done by modeling the current datagram_poll() and the
6406 * tcp_poll(). Note that, based on these implementations, we don't
6407 * lock the socket in this function, even though it seems that,
6408 * ideally, locking or some other mechanisms can be used to ensure
6409 * the integrity of the counters (sndbuf and wmem_alloc) used
6410 * in this place. We assume that we don't need locks either until proven
6413 * Another thing to note is that we include the Async I/O support
6414 * here, again, by modeling the current TCP/UDP code. We don't have
6415 * a good way to test with it yet.
6417 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6419 struct sock
*sk
= sock
->sk
;
6420 struct sctp_sock
*sp
= sctp_sk(sk
);
6423 poll_wait(file
, sk_sleep(sk
), wait
);
6425 /* A TCP-style listening socket becomes readable when the accept queue
6428 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6429 return (!list_empty(&sp
->ep
->asocs
)) ?
6430 (POLLIN
| POLLRDNORM
) : 0;
6434 /* Is there any exceptional events? */
6435 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6437 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6438 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6439 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6440 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6443 /* Is it readable? Reconsider this code with TCP-style support. */
6444 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6445 mask
|= POLLIN
| POLLRDNORM
;
6447 /* The association is either gone or not ready. */
6448 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6451 /* Is it writable? */
6452 if (sctp_writeable(sk
)) {
6453 mask
|= POLLOUT
| POLLWRNORM
;
6455 sk_set_bit(SOCKWQ_ASYNC_NOSPACE
, sk
);
6457 * Since the socket is not locked, the buffer
6458 * might be made available after the writeable check and
6459 * before the bit is set. This could cause a lost I/O
6460 * signal. tcp_poll() has a race breaker for this race
6461 * condition. Based on their implementation, we put
6462 * in the following code to cover it as well.
6464 if (sctp_writeable(sk
))
6465 mask
|= POLLOUT
| POLLWRNORM
;
6470 /********************************************************************
6471 * 2nd Level Abstractions
6472 ********************************************************************/
6474 static struct sctp_bind_bucket
*sctp_bucket_create(
6475 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6477 struct sctp_bind_bucket
*pp
;
6479 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6481 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6484 INIT_HLIST_HEAD(&pp
->owner
);
6486 hlist_add_head(&pp
->node
, &head
->chain
);
6491 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6492 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6494 if (pp
&& hlist_empty(&pp
->owner
)) {
6495 __hlist_del(&pp
->node
);
6496 kmem_cache_free(sctp_bucket_cachep
, pp
);
6497 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6501 /* Release this socket's reference to a local port. */
6502 static inline void __sctp_put_port(struct sock
*sk
)
6504 struct sctp_bind_hashbucket
*head
=
6505 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6506 inet_sk(sk
)->inet_num
)];
6507 struct sctp_bind_bucket
*pp
;
6509 spin_lock(&head
->lock
);
6510 pp
= sctp_sk(sk
)->bind_hash
;
6511 __sk_del_bind_node(sk
);
6512 sctp_sk(sk
)->bind_hash
= NULL
;
6513 inet_sk(sk
)->inet_num
= 0;
6514 sctp_bucket_destroy(pp
);
6515 spin_unlock(&head
->lock
);
6518 void sctp_put_port(struct sock
*sk
)
6521 __sctp_put_port(sk
);
6526 * The system picks an ephemeral port and choose an address set equivalent
6527 * to binding with a wildcard address.
6528 * One of those addresses will be the primary address for the association.
6529 * This automatically enables the multihoming capability of SCTP.
6531 static int sctp_autobind(struct sock
*sk
)
6533 union sctp_addr autoaddr
;
6537 /* Initialize a local sockaddr structure to INADDR_ANY. */
6538 af
= sctp_sk(sk
)->pf
->af
;
6540 port
= htons(inet_sk(sk
)->inet_num
);
6541 af
->inaddr_any(&autoaddr
, port
);
6543 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
6546 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6549 * 4.2 The cmsghdr Structure *
6551 * When ancillary data is sent or received, any number of ancillary data
6552 * objects can be specified by the msg_control and msg_controllen members of
6553 * the msghdr structure, because each object is preceded by
6554 * a cmsghdr structure defining the object's length (the cmsg_len member).
6555 * Historically Berkeley-derived implementations have passed only one object
6556 * at a time, but this API allows multiple objects to be
6557 * passed in a single call to sendmsg() or recvmsg(). The following example
6558 * shows two ancillary data objects in a control buffer.
6560 * |<--------------------------- msg_controllen -------------------------->|
6563 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6565 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6568 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6570 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6573 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6574 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6576 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6578 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6585 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
6587 struct cmsghdr
*cmsg
;
6588 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
6590 for_each_cmsghdr(cmsg
, my_msg
) {
6591 if (!CMSG_OK(my_msg
, cmsg
))
6594 /* Should we parse this header or ignore? */
6595 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
6598 /* Strictly check lengths following example in SCM code. */
6599 switch (cmsg
->cmsg_type
) {
6601 /* SCTP Socket API Extension
6602 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6604 * This cmsghdr structure provides information for
6605 * initializing new SCTP associations with sendmsg().
6606 * The SCTP_INITMSG socket option uses this same data
6607 * structure. This structure is not used for
6610 * cmsg_level cmsg_type cmsg_data[]
6611 * ------------ ------------ ----------------------
6612 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6614 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_initmsg
)))
6617 cmsgs
->init
= CMSG_DATA(cmsg
);
6621 /* SCTP Socket API Extension
6622 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6624 * This cmsghdr structure specifies SCTP options for
6625 * sendmsg() and describes SCTP header information
6626 * about a received message through recvmsg().
6628 * cmsg_level cmsg_type cmsg_data[]
6629 * ------------ ------------ ----------------------
6630 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6632 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
6635 cmsgs
->srinfo
= CMSG_DATA(cmsg
);
6637 if (cmsgs
->srinfo
->sinfo_flags
&
6638 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6639 SCTP_ABORT
| SCTP_EOF
))
6644 /* SCTP Socket API Extension
6645 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6647 * This cmsghdr structure specifies SCTP options for
6648 * sendmsg(). This structure and SCTP_RCVINFO replaces
6649 * SCTP_SNDRCV which has been deprecated.
6651 * cmsg_level cmsg_type cmsg_data[]
6652 * ------------ ------------ ---------------------
6653 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6655 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndinfo
)))
6658 cmsgs
->sinfo
= CMSG_DATA(cmsg
);
6660 if (cmsgs
->sinfo
->snd_flags
&
6661 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
6662 SCTP_ABORT
| SCTP_EOF
))
6674 * Wait for a packet..
6675 * Note: This function is the same function as in core/datagram.c
6676 * with a few modifications to make lksctp work.
6678 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
6683 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
6685 /* Socket errors? */
6686 error
= sock_error(sk
);
6690 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6693 /* Socket shut down? */
6694 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6697 /* Sequenced packets can come disconnected. If so we report the
6702 /* Is there a good reason to think that we may receive some data? */
6703 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
6706 /* Handle signals. */
6707 if (signal_pending(current
))
6710 /* Let another process have a go. Since we are going to sleep
6711 * anyway. Note: This may cause odd behaviors if the message
6712 * does not fit in the user's buffer, but this seems to be the
6713 * only way to honor MSG_DONTWAIT realistically.
6716 *timeo_p
= schedule_timeout(*timeo_p
);
6720 finish_wait(sk_sleep(sk
), &wait
);
6724 error
= sock_intr_errno(*timeo_p
);
6727 finish_wait(sk_sleep(sk
), &wait
);
6732 /* Receive a datagram.
6733 * Note: This is pretty much the same routine as in core/datagram.c
6734 * with a few changes to make lksctp work.
6736 struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
6737 int noblock
, int *err
)
6740 struct sk_buff
*skb
;
6743 timeo
= sock_rcvtimeo(sk
, noblock
);
6745 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
6746 MAX_SCHEDULE_TIMEOUT
);
6749 /* Again only user level code calls this function,
6750 * so nothing interrupt level
6751 * will suddenly eat the receive_queue.
6753 * Look at current nfs client by the way...
6754 * However, this function was correct in any case. 8)
6756 if (flags
& MSG_PEEK
) {
6757 spin_lock_bh(&sk
->sk_receive_queue
.lock
);
6758 skb
= skb_peek(&sk
->sk_receive_queue
);
6760 atomic_inc(&skb
->users
);
6761 spin_unlock_bh(&sk
->sk_receive_queue
.lock
);
6763 skb
= skb_dequeue(&sk
->sk_receive_queue
);
6769 /* Caller is allowed not to check sk->sk_err before calling. */
6770 error
= sock_error(sk
);
6774 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6777 if (sk_can_busy_loop(sk
) &&
6778 sk_busy_loop(sk
, noblock
))
6781 /* User doesn't want to wait. */
6785 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
6794 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6795 static void __sctp_write_space(struct sctp_association
*asoc
)
6797 struct sock
*sk
= asoc
->base
.sk
;
6799 if (sctp_wspace(asoc
) <= 0)
6802 if (waitqueue_active(&asoc
->wait
))
6803 wake_up_interruptible(&asoc
->wait
);
6805 if (sctp_writeable(sk
)) {
6806 struct socket_wq
*wq
;
6809 wq
= rcu_dereference(sk
->sk_wq
);
6811 if (waitqueue_active(&wq
->wait
))
6812 wake_up_interruptible(&wq
->wait
);
6814 /* Note that we try to include the Async I/O support
6815 * here by modeling from the current TCP/UDP code.
6816 * We have not tested with it yet.
6818 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
6819 sock_wake_async(wq
, SOCK_WAKE_SPACE
, POLL_OUT
);
6825 static void sctp_wake_up_waiters(struct sock
*sk
,
6826 struct sctp_association
*asoc
)
6828 struct sctp_association
*tmp
= asoc
;
6830 /* We do accounting for the sndbuf space per association,
6831 * so we only need to wake our own association.
6833 if (asoc
->ep
->sndbuf_policy
)
6834 return __sctp_write_space(asoc
);
6836 /* If association goes down and is just flushing its
6837 * outq, then just normally notify others.
6839 if (asoc
->base
.dead
)
6840 return sctp_write_space(sk
);
6842 /* Accounting for the sndbuf space is per socket, so we
6843 * need to wake up others, try to be fair and in case of
6844 * other associations, let them have a go first instead
6845 * of just doing a sctp_write_space() call.
6847 * Note that we reach sctp_wake_up_waiters() only when
6848 * associations free up queued chunks, thus we are under
6849 * lock and the list of associations on a socket is
6850 * guaranteed not to change.
6852 for (tmp
= list_next_entry(tmp
, asocs
); 1;
6853 tmp
= list_next_entry(tmp
, asocs
)) {
6854 /* Manually skip the head element. */
6855 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
6857 /* Wake up association. */
6858 __sctp_write_space(tmp
);
6859 /* We've reached the end. */
6865 /* Do accounting for the sndbuf space.
6866 * Decrement the used sndbuf space of the corresponding association by the
6867 * data size which was just transmitted(freed).
6869 static void sctp_wfree(struct sk_buff
*skb
)
6871 struct sctp_chunk
*chunk
= skb_shinfo(skb
)->destructor_arg
;
6872 struct sctp_association
*asoc
= chunk
->asoc
;
6873 struct sock
*sk
= asoc
->base
.sk
;
6875 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
6876 sizeof(struct sk_buff
) +
6877 sizeof(struct sctp_chunk
);
6879 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
6882 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6884 sk
->sk_wmem_queued
-= skb
->truesize
;
6885 sk_mem_uncharge(sk
, skb
->truesize
);
6888 sctp_wake_up_waiters(sk
, asoc
);
6890 sctp_association_put(asoc
);
6893 /* Do accounting for the receive space on the socket.
6894 * Accounting for the association is done in ulpevent.c
6895 * We set this as a destructor for the cloned data skbs so that
6896 * accounting is done at the correct time.
6898 void sctp_sock_rfree(struct sk_buff
*skb
)
6900 struct sock
*sk
= skb
->sk
;
6901 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
6903 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
6906 * Mimic the behavior of sock_rfree
6908 sk_mem_uncharge(sk
, event
->rmem_len
);
6912 /* Helper function to wait for space in the sndbuf. */
6913 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
6916 struct sock
*sk
= asoc
->base
.sk
;
6918 long current_timeo
= *timeo_p
;
6921 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
6924 /* Increment the association's refcnt. */
6925 sctp_association_hold(asoc
);
6927 /* Wait on the association specific sndbuf space. */
6929 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
6930 TASK_INTERRUPTIBLE
);
6933 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
6936 if (signal_pending(current
))
6937 goto do_interrupted
;
6938 if (msg_len
<= sctp_wspace(asoc
))
6941 /* Let another process have a go. Since we are going
6945 current_timeo
= schedule_timeout(current_timeo
);
6946 BUG_ON(sk
!= asoc
->base
.sk
);
6949 *timeo_p
= current_timeo
;
6953 finish_wait(&asoc
->wait
, &wait
);
6955 /* Release the association's refcnt. */
6956 sctp_association_put(asoc
);
6965 err
= sock_intr_errno(*timeo_p
);
6973 void sctp_data_ready(struct sock
*sk
)
6975 struct socket_wq
*wq
;
6978 wq
= rcu_dereference(sk
->sk_wq
);
6979 if (skwq_has_sleeper(wq
))
6980 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
6981 POLLRDNORM
| POLLRDBAND
);
6982 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
6986 /* If socket sndbuf has changed, wake up all per association waiters. */
6987 void sctp_write_space(struct sock
*sk
)
6989 struct sctp_association
*asoc
;
6991 /* Wake up the tasks in each wait queue. */
6992 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
6993 __sctp_write_space(asoc
);
6997 /* Is there any sndbuf space available on the socket?
6999 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7000 * associations on the same socket. For a UDP-style socket with
7001 * multiple associations, it is possible for it to be "unwriteable"
7002 * prematurely. I assume that this is acceptable because
7003 * a premature "unwriteable" is better than an accidental "writeable" which
7004 * would cause an unwanted block under certain circumstances. For the 1-1
7005 * UDP-style sockets or TCP-style sockets, this code should work.
7008 static int sctp_writeable(struct sock
*sk
)
7012 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
7018 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7019 * returns immediately with EINPROGRESS.
7021 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
7023 struct sock
*sk
= asoc
->base
.sk
;
7025 long current_timeo
= *timeo_p
;
7028 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
7030 /* Increment the association's refcnt. */
7031 sctp_association_hold(asoc
);
7034 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7035 TASK_INTERRUPTIBLE
);
7038 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7040 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7043 if (signal_pending(current
))
7044 goto do_interrupted
;
7046 if (sctp_state(asoc
, ESTABLISHED
))
7049 /* Let another process have a go. Since we are going
7053 current_timeo
= schedule_timeout(current_timeo
);
7056 *timeo_p
= current_timeo
;
7060 finish_wait(&asoc
->wait
, &wait
);
7062 /* Release the association's refcnt. */
7063 sctp_association_put(asoc
);
7068 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
7071 err
= -ECONNREFUSED
;
7075 err
= sock_intr_errno(*timeo_p
);
7083 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
7085 struct sctp_endpoint
*ep
;
7089 ep
= sctp_sk(sk
)->ep
;
7093 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
7094 TASK_INTERRUPTIBLE
);
7096 if (list_empty(&ep
->asocs
)) {
7098 timeo
= schedule_timeout(timeo
);
7103 if (!sctp_sstate(sk
, LISTENING
))
7107 if (!list_empty(&ep
->asocs
))
7110 err
= sock_intr_errno(timeo
);
7111 if (signal_pending(current
))
7119 finish_wait(sk_sleep(sk
), &wait
);
7124 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
7129 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7130 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
7133 timeout
= schedule_timeout(timeout
);
7135 } while (!signal_pending(current
) && timeout
);
7137 finish_wait(sk_sleep(sk
), &wait
);
7140 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
7142 struct sk_buff
*frag
;
7147 /* Don't forget the fragments. */
7148 skb_walk_frags(skb
, frag
)
7149 sctp_skb_set_owner_r_frag(frag
, sk
);
7152 sctp_skb_set_owner_r(skb
, sk
);
7155 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
7156 struct sctp_association
*asoc
)
7158 struct inet_sock
*inet
= inet_sk(sk
);
7159 struct inet_sock
*newinet
;
7161 newsk
->sk_type
= sk
->sk_type
;
7162 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
7163 newsk
->sk_flags
= sk
->sk_flags
;
7164 newsk
->sk_tsflags
= sk
->sk_tsflags
;
7165 newsk
->sk_no_check_tx
= sk
->sk_no_check_tx
;
7166 newsk
->sk_no_check_rx
= sk
->sk_no_check_rx
;
7167 newsk
->sk_reuse
= sk
->sk_reuse
;
7169 newsk
->sk_shutdown
= sk
->sk_shutdown
;
7170 newsk
->sk_destruct
= sctp_destruct_sock
;
7171 newsk
->sk_family
= sk
->sk_family
;
7172 newsk
->sk_protocol
= IPPROTO_SCTP
;
7173 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
7174 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
7175 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
7176 newsk
->sk_lingertime
= sk
->sk_lingertime
;
7177 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
7178 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
7180 newinet
= inet_sk(newsk
);
7182 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7183 * getsockname() and getpeername()
7185 newinet
->inet_sport
= inet
->inet_sport
;
7186 newinet
->inet_saddr
= inet
->inet_saddr
;
7187 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
7188 newinet
->inet_dport
= htons(asoc
->peer
.port
);
7189 newinet
->pmtudisc
= inet
->pmtudisc
;
7190 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
7192 newinet
->uc_ttl
= inet
->uc_ttl
;
7193 newinet
->mc_loop
= 1;
7194 newinet
->mc_ttl
= 1;
7195 newinet
->mc_index
= 0;
7196 newinet
->mc_list
= NULL
;
7198 if (newsk
->sk_flags
& SK_FLAGS_TIMESTAMP
)
7199 net_enable_timestamp();
7201 security_sk_clone(sk
, newsk
);
7204 static inline void sctp_copy_descendant(struct sock
*sk_to
,
7205 const struct sock
*sk_from
)
7207 int ancestor_size
= sizeof(struct inet_sock
) +
7208 sizeof(struct sctp_sock
) -
7209 offsetof(struct sctp_sock
, auto_asconf_list
);
7211 if (sk_from
->sk_family
== PF_INET6
)
7212 ancestor_size
+= sizeof(struct ipv6_pinfo
);
7214 __inet_sk_copy_descendant(sk_to
, sk_from
, ancestor_size
);
7217 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7218 * and its messages to the newsk.
7220 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
7221 struct sctp_association
*assoc
,
7222 sctp_socket_type_t type
)
7224 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
7225 struct sctp_sock
*newsp
= sctp_sk(newsk
);
7226 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
7227 struct sctp_endpoint
*newep
= newsp
->ep
;
7228 struct sk_buff
*skb
, *tmp
;
7229 struct sctp_ulpevent
*event
;
7230 struct sctp_bind_hashbucket
*head
;
7232 /* Migrate socket buffer sizes and all the socket level options to the
7235 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
7236 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
7237 /* Brute force copy old sctp opt. */
7238 sctp_copy_descendant(newsk
, oldsk
);
7240 /* Restore the ep value that was overwritten with the above structure
7246 /* Hook this new socket in to the bind_hash list. */
7247 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7248 inet_sk(oldsk
)->inet_num
)];
7250 spin_lock(&head
->lock
);
7251 pp
= sctp_sk(oldsk
)->bind_hash
;
7252 sk_add_bind_node(newsk
, &pp
->owner
);
7253 sctp_sk(newsk
)->bind_hash
= pp
;
7254 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7255 spin_unlock(&head
->lock
);
7258 /* Copy the bind_addr list from the original endpoint to the new
7259 * endpoint so that we can handle restarts properly
7261 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7262 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7264 /* Move any messages in the old socket's receive queue that are for the
7265 * peeled off association to the new socket's receive queue.
7267 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7268 event
= sctp_skb2event(skb
);
7269 if (event
->asoc
== assoc
) {
7270 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7271 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7272 sctp_skb_set_owner_r_frag(skb
, newsk
);
7276 /* Clean up any messages pending delivery due to partial
7277 * delivery. Three cases:
7278 * 1) No partial deliver; no work.
7279 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7280 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7282 skb_queue_head_init(&newsp
->pd_lobby
);
7283 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7285 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7286 struct sk_buff_head
*queue
;
7288 /* Decide which queue to move pd_lobby skbs to. */
7289 if (assoc
->ulpq
.pd_mode
) {
7290 queue
= &newsp
->pd_lobby
;
7292 queue
= &newsk
->sk_receive_queue
;
7294 /* Walk through the pd_lobby, looking for skbs that
7295 * need moved to the new socket.
7297 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7298 event
= sctp_skb2event(skb
);
7299 if (event
->asoc
== assoc
) {
7300 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7301 __skb_queue_tail(queue
, skb
);
7302 sctp_skb_set_owner_r_frag(skb
, newsk
);
7306 /* Clear up any skbs waiting for the partial
7307 * delivery to finish.
7309 if (assoc
->ulpq
.pd_mode
)
7310 sctp_clear_pd(oldsk
, NULL
);
7314 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7315 sctp_skb_set_owner_r_frag(skb
, newsk
);
7317 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7318 sctp_skb_set_owner_r_frag(skb
, newsk
);
7320 /* Set the type of socket to indicate that it is peeled off from the
7321 * original UDP-style socket or created with the accept() call on a
7322 * TCP-style socket..
7326 /* Mark the new socket "in-use" by the user so that any packets
7327 * that may arrive on the association after we've moved it are
7328 * queued to the backlog. This prevents a potential race between
7329 * backlog processing on the old socket and new-packet processing
7330 * on the new socket.
7332 * The caller has just allocated newsk so we can guarantee that other
7333 * paths won't try to lock it and then oldsk.
7335 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7336 sctp_assoc_migrate(assoc
, newsk
);
7338 /* If the association on the newsk is already closed before accept()
7339 * is called, set RCV_SHUTDOWN flag.
7341 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
))
7342 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7344 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7345 release_sock(newsk
);
7349 /* This proto struct describes the ULP interface for SCTP. */
7350 struct proto sctp_prot
= {
7352 .owner
= THIS_MODULE
,
7353 .close
= sctp_close
,
7354 .connect
= sctp_connect
,
7355 .disconnect
= sctp_disconnect
,
7356 .accept
= sctp_accept
,
7357 .ioctl
= sctp_ioctl
,
7358 .init
= sctp_init_sock
,
7359 .destroy
= sctp_destroy_sock
,
7360 .shutdown
= sctp_shutdown
,
7361 .setsockopt
= sctp_setsockopt
,
7362 .getsockopt
= sctp_getsockopt
,
7363 .sendmsg
= sctp_sendmsg
,
7364 .recvmsg
= sctp_recvmsg
,
7366 .backlog_rcv
= sctp_backlog_rcv
,
7368 .unhash
= sctp_unhash
,
7369 .get_port
= sctp_get_port
,
7370 .obj_size
= sizeof(struct sctp_sock
),
7371 .sysctl_mem
= sysctl_sctp_mem
,
7372 .sysctl_rmem
= sysctl_sctp_rmem
,
7373 .sysctl_wmem
= sysctl_sctp_wmem
,
7374 .memory_pressure
= &sctp_memory_pressure
,
7375 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7376 .memory_allocated
= &sctp_memory_allocated
,
7377 .sockets_allocated
= &sctp_sockets_allocated
,
7380 #if IS_ENABLED(CONFIG_IPV6)
7382 #include <net/transp_v6.h>
7383 static void sctp_v6_destroy_sock(struct sock
*sk
)
7385 sctp_destroy_sock(sk
);
7386 inet6_destroy_sock(sk
);
7389 struct proto sctpv6_prot
= {
7391 .owner
= THIS_MODULE
,
7392 .close
= sctp_close
,
7393 .connect
= sctp_connect
,
7394 .disconnect
= sctp_disconnect
,
7395 .accept
= sctp_accept
,
7396 .ioctl
= sctp_ioctl
,
7397 .init
= sctp_init_sock
,
7398 .destroy
= sctp_v6_destroy_sock
,
7399 .shutdown
= sctp_shutdown
,
7400 .setsockopt
= sctp_setsockopt
,
7401 .getsockopt
= sctp_getsockopt
,
7402 .sendmsg
= sctp_sendmsg
,
7403 .recvmsg
= sctp_recvmsg
,
7405 .backlog_rcv
= sctp_backlog_rcv
,
7407 .unhash
= sctp_unhash
,
7408 .get_port
= sctp_get_port
,
7409 .obj_size
= sizeof(struct sctp6_sock
),
7410 .sysctl_mem
= sysctl_sctp_mem
,
7411 .sysctl_rmem
= sysctl_sctp_rmem
,
7412 .sysctl_wmem
= sysctl_sctp_wmem
,
7413 .memory_pressure
= &sctp_memory_pressure
,
7414 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7415 .memory_allocated
= &sctp_memory_allocated
,
7416 .sockets_allocated
= &sctp_sockets_allocated
,
7418 #endif /* IS_ENABLED(CONFIG_IPV6) */