1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
71 #include <net/route.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock
*sk
);
84 static void sctp_wfree(struct sk_buff
*skb
);
85 static int sctp_wait_for_sndbuf(struct sctp_association
*, long *timeo_p
,
87 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
);
88 static int sctp_wait_for_connect(struct sctp_association
*, long *timeo_p
);
89 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
);
90 static void sctp_wait_for_close(struct sock
*sk
, long timeo
);
91 static void sctp_destruct_sock(struct sock
*sk
);
92 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
93 union sctp_addr
*addr
, int len
);
94 static int sctp_bindx_add(struct sock
*, struct sockaddr
*, int);
95 static int sctp_bindx_rem(struct sock
*, struct sockaddr
*, int);
96 static int sctp_send_asconf_add_ip(struct sock
*, struct sockaddr
*, int);
97 static int sctp_send_asconf_del_ip(struct sock
*, struct sockaddr
*, int);
98 static int sctp_send_asconf(struct sctp_association
*asoc
,
99 struct sctp_chunk
*chunk
);
100 static int sctp_do_bind(struct sock
*, union sctp_addr
*, int);
101 static int sctp_autobind(struct sock
*sk
);
102 static void sctp_sock_migrate(struct sock
*, struct sock
*,
103 struct sctp_association
*, sctp_socket_type_t
);
105 static int sctp_memory_pressure
;
106 static atomic_long_t sctp_memory_allocated
;
107 struct percpu_counter sctp_sockets_allocated
;
109 static void sctp_enter_memory_pressure(struct sock
*sk
)
111 sctp_memory_pressure
= 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association
*asoc
)
120 if (asoc
->ep
->sndbuf_policy
)
121 amt
= asoc
->sndbuf_used
;
123 amt
= sk_wmem_alloc_get(asoc
->base
.sk
);
125 if (amt
>= asoc
->base
.sk
->sk_sndbuf
) {
126 if (asoc
->base
.sk
->sk_userlocks
& SOCK_SNDBUF_LOCK
)
129 amt
= sk_stream_wspace(asoc
->base
.sk
);
134 amt
= asoc
->base
.sk
->sk_sndbuf
- amt
;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
148 static inline void sctp_set_owner_w(struct sctp_chunk
*chunk
)
150 struct sctp_association
*asoc
= chunk
->asoc
;
151 struct sock
*sk
= asoc
->base
.sk
;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc
);
156 skb_set_owner_w(chunk
->skb
, sk
);
158 chunk
->skb
->destructor
= sctp_wfree
;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk
->skb
)->destructor_arg
= chunk
;
162 asoc
->sndbuf_used
+= SCTP_DATA_SNDSIZE(chunk
) +
163 sizeof(struct sk_buff
) +
164 sizeof(struct sctp_chunk
);
166 atomic_add(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
167 sk
->sk_wmem_queued
+= chunk
->skb
->truesize
;
168 sk_mem_charge(sk
, chunk
->skb
->truesize
);
171 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock
*sk
, union sctp_addr
*addr
,
177 /* Verify basic sockaddr. */
178 af
= sctp_sockaddr_af(sctp_sk(sk
), addr
, len
);
182 /* Is this a valid SCTP address? */
183 if (!af
->addr_valid(addr
, sctp_sk(sk
), NULL
))
186 if (!sctp_sk(sk
)->pf
->send_verify(sctp_sk(sk
), (addr
)))
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
195 struct sctp_association
*sctp_id2assoc(struct sock
*sk
, sctp_assoc_t id
)
197 struct sctp_association
*asoc
= NULL
;
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk
, UDP
)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
205 if (!sctp_sstate(sk
, ESTABLISHED
) && !sctp_sstate(sk
, CLOSING
))
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_af
*af
= sctp_get_af_specific(addr
->ss_family
);
239 union sctp_addr
*laddr
= (union sctp_addr
*)addr
;
240 struct sctp_transport
*transport
;
242 if (!af
|| sctp_verify_addr(sk
, laddr
, af
->sockaddr_len
))
245 addr_asoc
= sctp_endpoint_lookup_assoc(sctp_sk(sk
)->ep
,
252 id_asoc
= sctp_id2assoc(sk
, id
);
253 if (id_asoc
&& (id_asoc
!= addr_asoc
))
256 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
257 (union sctp_addr
*)addr
);
262 /* API 3.1.2 bind() - UDP Style Syntax
263 * The syntax of bind() is,
265 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
267 * sd - the socket descriptor returned by socket().
268 * addr - the address structure (struct sockaddr_in or struct
269 * sockaddr_in6 [RFC 2553]),
270 * addr_len - the size of the address structure.
272 static int sctp_bind(struct sock
*sk
, struct sockaddr
*addr
, int addr_len
)
278 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__
, sk
,
281 /* Disallow binding twice. */
282 if (!sctp_sk(sk
)->ep
->base
.bind_addr
.port
)
283 retval
= sctp_do_bind(sk
, (union sctp_addr
*)addr
,
293 static long sctp_get_port_local(struct sock
*, union sctp_addr
*);
295 /* Verify this is a valid sockaddr. */
296 static struct sctp_af
*sctp_sockaddr_af(struct sctp_sock
*opt
,
297 union sctp_addr
*addr
, int len
)
301 /* Check minimum size. */
302 if (len
< sizeof (struct sockaddr
))
305 /* V4 mapped address are really of AF_INET family */
306 if (addr
->sa
.sa_family
== AF_INET6
&&
307 ipv6_addr_v4mapped(&addr
->v6
.sin6_addr
)) {
308 if (!opt
->pf
->af_supported(AF_INET
, opt
))
311 /* Does this PF support this AF? */
312 if (!opt
->pf
->af_supported(addr
->sa
.sa_family
, opt
))
316 /* If we get this far, af is valid. */
317 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
319 if (len
< af
->sockaddr_len
)
325 /* Bind a local address either to an endpoint or to an association. */
326 static int sctp_do_bind(struct sock
*sk
, union sctp_addr
*addr
, int len
)
328 struct net
*net
= sock_net(sk
);
329 struct sctp_sock
*sp
= sctp_sk(sk
);
330 struct sctp_endpoint
*ep
= sp
->ep
;
331 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
336 /* Common sockaddr verification. */
337 af
= sctp_sockaddr_af(sp
, addr
, len
);
339 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
340 __func__
, sk
, addr
, len
);
344 snum
= ntohs(addr
->v4
.sin_port
);
346 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
347 __func__
, sk
, &addr
->sa
, bp
->port
, snum
, len
);
349 /* PF specific bind() address verification. */
350 if (!sp
->pf
->bind_verify(sp
, addr
))
351 return -EADDRNOTAVAIL
;
353 /* We must either be unbound, or bind to the same port.
354 * It's OK to allow 0 ports if we are already bound.
355 * We'll just inhert an already bound port in this case
360 else if (snum
!= bp
->port
) {
361 pr_debug("%s: new port %d doesn't match existing port "
362 "%d\n", __func__
, snum
, bp
->port
);
367 if (snum
&& snum
< PROT_SOCK
&&
368 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
))
371 /* See if the address matches any of the addresses we may have
372 * already bound before checking against other endpoints.
374 if (sctp_bind_addr_match(bp
, addr
, sp
))
377 /* Make sure we are allowed to bind here.
378 * The function sctp_get_port_local() does duplicate address
381 addr
->v4
.sin_port
= htons(snum
);
382 if ((ret
= sctp_get_port_local(sk
, addr
))) {
386 /* Refresh ephemeral port. */
388 bp
->port
= inet_sk(sk
)->inet_num
;
390 /* Add the address to the bind address list.
391 * Use GFP_ATOMIC since BHs will be disabled.
393 ret
= sctp_add_bind_addr(bp
, addr
, af
->sockaddr_len
,
394 SCTP_ADDR_SRC
, GFP_ATOMIC
);
396 /* Copy back into socket for getsockname() use. */
398 inet_sk(sk
)->inet_sport
= htons(inet_sk(sk
)->inet_num
);
399 sp
->pf
->to_sk_saddr(addr
, sk
);
405 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
407 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408 * at any one time. If a sender, after sending an ASCONF chunk, decides
409 * it needs to transfer another ASCONF Chunk, it MUST wait until the
410 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411 * subsequent ASCONF. Note this restriction binds each side, so at any
412 * time two ASCONF may be in-transit on any given association (one sent
413 * from each endpoint).
415 static int sctp_send_asconf(struct sctp_association
*asoc
,
416 struct sctp_chunk
*chunk
)
418 struct net
*net
= sock_net(asoc
->base
.sk
);
421 /* If there is an outstanding ASCONF chunk, queue it for later
424 if (asoc
->addip_last_asconf
) {
425 list_add_tail(&chunk
->list
, &asoc
->addip_chunk_list
);
429 /* Hold the chunk until an ASCONF_ACK is received. */
430 sctp_chunk_hold(chunk
);
431 retval
= sctp_primitive_ASCONF(net
, asoc
, chunk
);
433 sctp_chunk_free(chunk
);
435 asoc
->addip_last_asconf
= chunk
;
441 /* Add a list of addresses as bind addresses to local endpoint or
444 * Basically run through each address specified in the addrs/addrcnt
445 * array/length pair, determine if it is IPv6 or IPv4 and call
446 * sctp_do_bind() on it.
448 * If any of them fails, then the operation will be reversed and the
449 * ones that were added will be removed.
451 * Only sctp_setsockopt_bindx() is supposed to call this function.
453 static int sctp_bindx_add(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
458 struct sockaddr
*sa_addr
;
461 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__
, sk
,
465 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
466 /* The list may contain either IPv4 or IPv6 address;
467 * determine the address length for walking thru the list.
470 af
= sctp_get_af_specific(sa_addr
->sa_family
);
476 retval
= sctp_do_bind(sk
, (union sctp_addr
*)sa_addr
,
479 addr_buf
+= af
->sockaddr_len
;
483 /* Failed. Cleanup the ones that have been added */
485 sctp_bindx_rem(sk
, addrs
, cnt
);
493 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
494 * associations that are part of the endpoint indicating that a list of local
495 * addresses are added to the endpoint.
497 * If any of the addresses is already in the bind address list of the
498 * association, we do not send the chunk for that association. But it will not
499 * affect other associations.
501 * Only sctp_setsockopt_bindx() is supposed to call this function.
503 static int sctp_send_asconf_add_ip(struct sock
*sk
,
504 struct sockaddr
*addrs
,
507 struct net
*net
= sock_net(sk
);
508 struct sctp_sock
*sp
;
509 struct sctp_endpoint
*ep
;
510 struct sctp_association
*asoc
;
511 struct sctp_bind_addr
*bp
;
512 struct sctp_chunk
*chunk
;
513 struct sctp_sockaddr_entry
*laddr
;
514 union sctp_addr
*addr
;
515 union sctp_addr saveaddr
;
522 if (!net
->sctp
.addip_enable
)
528 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
529 __func__
, sk
, addrs
, addrcnt
);
531 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
532 if (!asoc
->peer
.asconf_capable
)
535 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_ADD_IP
)
538 if (!sctp_state(asoc
, ESTABLISHED
))
541 /* Check if any address in the packed array of addresses is
542 * in the bind address list of the association. If so,
543 * do not send the asconf chunk to its peer, but continue with
544 * other associations.
547 for (i
= 0; i
< addrcnt
; i
++) {
549 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
555 if (sctp_assoc_lookup_laddr(asoc
, addr
))
558 addr_buf
+= af
->sockaddr_len
;
563 /* Use the first valid address in bind addr list of
564 * association as Address Parameter of ASCONF CHUNK.
566 bp
= &asoc
->base
.bind_addr
;
567 p
= bp
->address_list
.next
;
568 laddr
= list_entry(p
, struct sctp_sockaddr_entry
, list
);
569 chunk
= sctp_make_asconf_update_ip(asoc
, &laddr
->a
, addrs
,
570 addrcnt
, SCTP_PARAM_ADD_IP
);
576 /* Add the new addresses to the bind address list with
577 * use_as_src set to 0.
580 for (i
= 0; i
< addrcnt
; i
++) {
582 af
= sctp_get_af_specific(addr
->v4
.sin_family
);
583 memcpy(&saveaddr
, addr
, af
->sockaddr_len
);
584 retval
= sctp_add_bind_addr(bp
, &saveaddr
,
586 SCTP_ADDR_NEW
, GFP_ATOMIC
);
587 addr_buf
+= af
->sockaddr_len
;
589 if (asoc
->src_out_of_asoc_ok
) {
590 struct sctp_transport
*trans
;
592 list_for_each_entry(trans
,
593 &asoc
->peer
.transport_addr_list
, transports
) {
594 /* Clear the source and route cache */
595 dst_release(trans
->dst
);
596 trans
->cwnd
= min(4*asoc
->pathmtu
, max_t(__u32
,
597 2*asoc
->pathmtu
, 4380));
598 trans
->ssthresh
= asoc
->peer
.i
.a_rwnd
;
599 trans
->rto
= asoc
->rto_initial
;
600 sctp_max_rto(asoc
, trans
);
601 trans
->rtt
= trans
->srtt
= trans
->rttvar
= 0;
602 sctp_transport_route(trans
, NULL
,
603 sctp_sk(asoc
->base
.sk
));
606 retval
= sctp_send_asconf(asoc
, chunk
);
613 /* Remove a list of addresses from bind addresses list. Do not remove the
616 * Basically run through each address specified in the addrs/addrcnt
617 * array/length pair, determine if it is IPv6 or IPv4 and call
618 * sctp_del_bind() on it.
620 * If any of them fails, then the operation will be reversed and the
621 * ones that were removed will be added back.
623 * At least one address has to be left; if only one address is
624 * available, the operation will return -EBUSY.
626 * Only sctp_setsockopt_bindx() is supposed to call this function.
628 static int sctp_bindx_rem(struct sock
*sk
, struct sockaddr
*addrs
, int addrcnt
)
630 struct sctp_sock
*sp
= sctp_sk(sk
);
631 struct sctp_endpoint
*ep
= sp
->ep
;
633 struct sctp_bind_addr
*bp
= &ep
->base
.bind_addr
;
636 union sctp_addr
*sa_addr
;
639 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
640 __func__
, sk
, addrs
, addrcnt
);
643 for (cnt
= 0; cnt
< addrcnt
; cnt
++) {
644 /* If the bind address list is empty or if there is only one
645 * bind address, there is nothing more to be removed (we need
646 * at least one address here).
648 if (list_empty(&bp
->address_list
) ||
649 (sctp_list_single_entry(&bp
->address_list
))) {
655 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
661 if (!af
->addr_valid(sa_addr
, sp
, NULL
)) {
662 retval
= -EADDRNOTAVAIL
;
666 if (sa_addr
->v4
.sin_port
&&
667 sa_addr
->v4
.sin_port
!= htons(bp
->port
)) {
672 if (!sa_addr
->v4
.sin_port
)
673 sa_addr
->v4
.sin_port
= htons(bp
->port
);
675 /* FIXME - There is probably a need to check if sk->sk_saddr and
676 * sk->sk_rcv_addr are currently set to one of the addresses to
677 * be removed. This is something which needs to be looked into
678 * when we are fixing the outstanding issues with multi-homing
679 * socket routing and failover schemes. Refer to comments in
680 * sctp_do_bind(). -daisy
682 retval
= sctp_del_bind_addr(bp
, sa_addr
);
684 addr_buf
+= af
->sockaddr_len
;
687 /* Failed. Add the ones that has been removed back */
689 sctp_bindx_add(sk
, addrs
, cnt
);
697 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
698 * the associations that are part of the endpoint indicating that a list of
699 * local addresses are removed from the endpoint.
701 * If any of the addresses is already in the bind address list of the
702 * association, we do not send the chunk for that association. But it will not
703 * affect other associations.
705 * Only sctp_setsockopt_bindx() is supposed to call this function.
707 static int sctp_send_asconf_del_ip(struct sock
*sk
,
708 struct sockaddr
*addrs
,
711 struct net
*net
= sock_net(sk
);
712 struct sctp_sock
*sp
;
713 struct sctp_endpoint
*ep
;
714 struct sctp_association
*asoc
;
715 struct sctp_transport
*transport
;
716 struct sctp_bind_addr
*bp
;
717 struct sctp_chunk
*chunk
;
718 union sctp_addr
*laddr
;
721 struct sctp_sockaddr_entry
*saddr
;
727 if (!net
->sctp
.addip_enable
)
733 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
734 __func__
, sk
, addrs
, addrcnt
);
736 list_for_each_entry(asoc
, &ep
->asocs
, asocs
) {
738 if (!asoc
->peer
.asconf_capable
)
741 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_DEL_IP
)
744 if (!sctp_state(asoc
, ESTABLISHED
))
747 /* Check if any address in the packed array of addresses is
748 * not present in the bind address list of the association.
749 * If so, do not send the asconf chunk to its peer, but
750 * continue with other associations.
753 for (i
= 0; i
< addrcnt
; i
++) {
755 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
761 if (!sctp_assoc_lookup_laddr(asoc
, laddr
))
764 addr_buf
+= af
->sockaddr_len
;
769 /* Find one address in the association's bind address list
770 * that is not in the packed array of addresses. This is to
771 * make sure that we do not delete all the addresses in the
774 bp
= &asoc
->base
.bind_addr
;
775 laddr
= sctp_find_unmatch_addr(bp
, (union sctp_addr
*)addrs
,
777 if ((laddr
== NULL
) && (addrcnt
== 1)) {
778 if (asoc
->asconf_addr_del_pending
)
780 asoc
->asconf_addr_del_pending
=
781 kzalloc(sizeof(union sctp_addr
), GFP_ATOMIC
);
782 if (asoc
->asconf_addr_del_pending
== NULL
) {
786 asoc
->asconf_addr_del_pending
->sa
.sa_family
=
788 asoc
->asconf_addr_del_pending
->v4
.sin_port
=
790 if (addrs
->sa_family
== AF_INET
) {
791 struct sockaddr_in
*sin
;
793 sin
= (struct sockaddr_in
*)addrs
;
794 asoc
->asconf_addr_del_pending
->v4
.sin_addr
.s_addr
= sin
->sin_addr
.s_addr
;
795 } else if (addrs
->sa_family
== AF_INET6
) {
796 struct sockaddr_in6
*sin6
;
798 sin6
= (struct sockaddr_in6
*)addrs
;
799 asoc
->asconf_addr_del_pending
->v6
.sin6_addr
= sin6
->sin6_addr
;
802 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
803 __func__
, asoc
, &asoc
->asconf_addr_del_pending
->sa
,
804 asoc
->asconf_addr_del_pending
);
806 asoc
->src_out_of_asoc_ok
= 1;
814 /* We do not need RCU protection throughout this loop
815 * because this is done under a socket lock from the
818 chunk
= sctp_make_asconf_update_ip(asoc
, laddr
, addrs
, addrcnt
,
826 /* Reset use_as_src flag for the addresses in the bind address
827 * list that are to be deleted.
830 for (i
= 0; i
< addrcnt
; i
++) {
832 af
= sctp_get_af_specific(laddr
->v4
.sin_family
);
833 list_for_each_entry(saddr
, &bp
->address_list
, list
) {
834 if (sctp_cmp_addr_exact(&saddr
->a
, laddr
))
835 saddr
->state
= SCTP_ADDR_DEL
;
837 addr_buf
+= af
->sockaddr_len
;
840 /* Update the route and saddr entries for all the transports
841 * as some of the addresses in the bind address list are
842 * about to be deleted and cannot be used as source addresses.
844 list_for_each_entry(transport
, &asoc
->peer
.transport_addr_list
,
846 dst_release(transport
->dst
);
847 sctp_transport_route(transport
, NULL
,
848 sctp_sk(asoc
->base
.sk
));
852 /* We don't need to transmit ASCONF */
854 retval
= sctp_send_asconf(asoc
, chunk
);
860 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
861 int sctp_asconf_mgmt(struct sctp_sock
*sp
, struct sctp_sockaddr_entry
*addrw
)
863 struct sock
*sk
= sctp_opt2sk(sp
);
864 union sctp_addr
*addr
;
867 /* It is safe to write port space in caller. */
869 addr
->v4
.sin_port
= htons(sp
->ep
->base
.bind_addr
.port
);
870 af
= sctp_get_af_specific(addr
->sa
.sa_family
);
873 if (sctp_verify_addr(sk
, addr
, af
->sockaddr_len
))
876 if (addrw
->state
== SCTP_ADDR_NEW
)
877 return sctp_send_asconf_add_ip(sk
, (struct sockaddr
*)addr
, 1);
879 return sctp_send_asconf_del_ip(sk
, (struct sockaddr
*)addr
, 1);
882 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
885 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
888 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
889 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
892 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
893 * Section 3.1.2 for this usage.
895 * addrs is a pointer to an array of one or more socket addresses. Each
896 * address is contained in its appropriate structure (i.e. struct
897 * sockaddr_in or struct sockaddr_in6) the family of the address type
898 * must be used to distinguish the address length (note that this
899 * representation is termed a "packed array" of addresses). The caller
900 * specifies the number of addresses in the array with addrcnt.
902 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
903 * -1, and sets errno to the appropriate error code.
905 * For SCTP, the port given in each socket address must be the same, or
906 * sctp_bindx() will fail, setting errno to EINVAL.
908 * The flags parameter is formed from the bitwise OR of zero or more of
909 * the following currently defined flags:
911 * SCTP_BINDX_ADD_ADDR
913 * SCTP_BINDX_REM_ADDR
915 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
916 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
917 * addresses from the association. The two flags are mutually exclusive;
918 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
919 * not remove all addresses from an association; sctp_bindx() will
920 * reject such an attempt with EINVAL.
922 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
923 * additional addresses with an endpoint after calling bind(). Or use
924 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
925 * socket is associated with so that no new association accepted will be
926 * associated with those addresses. If the endpoint supports dynamic
927 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
928 * endpoint to send the appropriate message to the peer to change the
929 * peers address lists.
931 * Adding and removing addresses from a connected association is
932 * optional functionality. Implementations that do not support this
933 * functionality should return EOPNOTSUPP.
935 * Basically do nothing but copying the addresses from user to kernel
936 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
937 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
940 * We don't use copy_from_user() for optimization: we first do the
941 * sanity checks (buffer size -fast- and access check-healthy
942 * pointer); if all of those succeed, then we can alloc the memory
943 * (expensive operation) needed to copy the data to kernel. Then we do
944 * the copying without checking the user space area
945 * (__copy_from_user()).
947 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
950 * sk The sk of the socket
951 * addrs The pointer to the addresses in user land
952 * addrssize Size of the addrs buffer
953 * op Operation to perform (add or remove, see the flags of
956 * Returns 0 if ok, <0 errno code on error.
958 static int sctp_setsockopt_bindx(struct sock
*sk
,
959 struct sockaddr __user
*addrs
,
960 int addrs_size
, int op
)
962 struct sockaddr
*kaddrs
;
966 struct sockaddr
*sa_addr
;
970 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
971 __func__
, sk
, addrs
, addrs_size
, op
);
973 if (unlikely(addrs_size
<= 0))
976 /* Check the user passed a healthy pointer. */
977 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
980 /* Alloc space for the address array in kernel memory. */
981 kaddrs
= kmalloc(addrs_size
, GFP_USER
| __GFP_NOWARN
);
982 if (unlikely(!kaddrs
))
985 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
990 /* Walk through the addrs buffer and count the number of addresses. */
992 while (walk_size
< addrs_size
) {
993 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
999 af
= sctp_get_af_specific(sa_addr
->sa_family
);
1001 /* If the address family is not supported or if this address
1002 * causes the address buffer to overflow return EINVAL.
1004 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1009 addr_buf
+= af
->sockaddr_len
;
1010 walk_size
+= af
->sockaddr_len
;
1015 case SCTP_BINDX_ADD_ADDR
:
1016 err
= sctp_bindx_add(sk
, kaddrs
, addrcnt
);
1019 err
= sctp_send_asconf_add_ip(sk
, kaddrs
, addrcnt
);
1022 case SCTP_BINDX_REM_ADDR
:
1023 err
= sctp_bindx_rem(sk
, kaddrs
, addrcnt
);
1026 err
= sctp_send_asconf_del_ip(sk
, kaddrs
, addrcnt
);
1040 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1042 * Common routine for handling connect() and sctp_connectx().
1043 * Connect will come in with just a single address.
1045 static int __sctp_connect(struct sock
*sk
,
1046 struct sockaddr
*kaddrs
,
1048 sctp_assoc_t
*assoc_id
)
1050 struct net
*net
= sock_net(sk
);
1051 struct sctp_sock
*sp
;
1052 struct sctp_endpoint
*ep
;
1053 struct sctp_association
*asoc
= NULL
;
1054 struct sctp_association
*asoc2
;
1055 struct sctp_transport
*transport
;
1062 union sctp_addr
*sa_addr
= NULL
;
1064 unsigned short port
;
1065 unsigned int f_flags
= 0;
1070 /* connect() cannot be done on a socket that is already in ESTABLISHED
1071 * state - UDP-style peeled off socket or a TCP-style socket that
1072 * is already connected.
1073 * It cannot be done even on a TCP-style listening socket.
1075 if (sctp_sstate(sk
, ESTABLISHED
) || sctp_sstate(sk
, CLOSING
) ||
1076 (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))) {
1081 /* Walk through the addrs buffer and count the number of addresses. */
1083 while (walk_size
< addrs_size
) {
1086 if (walk_size
+ sizeof(sa_family_t
) > addrs_size
) {
1092 af
= sctp_get_af_specific(sa_addr
->sa
.sa_family
);
1094 /* If the address family is not supported or if this address
1095 * causes the address buffer to overflow return EINVAL.
1097 if (!af
|| (walk_size
+ af
->sockaddr_len
) > addrs_size
) {
1102 port
= ntohs(sa_addr
->v4
.sin_port
);
1104 /* Save current address so we can work with it */
1105 memcpy(&to
, sa_addr
, af
->sockaddr_len
);
1107 err
= sctp_verify_addr(sk
, &to
, af
->sockaddr_len
);
1111 /* Make sure the destination port is correctly set
1114 if (asoc
&& asoc
->peer
.port
&& asoc
->peer
.port
!= port
) {
1119 /* Check if there already is a matching association on the
1120 * endpoint (other than the one created here).
1122 asoc2
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1123 if (asoc2
&& asoc2
!= asoc
) {
1124 if (asoc2
->state
>= SCTP_STATE_ESTABLISHED
)
1131 /* If we could not find a matching association on the endpoint,
1132 * make sure that there is no peeled-off association matching
1133 * the peer address even on another socket.
1135 if (sctp_endpoint_is_peeled_off(ep
, &to
)) {
1136 err
= -EADDRNOTAVAIL
;
1141 /* If a bind() or sctp_bindx() is not called prior to
1142 * an sctp_connectx() call, the system picks an
1143 * ephemeral port and will choose an address set
1144 * equivalent to binding with a wildcard address.
1146 if (!ep
->base
.bind_addr
.port
) {
1147 if (sctp_autobind(sk
)) {
1153 * If an unprivileged user inherits a 1-many
1154 * style socket with open associations on a
1155 * privileged port, it MAY be permitted to
1156 * accept new associations, but it SHOULD NOT
1157 * be permitted to open new associations.
1159 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1160 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1166 scope
= sctp_scope(&to
);
1167 asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1173 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
,
1181 /* Prime the peer's transport structures. */
1182 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
,
1190 addr_buf
+= af
->sockaddr_len
;
1191 walk_size
+= af
->sockaddr_len
;
1194 /* In case the user of sctp_connectx() wants an association
1195 * id back, assign one now.
1198 err
= sctp_assoc_set_id(asoc
, GFP_KERNEL
);
1203 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1208 /* Initialize sk's dport and daddr for getpeername() */
1209 inet_sk(sk
)->inet_dport
= htons(asoc
->peer
.port
);
1210 sp
->pf
->to_sk_daddr(sa_addr
, sk
);
1213 /* in-kernel sockets don't generally have a file allocated to them
1214 * if all they do is call sock_create_kern().
1216 if (sk
->sk_socket
->file
)
1217 f_flags
= sk
->sk_socket
->file
->f_flags
;
1219 timeo
= sock_sndtimeo(sk
, f_flags
& O_NONBLOCK
);
1222 *assoc_id
= asoc
->assoc_id
;
1223 err
= sctp_wait_for_connect(asoc
, &timeo
);
1224 /* Note: the asoc may be freed after the return of
1225 * sctp_wait_for_connect.
1228 /* Don't free association on exit. */
1232 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1233 __func__
, asoc
, kaddrs
, err
);
1236 /* sctp_primitive_ASSOCIATE may have added this association
1237 * To the hash table, try to unhash it, just in case, its a noop
1238 * if it wasn't hashed so we're safe
1240 sctp_association_free(asoc
);
1245 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1248 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1249 * sctp_assoc_t *asoc);
1251 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1252 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1253 * or IPv6 addresses.
1255 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1256 * Section 3.1.2 for this usage.
1258 * addrs is a pointer to an array of one or more socket addresses. Each
1259 * address is contained in its appropriate structure (i.e. struct
1260 * sockaddr_in or struct sockaddr_in6) the family of the address type
1261 * must be used to distengish the address length (note that this
1262 * representation is termed a "packed array" of addresses). The caller
1263 * specifies the number of addresses in the array with addrcnt.
1265 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1266 * the association id of the new association. On failure, sctp_connectx()
1267 * returns -1, and sets errno to the appropriate error code. The assoc_id
1268 * is not touched by the kernel.
1270 * For SCTP, the port given in each socket address must be the same, or
1271 * sctp_connectx() will fail, setting errno to EINVAL.
1273 * An application can use sctp_connectx to initiate an association with
1274 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1275 * allows a caller to specify multiple addresses at which a peer can be
1276 * reached. The way the SCTP stack uses the list of addresses to set up
1277 * the association is implementation dependent. This function only
1278 * specifies that the stack will try to make use of all the addresses in
1279 * the list when needed.
1281 * Note that the list of addresses passed in is only used for setting up
1282 * the association. It does not necessarily equal the set of addresses
1283 * the peer uses for the resulting association. If the caller wants to
1284 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1285 * retrieve them after the association has been set up.
1287 * Basically do nothing but copying the addresses from user to kernel
1288 * land and invoking either sctp_connectx(). This is used for tunneling
1289 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1291 * We don't use copy_from_user() for optimization: we first do the
1292 * sanity checks (buffer size -fast- and access check-healthy
1293 * pointer); if all of those succeed, then we can alloc the memory
1294 * (expensive operation) needed to copy the data to kernel. Then we do
1295 * the copying without checking the user space area
1296 * (__copy_from_user()).
1298 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1301 * sk The sk of the socket
1302 * addrs The pointer to the addresses in user land
1303 * addrssize Size of the addrs buffer
1305 * Returns >=0 if ok, <0 errno code on error.
1307 static int __sctp_setsockopt_connectx(struct sock
*sk
,
1308 struct sockaddr __user
*addrs
,
1310 sctp_assoc_t
*assoc_id
)
1312 struct sockaddr
*kaddrs
;
1313 gfp_t gfp
= GFP_KERNEL
;
1316 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1317 __func__
, sk
, addrs
, addrs_size
);
1319 if (unlikely(addrs_size
<= 0))
1322 /* Check the user passed a healthy pointer. */
1323 if (unlikely(!access_ok(VERIFY_READ
, addrs
, addrs_size
)))
1326 /* Alloc space for the address array in kernel memory. */
1327 if (sk
->sk_socket
->file
)
1328 gfp
= GFP_USER
| __GFP_NOWARN
;
1329 kaddrs
= kmalloc(addrs_size
, gfp
);
1330 if (unlikely(!kaddrs
))
1333 if (__copy_from_user(kaddrs
, addrs
, addrs_size
)) {
1336 err
= __sctp_connect(sk
, kaddrs
, addrs_size
, assoc_id
);
1345 * This is an older interface. It's kept for backward compatibility
1346 * to the option that doesn't provide association id.
1348 static int sctp_setsockopt_connectx_old(struct sock
*sk
,
1349 struct sockaddr __user
*addrs
,
1352 return __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, NULL
);
1356 * New interface for the API. The since the API is done with a socket
1357 * option, to make it simple we feed back the association id is as a return
1358 * indication to the call. Error is always negative and association id is
1361 static int sctp_setsockopt_connectx(struct sock
*sk
,
1362 struct sockaddr __user
*addrs
,
1365 sctp_assoc_t assoc_id
= 0;
1368 err
= __sctp_setsockopt_connectx(sk
, addrs
, addrs_size
, &assoc_id
);
1377 * New (hopefully final) interface for the API.
1378 * We use the sctp_getaddrs_old structure so that use-space library
1379 * can avoid any unnecessary allocations. The only different part
1380 * is that we store the actual length of the address buffer into the
1381 * addrs_num structure member. That way we can re-use the existing
1384 #ifdef CONFIG_COMPAT
1385 struct compat_sctp_getaddrs_old
{
1386 sctp_assoc_t assoc_id
;
1388 compat_uptr_t addrs
; /* struct sockaddr * */
1392 static int sctp_getsockopt_connectx3(struct sock
*sk
, int len
,
1393 char __user
*optval
,
1396 struct sctp_getaddrs_old param
;
1397 sctp_assoc_t assoc_id
= 0;
1400 #ifdef CONFIG_COMPAT
1401 if (in_compat_syscall()) {
1402 struct compat_sctp_getaddrs_old param32
;
1404 if (len
< sizeof(param32
))
1406 if (copy_from_user(¶m32
, optval
, sizeof(param32
)))
1409 param
.assoc_id
= param32
.assoc_id
;
1410 param
.addr_num
= param32
.addr_num
;
1411 param
.addrs
= compat_ptr(param32
.addrs
);
1415 if (len
< sizeof(param
))
1417 if (copy_from_user(¶m
, optval
, sizeof(param
)))
1421 err
= __sctp_setsockopt_connectx(sk
, (struct sockaddr __user
*)
1422 param
.addrs
, param
.addr_num
,
1424 if (err
== 0 || err
== -EINPROGRESS
) {
1425 if (copy_to_user(optval
, &assoc_id
, sizeof(assoc_id
)))
1427 if (put_user(sizeof(assoc_id
), optlen
))
1434 /* API 3.1.4 close() - UDP Style Syntax
1435 * Applications use close() to perform graceful shutdown (as described in
1436 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1437 * by a UDP-style socket.
1441 * ret = close(int sd);
1443 * sd - the socket descriptor of the associations to be closed.
1445 * To gracefully shutdown a specific association represented by the
1446 * UDP-style socket, an application should use the sendmsg() call,
1447 * passing no user data, but including the appropriate flag in the
1448 * ancillary data (see Section xxxx).
1450 * If sd in the close() call is a branched-off socket representing only
1451 * one association, the shutdown is performed on that association only.
1453 * 4.1.6 close() - TCP Style Syntax
1455 * Applications use close() to gracefully close down an association.
1459 * int close(int sd);
1461 * sd - the socket descriptor of the association to be closed.
1463 * After an application calls close() on a socket descriptor, no further
1464 * socket operations will succeed on that descriptor.
1466 * API 7.1.4 SO_LINGER
1468 * An application using the TCP-style socket can use this option to
1469 * perform the SCTP ABORT primitive. The linger option structure is:
1472 * int l_onoff; // option on/off
1473 * int l_linger; // linger time
1476 * To enable the option, set l_onoff to 1. If the l_linger value is set
1477 * to 0, calling close() is the same as the ABORT primitive. If the
1478 * value is set to a negative value, the setsockopt() call will return
1479 * an error. If the value is set to a positive value linger_time, the
1480 * close() can be blocked for at most linger_time ms. If the graceful
1481 * shutdown phase does not finish during this period, close() will
1482 * return but the graceful shutdown phase continues in the system.
1484 static void sctp_close(struct sock
*sk
, long timeout
)
1486 struct net
*net
= sock_net(sk
);
1487 struct sctp_endpoint
*ep
;
1488 struct sctp_association
*asoc
;
1489 struct list_head
*pos
, *temp
;
1490 unsigned int data_was_unread
;
1492 pr_debug("%s: sk:%p, timeout:%ld\n", __func__
, sk
, timeout
);
1495 sk
->sk_shutdown
= SHUTDOWN_MASK
;
1496 sk
->sk_state
= SCTP_SS_CLOSING
;
1498 ep
= sctp_sk(sk
)->ep
;
1500 /* Clean up any skbs sitting on the receive queue. */
1501 data_was_unread
= sctp_queue_purge_ulpevents(&sk
->sk_receive_queue
);
1502 data_was_unread
+= sctp_queue_purge_ulpevents(&sctp_sk(sk
)->pd_lobby
);
1504 /* Walk all associations on an endpoint. */
1505 list_for_each_safe(pos
, temp
, &ep
->asocs
) {
1506 asoc
= list_entry(pos
, struct sctp_association
, asocs
);
1508 if (sctp_style(sk
, TCP
)) {
1509 /* A closed association can still be in the list if
1510 * it belongs to a TCP-style listening socket that is
1511 * not yet accepted. If so, free it. If not, send an
1512 * ABORT or SHUTDOWN based on the linger options.
1514 if (sctp_state(asoc
, CLOSED
)) {
1515 sctp_association_free(asoc
);
1520 if (data_was_unread
|| !skb_queue_empty(&asoc
->ulpq
.lobby
) ||
1521 !skb_queue_empty(&asoc
->ulpq
.reasm
) ||
1522 (sock_flag(sk
, SOCK_LINGER
) && !sk
->sk_lingertime
)) {
1523 struct sctp_chunk
*chunk
;
1525 chunk
= sctp_make_abort_user(asoc
, NULL
, 0);
1526 sctp_primitive_ABORT(net
, asoc
, chunk
);
1528 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1531 /* On a TCP-style socket, block for at most linger_time if set. */
1532 if (sctp_style(sk
, TCP
) && timeout
)
1533 sctp_wait_for_close(sk
, timeout
);
1535 /* This will run the backlog queue. */
1538 /* Supposedly, no process has access to the socket, but
1539 * the net layers still may.
1540 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1541 * held and that should be grabbed before socket lock.
1543 spin_lock_bh(&net
->sctp
.addr_wq_lock
);
1546 /* Hold the sock, since sk_common_release() will put sock_put()
1547 * and we have just a little more cleanup.
1550 sk_common_release(sk
);
1553 spin_unlock_bh(&net
->sctp
.addr_wq_lock
);
1557 SCTP_DBG_OBJCNT_DEC(sock
);
1560 /* Handle EPIPE error. */
1561 static int sctp_error(struct sock
*sk
, int flags
, int err
)
1564 err
= sock_error(sk
) ? : -EPIPE
;
1565 if (err
== -EPIPE
&& !(flags
& MSG_NOSIGNAL
))
1566 send_sig(SIGPIPE
, current
, 0);
1570 /* API 3.1.3 sendmsg() - UDP Style Syntax
1572 * An application uses sendmsg() and recvmsg() calls to transmit data to
1573 * and receive data from its peer.
1575 * ssize_t sendmsg(int socket, const struct msghdr *message,
1578 * socket - the socket descriptor of the endpoint.
1579 * message - pointer to the msghdr structure which contains a single
1580 * user message and possibly some ancillary data.
1582 * See Section 5 for complete description of the data
1585 * flags - flags sent or received with the user message, see Section
1586 * 5 for complete description of the flags.
1588 * Note: This function could use a rewrite especially when explicit
1589 * connect support comes in.
1591 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1593 static int sctp_msghdr_parse(const struct msghdr
*, sctp_cmsgs_t
*);
1595 static int sctp_sendmsg(struct sock
*sk
, struct msghdr
*msg
, size_t msg_len
)
1597 struct net
*net
= sock_net(sk
);
1598 struct sctp_sock
*sp
;
1599 struct sctp_endpoint
*ep
;
1600 struct sctp_association
*new_asoc
= NULL
, *asoc
= NULL
;
1601 struct sctp_transport
*transport
, *chunk_tp
;
1602 struct sctp_chunk
*chunk
;
1604 struct sockaddr
*msg_name
= NULL
;
1605 struct sctp_sndrcvinfo default_sinfo
;
1606 struct sctp_sndrcvinfo
*sinfo
;
1607 struct sctp_initmsg
*sinit
;
1608 sctp_assoc_t associd
= 0;
1609 sctp_cmsgs_t cmsgs
= { NULL
};
1611 bool fill_sinfo_ttl
= false, wait_connect
= false;
1612 struct sctp_datamsg
*datamsg
;
1613 int msg_flags
= msg
->msg_flags
;
1614 __u16 sinfo_flags
= 0;
1622 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__
, sk
,
1625 /* We cannot send a message over a TCP-style listening socket. */
1626 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
)) {
1631 /* Parse out the SCTP CMSGs. */
1632 err
= sctp_msghdr_parse(msg
, &cmsgs
);
1634 pr_debug("%s: msghdr parse err:%x\n", __func__
, err
);
1638 /* Fetch the destination address for this packet. This
1639 * address only selects the association--it is not necessarily
1640 * the address we will send to.
1641 * For a peeled-off socket, msg_name is ignored.
1643 if (!sctp_style(sk
, UDP_HIGH_BANDWIDTH
) && msg
->msg_name
) {
1644 int msg_namelen
= msg
->msg_namelen
;
1646 err
= sctp_verify_addr(sk
, (union sctp_addr
*)msg
->msg_name
,
1651 if (msg_namelen
> sizeof(to
))
1652 msg_namelen
= sizeof(to
);
1653 memcpy(&to
, msg
->msg_name
, msg_namelen
);
1654 msg_name
= msg
->msg_name
;
1658 if (cmsgs
.sinfo
!= NULL
) {
1659 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1660 default_sinfo
.sinfo_stream
= cmsgs
.sinfo
->snd_sid
;
1661 default_sinfo
.sinfo_flags
= cmsgs
.sinfo
->snd_flags
;
1662 default_sinfo
.sinfo_ppid
= cmsgs
.sinfo
->snd_ppid
;
1663 default_sinfo
.sinfo_context
= cmsgs
.sinfo
->snd_context
;
1664 default_sinfo
.sinfo_assoc_id
= cmsgs
.sinfo
->snd_assoc_id
;
1666 sinfo
= &default_sinfo
;
1667 fill_sinfo_ttl
= true;
1669 sinfo
= cmsgs
.srinfo
;
1671 /* Did the user specify SNDINFO/SNDRCVINFO? */
1673 sinfo_flags
= sinfo
->sinfo_flags
;
1674 associd
= sinfo
->sinfo_assoc_id
;
1677 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__
,
1678 msg_len
, sinfo_flags
);
1680 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1681 if (sctp_style(sk
, TCP
) && (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
))) {
1686 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1687 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1688 * If SCTP_ABORT is set, the message length could be non zero with
1689 * the msg_iov set to the user abort reason.
1691 if (((sinfo_flags
& SCTP_EOF
) && (msg_len
> 0)) ||
1692 (!(sinfo_flags
& (SCTP_EOF
|SCTP_ABORT
)) && (msg_len
== 0))) {
1697 /* If SCTP_ADDR_OVER is set, there must be an address
1698 * specified in msg_name.
1700 if ((sinfo_flags
& SCTP_ADDR_OVER
) && (!msg
->msg_name
)) {
1707 pr_debug("%s: about to look up association\n", __func__
);
1711 /* If a msg_name has been specified, assume this is to be used. */
1713 /* Look for a matching association on the endpoint. */
1714 asoc
= sctp_endpoint_lookup_assoc(ep
, &to
, &transport
);
1716 /* If we could not find a matching association on the
1717 * endpoint, make sure that it is not a TCP-style
1718 * socket that already has an association or there is
1719 * no peeled-off association on another socket.
1722 ((sctp_style(sk
, TCP
) &&
1723 (sctp_sstate(sk
, ESTABLISHED
) ||
1724 sctp_sstate(sk
, CLOSING
))) ||
1725 sctp_endpoint_is_peeled_off(ep
, &to
))) {
1726 err
= -EADDRNOTAVAIL
;
1730 asoc
= sctp_id2assoc(sk
, associd
);
1738 pr_debug("%s: just looked up association:%p\n", __func__
, asoc
);
1740 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1741 * socket that has an association in CLOSED state. This can
1742 * happen when an accepted socket has an association that is
1745 if (sctp_state(asoc
, CLOSED
) && sctp_style(sk
, TCP
)) {
1750 if (sinfo_flags
& SCTP_EOF
) {
1751 pr_debug("%s: shutting down association:%p\n",
1754 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
1758 if (sinfo_flags
& SCTP_ABORT
) {
1760 chunk
= sctp_make_abort_user(asoc
, msg
, msg_len
);
1766 pr_debug("%s: aborting association:%p\n",
1769 sctp_primitive_ABORT(net
, asoc
, chunk
);
1775 /* Do we need to create the association? */
1777 pr_debug("%s: there is no association yet\n", __func__
);
1779 if (sinfo_flags
& (SCTP_EOF
| SCTP_ABORT
)) {
1784 /* Check for invalid stream against the stream counts,
1785 * either the default or the user specified stream counts.
1788 if (!sinit
|| !sinit
->sinit_num_ostreams
) {
1789 /* Check against the defaults. */
1790 if (sinfo
->sinfo_stream
>=
1791 sp
->initmsg
.sinit_num_ostreams
) {
1796 /* Check against the requested. */
1797 if (sinfo
->sinfo_stream
>=
1798 sinit
->sinit_num_ostreams
) {
1806 * API 3.1.2 bind() - UDP Style Syntax
1807 * If a bind() or sctp_bindx() is not called prior to a
1808 * sendmsg() call that initiates a new association, the
1809 * system picks an ephemeral port and will choose an address
1810 * set equivalent to binding with a wildcard address.
1812 if (!ep
->base
.bind_addr
.port
) {
1813 if (sctp_autobind(sk
)) {
1819 * If an unprivileged user inherits a one-to-many
1820 * style socket with open associations on a privileged
1821 * port, it MAY be permitted to accept new associations,
1822 * but it SHOULD NOT be permitted to open new
1825 if (ep
->base
.bind_addr
.port
< PROT_SOCK
&&
1826 !ns_capable(net
->user_ns
, CAP_NET_BIND_SERVICE
)) {
1832 scope
= sctp_scope(&to
);
1833 new_asoc
= sctp_association_new(ep
, sk
, scope
, GFP_KERNEL
);
1839 err
= sctp_assoc_set_bind_addr_from_ep(asoc
, scope
, GFP_KERNEL
);
1845 /* If the SCTP_INIT ancillary data is specified, set all
1846 * the association init values accordingly.
1849 if (sinit
->sinit_num_ostreams
) {
1850 asoc
->c
.sinit_num_ostreams
=
1851 sinit
->sinit_num_ostreams
;
1853 if (sinit
->sinit_max_instreams
) {
1854 asoc
->c
.sinit_max_instreams
=
1855 sinit
->sinit_max_instreams
;
1857 if (sinit
->sinit_max_attempts
) {
1858 asoc
->max_init_attempts
1859 = sinit
->sinit_max_attempts
;
1861 if (sinit
->sinit_max_init_timeo
) {
1862 asoc
->max_init_timeo
=
1863 msecs_to_jiffies(sinit
->sinit_max_init_timeo
);
1867 /* Prime the peer's transport structures. */
1868 transport
= sctp_assoc_add_peer(asoc
, &to
, GFP_KERNEL
, SCTP_UNKNOWN
);
1875 /* ASSERT: we have a valid association at this point. */
1876 pr_debug("%s: we have a valid association\n", __func__
);
1879 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1880 * one with some defaults.
1882 memset(&default_sinfo
, 0, sizeof(default_sinfo
));
1883 default_sinfo
.sinfo_stream
= asoc
->default_stream
;
1884 default_sinfo
.sinfo_flags
= asoc
->default_flags
;
1885 default_sinfo
.sinfo_ppid
= asoc
->default_ppid
;
1886 default_sinfo
.sinfo_context
= asoc
->default_context
;
1887 default_sinfo
.sinfo_timetolive
= asoc
->default_timetolive
;
1888 default_sinfo
.sinfo_assoc_id
= sctp_assoc2id(asoc
);
1890 sinfo
= &default_sinfo
;
1891 } else if (fill_sinfo_ttl
) {
1892 /* In case SNDINFO was specified, we still need to fill
1893 * it with a default ttl from the assoc here.
1895 sinfo
->sinfo_timetolive
= asoc
->default_timetolive
;
1898 /* API 7.1.7, the sndbuf size per association bounds the
1899 * maximum size of data that can be sent in a single send call.
1901 if (msg_len
> sk
->sk_sndbuf
) {
1906 if (asoc
->pmtu_pending
)
1907 sctp_assoc_pending_pmtu(sk
, asoc
);
1909 /* If fragmentation is disabled and the message length exceeds the
1910 * association fragmentation point, return EMSGSIZE. The I-D
1911 * does not specify what this error is, but this looks like
1914 if (sctp_sk(sk
)->disable_fragments
&& (msg_len
> asoc
->frag_point
)) {
1919 /* Check for invalid stream. */
1920 if (sinfo
->sinfo_stream
>= asoc
->c
.sinit_num_ostreams
) {
1925 if (sctp_wspace(asoc
) < msg_len
)
1926 sctp_prsctp_prune(asoc
, sinfo
, msg_len
- sctp_wspace(asoc
));
1928 timeo
= sock_sndtimeo(sk
, msg
->msg_flags
& MSG_DONTWAIT
);
1929 if (!sctp_wspace(asoc
)) {
1930 err
= sctp_wait_for_sndbuf(asoc
, &timeo
, msg_len
);
1935 /* If an address is passed with the sendto/sendmsg call, it is used
1936 * to override the primary destination address in the TCP model, or
1937 * when SCTP_ADDR_OVER flag is set in the UDP model.
1939 if ((sctp_style(sk
, TCP
) && msg_name
) ||
1940 (sinfo_flags
& SCTP_ADDR_OVER
)) {
1941 chunk_tp
= sctp_assoc_lookup_paddr(asoc
, &to
);
1949 /* Auto-connect, if we aren't connected already. */
1950 if (sctp_state(asoc
, CLOSED
)) {
1951 err
= sctp_primitive_ASSOCIATE(net
, asoc
, NULL
);
1955 wait_connect
= true;
1956 pr_debug("%s: we associated primitively\n", __func__
);
1959 /* Break the message into multiple chunks of maximum size. */
1960 datamsg
= sctp_datamsg_from_user(asoc
, sinfo
, &msg
->msg_iter
);
1961 if (IS_ERR(datamsg
)) {
1962 err
= PTR_ERR(datamsg
);
1966 /* Now send the (possibly) fragmented message. */
1967 list_for_each_entry(chunk
, &datamsg
->chunks
, frag_list
) {
1968 sctp_chunk_hold(chunk
);
1970 /* Do accounting for the write space. */
1971 sctp_set_owner_w(chunk
);
1973 chunk
->transport
= chunk_tp
;
1976 /* Send it to the lower layers. Note: all chunks
1977 * must either fail or succeed. The lower layer
1978 * works that way today. Keep it that way or this
1981 err
= sctp_primitive_SEND(net
, asoc
, datamsg
);
1982 /* Did the lower layer accept the chunk? */
1984 sctp_datamsg_free(datamsg
);
1988 pr_debug("%s: we sent primitively\n", __func__
);
1990 sctp_datamsg_put(datamsg
);
1993 if (unlikely(wait_connect
)) {
1994 timeo
= sock_sndtimeo(sk
, msg_flags
& MSG_DONTWAIT
);
1995 sctp_wait_for_connect(asoc
, &timeo
);
1998 /* If we are already past ASSOCIATE, the lower
1999 * layers are responsible for association cleanup.
2005 sctp_association_free(asoc
);
2010 return sctp_error(sk
, msg_flags
, err
);
2017 err
= sock_error(sk
);
2027 /* This is an extended version of skb_pull() that removes the data from the
2028 * start of a skb even when data is spread across the list of skb's in the
2029 * frag_list. len specifies the total amount of data that needs to be removed.
2030 * when 'len' bytes could be removed from the skb, it returns 0.
2031 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2032 * could not be removed.
2034 static int sctp_skb_pull(struct sk_buff
*skb
, int len
)
2036 struct sk_buff
*list
;
2037 int skb_len
= skb_headlen(skb
);
2040 if (len
<= skb_len
) {
2041 __skb_pull(skb
, len
);
2045 __skb_pull(skb
, skb_len
);
2047 skb_walk_frags(skb
, list
) {
2048 rlen
= sctp_skb_pull(list
, len
);
2049 skb
->len
-= (len
-rlen
);
2050 skb
->data_len
-= (len
-rlen
);
2061 /* API 3.1.3 recvmsg() - UDP Style Syntax
2063 * ssize_t recvmsg(int socket, struct msghdr *message,
2066 * socket - the socket descriptor of the endpoint.
2067 * message - pointer to the msghdr structure which contains a single
2068 * user message and possibly some ancillary data.
2070 * See Section 5 for complete description of the data
2073 * flags - flags sent or received with the user message, see Section
2074 * 5 for complete description of the flags.
2076 static int sctp_recvmsg(struct sock
*sk
, struct msghdr
*msg
, size_t len
,
2077 int noblock
, int flags
, int *addr_len
)
2079 struct sctp_ulpevent
*event
= NULL
;
2080 struct sctp_sock
*sp
= sctp_sk(sk
);
2081 struct sk_buff
*skb
, *head_skb
;
2086 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2087 "addr_len:%p)\n", __func__
, sk
, msg
, len
, noblock
, flags
,
2092 if (sctp_style(sk
, TCP
) && !sctp_sstate(sk
, ESTABLISHED
) &&
2093 !sctp_sstate(sk
, CLOSING
) && !sctp_sstate(sk
, CLOSED
)) {
2098 skb
= sctp_skb_recv_datagram(sk
, flags
, noblock
, &err
);
2102 /* Get the total length of the skb including any skb's in the
2111 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
2113 event
= sctp_skb2event(skb
);
2118 if (event
->chunk
&& event
->chunk
->head_skb
)
2119 head_skb
= event
->chunk
->head_skb
;
2122 sock_recv_ts_and_drops(msg
, sk
, head_skb
);
2123 if (sctp_ulpevent_is_notification(event
)) {
2124 msg
->msg_flags
|= MSG_NOTIFICATION
;
2125 sp
->pf
->event_msgname(event
, msg
->msg_name
, addr_len
);
2127 sp
->pf
->skb_msgname(head_skb
, msg
->msg_name
, addr_len
);
2130 /* Check if we allow SCTP_NXTINFO. */
2131 if (sp
->recvnxtinfo
)
2132 sctp_ulpevent_read_nxtinfo(event
, msg
, sk
);
2133 /* Check if we allow SCTP_RCVINFO. */
2134 if (sp
->recvrcvinfo
)
2135 sctp_ulpevent_read_rcvinfo(event
, msg
);
2136 /* Check if we allow SCTP_SNDRCVINFO. */
2137 if (sp
->subscribe
.sctp_data_io_event
)
2138 sctp_ulpevent_read_sndrcvinfo(event
, msg
);
2142 /* If skb's length exceeds the user's buffer, update the skb and
2143 * push it back to the receive_queue so that the next call to
2144 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2146 if (skb_len
> copied
) {
2147 msg
->msg_flags
&= ~MSG_EOR
;
2148 if (flags
& MSG_PEEK
)
2150 sctp_skb_pull(skb
, copied
);
2151 skb_queue_head(&sk
->sk_receive_queue
, skb
);
2153 /* When only partial message is copied to the user, increase
2154 * rwnd by that amount. If all the data in the skb is read,
2155 * rwnd is updated when the event is freed.
2157 if (!sctp_ulpevent_is_notification(event
))
2158 sctp_assoc_rwnd_increase(event
->asoc
, copied
);
2160 } else if ((event
->msg_flags
& MSG_NOTIFICATION
) ||
2161 (event
->msg_flags
& MSG_EOR
))
2162 msg
->msg_flags
|= MSG_EOR
;
2164 msg
->msg_flags
&= ~MSG_EOR
;
2167 if (flags
& MSG_PEEK
) {
2168 /* Release the skb reference acquired after peeking the skb in
2169 * sctp_skb_recv_datagram().
2173 /* Free the event which includes releasing the reference to
2174 * the owner of the skb, freeing the skb and updating the
2177 sctp_ulpevent_free(event
);
2184 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2186 * This option is a on/off flag. If enabled no SCTP message
2187 * fragmentation will be performed. Instead if a message being sent
2188 * exceeds the current PMTU size, the message will NOT be sent and
2189 * instead a error will be indicated to the user.
2191 static int sctp_setsockopt_disable_fragments(struct sock
*sk
,
2192 char __user
*optval
,
2193 unsigned int optlen
)
2197 if (optlen
< sizeof(int))
2200 if (get_user(val
, (int __user
*)optval
))
2203 sctp_sk(sk
)->disable_fragments
= (val
== 0) ? 0 : 1;
2208 static int sctp_setsockopt_events(struct sock
*sk
, char __user
*optval
,
2209 unsigned int optlen
)
2211 struct sctp_association
*asoc
;
2212 struct sctp_ulpevent
*event
;
2214 if (optlen
> sizeof(struct sctp_event_subscribe
))
2216 if (copy_from_user(&sctp_sk(sk
)->subscribe
, optval
, optlen
))
2219 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2220 * if there is no data to be sent or retransmit, the stack will
2221 * immediately send up this notification.
2223 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT
,
2224 &sctp_sk(sk
)->subscribe
)) {
2225 asoc
= sctp_id2assoc(sk
, 0);
2227 if (asoc
&& sctp_outq_is_empty(&asoc
->outqueue
)) {
2228 event
= sctp_ulpevent_make_sender_dry_event(asoc
,
2233 sctp_ulpq_tail_event(&asoc
->ulpq
, event
);
2240 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2242 * This socket option is applicable to the UDP-style socket only. When
2243 * set it will cause associations that are idle for more than the
2244 * specified number of seconds to automatically close. An association
2245 * being idle is defined an association that has NOT sent or received
2246 * user data. The special value of '0' indicates that no automatic
2247 * close of any associations should be performed. The option expects an
2248 * integer defining the number of seconds of idle time before an
2249 * association is closed.
2251 static int sctp_setsockopt_autoclose(struct sock
*sk
, char __user
*optval
,
2252 unsigned int optlen
)
2254 struct sctp_sock
*sp
= sctp_sk(sk
);
2255 struct net
*net
= sock_net(sk
);
2257 /* Applicable to UDP-style socket only */
2258 if (sctp_style(sk
, TCP
))
2260 if (optlen
!= sizeof(int))
2262 if (copy_from_user(&sp
->autoclose
, optval
, optlen
))
2265 if (sp
->autoclose
> net
->sctp
.max_autoclose
)
2266 sp
->autoclose
= net
->sctp
.max_autoclose
;
2271 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2273 * Applications can enable or disable heartbeats for any peer address of
2274 * an association, modify an address's heartbeat interval, force a
2275 * heartbeat to be sent immediately, and adjust the address's maximum
2276 * number of retransmissions sent before an address is considered
2277 * unreachable. The following structure is used to access and modify an
2278 * address's parameters:
2280 * struct sctp_paddrparams {
2281 * sctp_assoc_t spp_assoc_id;
2282 * struct sockaddr_storage spp_address;
2283 * uint32_t spp_hbinterval;
2284 * uint16_t spp_pathmaxrxt;
2285 * uint32_t spp_pathmtu;
2286 * uint32_t spp_sackdelay;
2287 * uint32_t spp_flags;
2290 * spp_assoc_id - (one-to-many style socket) This is filled in the
2291 * application, and identifies the association for
2293 * spp_address - This specifies which address is of interest.
2294 * spp_hbinterval - This contains the value of the heartbeat interval,
2295 * in milliseconds. If a value of zero
2296 * is present in this field then no changes are to
2297 * be made to this parameter.
2298 * spp_pathmaxrxt - This contains the maximum number of
2299 * retransmissions before this address shall be
2300 * considered unreachable. If a value of zero
2301 * is present in this field then no changes are to
2302 * be made to this parameter.
2303 * spp_pathmtu - When Path MTU discovery is disabled the value
2304 * specified here will be the "fixed" path mtu.
2305 * Note that if the spp_address field is empty
2306 * then all associations on this address will
2307 * have this fixed path mtu set upon them.
2309 * spp_sackdelay - When delayed sack is enabled, this value specifies
2310 * the number of milliseconds that sacks will be delayed
2311 * for. This value will apply to all addresses of an
2312 * association if the spp_address field is empty. Note
2313 * also, that if delayed sack is enabled and this
2314 * value is set to 0, no change is made to the last
2315 * recorded delayed sack timer value.
2317 * spp_flags - These flags are used to control various features
2318 * on an association. The flag field may contain
2319 * zero or more of the following options.
2321 * SPP_HB_ENABLE - Enable heartbeats on the
2322 * specified address. Note that if the address
2323 * field is empty all addresses for the association
2324 * have heartbeats enabled upon them.
2326 * SPP_HB_DISABLE - Disable heartbeats on the
2327 * speicifed address. Note that if the address
2328 * field is empty all addresses for the association
2329 * will have their heartbeats disabled. Note also
2330 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2331 * mutually exclusive, only one of these two should
2332 * be specified. Enabling both fields will have
2333 * undetermined results.
2335 * SPP_HB_DEMAND - Request a user initiated heartbeat
2336 * to be made immediately.
2338 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2339 * heartbeat delayis to be set to the value of 0
2342 * SPP_PMTUD_ENABLE - This field will enable PMTU
2343 * discovery upon the specified address. Note that
2344 * if the address feild is empty then all addresses
2345 * on the association are effected.
2347 * SPP_PMTUD_DISABLE - This field will disable PMTU
2348 * discovery upon the specified address. Note that
2349 * if the address feild is empty then all addresses
2350 * on the association are effected. Not also that
2351 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2352 * exclusive. Enabling both will have undetermined
2355 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2356 * on delayed sack. The time specified in spp_sackdelay
2357 * is used to specify the sack delay for this address. Note
2358 * that if spp_address is empty then all addresses will
2359 * enable delayed sack and take on the sack delay
2360 * value specified in spp_sackdelay.
2361 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2362 * off delayed sack. If the spp_address field is blank then
2363 * delayed sack is disabled for the entire association. Note
2364 * also that this field is mutually exclusive to
2365 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2368 static int sctp_apply_peer_addr_params(struct sctp_paddrparams
*params
,
2369 struct sctp_transport
*trans
,
2370 struct sctp_association
*asoc
,
2371 struct sctp_sock
*sp
,
2374 int sackdelay_change
)
2378 if (params
->spp_flags
& SPP_HB_DEMAND
&& trans
) {
2379 struct net
*net
= sock_net(trans
->asoc
->base
.sk
);
2381 error
= sctp_primitive_REQUESTHEARTBEAT(net
, trans
->asoc
, trans
);
2386 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2387 * this field is ignored. Note also that a value of zero indicates
2388 * the current setting should be left unchanged.
2390 if (params
->spp_flags
& SPP_HB_ENABLE
) {
2392 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2393 * set. This lets us use 0 value when this flag
2396 if (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)
2397 params
->spp_hbinterval
= 0;
2399 if (params
->spp_hbinterval
||
2400 (params
->spp_flags
& SPP_HB_TIME_IS_ZERO
)) {
2403 msecs_to_jiffies(params
->spp_hbinterval
);
2406 msecs_to_jiffies(params
->spp_hbinterval
);
2408 sp
->hbinterval
= params
->spp_hbinterval
;
2415 trans
->param_flags
=
2416 (trans
->param_flags
& ~SPP_HB
) | hb_change
;
2419 (asoc
->param_flags
& ~SPP_HB
) | hb_change
;
2422 (sp
->param_flags
& ~SPP_HB
) | hb_change
;
2426 /* When Path MTU discovery is disabled the value specified here will
2427 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2428 * include the flag SPP_PMTUD_DISABLE for this field to have any
2431 if ((params
->spp_flags
& SPP_PMTUD_DISABLE
) && params
->spp_pathmtu
) {
2433 trans
->pathmtu
= params
->spp_pathmtu
;
2434 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2436 asoc
->pathmtu
= params
->spp_pathmtu
;
2437 sctp_frag_point(asoc
, params
->spp_pathmtu
);
2439 sp
->pathmtu
= params
->spp_pathmtu
;
2445 int update
= (trans
->param_flags
& SPP_PMTUD_DISABLE
) &&
2446 (params
->spp_flags
& SPP_PMTUD_ENABLE
);
2447 trans
->param_flags
=
2448 (trans
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2450 sctp_transport_pmtu(trans
, sctp_opt2sk(sp
));
2451 sctp_assoc_sync_pmtu(sctp_opt2sk(sp
), asoc
);
2455 (asoc
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2458 (sp
->param_flags
& ~SPP_PMTUD
) | pmtud_change
;
2462 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2463 * value of this field is ignored. Note also that a value of zero
2464 * indicates the current setting should be left unchanged.
2466 if ((params
->spp_flags
& SPP_SACKDELAY_ENABLE
) && params
->spp_sackdelay
) {
2469 msecs_to_jiffies(params
->spp_sackdelay
);
2472 msecs_to_jiffies(params
->spp_sackdelay
);
2474 sp
->sackdelay
= params
->spp_sackdelay
;
2478 if (sackdelay_change
) {
2480 trans
->param_flags
=
2481 (trans
->param_flags
& ~SPP_SACKDELAY
) |
2485 (asoc
->param_flags
& ~SPP_SACKDELAY
) |
2489 (sp
->param_flags
& ~SPP_SACKDELAY
) |
2494 /* Note that a value of zero indicates the current setting should be
2497 if (params
->spp_pathmaxrxt
) {
2499 trans
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2501 asoc
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2503 sp
->pathmaxrxt
= params
->spp_pathmaxrxt
;
2510 static int sctp_setsockopt_peer_addr_params(struct sock
*sk
,
2511 char __user
*optval
,
2512 unsigned int optlen
)
2514 struct sctp_paddrparams params
;
2515 struct sctp_transport
*trans
= NULL
;
2516 struct sctp_association
*asoc
= NULL
;
2517 struct sctp_sock
*sp
= sctp_sk(sk
);
2519 int hb_change
, pmtud_change
, sackdelay_change
;
2521 if (optlen
!= sizeof(struct sctp_paddrparams
))
2524 if (copy_from_user(¶ms
, optval
, optlen
))
2527 /* Validate flags and value parameters. */
2528 hb_change
= params
.spp_flags
& SPP_HB
;
2529 pmtud_change
= params
.spp_flags
& SPP_PMTUD
;
2530 sackdelay_change
= params
.spp_flags
& SPP_SACKDELAY
;
2532 if (hb_change
== SPP_HB
||
2533 pmtud_change
== SPP_PMTUD
||
2534 sackdelay_change
== SPP_SACKDELAY
||
2535 params
.spp_sackdelay
> 500 ||
2536 (params
.spp_pathmtu
&&
2537 params
.spp_pathmtu
< SCTP_DEFAULT_MINSEGMENT
))
2540 /* If an address other than INADDR_ANY is specified, and
2541 * no transport is found, then the request is invalid.
2543 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
2544 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
2545 params
.spp_assoc_id
);
2550 /* Get association, if assoc_id != 0 and the socket is a one
2551 * to many style socket, and an association was not found, then
2552 * the id was invalid.
2554 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
2555 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
))
2558 /* Heartbeat demand can only be sent on a transport or
2559 * association, but not a socket.
2561 if (params
.spp_flags
& SPP_HB_DEMAND
&& !trans
&& !asoc
)
2564 /* Process parameters. */
2565 error
= sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2566 hb_change
, pmtud_change
,
2572 /* If changes are for association, also apply parameters to each
2575 if (!trans
&& asoc
) {
2576 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2578 sctp_apply_peer_addr_params(¶ms
, trans
, asoc
, sp
,
2579 hb_change
, pmtud_change
,
2587 static inline __u32
sctp_spp_sackdelay_enable(__u32 param_flags
)
2589 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_ENABLE
;
2592 static inline __u32
sctp_spp_sackdelay_disable(__u32 param_flags
)
2594 return (param_flags
& ~SPP_SACKDELAY
) | SPP_SACKDELAY_DISABLE
;
2598 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2600 * This option will effect the way delayed acks are performed. This
2601 * option allows you to get or set the delayed ack time, in
2602 * milliseconds. It also allows changing the delayed ack frequency.
2603 * Changing the frequency to 1 disables the delayed sack algorithm. If
2604 * the assoc_id is 0, then this sets or gets the endpoints default
2605 * values. If the assoc_id field is non-zero, then the set or get
2606 * effects the specified association for the one to many model (the
2607 * assoc_id field is ignored by the one to one model). Note that if
2608 * sack_delay or sack_freq are 0 when setting this option, then the
2609 * current values will remain unchanged.
2611 * struct sctp_sack_info {
2612 * sctp_assoc_t sack_assoc_id;
2613 * uint32_t sack_delay;
2614 * uint32_t sack_freq;
2617 * sack_assoc_id - This parameter, indicates which association the user
2618 * is performing an action upon. Note that if this field's value is
2619 * zero then the endpoints default value is changed (effecting future
2620 * associations only).
2622 * sack_delay - This parameter contains the number of milliseconds that
2623 * the user is requesting the delayed ACK timer be set to. Note that
2624 * this value is defined in the standard to be between 200 and 500
2627 * sack_freq - This parameter contains the number of packets that must
2628 * be received before a sack is sent without waiting for the delay
2629 * timer to expire. The default value for this is 2, setting this
2630 * value to 1 will disable the delayed sack algorithm.
2633 static int sctp_setsockopt_delayed_ack(struct sock
*sk
,
2634 char __user
*optval
, unsigned int optlen
)
2636 struct sctp_sack_info params
;
2637 struct sctp_transport
*trans
= NULL
;
2638 struct sctp_association
*asoc
= NULL
;
2639 struct sctp_sock
*sp
= sctp_sk(sk
);
2641 if (optlen
== sizeof(struct sctp_sack_info
)) {
2642 if (copy_from_user(¶ms
, optval
, optlen
))
2645 if (params
.sack_delay
== 0 && params
.sack_freq
== 0)
2647 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
2648 pr_warn_ratelimited(DEPRECATED
2650 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2651 "Use struct sctp_sack_info instead\n",
2652 current
->comm
, task_pid_nr(current
));
2653 if (copy_from_user(¶ms
, optval
, optlen
))
2656 if (params
.sack_delay
== 0)
2657 params
.sack_freq
= 1;
2659 params
.sack_freq
= 0;
2663 /* Validate value parameter. */
2664 if (params
.sack_delay
> 500)
2667 /* Get association, if sack_assoc_id != 0 and the socket is a one
2668 * to many style socket, and an association was not found, then
2669 * the id was invalid.
2671 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
2672 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
2675 if (params
.sack_delay
) {
2678 msecs_to_jiffies(params
.sack_delay
);
2680 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2682 sp
->sackdelay
= params
.sack_delay
;
2684 sctp_spp_sackdelay_enable(sp
->param_flags
);
2688 if (params
.sack_freq
== 1) {
2691 sctp_spp_sackdelay_disable(asoc
->param_flags
);
2694 sctp_spp_sackdelay_disable(sp
->param_flags
);
2696 } else if (params
.sack_freq
> 1) {
2698 asoc
->sackfreq
= params
.sack_freq
;
2700 sctp_spp_sackdelay_enable(asoc
->param_flags
);
2702 sp
->sackfreq
= params
.sack_freq
;
2704 sctp_spp_sackdelay_enable(sp
->param_flags
);
2708 /* If change is for association, also apply to each transport. */
2710 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
2712 if (params
.sack_delay
) {
2714 msecs_to_jiffies(params
.sack_delay
);
2715 trans
->param_flags
=
2716 sctp_spp_sackdelay_enable(trans
->param_flags
);
2718 if (params
.sack_freq
== 1) {
2719 trans
->param_flags
=
2720 sctp_spp_sackdelay_disable(trans
->param_flags
);
2721 } else if (params
.sack_freq
> 1) {
2722 trans
->sackfreq
= params
.sack_freq
;
2723 trans
->param_flags
=
2724 sctp_spp_sackdelay_enable(trans
->param_flags
);
2732 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2734 * Applications can specify protocol parameters for the default association
2735 * initialization. The option name argument to setsockopt() and getsockopt()
2738 * Setting initialization parameters is effective only on an unconnected
2739 * socket (for UDP-style sockets only future associations are effected
2740 * by the change). With TCP-style sockets, this option is inherited by
2741 * sockets derived from a listener socket.
2743 static int sctp_setsockopt_initmsg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2745 struct sctp_initmsg sinit
;
2746 struct sctp_sock
*sp
= sctp_sk(sk
);
2748 if (optlen
!= sizeof(struct sctp_initmsg
))
2750 if (copy_from_user(&sinit
, optval
, optlen
))
2753 if (sinit
.sinit_num_ostreams
)
2754 sp
->initmsg
.sinit_num_ostreams
= sinit
.sinit_num_ostreams
;
2755 if (sinit
.sinit_max_instreams
)
2756 sp
->initmsg
.sinit_max_instreams
= sinit
.sinit_max_instreams
;
2757 if (sinit
.sinit_max_attempts
)
2758 sp
->initmsg
.sinit_max_attempts
= sinit
.sinit_max_attempts
;
2759 if (sinit
.sinit_max_init_timeo
)
2760 sp
->initmsg
.sinit_max_init_timeo
= sinit
.sinit_max_init_timeo
;
2766 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2768 * Applications that wish to use the sendto() system call may wish to
2769 * specify a default set of parameters that would normally be supplied
2770 * through the inclusion of ancillary data. This socket option allows
2771 * such an application to set the default sctp_sndrcvinfo structure.
2772 * The application that wishes to use this socket option simply passes
2773 * in to this call the sctp_sndrcvinfo structure defined in Section
2774 * 5.2.2) The input parameters accepted by this call include
2775 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2776 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2777 * to this call if the caller is using the UDP model.
2779 static int sctp_setsockopt_default_send_param(struct sock
*sk
,
2780 char __user
*optval
,
2781 unsigned int optlen
)
2783 struct sctp_sock
*sp
= sctp_sk(sk
);
2784 struct sctp_association
*asoc
;
2785 struct sctp_sndrcvinfo info
;
2787 if (optlen
!= sizeof(info
))
2789 if (copy_from_user(&info
, optval
, optlen
))
2791 if (info
.sinfo_flags
&
2792 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2793 SCTP_ABORT
| SCTP_EOF
))
2796 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
2797 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
2800 asoc
->default_stream
= info
.sinfo_stream
;
2801 asoc
->default_flags
= info
.sinfo_flags
;
2802 asoc
->default_ppid
= info
.sinfo_ppid
;
2803 asoc
->default_context
= info
.sinfo_context
;
2804 asoc
->default_timetolive
= info
.sinfo_timetolive
;
2806 sp
->default_stream
= info
.sinfo_stream
;
2807 sp
->default_flags
= info
.sinfo_flags
;
2808 sp
->default_ppid
= info
.sinfo_ppid
;
2809 sp
->default_context
= info
.sinfo_context
;
2810 sp
->default_timetolive
= info
.sinfo_timetolive
;
2816 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2817 * (SCTP_DEFAULT_SNDINFO)
2819 static int sctp_setsockopt_default_sndinfo(struct sock
*sk
,
2820 char __user
*optval
,
2821 unsigned int optlen
)
2823 struct sctp_sock
*sp
= sctp_sk(sk
);
2824 struct sctp_association
*asoc
;
2825 struct sctp_sndinfo info
;
2827 if (optlen
!= sizeof(info
))
2829 if (copy_from_user(&info
, optval
, optlen
))
2831 if (info
.snd_flags
&
2832 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
2833 SCTP_ABORT
| SCTP_EOF
))
2836 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
2837 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
2840 asoc
->default_stream
= info
.snd_sid
;
2841 asoc
->default_flags
= info
.snd_flags
;
2842 asoc
->default_ppid
= info
.snd_ppid
;
2843 asoc
->default_context
= info
.snd_context
;
2845 sp
->default_stream
= info
.snd_sid
;
2846 sp
->default_flags
= info
.snd_flags
;
2847 sp
->default_ppid
= info
.snd_ppid
;
2848 sp
->default_context
= info
.snd_context
;
2854 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2856 * Requests that the local SCTP stack use the enclosed peer address as
2857 * the association primary. The enclosed address must be one of the
2858 * association peer's addresses.
2860 static int sctp_setsockopt_primary_addr(struct sock
*sk
, char __user
*optval
,
2861 unsigned int optlen
)
2863 struct sctp_prim prim
;
2864 struct sctp_transport
*trans
;
2866 if (optlen
!= sizeof(struct sctp_prim
))
2869 if (copy_from_user(&prim
, optval
, sizeof(struct sctp_prim
)))
2872 trans
= sctp_addr_id2transport(sk
, &prim
.ssp_addr
, prim
.ssp_assoc_id
);
2876 sctp_assoc_set_primary(trans
->asoc
, trans
);
2882 * 7.1.5 SCTP_NODELAY
2884 * Turn on/off any Nagle-like algorithm. This means that packets are
2885 * generally sent as soon as possible and no unnecessary delays are
2886 * introduced, at the cost of more packets in the network. Expects an
2887 * integer boolean flag.
2889 static int sctp_setsockopt_nodelay(struct sock
*sk
, char __user
*optval
,
2890 unsigned int optlen
)
2894 if (optlen
< sizeof(int))
2896 if (get_user(val
, (int __user
*)optval
))
2899 sctp_sk(sk
)->nodelay
= (val
== 0) ? 0 : 1;
2905 * 7.1.1 SCTP_RTOINFO
2907 * The protocol parameters used to initialize and bound retransmission
2908 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2909 * and modify these parameters.
2910 * All parameters are time values, in milliseconds. A value of 0, when
2911 * modifying the parameters, indicates that the current value should not
2915 static int sctp_setsockopt_rtoinfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2917 struct sctp_rtoinfo rtoinfo
;
2918 struct sctp_association
*asoc
;
2919 unsigned long rto_min
, rto_max
;
2920 struct sctp_sock
*sp
= sctp_sk(sk
);
2922 if (optlen
!= sizeof (struct sctp_rtoinfo
))
2925 if (copy_from_user(&rtoinfo
, optval
, optlen
))
2928 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
2930 /* Set the values to the specific association */
2931 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
2934 rto_max
= rtoinfo
.srto_max
;
2935 rto_min
= rtoinfo
.srto_min
;
2938 rto_max
= asoc
? msecs_to_jiffies(rto_max
) : rto_max
;
2940 rto_max
= asoc
? asoc
->rto_max
: sp
->rtoinfo
.srto_max
;
2943 rto_min
= asoc
? msecs_to_jiffies(rto_min
) : rto_min
;
2945 rto_min
= asoc
? asoc
->rto_min
: sp
->rtoinfo
.srto_min
;
2947 if (rto_min
> rto_max
)
2951 if (rtoinfo
.srto_initial
!= 0)
2953 msecs_to_jiffies(rtoinfo
.srto_initial
);
2954 asoc
->rto_max
= rto_max
;
2955 asoc
->rto_min
= rto_min
;
2957 /* If there is no association or the association-id = 0
2958 * set the values to the endpoint.
2960 if (rtoinfo
.srto_initial
!= 0)
2961 sp
->rtoinfo
.srto_initial
= rtoinfo
.srto_initial
;
2962 sp
->rtoinfo
.srto_max
= rto_max
;
2963 sp
->rtoinfo
.srto_min
= rto_min
;
2971 * 7.1.2 SCTP_ASSOCINFO
2973 * This option is used to tune the maximum retransmission attempts
2974 * of the association.
2975 * Returns an error if the new association retransmission value is
2976 * greater than the sum of the retransmission value of the peer.
2977 * See [SCTP] for more information.
2980 static int sctp_setsockopt_associnfo(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
2983 struct sctp_assocparams assocparams
;
2984 struct sctp_association
*asoc
;
2986 if (optlen
!= sizeof(struct sctp_assocparams
))
2988 if (copy_from_user(&assocparams
, optval
, optlen
))
2991 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
2993 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
2996 /* Set the values to the specific association */
2998 if (assocparams
.sasoc_asocmaxrxt
!= 0) {
3001 struct sctp_transport
*peer_addr
;
3003 list_for_each_entry(peer_addr
, &asoc
->peer
.transport_addr_list
,
3005 path_sum
+= peer_addr
->pathmaxrxt
;
3009 /* Only validate asocmaxrxt if we have more than
3010 * one path/transport. We do this because path
3011 * retransmissions are only counted when we have more
3015 assocparams
.sasoc_asocmaxrxt
> path_sum
)
3018 asoc
->max_retrans
= assocparams
.sasoc_asocmaxrxt
;
3021 if (assocparams
.sasoc_cookie_life
!= 0)
3022 asoc
->cookie_life
= ms_to_ktime(assocparams
.sasoc_cookie_life
);
3024 /* Set the values to the endpoint */
3025 struct sctp_sock
*sp
= sctp_sk(sk
);
3027 if (assocparams
.sasoc_asocmaxrxt
!= 0)
3028 sp
->assocparams
.sasoc_asocmaxrxt
=
3029 assocparams
.sasoc_asocmaxrxt
;
3030 if (assocparams
.sasoc_cookie_life
!= 0)
3031 sp
->assocparams
.sasoc_cookie_life
=
3032 assocparams
.sasoc_cookie_life
;
3038 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3040 * This socket option is a boolean flag which turns on or off mapped V4
3041 * addresses. If this option is turned on and the socket is type
3042 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3043 * If this option is turned off, then no mapping will be done of V4
3044 * addresses and a user will receive both PF_INET6 and PF_INET type
3045 * addresses on the socket.
3047 static int sctp_setsockopt_mappedv4(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3050 struct sctp_sock
*sp
= sctp_sk(sk
);
3052 if (optlen
< sizeof(int))
3054 if (get_user(val
, (int __user
*)optval
))
3065 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3066 * This option will get or set the maximum size to put in any outgoing
3067 * SCTP DATA chunk. If a message is larger than this size it will be
3068 * fragmented by SCTP into the specified size. Note that the underlying
3069 * SCTP implementation may fragment into smaller sized chunks when the
3070 * PMTU of the underlying association is smaller than the value set by
3071 * the user. The default value for this option is '0' which indicates
3072 * the user is NOT limiting fragmentation and only the PMTU will effect
3073 * SCTP's choice of DATA chunk size. Note also that values set larger
3074 * than the maximum size of an IP datagram will effectively let SCTP
3075 * control fragmentation (i.e. the same as setting this option to 0).
3077 * The following structure is used to access and modify this parameter:
3079 * struct sctp_assoc_value {
3080 * sctp_assoc_t assoc_id;
3081 * uint32_t assoc_value;
3084 * assoc_id: This parameter is ignored for one-to-one style sockets.
3085 * For one-to-many style sockets this parameter indicates which
3086 * association the user is performing an action upon. Note that if
3087 * this field's value is zero then the endpoints default value is
3088 * changed (effecting future associations only).
3089 * assoc_value: This parameter specifies the maximum size in bytes.
3091 static int sctp_setsockopt_maxseg(struct sock
*sk
, char __user
*optval
, unsigned int optlen
)
3093 struct sctp_assoc_value params
;
3094 struct sctp_association
*asoc
;
3095 struct sctp_sock
*sp
= sctp_sk(sk
);
3098 if (optlen
== sizeof(int)) {
3099 pr_warn_ratelimited(DEPRECATED
3101 "Use of int in maxseg socket option.\n"
3102 "Use struct sctp_assoc_value instead\n",
3103 current
->comm
, task_pid_nr(current
));
3104 if (copy_from_user(&val
, optval
, optlen
))
3106 params
.assoc_id
= 0;
3107 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3108 if (copy_from_user(¶ms
, optval
, optlen
))
3110 val
= params
.assoc_value
;
3114 if ((val
!= 0) && ((val
< 8) || (val
> SCTP_MAX_CHUNK_LEN
)))
3117 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3118 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
3123 val
= asoc
->pathmtu
;
3124 val
-= sp
->pf
->af
->net_header_len
;
3125 val
-= sizeof(struct sctphdr
) +
3126 sizeof(struct sctp_data_chunk
);
3128 asoc
->user_frag
= val
;
3129 asoc
->frag_point
= sctp_frag_point(asoc
, asoc
->pathmtu
);
3131 sp
->user_frag
= val
;
3139 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3141 * Requests that the peer mark the enclosed address as the association
3142 * primary. The enclosed address must be one of the association's
3143 * locally bound addresses. The following structure is used to make a
3144 * set primary request:
3146 static int sctp_setsockopt_peer_primary_addr(struct sock
*sk
, char __user
*optval
,
3147 unsigned int optlen
)
3149 struct net
*net
= sock_net(sk
);
3150 struct sctp_sock
*sp
;
3151 struct sctp_association
*asoc
= NULL
;
3152 struct sctp_setpeerprim prim
;
3153 struct sctp_chunk
*chunk
;
3159 if (!net
->sctp
.addip_enable
)
3162 if (optlen
!= sizeof(struct sctp_setpeerprim
))
3165 if (copy_from_user(&prim
, optval
, optlen
))
3168 asoc
= sctp_id2assoc(sk
, prim
.sspp_assoc_id
);
3172 if (!asoc
->peer
.asconf_capable
)
3175 if (asoc
->peer
.addip_disabled_mask
& SCTP_PARAM_SET_PRIMARY
)
3178 if (!sctp_state(asoc
, ESTABLISHED
))
3181 af
= sctp_get_af_specific(prim
.sspp_addr
.ss_family
);
3185 if (!af
->addr_valid((union sctp_addr
*)&prim
.sspp_addr
, sp
, NULL
))
3186 return -EADDRNOTAVAIL
;
3188 if (!sctp_assoc_lookup_laddr(asoc
, (union sctp_addr
*)&prim
.sspp_addr
))
3189 return -EADDRNOTAVAIL
;
3191 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3192 chunk
= sctp_make_asconf_set_prim(asoc
,
3193 (union sctp_addr
*)&prim
.sspp_addr
);
3197 err
= sctp_send_asconf(asoc
, chunk
);
3199 pr_debug("%s: we set peer primary addr primitively\n", __func__
);
3204 static int sctp_setsockopt_adaptation_layer(struct sock
*sk
, char __user
*optval
,
3205 unsigned int optlen
)
3207 struct sctp_setadaptation adaptation
;
3209 if (optlen
!= sizeof(struct sctp_setadaptation
))
3211 if (copy_from_user(&adaptation
, optval
, optlen
))
3214 sctp_sk(sk
)->adaptation_ind
= adaptation
.ssb_adaptation_ind
;
3220 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3222 * The context field in the sctp_sndrcvinfo structure is normally only
3223 * used when a failed message is retrieved holding the value that was
3224 * sent down on the actual send call. This option allows the setting of
3225 * a default context on an association basis that will be received on
3226 * reading messages from the peer. This is especially helpful in the
3227 * one-2-many model for an application to keep some reference to an
3228 * internal state machine that is processing messages on the
3229 * association. Note that the setting of this value only effects
3230 * received messages from the peer and does not effect the value that is
3231 * saved with outbound messages.
3233 static int sctp_setsockopt_context(struct sock
*sk
, char __user
*optval
,
3234 unsigned int optlen
)
3236 struct sctp_assoc_value params
;
3237 struct sctp_sock
*sp
;
3238 struct sctp_association
*asoc
;
3240 if (optlen
!= sizeof(struct sctp_assoc_value
))
3242 if (copy_from_user(¶ms
, optval
, optlen
))
3247 if (params
.assoc_id
!= 0) {
3248 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3251 asoc
->default_rcv_context
= params
.assoc_value
;
3253 sp
->default_rcv_context
= params
.assoc_value
;
3260 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3262 * This options will at a minimum specify if the implementation is doing
3263 * fragmented interleave. Fragmented interleave, for a one to many
3264 * socket, is when subsequent calls to receive a message may return
3265 * parts of messages from different associations. Some implementations
3266 * may allow you to turn this value on or off. If so, when turned off,
3267 * no fragment interleave will occur (which will cause a head of line
3268 * blocking amongst multiple associations sharing the same one to many
3269 * socket). When this option is turned on, then each receive call may
3270 * come from a different association (thus the user must receive data
3271 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3272 * association each receive belongs to.
3274 * This option takes a boolean value. A non-zero value indicates that
3275 * fragmented interleave is on. A value of zero indicates that
3276 * fragmented interleave is off.
3278 * Note that it is important that an implementation that allows this
3279 * option to be turned on, have it off by default. Otherwise an unaware
3280 * application using the one to many model may become confused and act
3283 static int sctp_setsockopt_fragment_interleave(struct sock
*sk
,
3284 char __user
*optval
,
3285 unsigned int optlen
)
3289 if (optlen
!= sizeof(int))
3291 if (get_user(val
, (int __user
*)optval
))
3294 sctp_sk(sk
)->frag_interleave
= (val
== 0) ? 0 : 1;
3300 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3301 * (SCTP_PARTIAL_DELIVERY_POINT)
3303 * This option will set or get the SCTP partial delivery point. This
3304 * point is the size of a message where the partial delivery API will be
3305 * invoked to help free up rwnd space for the peer. Setting this to a
3306 * lower value will cause partial deliveries to happen more often. The
3307 * calls argument is an integer that sets or gets the partial delivery
3308 * point. Note also that the call will fail if the user attempts to set
3309 * this value larger than the socket receive buffer size.
3311 * Note that any single message having a length smaller than or equal to
3312 * the SCTP partial delivery point will be delivered in one single read
3313 * call as long as the user provided buffer is large enough to hold the
3316 static int sctp_setsockopt_partial_delivery_point(struct sock
*sk
,
3317 char __user
*optval
,
3318 unsigned int optlen
)
3322 if (optlen
!= sizeof(u32
))
3324 if (get_user(val
, (int __user
*)optval
))
3327 /* Note: We double the receive buffer from what the user sets
3328 * it to be, also initial rwnd is based on rcvbuf/2.
3330 if (val
> (sk
->sk_rcvbuf
>> 1))
3333 sctp_sk(sk
)->pd_point
= val
;
3335 return 0; /* is this the right error code? */
3339 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3341 * This option will allow a user to change the maximum burst of packets
3342 * that can be emitted by this association. Note that the default value
3343 * is 4, and some implementations may restrict this setting so that it
3344 * can only be lowered.
3346 * NOTE: This text doesn't seem right. Do this on a socket basis with
3347 * future associations inheriting the socket value.
3349 static int sctp_setsockopt_maxburst(struct sock
*sk
,
3350 char __user
*optval
,
3351 unsigned int optlen
)
3353 struct sctp_assoc_value params
;
3354 struct sctp_sock
*sp
;
3355 struct sctp_association
*asoc
;
3359 if (optlen
== sizeof(int)) {
3360 pr_warn_ratelimited(DEPRECATED
3362 "Use of int in max_burst socket option deprecated.\n"
3363 "Use struct sctp_assoc_value instead\n",
3364 current
->comm
, task_pid_nr(current
));
3365 if (copy_from_user(&val
, optval
, optlen
))
3367 } else if (optlen
== sizeof(struct sctp_assoc_value
)) {
3368 if (copy_from_user(¶ms
, optval
, optlen
))
3370 val
= params
.assoc_value
;
3371 assoc_id
= params
.assoc_id
;
3377 if (assoc_id
!= 0) {
3378 asoc
= sctp_id2assoc(sk
, assoc_id
);
3381 asoc
->max_burst
= val
;
3383 sp
->max_burst
= val
;
3389 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3391 * This set option adds a chunk type that the user is requesting to be
3392 * received only in an authenticated way. Changes to the list of chunks
3393 * will only effect future associations on the socket.
3395 static int sctp_setsockopt_auth_chunk(struct sock
*sk
,
3396 char __user
*optval
,
3397 unsigned int optlen
)
3399 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3400 struct sctp_authchunk val
;
3402 if (!ep
->auth_enable
)
3405 if (optlen
!= sizeof(struct sctp_authchunk
))
3407 if (copy_from_user(&val
, optval
, optlen
))
3410 switch (val
.sauth_chunk
) {
3412 case SCTP_CID_INIT_ACK
:
3413 case SCTP_CID_SHUTDOWN_COMPLETE
:
3418 /* add this chunk id to the endpoint */
3419 return sctp_auth_ep_add_chunkid(ep
, val
.sauth_chunk
);
3423 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3425 * This option gets or sets the list of HMAC algorithms that the local
3426 * endpoint requires the peer to use.
3428 static int sctp_setsockopt_hmac_ident(struct sock
*sk
,
3429 char __user
*optval
,
3430 unsigned int optlen
)
3432 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3433 struct sctp_hmacalgo
*hmacs
;
3437 if (!ep
->auth_enable
)
3440 if (optlen
< sizeof(struct sctp_hmacalgo
))
3443 hmacs
= memdup_user(optval
, optlen
);
3445 return PTR_ERR(hmacs
);
3447 idents
= hmacs
->shmac_num_idents
;
3448 if (idents
== 0 || idents
> SCTP_AUTH_NUM_HMACS
||
3449 (idents
* sizeof(u16
)) > (optlen
- sizeof(struct sctp_hmacalgo
))) {
3454 err
= sctp_auth_ep_set_hmacs(ep
, hmacs
);
3461 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3463 * This option will set a shared secret key which is used to build an
3464 * association shared key.
3466 static int sctp_setsockopt_auth_key(struct sock
*sk
,
3467 char __user
*optval
,
3468 unsigned int optlen
)
3470 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3471 struct sctp_authkey
*authkey
;
3472 struct sctp_association
*asoc
;
3475 if (!ep
->auth_enable
)
3478 if (optlen
<= sizeof(struct sctp_authkey
))
3481 authkey
= memdup_user(optval
, optlen
);
3482 if (IS_ERR(authkey
))
3483 return PTR_ERR(authkey
);
3485 if (authkey
->sca_keylength
> optlen
- sizeof(struct sctp_authkey
)) {
3490 asoc
= sctp_id2assoc(sk
, authkey
->sca_assoc_id
);
3491 if (!asoc
&& authkey
->sca_assoc_id
&& sctp_style(sk
, UDP
)) {
3496 ret
= sctp_auth_set_key(ep
, asoc
, authkey
);
3503 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3505 * This option will get or set the active shared key to be used to build
3506 * the association shared key.
3508 static int sctp_setsockopt_active_key(struct sock
*sk
,
3509 char __user
*optval
,
3510 unsigned int optlen
)
3512 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3513 struct sctp_authkeyid val
;
3514 struct sctp_association
*asoc
;
3516 if (!ep
->auth_enable
)
3519 if (optlen
!= sizeof(struct sctp_authkeyid
))
3521 if (copy_from_user(&val
, optval
, optlen
))
3524 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3525 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3528 return sctp_auth_set_active_key(ep
, asoc
, val
.scact_keynumber
);
3532 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3534 * This set option will delete a shared secret key from use.
3536 static int sctp_setsockopt_del_key(struct sock
*sk
,
3537 char __user
*optval
,
3538 unsigned int optlen
)
3540 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
3541 struct sctp_authkeyid val
;
3542 struct sctp_association
*asoc
;
3544 if (!ep
->auth_enable
)
3547 if (optlen
!= sizeof(struct sctp_authkeyid
))
3549 if (copy_from_user(&val
, optval
, optlen
))
3552 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
3553 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
3556 return sctp_auth_del_key_id(ep
, asoc
, val
.scact_keynumber
);
3561 * 8.1.23 SCTP_AUTO_ASCONF
3563 * This option will enable or disable the use of the automatic generation of
3564 * ASCONF chunks to add and delete addresses to an existing association. Note
3565 * that this option has two caveats namely: a) it only affects sockets that
3566 * are bound to all addresses available to the SCTP stack, and b) the system
3567 * administrator may have an overriding control that turns the ASCONF feature
3568 * off no matter what setting the socket option may have.
3569 * This option expects an integer boolean flag, where a non-zero value turns on
3570 * the option, and a zero value turns off the option.
3571 * Note. In this implementation, socket operation overrides default parameter
3572 * being set by sysctl as well as FreeBSD implementation
3574 static int sctp_setsockopt_auto_asconf(struct sock
*sk
, char __user
*optval
,
3575 unsigned int optlen
)
3578 struct sctp_sock
*sp
= sctp_sk(sk
);
3580 if (optlen
< sizeof(int))
3582 if (get_user(val
, (int __user
*)optval
))
3584 if (!sctp_is_ep_boundall(sk
) && val
)
3586 if ((val
&& sp
->do_auto_asconf
) || (!val
&& !sp
->do_auto_asconf
))
3589 spin_lock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3590 if (val
== 0 && sp
->do_auto_asconf
) {
3591 list_del(&sp
->auto_asconf_list
);
3592 sp
->do_auto_asconf
= 0;
3593 } else if (val
&& !sp
->do_auto_asconf
) {
3594 list_add_tail(&sp
->auto_asconf_list
,
3595 &sock_net(sk
)->sctp
.auto_asconf_splist
);
3596 sp
->do_auto_asconf
= 1;
3598 spin_unlock_bh(&sock_net(sk
)->sctp
.addr_wq_lock
);
3603 * SCTP_PEER_ADDR_THLDS
3605 * This option allows us to alter the partially failed threshold for one or all
3606 * transports in an association. See Section 6.1 of:
3607 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3609 static int sctp_setsockopt_paddr_thresholds(struct sock
*sk
,
3610 char __user
*optval
,
3611 unsigned int optlen
)
3613 struct sctp_paddrthlds val
;
3614 struct sctp_transport
*trans
;
3615 struct sctp_association
*asoc
;
3617 if (optlen
< sizeof(struct sctp_paddrthlds
))
3619 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
,
3620 sizeof(struct sctp_paddrthlds
)))
3624 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
3625 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
3628 list_for_each_entry(trans
, &asoc
->peer
.transport_addr_list
,
3630 if (val
.spt_pathmaxrxt
)
3631 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3632 trans
->pf_retrans
= val
.spt_pathpfthld
;
3635 if (val
.spt_pathmaxrxt
)
3636 asoc
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3637 asoc
->pf_retrans
= val
.spt_pathpfthld
;
3639 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
3644 if (val
.spt_pathmaxrxt
)
3645 trans
->pathmaxrxt
= val
.spt_pathmaxrxt
;
3646 trans
->pf_retrans
= val
.spt_pathpfthld
;
3652 static int sctp_setsockopt_recvrcvinfo(struct sock
*sk
,
3653 char __user
*optval
,
3654 unsigned int optlen
)
3658 if (optlen
< sizeof(int))
3660 if (get_user(val
, (int __user
*) optval
))
3663 sctp_sk(sk
)->recvrcvinfo
= (val
== 0) ? 0 : 1;
3668 static int sctp_setsockopt_recvnxtinfo(struct sock
*sk
,
3669 char __user
*optval
,
3670 unsigned int optlen
)
3674 if (optlen
< sizeof(int))
3676 if (get_user(val
, (int __user
*) optval
))
3679 sctp_sk(sk
)->recvnxtinfo
= (val
== 0) ? 0 : 1;
3684 static int sctp_setsockopt_pr_supported(struct sock
*sk
,
3685 char __user
*optval
,
3686 unsigned int optlen
)
3688 struct sctp_assoc_value params
;
3689 struct sctp_association
*asoc
;
3690 int retval
= -EINVAL
;
3692 if (optlen
!= sizeof(params
))
3695 if (copy_from_user(¶ms
, optval
, optlen
)) {
3700 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
3702 asoc
->prsctp_enable
= !!params
.assoc_value
;
3703 } else if (!params
.assoc_id
) {
3704 struct sctp_sock
*sp
= sctp_sk(sk
);
3706 sp
->ep
->prsctp_enable
= !!params
.assoc_value
;
3717 static int sctp_setsockopt_default_prinfo(struct sock
*sk
,
3718 char __user
*optval
,
3719 unsigned int optlen
)
3721 struct sctp_default_prinfo info
;
3722 struct sctp_association
*asoc
;
3723 int retval
= -EINVAL
;
3725 if (optlen
!= sizeof(info
))
3728 if (copy_from_user(&info
, optval
, sizeof(info
))) {
3733 if (info
.pr_policy
& ~SCTP_PR_SCTP_MASK
)
3736 if (info
.pr_policy
== SCTP_PR_SCTP_NONE
)
3739 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
3741 SCTP_PR_SET_POLICY(asoc
->default_flags
, info
.pr_policy
);
3742 asoc
->default_timetolive
= info
.pr_value
;
3743 } else if (!info
.pr_assoc_id
) {
3744 struct sctp_sock
*sp
= sctp_sk(sk
);
3746 SCTP_PR_SET_POLICY(sp
->default_flags
, info
.pr_policy
);
3747 sp
->default_timetolive
= info
.pr_value
;
3758 /* API 6.2 setsockopt(), getsockopt()
3760 * Applications use setsockopt() and getsockopt() to set or retrieve
3761 * socket options. Socket options are used to change the default
3762 * behavior of sockets calls. They are described in Section 7.
3766 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3767 * int __user *optlen);
3768 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3771 * sd - the socket descript.
3772 * level - set to IPPROTO_SCTP for all SCTP options.
3773 * optname - the option name.
3774 * optval - the buffer to store the value of the option.
3775 * optlen - the size of the buffer.
3777 static int sctp_setsockopt(struct sock
*sk
, int level
, int optname
,
3778 char __user
*optval
, unsigned int optlen
)
3782 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
3784 /* I can hardly begin to describe how wrong this is. This is
3785 * so broken as to be worse than useless. The API draft
3786 * REALLY is NOT helpful here... I am not convinced that the
3787 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3788 * are at all well-founded.
3790 if (level
!= SOL_SCTP
) {
3791 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
3792 retval
= af
->setsockopt(sk
, level
, optname
, optval
, optlen
);
3799 case SCTP_SOCKOPT_BINDX_ADD
:
3800 /* 'optlen' is the size of the addresses buffer. */
3801 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3802 optlen
, SCTP_BINDX_ADD_ADDR
);
3805 case SCTP_SOCKOPT_BINDX_REM
:
3806 /* 'optlen' is the size of the addresses buffer. */
3807 retval
= sctp_setsockopt_bindx(sk
, (struct sockaddr __user
*)optval
,
3808 optlen
, SCTP_BINDX_REM_ADDR
);
3811 case SCTP_SOCKOPT_CONNECTX_OLD
:
3812 /* 'optlen' is the size of the addresses buffer. */
3813 retval
= sctp_setsockopt_connectx_old(sk
,
3814 (struct sockaddr __user
*)optval
,
3818 case SCTP_SOCKOPT_CONNECTX
:
3819 /* 'optlen' is the size of the addresses buffer. */
3820 retval
= sctp_setsockopt_connectx(sk
,
3821 (struct sockaddr __user
*)optval
,
3825 case SCTP_DISABLE_FRAGMENTS
:
3826 retval
= sctp_setsockopt_disable_fragments(sk
, optval
, optlen
);
3830 retval
= sctp_setsockopt_events(sk
, optval
, optlen
);
3833 case SCTP_AUTOCLOSE
:
3834 retval
= sctp_setsockopt_autoclose(sk
, optval
, optlen
);
3837 case SCTP_PEER_ADDR_PARAMS
:
3838 retval
= sctp_setsockopt_peer_addr_params(sk
, optval
, optlen
);
3841 case SCTP_DELAYED_SACK
:
3842 retval
= sctp_setsockopt_delayed_ack(sk
, optval
, optlen
);
3844 case SCTP_PARTIAL_DELIVERY_POINT
:
3845 retval
= sctp_setsockopt_partial_delivery_point(sk
, optval
, optlen
);
3849 retval
= sctp_setsockopt_initmsg(sk
, optval
, optlen
);
3851 case SCTP_DEFAULT_SEND_PARAM
:
3852 retval
= sctp_setsockopt_default_send_param(sk
, optval
,
3855 case SCTP_DEFAULT_SNDINFO
:
3856 retval
= sctp_setsockopt_default_sndinfo(sk
, optval
, optlen
);
3858 case SCTP_PRIMARY_ADDR
:
3859 retval
= sctp_setsockopt_primary_addr(sk
, optval
, optlen
);
3861 case SCTP_SET_PEER_PRIMARY_ADDR
:
3862 retval
= sctp_setsockopt_peer_primary_addr(sk
, optval
, optlen
);
3865 retval
= sctp_setsockopt_nodelay(sk
, optval
, optlen
);
3868 retval
= sctp_setsockopt_rtoinfo(sk
, optval
, optlen
);
3870 case SCTP_ASSOCINFO
:
3871 retval
= sctp_setsockopt_associnfo(sk
, optval
, optlen
);
3873 case SCTP_I_WANT_MAPPED_V4_ADDR
:
3874 retval
= sctp_setsockopt_mappedv4(sk
, optval
, optlen
);
3877 retval
= sctp_setsockopt_maxseg(sk
, optval
, optlen
);
3879 case SCTP_ADAPTATION_LAYER
:
3880 retval
= sctp_setsockopt_adaptation_layer(sk
, optval
, optlen
);
3883 retval
= sctp_setsockopt_context(sk
, optval
, optlen
);
3885 case SCTP_FRAGMENT_INTERLEAVE
:
3886 retval
= sctp_setsockopt_fragment_interleave(sk
, optval
, optlen
);
3888 case SCTP_MAX_BURST
:
3889 retval
= sctp_setsockopt_maxburst(sk
, optval
, optlen
);
3891 case SCTP_AUTH_CHUNK
:
3892 retval
= sctp_setsockopt_auth_chunk(sk
, optval
, optlen
);
3894 case SCTP_HMAC_IDENT
:
3895 retval
= sctp_setsockopt_hmac_ident(sk
, optval
, optlen
);
3898 retval
= sctp_setsockopt_auth_key(sk
, optval
, optlen
);
3900 case SCTP_AUTH_ACTIVE_KEY
:
3901 retval
= sctp_setsockopt_active_key(sk
, optval
, optlen
);
3903 case SCTP_AUTH_DELETE_KEY
:
3904 retval
= sctp_setsockopt_del_key(sk
, optval
, optlen
);
3906 case SCTP_AUTO_ASCONF
:
3907 retval
= sctp_setsockopt_auto_asconf(sk
, optval
, optlen
);
3909 case SCTP_PEER_ADDR_THLDS
:
3910 retval
= sctp_setsockopt_paddr_thresholds(sk
, optval
, optlen
);
3912 case SCTP_RECVRCVINFO
:
3913 retval
= sctp_setsockopt_recvrcvinfo(sk
, optval
, optlen
);
3915 case SCTP_RECVNXTINFO
:
3916 retval
= sctp_setsockopt_recvnxtinfo(sk
, optval
, optlen
);
3918 case SCTP_PR_SUPPORTED
:
3919 retval
= sctp_setsockopt_pr_supported(sk
, optval
, optlen
);
3921 case SCTP_DEFAULT_PRINFO
:
3922 retval
= sctp_setsockopt_default_prinfo(sk
, optval
, optlen
);
3925 retval
= -ENOPROTOOPT
;
3935 /* API 3.1.6 connect() - UDP Style Syntax
3937 * An application may use the connect() call in the UDP model to initiate an
3938 * association without sending data.
3942 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3944 * sd: the socket descriptor to have a new association added to.
3946 * nam: the address structure (either struct sockaddr_in or struct
3947 * sockaddr_in6 defined in RFC2553 [7]).
3949 * len: the size of the address.
3951 static int sctp_connect(struct sock
*sk
, struct sockaddr
*addr
,
3959 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__
, sk
,
3962 /* Validate addr_len before calling common connect/connectx routine. */
3963 af
= sctp_get_af_specific(addr
->sa_family
);
3964 if (!af
|| addr_len
< af
->sockaddr_len
) {
3967 /* Pass correct addr len to common routine (so it knows there
3968 * is only one address being passed.
3970 err
= __sctp_connect(sk
, addr
, af
->sockaddr_len
, NULL
);
3977 /* FIXME: Write comments. */
3978 static int sctp_disconnect(struct sock
*sk
, int flags
)
3980 return -EOPNOTSUPP
; /* STUB */
3983 /* 4.1.4 accept() - TCP Style Syntax
3985 * Applications use accept() call to remove an established SCTP
3986 * association from the accept queue of the endpoint. A new socket
3987 * descriptor will be returned from accept() to represent the newly
3988 * formed association.
3990 static struct sock
*sctp_accept(struct sock
*sk
, int flags
, int *err
)
3992 struct sctp_sock
*sp
;
3993 struct sctp_endpoint
*ep
;
3994 struct sock
*newsk
= NULL
;
3995 struct sctp_association
*asoc
;
4004 if (!sctp_style(sk
, TCP
)) {
4005 error
= -EOPNOTSUPP
;
4009 if (!sctp_sstate(sk
, LISTENING
)) {
4014 timeo
= sock_rcvtimeo(sk
, flags
& O_NONBLOCK
);
4016 error
= sctp_wait_for_accept(sk
, timeo
);
4020 /* We treat the list of associations on the endpoint as the accept
4021 * queue and pick the first association on the list.
4023 asoc
= list_entry(ep
->asocs
.next
, struct sctp_association
, asocs
);
4025 newsk
= sp
->pf
->create_accept_sk(sk
, asoc
);
4031 /* Populate the fields of the newsk from the oldsk and migrate the
4032 * asoc to the newsk.
4034 sctp_sock_migrate(sk
, newsk
, asoc
, SCTP_SOCKET_TCP
);
4042 /* The SCTP ioctl handler. */
4043 static int sctp_ioctl(struct sock
*sk
, int cmd
, unsigned long arg
)
4050 * SEQPACKET-style sockets in LISTENING state are valid, for
4051 * SCTP, so only discard TCP-style sockets in LISTENING state.
4053 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
4058 struct sk_buff
*skb
;
4059 unsigned int amount
= 0;
4061 skb
= skb_peek(&sk
->sk_receive_queue
);
4064 * We will only return the amount of this packet since
4065 * that is all that will be read.
4069 rc
= put_user(amount
, (int __user
*)arg
);
4081 /* This is the function which gets called during socket creation to
4082 * initialized the SCTP-specific portion of the sock.
4083 * The sock structure should already be zero-filled memory.
4085 static int sctp_init_sock(struct sock
*sk
)
4087 struct net
*net
= sock_net(sk
);
4088 struct sctp_sock
*sp
;
4090 pr_debug("%s: sk:%p\n", __func__
, sk
);
4094 /* Initialize the SCTP per socket area. */
4095 switch (sk
->sk_type
) {
4096 case SOCK_SEQPACKET
:
4097 sp
->type
= SCTP_SOCKET_UDP
;
4100 sp
->type
= SCTP_SOCKET_TCP
;
4103 return -ESOCKTNOSUPPORT
;
4106 sk
->sk_gso_type
= SKB_GSO_SCTP
;
4108 /* Initialize default send parameters. These parameters can be
4109 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4111 sp
->default_stream
= 0;
4112 sp
->default_ppid
= 0;
4113 sp
->default_flags
= 0;
4114 sp
->default_context
= 0;
4115 sp
->default_timetolive
= 0;
4117 sp
->default_rcv_context
= 0;
4118 sp
->max_burst
= net
->sctp
.max_burst
;
4120 sp
->sctp_hmac_alg
= net
->sctp
.sctp_hmac_alg
;
4122 /* Initialize default setup parameters. These parameters
4123 * can be modified with the SCTP_INITMSG socket option or
4124 * overridden by the SCTP_INIT CMSG.
4126 sp
->initmsg
.sinit_num_ostreams
= sctp_max_outstreams
;
4127 sp
->initmsg
.sinit_max_instreams
= sctp_max_instreams
;
4128 sp
->initmsg
.sinit_max_attempts
= net
->sctp
.max_retrans_init
;
4129 sp
->initmsg
.sinit_max_init_timeo
= net
->sctp
.rto_max
;
4131 /* Initialize default RTO related parameters. These parameters can
4132 * be modified for with the SCTP_RTOINFO socket option.
4134 sp
->rtoinfo
.srto_initial
= net
->sctp
.rto_initial
;
4135 sp
->rtoinfo
.srto_max
= net
->sctp
.rto_max
;
4136 sp
->rtoinfo
.srto_min
= net
->sctp
.rto_min
;
4138 /* Initialize default association related parameters. These parameters
4139 * can be modified with the SCTP_ASSOCINFO socket option.
4141 sp
->assocparams
.sasoc_asocmaxrxt
= net
->sctp
.max_retrans_association
;
4142 sp
->assocparams
.sasoc_number_peer_destinations
= 0;
4143 sp
->assocparams
.sasoc_peer_rwnd
= 0;
4144 sp
->assocparams
.sasoc_local_rwnd
= 0;
4145 sp
->assocparams
.sasoc_cookie_life
= net
->sctp
.valid_cookie_life
;
4147 /* Initialize default event subscriptions. By default, all the
4150 memset(&sp
->subscribe
, 0, sizeof(struct sctp_event_subscribe
));
4152 /* Default Peer Address Parameters. These defaults can
4153 * be modified via SCTP_PEER_ADDR_PARAMS
4155 sp
->hbinterval
= net
->sctp
.hb_interval
;
4156 sp
->pathmaxrxt
= net
->sctp
.max_retrans_path
;
4157 sp
->pathmtu
= 0; /* allow default discovery */
4158 sp
->sackdelay
= net
->sctp
.sack_timeout
;
4160 sp
->param_flags
= SPP_HB_ENABLE
|
4162 SPP_SACKDELAY_ENABLE
;
4164 /* If enabled no SCTP message fragmentation will be performed.
4165 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4167 sp
->disable_fragments
= 0;
4169 /* Enable Nagle algorithm by default. */
4172 sp
->recvrcvinfo
= 0;
4173 sp
->recvnxtinfo
= 0;
4175 /* Enable by default. */
4178 /* Auto-close idle associations after the configured
4179 * number of seconds. A value of 0 disables this
4180 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4181 * for UDP-style sockets only.
4185 /* User specified fragmentation limit. */
4188 sp
->adaptation_ind
= 0;
4190 sp
->pf
= sctp_get_pf_specific(sk
->sk_family
);
4192 /* Control variables for partial data delivery. */
4193 atomic_set(&sp
->pd_mode
, 0);
4194 skb_queue_head_init(&sp
->pd_lobby
);
4195 sp
->frag_interleave
= 0;
4197 /* Create a per socket endpoint structure. Even if we
4198 * change the data structure relationships, this may still
4199 * be useful for storing pre-connect address information.
4201 sp
->ep
= sctp_endpoint_new(sk
, GFP_KERNEL
);
4207 sk
->sk_destruct
= sctp_destruct_sock
;
4209 SCTP_DBG_OBJCNT_INC(sock
);
4212 percpu_counter_inc(&sctp_sockets_allocated
);
4213 sock_prot_inuse_add(net
, sk
->sk_prot
, 1);
4215 /* Nothing can fail after this block, otherwise
4216 * sctp_destroy_sock() will be called without addr_wq_lock held
4218 if (net
->sctp
.default_auto_asconf
) {
4219 spin_lock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4220 list_add_tail(&sp
->auto_asconf_list
,
4221 &net
->sctp
.auto_asconf_splist
);
4222 sp
->do_auto_asconf
= 1;
4223 spin_unlock(&sock_net(sk
)->sctp
.addr_wq_lock
);
4225 sp
->do_auto_asconf
= 0;
4233 /* Cleanup any SCTP per socket resources. Must be called with
4234 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4236 static void sctp_destroy_sock(struct sock
*sk
)
4238 struct sctp_sock
*sp
;
4240 pr_debug("%s: sk:%p\n", __func__
, sk
);
4242 /* Release our hold on the endpoint. */
4244 /* This could happen during socket init, thus we bail out
4245 * early, since the rest of the below is not setup either.
4250 if (sp
->do_auto_asconf
) {
4251 sp
->do_auto_asconf
= 0;
4252 list_del(&sp
->auto_asconf_list
);
4254 sctp_endpoint_free(sp
->ep
);
4256 percpu_counter_dec(&sctp_sockets_allocated
);
4257 sock_prot_inuse_add(sock_net(sk
), sk
->sk_prot
, -1);
4261 /* Triggered when there are no references on the socket anymore */
4262 static void sctp_destruct_sock(struct sock
*sk
)
4264 struct sctp_sock
*sp
= sctp_sk(sk
);
4266 /* Free up the HMAC transform. */
4267 crypto_free_shash(sp
->hmac
);
4269 inet_sock_destruct(sk
);
4272 /* API 4.1.7 shutdown() - TCP Style Syntax
4273 * int shutdown(int socket, int how);
4275 * sd - the socket descriptor of the association to be closed.
4276 * how - Specifies the type of shutdown. The values are
4279 * Disables further receive operations. No SCTP
4280 * protocol action is taken.
4282 * Disables further send operations, and initiates
4283 * the SCTP shutdown sequence.
4285 * Disables further send and receive operations
4286 * and initiates the SCTP shutdown sequence.
4288 static void sctp_shutdown(struct sock
*sk
, int how
)
4290 struct net
*net
= sock_net(sk
);
4291 struct sctp_endpoint
*ep
;
4293 if (!sctp_style(sk
, TCP
))
4296 ep
= sctp_sk(sk
)->ep
;
4297 if (how
& SEND_SHUTDOWN
&& !list_empty(&ep
->asocs
)) {
4298 struct sctp_association
*asoc
;
4300 sk
->sk_state
= SCTP_SS_CLOSING
;
4301 asoc
= list_entry(ep
->asocs
.next
,
4302 struct sctp_association
, asocs
);
4303 sctp_primitive_SHUTDOWN(net
, asoc
, NULL
);
4307 int sctp_get_sctp_info(struct sock
*sk
, struct sctp_association
*asoc
,
4308 struct sctp_info
*info
)
4310 struct sctp_transport
*prim
;
4311 struct list_head
*pos
;
4314 memset(info
, 0, sizeof(*info
));
4316 struct sctp_sock
*sp
= sctp_sk(sk
);
4318 info
->sctpi_s_autoclose
= sp
->autoclose
;
4319 info
->sctpi_s_adaptation_ind
= sp
->adaptation_ind
;
4320 info
->sctpi_s_pd_point
= sp
->pd_point
;
4321 info
->sctpi_s_nodelay
= sp
->nodelay
;
4322 info
->sctpi_s_disable_fragments
= sp
->disable_fragments
;
4323 info
->sctpi_s_v4mapped
= sp
->v4mapped
;
4324 info
->sctpi_s_frag_interleave
= sp
->frag_interleave
;
4325 info
->sctpi_s_type
= sp
->type
;
4330 info
->sctpi_tag
= asoc
->c
.my_vtag
;
4331 info
->sctpi_state
= asoc
->state
;
4332 info
->sctpi_rwnd
= asoc
->a_rwnd
;
4333 info
->sctpi_unackdata
= asoc
->unack_data
;
4334 info
->sctpi_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4335 info
->sctpi_instrms
= asoc
->c
.sinit_max_instreams
;
4336 info
->sctpi_outstrms
= asoc
->c
.sinit_num_ostreams
;
4337 list_for_each(pos
, &asoc
->base
.inqueue
.in_chunk_list
)
4338 info
->sctpi_inqueue
++;
4339 list_for_each(pos
, &asoc
->outqueue
.out_chunk_list
)
4340 info
->sctpi_outqueue
++;
4341 info
->sctpi_overall_error
= asoc
->overall_error_count
;
4342 info
->sctpi_max_burst
= asoc
->max_burst
;
4343 info
->sctpi_maxseg
= asoc
->frag_point
;
4344 info
->sctpi_peer_rwnd
= asoc
->peer
.rwnd
;
4345 info
->sctpi_peer_tag
= asoc
->c
.peer_vtag
;
4347 mask
= asoc
->peer
.ecn_capable
<< 1;
4348 mask
= (mask
| asoc
->peer
.ipv4_address
) << 1;
4349 mask
= (mask
| asoc
->peer
.ipv6_address
) << 1;
4350 mask
= (mask
| asoc
->peer
.hostname_address
) << 1;
4351 mask
= (mask
| asoc
->peer
.asconf_capable
) << 1;
4352 mask
= (mask
| asoc
->peer
.prsctp_capable
) << 1;
4353 mask
= (mask
| asoc
->peer
.auth_capable
);
4354 info
->sctpi_peer_capable
= mask
;
4355 mask
= asoc
->peer
.sack_needed
<< 1;
4356 mask
= (mask
| asoc
->peer
.sack_generation
) << 1;
4357 mask
= (mask
| asoc
->peer
.zero_window_announced
);
4358 info
->sctpi_peer_sack
= mask
;
4360 info
->sctpi_isacks
= asoc
->stats
.isacks
;
4361 info
->sctpi_osacks
= asoc
->stats
.osacks
;
4362 info
->sctpi_opackets
= asoc
->stats
.opackets
;
4363 info
->sctpi_ipackets
= asoc
->stats
.ipackets
;
4364 info
->sctpi_rtxchunks
= asoc
->stats
.rtxchunks
;
4365 info
->sctpi_outofseqtsns
= asoc
->stats
.outofseqtsns
;
4366 info
->sctpi_idupchunks
= asoc
->stats
.idupchunks
;
4367 info
->sctpi_gapcnt
= asoc
->stats
.gapcnt
;
4368 info
->sctpi_ouodchunks
= asoc
->stats
.ouodchunks
;
4369 info
->sctpi_iuodchunks
= asoc
->stats
.iuodchunks
;
4370 info
->sctpi_oodchunks
= asoc
->stats
.oodchunks
;
4371 info
->sctpi_iodchunks
= asoc
->stats
.iodchunks
;
4372 info
->sctpi_octrlchunks
= asoc
->stats
.octrlchunks
;
4373 info
->sctpi_ictrlchunks
= asoc
->stats
.ictrlchunks
;
4375 prim
= asoc
->peer
.primary_path
;
4376 memcpy(&info
->sctpi_p_address
, &prim
->ipaddr
,
4377 sizeof(struct sockaddr_storage
));
4378 info
->sctpi_p_state
= prim
->state
;
4379 info
->sctpi_p_cwnd
= prim
->cwnd
;
4380 info
->sctpi_p_srtt
= prim
->srtt
;
4381 info
->sctpi_p_rto
= jiffies_to_msecs(prim
->rto
);
4382 info
->sctpi_p_hbinterval
= prim
->hbinterval
;
4383 info
->sctpi_p_pathmaxrxt
= prim
->pathmaxrxt
;
4384 info
->sctpi_p_sackdelay
= jiffies_to_msecs(prim
->sackdelay
);
4385 info
->sctpi_p_ssthresh
= prim
->ssthresh
;
4386 info
->sctpi_p_partial_bytes_acked
= prim
->partial_bytes_acked
;
4387 info
->sctpi_p_flight_size
= prim
->flight_size
;
4388 info
->sctpi_p_error
= prim
->error_count
;
4392 EXPORT_SYMBOL_GPL(sctp_get_sctp_info
);
4394 /* use callback to avoid exporting the core structure */
4395 int sctp_transport_walk_start(struct rhashtable_iter
*iter
)
4399 rhltable_walk_enter(&sctp_transport_hashtable
, iter
);
4401 err
= rhashtable_walk_start(iter
);
4402 if (err
&& err
!= -EAGAIN
) {
4403 rhashtable_walk_stop(iter
);
4404 rhashtable_walk_exit(iter
);
4411 void sctp_transport_walk_stop(struct rhashtable_iter
*iter
)
4413 rhashtable_walk_stop(iter
);
4414 rhashtable_walk_exit(iter
);
4417 struct sctp_transport
*sctp_transport_get_next(struct net
*net
,
4418 struct rhashtable_iter
*iter
)
4420 struct sctp_transport
*t
;
4422 t
= rhashtable_walk_next(iter
);
4423 for (; t
; t
= rhashtable_walk_next(iter
)) {
4425 if (PTR_ERR(t
) == -EAGAIN
)
4430 if (net_eq(sock_net(t
->asoc
->base
.sk
), net
) &&
4431 t
->asoc
->peer
.primary_path
== t
)
4438 struct sctp_transport
*sctp_transport_get_idx(struct net
*net
,
4439 struct rhashtable_iter
*iter
,
4442 void *obj
= SEQ_START_TOKEN
;
4444 while (pos
&& (obj
= sctp_transport_get_next(net
, iter
)) &&
4451 int sctp_for_each_endpoint(int (*cb
)(struct sctp_endpoint
*, void *),
4455 struct sctp_ep_common
*epb
;
4456 struct sctp_hashbucket
*head
;
4458 for (head
= sctp_ep_hashtable
; hash
< sctp_ep_hashsize
;
4460 read_lock(&head
->lock
);
4461 sctp_for_each_hentry(epb
, &head
->chain
) {
4462 err
= cb(sctp_ep(epb
), p
);
4466 read_unlock(&head
->lock
);
4471 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint
);
4473 int sctp_transport_lookup_process(int (*cb
)(struct sctp_transport
*, void *),
4475 const union sctp_addr
*laddr
,
4476 const union sctp_addr
*paddr
, void *p
)
4478 struct sctp_transport
*transport
;
4482 transport
= sctp_addrs_lookup_transport(net
, laddr
, paddr
);
4487 err
= cb(transport
, p
);
4488 sctp_transport_put(transport
);
4492 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process
);
4494 int sctp_for_each_transport(int (*cb
)(struct sctp_transport
*, void *),
4495 struct net
*net
, int pos
, void *p
) {
4496 struct rhashtable_iter hti
;
4500 err
= sctp_transport_walk_start(&hti
);
4504 sctp_transport_get_idx(net
, &hti
, pos
);
4505 obj
= sctp_transport_get_next(net
, &hti
);
4506 for (; obj
&& !IS_ERR(obj
); obj
= sctp_transport_get_next(net
, &hti
)) {
4507 struct sctp_transport
*transport
= obj
;
4509 if (!sctp_transport_hold(transport
))
4511 err
= cb(transport
, p
);
4512 sctp_transport_put(transport
);
4516 sctp_transport_walk_stop(&hti
);
4520 EXPORT_SYMBOL_GPL(sctp_for_each_transport
);
4522 /* 7.2.1 Association Status (SCTP_STATUS)
4524 * Applications can retrieve current status information about an
4525 * association, including association state, peer receiver window size,
4526 * number of unacked data chunks, and number of data chunks pending
4527 * receipt. This information is read-only.
4529 static int sctp_getsockopt_sctp_status(struct sock
*sk
, int len
,
4530 char __user
*optval
,
4533 struct sctp_status status
;
4534 struct sctp_association
*asoc
= NULL
;
4535 struct sctp_transport
*transport
;
4536 sctp_assoc_t associd
;
4539 if (len
< sizeof(status
)) {
4544 len
= sizeof(status
);
4545 if (copy_from_user(&status
, optval
, len
)) {
4550 associd
= status
.sstat_assoc_id
;
4551 asoc
= sctp_id2assoc(sk
, associd
);
4557 transport
= asoc
->peer
.primary_path
;
4559 status
.sstat_assoc_id
= sctp_assoc2id(asoc
);
4560 status
.sstat_state
= sctp_assoc_to_state(asoc
);
4561 status
.sstat_rwnd
= asoc
->peer
.rwnd
;
4562 status
.sstat_unackdata
= asoc
->unack_data
;
4564 status
.sstat_penddata
= sctp_tsnmap_pending(&asoc
->peer
.tsn_map
);
4565 status
.sstat_instrms
= asoc
->c
.sinit_max_instreams
;
4566 status
.sstat_outstrms
= asoc
->c
.sinit_num_ostreams
;
4567 status
.sstat_fragmentation_point
= asoc
->frag_point
;
4568 status
.sstat_primary
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4569 memcpy(&status
.sstat_primary
.spinfo_address
, &transport
->ipaddr
,
4570 transport
->af_specific
->sockaddr_len
);
4571 /* Map ipv4 address into v4-mapped-on-v6 address. */
4572 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sctp_sk(sk
),
4573 (union sctp_addr
*)&status
.sstat_primary
.spinfo_address
);
4574 status
.sstat_primary
.spinfo_state
= transport
->state
;
4575 status
.sstat_primary
.spinfo_cwnd
= transport
->cwnd
;
4576 status
.sstat_primary
.spinfo_srtt
= transport
->srtt
;
4577 status
.sstat_primary
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4578 status
.sstat_primary
.spinfo_mtu
= transport
->pathmtu
;
4580 if (status
.sstat_primary
.spinfo_state
== SCTP_UNKNOWN
)
4581 status
.sstat_primary
.spinfo_state
= SCTP_ACTIVE
;
4583 if (put_user(len
, optlen
)) {
4588 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4589 __func__
, len
, status
.sstat_state
, status
.sstat_rwnd
,
4590 status
.sstat_assoc_id
);
4592 if (copy_to_user(optval
, &status
, len
)) {
4602 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4604 * Applications can retrieve information about a specific peer address
4605 * of an association, including its reachability state, congestion
4606 * window, and retransmission timer values. This information is
4609 static int sctp_getsockopt_peer_addr_info(struct sock
*sk
, int len
,
4610 char __user
*optval
,
4613 struct sctp_paddrinfo pinfo
;
4614 struct sctp_transport
*transport
;
4617 if (len
< sizeof(pinfo
)) {
4622 len
= sizeof(pinfo
);
4623 if (copy_from_user(&pinfo
, optval
, len
)) {
4628 transport
= sctp_addr_id2transport(sk
, &pinfo
.spinfo_address
,
4629 pinfo
.spinfo_assoc_id
);
4633 pinfo
.spinfo_assoc_id
= sctp_assoc2id(transport
->asoc
);
4634 pinfo
.spinfo_state
= transport
->state
;
4635 pinfo
.spinfo_cwnd
= transport
->cwnd
;
4636 pinfo
.spinfo_srtt
= transport
->srtt
;
4637 pinfo
.spinfo_rto
= jiffies_to_msecs(transport
->rto
);
4638 pinfo
.spinfo_mtu
= transport
->pathmtu
;
4640 if (pinfo
.spinfo_state
== SCTP_UNKNOWN
)
4641 pinfo
.spinfo_state
= SCTP_ACTIVE
;
4643 if (put_user(len
, optlen
)) {
4648 if (copy_to_user(optval
, &pinfo
, len
)) {
4657 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4659 * This option is a on/off flag. If enabled no SCTP message
4660 * fragmentation will be performed. Instead if a message being sent
4661 * exceeds the current PMTU size, the message will NOT be sent and
4662 * instead a error will be indicated to the user.
4664 static int sctp_getsockopt_disable_fragments(struct sock
*sk
, int len
,
4665 char __user
*optval
, int __user
*optlen
)
4669 if (len
< sizeof(int))
4673 val
= (sctp_sk(sk
)->disable_fragments
== 1);
4674 if (put_user(len
, optlen
))
4676 if (copy_to_user(optval
, &val
, len
))
4681 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4683 * This socket option is used to specify various notifications and
4684 * ancillary data the user wishes to receive.
4686 static int sctp_getsockopt_events(struct sock
*sk
, int len
, char __user
*optval
,
4691 if (len
> sizeof(struct sctp_event_subscribe
))
4692 len
= sizeof(struct sctp_event_subscribe
);
4693 if (put_user(len
, optlen
))
4695 if (copy_to_user(optval
, &sctp_sk(sk
)->subscribe
, len
))
4700 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4702 * This socket option is applicable to the UDP-style socket only. When
4703 * set it will cause associations that are idle for more than the
4704 * specified number of seconds to automatically close. An association
4705 * being idle is defined an association that has NOT sent or received
4706 * user data. The special value of '0' indicates that no automatic
4707 * close of any associations should be performed. The option expects an
4708 * integer defining the number of seconds of idle time before an
4709 * association is closed.
4711 static int sctp_getsockopt_autoclose(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4713 /* Applicable to UDP-style socket only */
4714 if (sctp_style(sk
, TCP
))
4716 if (len
< sizeof(int))
4719 if (put_user(len
, optlen
))
4721 if (copy_to_user(optval
, &sctp_sk(sk
)->autoclose
, sizeof(int)))
4726 /* Helper routine to branch off an association to a new socket. */
4727 int sctp_do_peeloff(struct sock
*sk
, sctp_assoc_t id
, struct socket
**sockp
)
4729 struct sctp_association
*asoc
= sctp_id2assoc(sk
, id
);
4730 struct sctp_sock
*sp
= sctp_sk(sk
);
4731 struct socket
*sock
;
4737 /* If there is a thread waiting on more sndbuf space for
4738 * sending on this asoc, it cannot be peeled.
4740 if (waitqueue_active(&asoc
->wait
))
4743 /* An association cannot be branched off from an already peeled-off
4744 * socket, nor is this supported for tcp style sockets.
4746 if (!sctp_style(sk
, UDP
))
4749 /* Create a new socket. */
4750 err
= sock_create(sk
->sk_family
, SOCK_SEQPACKET
, IPPROTO_SCTP
, &sock
);
4754 sctp_copy_sock(sock
->sk
, sk
, asoc
);
4756 /* Make peeled-off sockets more like 1-1 accepted sockets.
4757 * Set the daddr and initialize id to something more random
4759 sp
->pf
->to_sk_daddr(&asoc
->peer
.primary_addr
, sk
);
4761 /* Populate the fields of the newsk from the oldsk and migrate the
4762 * asoc to the newsk.
4764 sctp_sock_migrate(sk
, sock
->sk
, asoc
, SCTP_SOCKET_UDP_HIGH_BANDWIDTH
);
4770 EXPORT_SYMBOL(sctp_do_peeloff
);
4772 static int sctp_getsockopt_peeloff(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
4774 sctp_peeloff_arg_t peeloff
;
4775 struct socket
*newsock
;
4776 struct file
*newfile
;
4779 if (len
< sizeof(sctp_peeloff_arg_t
))
4781 len
= sizeof(sctp_peeloff_arg_t
);
4782 if (copy_from_user(&peeloff
, optval
, len
))
4785 retval
= sctp_do_peeloff(sk
, peeloff
.associd
, &newsock
);
4789 /* Map the socket to an unused fd that can be returned to the user. */
4790 retval
= get_unused_fd_flags(0);
4792 sock_release(newsock
);
4796 newfile
= sock_alloc_file(newsock
, 0, NULL
);
4797 if (IS_ERR(newfile
)) {
4798 put_unused_fd(retval
);
4799 sock_release(newsock
);
4800 return PTR_ERR(newfile
);
4803 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__
, sk
, newsock
->sk
,
4806 /* Return the fd mapped to the new socket. */
4807 if (put_user(len
, optlen
)) {
4809 put_unused_fd(retval
);
4812 peeloff
.sd
= retval
;
4813 if (copy_to_user(optval
, &peeloff
, len
)) {
4815 put_unused_fd(retval
);
4818 fd_install(retval
, newfile
);
4823 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4825 * Applications can enable or disable heartbeats for any peer address of
4826 * an association, modify an address's heartbeat interval, force a
4827 * heartbeat to be sent immediately, and adjust the address's maximum
4828 * number of retransmissions sent before an address is considered
4829 * unreachable. The following structure is used to access and modify an
4830 * address's parameters:
4832 * struct sctp_paddrparams {
4833 * sctp_assoc_t spp_assoc_id;
4834 * struct sockaddr_storage spp_address;
4835 * uint32_t spp_hbinterval;
4836 * uint16_t spp_pathmaxrxt;
4837 * uint32_t spp_pathmtu;
4838 * uint32_t spp_sackdelay;
4839 * uint32_t spp_flags;
4842 * spp_assoc_id - (one-to-many style socket) This is filled in the
4843 * application, and identifies the association for
4845 * spp_address - This specifies which address is of interest.
4846 * spp_hbinterval - This contains the value of the heartbeat interval,
4847 * in milliseconds. If a value of zero
4848 * is present in this field then no changes are to
4849 * be made to this parameter.
4850 * spp_pathmaxrxt - This contains the maximum number of
4851 * retransmissions before this address shall be
4852 * considered unreachable. If a value of zero
4853 * is present in this field then no changes are to
4854 * be made to this parameter.
4855 * spp_pathmtu - When Path MTU discovery is disabled the value
4856 * specified here will be the "fixed" path mtu.
4857 * Note that if the spp_address field is empty
4858 * then all associations on this address will
4859 * have this fixed path mtu set upon them.
4861 * spp_sackdelay - When delayed sack is enabled, this value specifies
4862 * the number of milliseconds that sacks will be delayed
4863 * for. This value will apply to all addresses of an
4864 * association if the spp_address field is empty. Note
4865 * also, that if delayed sack is enabled and this
4866 * value is set to 0, no change is made to the last
4867 * recorded delayed sack timer value.
4869 * spp_flags - These flags are used to control various features
4870 * on an association. The flag field may contain
4871 * zero or more of the following options.
4873 * SPP_HB_ENABLE - Enable heartbeats on the
4874 * specified address. Note that if the address
4875 * field is empty all addresses for the association
4876 * have heartbeats enabled upon them.
4878 * SPP_HB_DISABLE - Disable heartbeats on the
4879 * speicifed address. Note that if the address
4880 * field is empty all addresses for the association
4881 * will have their heartbeats disabled. Note also
4882 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4883 * mutually exclusive, only one of these two should
4884 * be specified. Enabling both fields will have
4885 * undetermined results.
4887 * SPP_HB_DEMAND - Request a user initiated heartbeat
4888 * to be made immediately.
4890 * SPP_PMTUD_ENABLE - This field will enable PMTU
4891 * discovery upon the specified address. Note that
4892 * if the address feild is empty then all addresses
4893 * on the association are effected.
4895 * SPP_PMTUD_DISABLE - This field will disable PMTU
4896 * discovery upon the specified address. Note that
4897 * if the address feild is empty then all addresses
4898 * on the association are effected. Not also that
4899 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4900 * exclusive. Enabling both will have undetermined
4903 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4904 * on delayed sack. The time specified in spp_sackdelay
4905 * is used to specify the sack delay for this address. Note
4906 * that if spp_address is empty then all addresses will
4907 * enable delayed sack and take on the sack delay
4908 * value specified in spp_sackdelay.
4909 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4910 * off delayed sack. If the spp_address field is blank then
4911 * delayed sack is disabled for the entire association. Note
4912 * also that this field is mutually exclusive to
4913 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4916 static int sctp_getsockopt_peer_addr_params(struct sock
*sk
, int len
,
4917 char __user
*optval
, int __user
*optlen
)
4919 struct sctp_paddrparams params
;
4920 struct sctp_transport
*trans
= NULL
;
4921 struct sctp_association
*asoc
= NULL
;
4922 struct sctp_sock
*sp
= sctp_sk(sk
);
4924 if (len
< sizeof(struct sctp_paddrparams
))
4926 len
= sizeof(struct sctp_paddrparams
);
4927 if (copy_from_user(¶ms
, optval
, len
))
4930 /* If an address other than INADDR_ANY is specified, and
4931 * no transport is found, then the request is invalid.
4933 if (!sctp_is_any(sk
, (union sctp_addr
*)¶ms
.spp_address
)) {
4934 trans
= sctp_addr_id2transport(sk
, ¶ms
.spp_address
,
4935 params
.spp_assoc_id
);
4937 pr_debug("%s: failed no transport\n", __func__
);
4942 /* Get association, if assoc_id != 0 and the socket is a one
4943 * to many style socket, and an association was not found, then
4944 * the id was invalid.
4946 asoc
= sctp_id2assoc(sk
, params
.spp_assoc_id
);
4947 if (!asoc
&& params
.spp_assoc_id
&& sctp_style(sk
, UDP
)) {
4948 pr_debug("%s: failed no association\n", __func__
);
4953 /* Fetch transport values. */
4954 params
.spp_hbinterval
= jiffies_to_msecs(trans
->hbinterval
);
4955 params
.spp_pathmtu
= trans
->pathmtu
;
4956 params
.spp_pathmaxrxt
= trans
->pathmaxrxt
;
4957 params
.spp_sackdelay
= jiffies_to_msecs(trans
->sackdelay
);
4959 /*draft-11 doesn't say what to return in spp_flags*/
4960 params
.spp_flags
= trans
->param_flags
;
4962 /* Fetch association values. */
4963 params
.spp_hbinterval
= jiffies_to_msecs(asoc
->hbinterval
);
4964 params
.spp_pathmtu
= asoc
->pathmtu
;
4965 params
.spp_pathmaxrxt
= asoc
->pathmaxrxt
;
4966 params
.spp_sackdelay
= jiffies_to_msecs(asoc
->sackdelay
);
4968 /*draft-11 doesn't say what to return in spp_flags*/
4969 params
.spp_flags
= asoc
->param_flags
;
4971 /* Fetch socket values. */
4972 params
.spp_hbinterval
= sp
->hbinterval
;
4973 params
.spp_pathmtu
= sp
->pathmtu
;
4974 params
.spp_sackdelay
= sp
->sackdelay
;
4975 params
.spp_pathmaxrxt
= sp
->pathmaxrxt
;
4977 /*draft-11 doesn't say what to return in spp_flags*/
4978 params
.spp_flags
= sp
->param_flags
;
4981 if (copy_to_user(optval
, ¶ms
, len
))
4984 if (put_user(len
, optlen
))
4991 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4993 * This option will effect the way delayed acks are performed. This
4994 * option allows you to get or set the delayed ack time, in
4995 * milliseconds. It also allows changing the delayed ack frequency.
4996 * Changing the frequency to 1 disables the delayed sack algorithm. If
4997 * the assoc_id is 0, then this sets or gets the endpoints default
4998 * values. If the assoc_id field is non-zero, then the set or get
4999 * effects the specified association for the one to many model (the
5000 * assoc_id field is ignored by the one to one model). Note that if
5001 * sack_delay or sack_freq are 0 when setting this option, then the
5002 * current values will remain unchanged.
5004 * struct sctp_sack_info {
5005 * sctp_assoc_t sack_assoc_id;
5006 * uint32_t sack_delay;
5007 * uint32_t sack_freq;
5010 * sack_assoc_id - This parameter, indicates which association the user
5011 * is performing an action upon. Note that if this field's value is
5012 * zero then the endpoints default value is changed (effecting future
5013 * associations only).
5015 * sack_delay - This parameter contains the number of milliseconds that
5016 * the user is requesting the delayed ACK timer be set to. Note that
5017 * this value is defined in the standard to be between 200 and 500
5020 * sack_freq - This parameter contains the number of packets that must
5021 * be received before a sack is sent without waiting for the delay
5022 * timer to expire. The default value for this is 2, setting this
5023 * value to 1 will disable the delayed sack algorithm.
5025 static int sctp_getsockopt_delayed_ack(struct sock
*sk
, int len
,
5026 char __user
*optval
,
5029 struct sctp_sack_info params
;
5030 struct sctp_association
*asoc
= NULL
;
5031 struct sctp_sock
*sp
= sctp_sk(sk
);
5033 if (len
>= sizeof(struct sctp_sack_info
)) {
5034 len
= sizeof(struct sctp_sack_info
);
5036 if (copy_from_user(¶ms
, optval
, len
))
5038 } else if (len
== sizeof(struct sctp_assoc_value
)) {
5039 pr_warn_ratelimited(DEPRECATED
5041 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5042 "Use struct sctp_sack_info instead\n",
5043 current
->comm
, task_pid_nr(current
));
5044 if (copy_from_user(¶ms
, optval
, len
))
5049 /* Get association, if sack_assoc_id != 0 and the socket is a one
5050 * to many style socket, and an association was not found, then
5051 * the id was invalid.
5053 asoc
= sctp_id2assoc(sk
, params
.sack_assoc_id
);
5054 if (!asoc
&& params
.sack_assoc_id
&& sctp_style(sk
, UDP
))
5058 /* Fetch association values. */
5059 if (asoc
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5060 params
.sack_delay
= jiffies_to_msecs(
5062 params
.sack_freq
= asoc
->sackfreq
;
5065 params
.sack_delay
= 0;
5066 params
.sack_freq
= 1;
5069 /* Fetch socket values. */
5070 if (sp
->param_flags
& SPP_SACKDELAY_ENABLE
) {
5071 params
.sack_delay
= sp
->sackdelay
;
5072 params
.sack_freq
= sp
->sackfreq
;
5074 params
.sack_delay
= 0;
5075 params
.sack_freq
= 1;
5079 if (copy_to_user(optval
, ¶ms
, len
))
5082 if (put_user(len
, optlen
))
5088 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5090 * Applications can specify protocol parameters for the default association
5091 * initialization. The option name argument to setsockopt() and getsockopt()
5094 * Setting initialization parameters is effective only on an unconnected
5095 * socket (for UDP-style sockets only future associations are effected
5096 * by the change). With TCP-style sockets, this option is inherited by
5097 * sockets derived from a listener socket.
5099 static int sctp_getsockopt_initmsg(struct sock
*sk
, int len
, char __user
*optval
, int __user
*optlen
)
5101 if (len
< sizeof(struct sctp_initmsg
))
5103 len
= sizeof(struct sctp_initmsg
);
5104 if (put_user(len
, optlen
))
5106 if (copy_to_user(optval
, &sctp_sk(sk
)->initmsg
, len
))
5112 static int sctp_getsockopt_peer_addrs(struct sock
*sk
, int len
,
5113 char __user
*optval
, int __user
*optlen
)
5115 struct sctp_association
*asoc
;
5117 struct sctp_getaddrs getaddrs
;
5118 struct sctp_transport
*from
;
5120 union sctp_addr temp
;
5121 struct sctp_sock
*sp
= sctp_sk(sk
);
5126 if (len
< sizeof(struct sctp_getaddrs
))
5129 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5132 /* For UDP-style sockets, id specifies the association to query. */
5133 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5137 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5138 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5140 list_for_each_entry(from
, &asoc
->peer
.transport_addr_list
,
5142 memcpy(&temp
, &from
->ipaddr
, sizeof(temp
));
5143 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5144 ->addr_to_user(sp
, &temp
);
5145 if (space_left
< addrlen
)
5147 if (copy_to_user(to
, &temp
, addrlen
))
5151 space_left
-= addrlen
;
5154 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
))
5156 bytes_copied
= ((char __user
*)to
) - optval
;
5157 if (put_user(bytes_copied
, optlen
))
5163 static int sctp_copy_laddrs(struct sock
*sk
, __u16 port
, void *to
,
5164 size_t space_left
, int *bytes_copied
)
5166 struct sctp_sockaddr_entry
*addr
;
5167 union sctp_addr temp
;
5170 struct net
*net
= sock_net(sk
);
5173 list_for_each_entry_rcu(addr
, &net
->sctp
.local_addr_list
, list
) {
5177 if ((PF_INET
== sk
->sk_family
) &&
5178 (AF_INET6
== addr
->a
.sa
.sa_family
))
5180 if ((PF_INET6
== sk
->sk_family
) &&
5181 inet_v6_ipv6only(sk
) &&
5182 (AF_INET
== addr
->a
.sa
.sa_family
))
5184 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5185 if (!temp
.v4
.sin_port
)
5186 temp
.v4
.sin_port
= htons(port
);
5188 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5189 ->addr_to_user(sctp_sk(sk
), &temp
);
5191 if (space_left
< addrlen
) {
5195 memcpy(to
, &temp
, addrlen
);
5199 space_left
-= addrlen
;
5200 *bytes_copied
+= addrlen
;
5208 static int sctp_getsockopt_local_addrs(struct sock
*sk
, int len
,
5209 char __user
*optval
, int __user
*optlen
)
5211 struct sctp_bind_addr
*bp
;
5212 struct sctp_association
*asoc
;
5214 struct sctp_getaddrs getaddrs
;
5215 struct sctp_sockaddr_entry
*addr
;
5217 union sctp_addr temp
;
5218 struct sctp_sock
*sp
= sctp_sk(sk
);
5222 int bytes_copied
= 0;
5226 if (len
< sizeof(struct sctp_getaddrs
))
5229 if (copy_from_user(&getaddrs
, optval
, sizeof(struct sctp_getaddrs
)))
5233 * For UDP-style sockets, id specifies the association to query.
5234 * If the id field is set to the value '0' then the locally bound
5235 * addresses are returned without regard to any particular
5238 if (0 == getaddrs
.assoc_id
) {
5239 bp
= &sctp_sk(sk
)->ep
->base
.bind_addr
;
5241 asoc
= sctp_id2assoc(sk
, getaddrs
.assoc_id
);
5244 bp
= &asoc
->base
.bind_addr
;
5247 to
= optval
+ offsetof(struct sctp_getaddrs
, addrs
);
5248 space_left
= len
- offsetof(struct sctp_getaddrs
, addrs
);
5250 addrs
= kmalloc(space_left
, GFP_USER
| __GFP_NOWARN
);
5254 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5255 * addresses from the global local address list.
5257 if (sctp_list_single_entry(&bp
->address_list
)) {
5258 addr
= list_entry(bp
->address_list
.next
,
5259 struct sctp_sockaddr_entry
, list
);
5260 if (sctp_is_any(sk
, &addr
->a
)) {
5261 cnt
= sctp_copy_laddrs(sk
, bp
->port
, addrs
,
5262 space_left
, &bytes_copied
);
5272 /* Protection on the bound address list is not needed since
5273 * in the socket option context we hold a socket lock and
5274 * thus the bound address list can't change.
5276 list_for_each_entry(addr
, &bp
->address_list
, list
) {
5277 memcpy(&temp
, &addr
->a
, sizeof(temp
));
5278 addrlen
= sctp_get_pf_specific(sk
->sk_family
)
5279 ->addr_to_user(sp
, &temp
);
5280 if (space_left
< addrlen
) {
5281 err
= -ENOMEM
; /*fixme: right error?*/
5284 memcpy(buf
, &temp
, addrlen
);
5286 bytes_copied
+= addrlen
;
5288 space_left
-= addrlen
;
5292 if (copy_to_user(to
, addrs
, bytes_copied
)) {
5296 if (put_user(cnt
, &((struct sctp_getaddrs __user
*)optval
)->addr_num
)) {
5300 if (put_user(bytes_copied
, optlen
))
5307 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5309 * Requests that the local SCTP stack use the enclosed peer address as
5310 * the association primary. The enclosed address must be one of the
5311 * association peer's addresses.
5313 static int sctp_getsockopt_primary_addr(struct sock
*sk
, int len
,
5314 char __user
*optval
, int __user
*optlen
)
5316 struct sctp_prim prim
;
5317 struct sctp_association
*asoc
;
5318 struct sctp_sock
*sp
= sctp_sk(sk
);
5320 if (len
< sizeof(struct sctp_prim
))
5323 len
= sizeof(struct sctp_prim
);
5325 if (copy_from_user(&prim
, optval
, len
))
5328 asoc
= sctp_id2assoc(sk
, prim
.ssp_assoc_id
);
5332 if (!asoc
->peer
.primary_path
)
5335 memcpy(&prim
.ssp_addr
, &asoc
->peer
.primary_path
->ipaddr
,
5336 asoc
->peer
.primary_path
->af_specific
->sockaddr_len
);
5338 sctp_get_pf_specific(sk
->sk_family
)->addr_to_user(sp
,
5339 (union sctp_addr
*)&prim
.ssp_addr
);
5341 if (put_user(len
, optlen
))
5343 if (copy_to_user(optval
, &prim
, len
))
5350 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5352 * Requests that the local endpoint set the specified Adaptation Layer
5353 * Indication parameter for all future INIT and INIT-ACK exchanges.
5355 static int sctp_getsockopt_adaptation_layer(struct sock
*sk
, int len
,
5356 char __user
*optval
, int __user
*optlen
)
5358 struct sctp_setadaptation adaptation
;
5360 if (len
< sizeof(struct sctp_setadaptation
))
5363 len
= sizeof(struct sctp_setadaptation
);
5365 adaptation
.ssb_adaptation_ind
= sctp_sk(sk
)->adaptation_ind
;
5367 if (put_user(len
, optlen
))
5369 if (copy_to_user(optval
, &adaptation
, len
))
5377 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5379 * Applications that wish to use the sendto() system call may wish to
5380 * specify a default set of parameters that would normally be supplied
5381 * through the inclusion of ancillary data. This socket option allows
5382 * such an application to set the default sctp_sndrcvinfo structure.
5385 * The application that wishes to use this socket option simply passes
5386 * in to this call the sctp_sndrcvinfo structure defined in Section
5387 * 5.2.2) The input parameters accepted by this call include
5388 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5389 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5390 * to this call if the caller is using the UDP model.
5392 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5394 static int sctp_getsockopt_default_send_param(struct sock
*sk
,
5395 int len
, char __user
*optval
,
5398 struct sctp_sock
*sp
= sctp_sk(sk
);
5399 struct sctp_association
*asoc
;
5400 struct sctp_sndrcvinfo info
;
5402 if (len
< sizeof(info
))
5407 if (copy_from_user(&info
, optval
, len
))
5410 asoc
= sctp_id2assoc(sk
, info
.sinfo_assoc_id
);
5411 if (!asoc
&& info
.sinfo_assoc_id
&& sctp_style(sk
, UDP
))
5414 info
.sinfo_stream
= asoc
->default_stream
;
5415 info
.sinfo_flags
= asoc
->default_flags
;
5416 info
.sinfo_ppid
= asoc
->default_ppid
;
5417 info
.sinfo_context
= asoc
->default_context
;
5418 info
.sinfo_timetolive
= asoc
->default_timetolive
;
5420 info
.sinfo_stream
= sp
->default_stream
;
5421 info
.sinfo_flags
= sp
->default_flags
;
5422 info
.sinfo_ppid
= sp
->default_ppid
;
5423 info
.sinfo_context
= sp
->default_context
;
5424 info
.sinfo_timetolive
= sp
->default_timetolive
;
5427 if (put_user(len
, optlen
))
5429 if (copy_to_user(optval
, &info
, len
))
5435 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5436 * (SCTP_DEFAULT_SNDINFO)
5438 static int sctp_getsockopt_default_sndinfo(struct sock
*sk
, int len
,
5439 char __user
*optval
,
5442 struct sctp_sock
*sp
= sctp_sk(sk
);
5443 struct sctp_association
*asoc
;
5444 struct sctp_sndinfo info
;
5446 if (len
< sizeof(info
))
5451 if (copy_from_user(&info
, optval
, len
))
5454 asoc
= sctp_id2assoc(sk
, info
.snd_assoc_id
);
5455 if (!asoc
&& info
.snd_assoc_id
&& sctp_style(sk
, UDP
))
5458 info
.snd_sid
= asoc
->default_stream
;
5459 info
.snd_flags
= asoc
->default_flags
;
5460 info
.snd_ppid
= asoc
->default_ppid
;
5461 info
.snd_context
= asoc
->default_context
;
5463 info
.snd_sid
= sp
->default_stream
;
5464 info
.snd_flags
= sp
->default_flags
;
5465 info
.snd_ppid
= sp
->default_ppid
;
5466 info
.snd_context
= sp
->default_context
;
5469 if (put_user(len
, optlen
))
5471 if (copy_to_user(optval
, &info
, len
))
5479 * 7.1.5 SCTP_NODELAY
5481 * Turn on/off any Nagle-like algorithm. This means that packets are
5482 * generally sent as soon as possible and no unnecessary delays are
5483 * introduced, at the cost of more packets in the network. Expects an
5484 * integer boolean flag.
5487 static int sctp_getsockopt_nodelay(struct sock
*sk
, int len
,
5488 char __user
*optval
, int __user
*optlen
)
5492 if (len
< sizeof(int))
5496 val
= (sctp_sk(sk
)->nodelay
== 1);
5497 if (put_user(len
, optlen
))
5499 if (copy_to_user(optval
, &val
, len
))
5506 * 7.1.1 SCTP_RTOINFO
5508 * The protocol parameters used to initialize and bound retransmission
5509 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5510 * and modify these parameters.
5511 * All parameters are time values, in milliseconds. A value of 0, when
5512 * modifying the parameters, indicates that the current value should not
5516 static int sctp_getsockopt_rtoinfo(struct sock
*sk
, int len
,
5517 char __user
*optval
,
5518 int __user
*optlen
) {
5519 struct sctp_rtoinfo rtoinfo
;
5520 struct sctp_association
*asoc
;
5522 if (len
< sizeof (struct sctp_rtoinfo
))
5525 len
= sizeof(struct sctp_rtoinfo
);
5527 if (copy_from_user(&rtoinfo
, optval
, len
))
5530 asoc
= sctp_id2assoc(sk
, rtoinfo
.srto_assoc_id
);
5532 if (!asoc
&& rtoinfo
.srto_assoc_id
&& sctp_style(sk
, UDP
))
5535 /* Values corresponding to the specific association. */
5537 rtoinfo
.srto_initial
= jiffies_to_msecs(asoc
->rto_initial
);
5538 rtoinfo
.srto_max
= jiffies_to_msecs(asoc
->rto_max
);
5539 rtoinfo
.srto_min
= jiffies_to_msecs(asoc
->rto_min
);
5541 /* Values corresponding to the endpoint. */
5542 struct sctp_sock
*sp
= sctp_sk(sk
);
5544 rtoinfo
.srto_initial
= sp
->rtoinfo
.srto_initial
;
5545 rtoinfo
.srto_max
= sp
->rtoinfo
.srto_max
;
5546 rtoinfo
.srto_min
= sp
->rtoinfo
.srto_min
;
5549 if (put_user(len
, optlen
))
5552 if (copy_to_user(optval
, &rtoinfo
, len
))
5560 * 7.1.2 SCTP_ASSOCINFO
5562 * This option is used to tune the maximum retransmission attempts
5563 * of the association.
5564 * Returns an error if the new association retransmission value is
5565 * greater than the sum of the retransmission value of the peer.
5566 * See [SCTP] for more information.
5569 static int sctp_getsockopt_associnfo(struct sock
*sk
, int len
,
5570 char __user
*optval
,
5574 struct sctp_assocparams assocparams
;
5575 struct sctp_association
*asoc
;
5576 struct list_head
*pos
;
5579 if (len
< sizeof (struct sctp_assocparams
))
5582 len
= sizeof(struct sctp_assocparams
);
5584 if (copy_from_user(&assocparams
, optval
, len
))
5587 asoc
= sctp_id2assoc(sk
, assocparams
.sasoc_assoc_id
);
5589 if (!asoc
&& assocparams
.sasoc_assoc_id
&& sctp_style(sk
, UDP
))
5592 /* Values correspoinding to the specific association */
5594 assocparams
.sasoc_asocmaxrxt
= asoc
->max_retrans
;
5595 assocparams
.sasoc_peer_rwnd
= asoc
->peer
.rwnd
;
5596 assocparams
.sasoc_local_rwnd
= asoc
->a_rwnd
;
5597 assocparams
.sasoc_cookie_life
= ktime_to_ms(asoc
->cookie_life
);
5599 list_for_each(pos
, &asoc
->peer
.transport_addr_list
) {
5603 assocparams
.sasoc_number_peer_destinations
= cnt
;
5605 /* Values corresponding to the endpoint */
5606 struct sctp_sock
*sp
= sctp_sk(sk
);
5608 assocparams
.sasoc_asocmaxrxt
= sp
->assocparams
.sasoc_asocmaxrxt
;
5609 assocparams
.sasoc_peer_rwnd
= sp
->assocparams
.sasoc_peer_rwnd
;
5610 assocparams
.sasoc_local_rwnd
= sp
->assocparams
.sasoc_local_rwnd
;
5611 assocparams
.sasoc_cookie_life
=
5612 sp
->assocparams
.sasoc_cookie_life
;
5613 assocparams
.sasoc_number_peer_destinations
=
5615 sasoc_number_peer_destinations
;
5618 if (put_user(len
, optlen
))
5621 if (copy_to_user(optval
, &assocparams
, len
))
5628 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5630 * This socket option is a boolean flag which turns on or off mapped V4
5631 * addresses. If this option is turned on and the socket is type
5632 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5633 * If this option is turned off, then no mapping will be done of V4
5634 * addresses and a user will receive both PF_INET6 and PF_INET type
5635 * addresses on the socket.
5637 static int sctp_getsockopt_mappedv4(struct sock
*sk
, int len
,
5638 char __user
*optval
, int __user
*optlen
)
5641 struct sctp_sock
*sp
= sctp_sk(sk
);
5643 if (len
< sizeof(int))
5648 if (put_user(len
, optlen
))
5650 if (copy_to_user(optval
, &val
, len
))
5657 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5658 * (chapter and verse is quoted at sctp_setsockopt_context())
5660 static int sctp_getsockopt_context(struct sock
*sk
, int len
,
5661 char __user
*optval
, int __user
*optlen
)
5663 struct sctp_assoc_value params
;
5664 struct sctp_sock
*sp
;
5665 struct sctp_association
*asoc
;
5667 if (len
< sizeof(struct sctp_assoc_value
))
5670 len
= sizeof(struct sctp_assoc_value
);
5672 if (copy_from_user(¶ms
, optval
, len
))
5677 if (params
.assoc_id
!= 0) {
5678 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5681 params
.assoc_value
= asoc
->default_rcv_context
;
5683 params
.assoc_value
= sp
->default_rcv_context
;
5686 if (put_user(len
, optlen
))
5688 if (copy_to_user(optval
, ¶ms
, len
))
5695 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5696 * This option will get or set the maximum size to put in any outgoing
5697 * SCTP DATA chunk. If a message is larger than this size it will be
5698 * fragmented by SCTP into the specified size. Note that the underlying
5699 * SCTP implementation may fragment into smaller sized chunks when the
5700 * PMTU of the underlying association is smaller than the value set by
5701 * the user. The default value for this option is '0' which indicates
5702 * the user is NOT limiting fragmentation and only the PMTU will effect
5703 * SCTP's choice of DATA chunk size. Note also that values set larger
5704 * than the maximum size of an IP datagram will effectively let SCTP
5705 * control fragmentation (i.e. the same as setting this option to 0).
5707 * The following structure is used to access and modify this parameter:
5709 * struct sctp_assoc_value {
5710 * sctp_assoc_t assoc_id;
5711 * uint32_t assoc_value;
5714 * assoc_id: This parameter is ignored for one-to-one style sockets.
5715 * For one-to-many style sockets this parameter indicates which
5716 * association the user is performing an action upon. Note that if
5717 * this field's value is zero then the endpoints default value is
5718 * changed (effecting future associations only).
5719 * assoc_value: This parameter specifies the maximum size in bytes.
5721 static int sctp_getsockopt_maxseg(struct sock
*sk
, int len
,
5722 char __user
*optval
, int __user
*optlen
)
5724 struct sctp_assoc_value params
;
5725 struct sctp_association
*asoc
;
5727 if (len
== sizeof(int)) {
5728 pr_warn_ratelimited(DEPRECATED
5730 "Use of int in maxseg socket option.\n"
5731 "Use struct sctp_assoc_value instead\n",
5732 current
->comm
, task_pid_nr(current
));
5733 params
.assoc_id
= 0;
5734 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5735 len
= sizeof(struct sctp_assoc_value
);
5736 if (copy_from_user(¶ms
, optval
, sizeof(params
)))
5741 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5742 if (!asoc
&& params
.assoc_id
&& sctp_style(sk
, UDP
))
5746 params
.assoc_value
= asoc
->frag_point
;
5748 params
.assoc_value
= sctp_sk(sk
)->user_frag
;
5750 if (put_user(len
, optlen
))
5752 if (len
== sizeof(int)) {
5753 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5756 if (copy_to_user(optval
, ¶ms
, len
))
5764 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5765 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5767 static int sctp_getsockopt_fragment_interleave(struct sock
*sk
, int len
,
5768 char __user
*optval
, int __user
*optlen
)
5772 if (len
< sizeof(int))
5777 val
= sctp_sk(sk
)->frag_interleave
;
5778 if (put_user(len
, optlen
))
5780 if (copy_to_user(optval
, &val
, len
))
5787 * 7.1.25. Set or Get the sctp partial delivery point
5788 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5790 static int sctp_getsockopt_partial_delivery_point(struct sock
*sk
, int len
,
5791 char __user
*optval
,
5796 if (len
< sizeof(u32
))
5801 val
= sctp_sk(sk
)->pd_point
;
5802 if (put_user(len
, optlen
))
5804 if (copy_to_user(optval
, &val
, len
))
5811 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5812 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5814 static int sctp_getsockopt_maxburst(struct sock
*sk
, int len
,
5815 char __user
*optval
,
5818 struct sctp_assoc_value params
;
5819 struct sctp_sock
*sp
;
5820 struct sctp_association
*asoc
;
5822 if (len
== sizeof(int)) {
5823 pr_warn_ratelimited(DEPRECATED
5825 "Use of int in max_burst socket option.\n"
5826 "Use struct sctp_assoc_value instead\n",
5827 current
->comm
, task_pid_nr(current
));
5828 params
.assoc_id
= 0;
5829 } else if (len
>= sizeof(struct sctp_assoc_value
)) {
5830 len
= sizeof(struct sctp_assoc_value
);
5831 if (copy_from_user(¶ms
, optval
, len
))
5838 if (params
.assoc_id
!= 0) {
5839 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
5842 params
.assoc_value
= asoc
->max_burst
;
5844 params
.assoc_value
= sp
->max_burst
;
5846 if (len
== sizeof(int)) {
5847 if (copy_to_user(optval
, ¶ms
.assoc_value
, len
))
5850 if (copy_to_user(optval
, ¶ms
, len
))
5858 static int sctp_getsockopt_hmac_ident(struct sock
*sk
, int len
,
5859 char __user
*optval
, int __user
*optlen
)
5861 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5862 struct sctp_hmacalgo __user
*p
= (void __user
*)optval
;
5863 struct sctp_hmac_algo_param
*hmacs
;
5868 if (!ep
->auth_enable
)
5871 hmacs
= ep
->auth_hmacs_list
;
5872 data_len
= ntohs(hmacs
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5874 if (len
< sizeof(struct sctp_hmacalgo
) + data_len
)
5877 len
= sizeof(struct sctp_hmacalgo
) + data_len
;
5878 num_idents
= data_len
/ sizeof(u16
);
5880 if (put_user(len
, optlen
))
5882 if (put_user(num_idents
, &p
->shmac_num_idents
))
5884 for (i
= 0; i
< num_idents
; i
++) {
5885 __u16 hmacid
= ntohs(hmacs
->hmac_ids
[i
]);
5887 if (copy_to_user(&p
->shmac_idents
[i
], &hmacid
, sizeof(__u16
)))
5893 static int sctp_getsockopt_active_key(struct sock
*sk
, int len
,
5894 char __user
*optval
, int __user
*optlen
)
5896 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5897 struct sctp_authkeyid val
;
5898 struct sctp_association
*asoc
;
5900 if (!ep
->auth_enable
)
5903 if (len
< sizeof(struct sctp_authkeyid
))
5905 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authkeyid
)))
5908 asoc
= sctp_id2assoc(sk
, val
.scact_assoc_id
);
5909 if (!asoc
&& val
.scact_assoc_id
&& sctp_style(sk
, UDP
))
5913 val
.scact_keynumber
= asoc
->active_key_id
;
5915 val
.scact_keynumber
= ep
->active_key_id
;
5917 len
= sizeof(struct sctp_authkeyid
);
5918 if (put_user(len
, optlen
))
5920 if (copy_to_user(optval
, &val
, len
))
5926 static int sctp_getsockopt_peer_auth_chunks(struct sock
*sk
, int len
,
5927 char __user
*optval
, int __user
*optlen
)
5929 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5930 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5931 struct sctp_authchunks val
;
5932 struct sctp_association
*asoc
;
5933 struct sctp_chunks_param
*ch
;
5937 if (!ep
->auth_enable
)
5940 if (len
< sizeof(struct sctp_authchunks
))
5943 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5946 to
= p
->gauth_chunks
;
5947 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5951 ch
= asoc
->peer
.peer_chunks
;
5955 /* See if the user provided enough room for all the data */
5956 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
5957 if (len
< num_chunks
)
5960 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
5963 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
5964 if (put_user(len
, optlen
))
5966 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
5971 static int sctp_getsockopt_local_auth_chunks(struct sock
*sk
, int len
,
5972 char __user
*optval
, int __user
*optlen
)
5974 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
5975 struct sctp_authchunks __user
*p
= (void __user
*)optval
;
5976 struct sctp_authchunks val
;
5977 struct sctp_association
*asoc
;
5978 struct sctp_chunks_param
*ch
;
5982 if (!ep
->auth_enable
)
5985 if (len
< sizeof(struct sctp_authchunks
))
5988 if (copy_from_user(&val
, optval
, sizeof(struct sctp_authchunks
)))
5991 to
= p
->gauth_chunks
;
5992 asoc
= sctp_id2assoc(sk
, val
.gauth_assoc_id
);
5993 if (!asoc
&& val
.gauth_assoc_id
&& sctp_style(sk
, UDP
))
5997 ch
= (struct sctp_chunks_param
*)asoc
->c
.auth_chunks
;
5999 ch
= ep
->auth_chunk_list
;
6004 num_chunks
= ntohs(ch
->param_hdr
.length
) - sizeof(sctp_paramhdr_t
);
6005 if (len
< sizeof(struct sctp_authchunks
) + num_chunks
)
6008 if (copy_to_user(to
, ch
->chunks
, num_chunks
))
6011 len
= sizeof(struct sctp_authchunks
) + num_chunks
;
6012 if (put_user(len
, optlen
))
6014 if (put_user(num_chunks
, &p
->gauth_number_of_chunks
))
6021 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6022 * This option gets the current number of associations that are attached
6023 * to a one-to-many style socket. The option value is an uint32_t.
6025 static int sctp_getsockopt_assoc_number(struct sock
*sk
, int len
,
6026 char __user
*optval
, int __user
*optlen
)
6028 struct sctp_sock
*sp
= sctp_sk(sk
);
6029 struct sctp_association
*asoc
;
6032 if (sctp_style(sk
, TCP
))
6035 if (len
< sizeof(u32
))
6040 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6044 if (put_user(len
, optlen
))
6046 if (copy_to_user(optval
, &val
, len
))
6053 * 8.1.23 SCTP_AUTO_ASCONF
6054 * See the corresponding setsockopt entry as description
6056 static int sctp_getsockopt_auto_asconf(struct sock
*sk
, int len
,
6057 char __user
*optval
, int __user
*optlen
)
6061 if (len
< sizeof(int))
6065 if (sctp_sk(sk
)->do_auto_asconf
&& sctp_is_ep_boundall(sk
))
6067 if (put_user(len
, optlen
))
6069 if (copy_to_user(optval
, &val
, len
))
6075 * 8.2.6. Get the Current Identifiers of Associations
6076 * (SCTP_GET_ASSOC_ID_LIST)
6078 * This option gets the current list of SCTP association identifiers of
6079 * the SCTP associations handled by a one-to-many style socket.
6081 static int sctp_getsockopt_assoc_ids(struct sock
*sk
, int len
,
6082 char __user
*optval
, int __user
*optlen
)
6084 struct sctp_sock
*sp
= sctp_sk(sk
);
6085 struct sctp_association
*asoc
;
6086 struct sctp_assoc_ids
*ids
;
6089 if (sctp_style(sk
, TCP
))
6092 if (len
< sizeof(struct sctp_assoc_ids
))
6095 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6099 if (len
< sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
)
6102 len
= sizeof(struct sctp_assoc_ids
) + sizeof(sctp_assoc_t
) * num
;
6104 ids
= kmalloc(len
, GFP_USER
| __GFP_NOWARN
);
6108 ids
->gaids_number_of_ids
= num
;
6110 list_for_each_entry(asoc
, &(sp
->ep
->asocs
), asocs
) {
6111 ids
->gaids_assoc_id
[num
++] = asoc
->assoc_id
;
6114 if (put_user(len
, optlen
) || copy_to_user(optval
, ids
, len
)) {
6124 * SCTP_PEER_ADDR_THLDS
6126 * This option allows us to fetch the partially failed threshold for one or all
6127 * transports in an association. See Section 6.1 of:
6128 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6130 static int sctp_getsockopt_paddr_thresholds(struct sock
*sk
,
6131 char __user
*optval
,
6135 struct sctp_paddrthlds val
;
6136 struct sctp_transport
*trans
;
6137 struct sctp_association
*asoc
;
6139 if (len
< sizeof(struct sctp_paddrthlds
))
6141 len
= sizeof(struct sctp_paddrthlds
);
6142 if (copy_from_user(&val
, (struct sctp_paddrthlds __user
*)optval
, len
))
6145 if (sctp_is_any(sk
, (const union sctp_addr
*)&val
.spt_address
)) {
6146 asoc
= sctp_id2assoc(sk
, val
.spt_assoc_id
);
6150 val
.spt_pathpfthld
= asoc
->pf_retrans
;
6151 val
.spt_pathmaxrxt
= asoc
->pathmaxrxt
;
6153 trans
= sctp_addr_id2transport(sk
, &val
.spt_address
,
6158 val
.spt_pathmaxrxt
= trans
->pathmaxrxt
;
6159 val
.spt_pathpfthld
= trans
->pf_retrans
;
6162 if (put_user(len
, optlen
) || copy_to_user(optval
, &val
, len
))
6169 * SCTP_GET_ASSOC_STATS
6171 * This option retrieves local per endpoint statistics. It is modeled
6172 * after OpenSolaris' implementation
6174 static int sctp_getsockopt_assoc_stats(struct sock
*sk
, int len
,
6175 char __user
*optval
,
6178 struct sctp_assoc_stats sas
;
6179 struct sctp_association
*asoc
= NULL
;
6181 /* User must provide at least the assoc id */
6182 if (len
< sizeof(sctp_assoc_t
))
6185 /* Allow the struct to grow and fill in as much as possible */
6186 len
= min_t(size_t, len
, sizeof(sas
));
6188 if (copy_from_user(&sas
, optval
, len
))
6191 asoc
= sctp_id2assoc(sk
, sas
.sas_assoc_id
);
6195 sas
.sas_rtxchunks
= asoc
->stats
.rtxchunks
;
6196 sas
.sas_gapcnt
= asoc
->stats
.gapcnt
;
6197 sas
.sas_outofseqtsns
= asoc
->stats
.outofseqtsns
;
6198 sas
.sas_osacks
= asoc
->stats
.osacks
;
6199 sas
.sas_isacks
= asoc
->stats
.isacks
;
6200 sas
.sas_octrlchunks
= asoc
->stats
.octrlchunks
;
6201 sas
.sas_ictrlchunks
= asoc
->stats
.ictrlchunks
;
6202 sas
.sas_oodchunks
= asoc
->stats
.oodchunks
;
6203 sas
.sas_iodchunks
= asoc
->stats
.iodchunks
;
6204 sas
.sas_ouodchunks
= asoc
->stats
.ouodchunks
;
6205 sas
.sas_iuodchunks
= asoc
->stats
.iuodchunks
;
6206 sas
.sas_idupchunks
= asoc
->stats
.idupchunks
;
6207 sas
.sas_opackets
= asoc
->stats
.opackets
;
6208 sas
.sas_ipackets
= asoc
->stats
.ipackets
;
6210 /* New high max rto observed, will return 0 if not a single
6211 * RTO update took place. obs_rto_ipaddr will be bogus
6214 sas
.sas_maxrto
= asoc
->stats
.max_obs_rto
;
6215 memcpy(&sas
.sas_obs_rto_ipaddr
, &asoc
->stats
.obs_rto_ipaddr
,
6216 sizeof(struct sockaddr_storage
));
6218 /* Mark beginning of a new observation period */
6219 asoc
->stats
.max_obs_rto
= asoc
->rto_min
;
6221 if (put_user(len
, optlen
))
6224 pr_debug("%s: len:%d, assoc_id:%d\n", __func__
, len
, sas
.sas_assoc_id
);
6226 if (copy_to_user(optval
, &sas
, len
))
6232 static int sctp_getsockopt_recvrcvinfo(struct sock
*sk
, int len
,
6233 char __user
*optval
,
6238 if (len
< sizeof(int))
6242 if (sctp_sk(sk
)->recvrcvinfo
)
6244 if (put_user(len
, optlen
))
6246 if (copy_to_user(optval
, &val
, len
))
6252 static int sctp_getsockopt_recvnxtinfo(struct sock
*sk
, int len
,
6253 char __user
*optval
,
6258 if (len
< sizeof(int))
6262 if (sctp_sk(sk
)->recvnxtinfo
)
6264 if (put_user(len
, optlen
))
6266 if (copy_to_user(optval
, &val
, len
))
6272 static int sctp_getsockopt_pr_supported(struct sock
*sk
, int len
,
6273 char __user
*optval
,
6276 struct sctp_assoc_value params
;
6277 struct sctp_association
*asoc
;
6278 int retval
= -EFAULT
;
6280 if (len
< sizeof(params
)) {
6285 len
= sizeof(params
);
6286 if (copy_from_user(¶ms
, optval
, len
))
6289 asoc
= sctp_id2assoc(sk
, params
.assoc_id
);
6291 params
.assoc_value
= asoc
->prsctp_enable
;
6292 } else if (!params
.assoc_id
) {
6293 struct sctp_sock
*sp
= sctp_sk(sk
);
6295 params
.assoc_value
= sp
->ep
->prsctp_enable
;
6301 if (put_user(len
, optlen
))
6304 if (copy_to_user(optval
, ¶ms
, len
))
6313 static int sctp_getsockopt_default_prinfo(struct sock
*sk
, int len
,
6314 char __user
*optval
,
6317 struct sctp_default_prinfo info
;
6318 struct sctp_association
*asoc
;
6319 int retval
= -EFAULT
;
6321 if (len
< sizeof(info
)) {
6327 if (copy_from_user(&info
, optval
, len
))
6330 asoc
= sctp_id2assoc(sk
, info
.pr_assoc_id
);
6332 info
.pr_policy
= SCTP_PR_POLICY(asoc
->default_flags
);
6333 info
.pr_value
= asoc
->default_timetolive
;
6334 } else if (!info
.pr_assoc_id
) {
6335 struct sctp_sock
*sp
= sctp_sk(sk
);
6337 info
.pr_policy
= SCTP_PR_POLICY(sp
->default_flags
);
6338 info
.pr_value
= sp
->default_timetolive
;
6344 if (put_user(len
, optlen
))
6347 if (copy_to_user(optval
, &info
, len
))
6356 static int sctp_getsockopt_pr_assocstatus(struct sock
*sk
, int len
,
6357 char __user
*optval
,
6360 struct sctp_prstatus params
;
6361 struct sctp_association
*asoc
;
6363 int retval
= -EINVAL
;
6365 if (len
< sizeof(params
))
6368 len
= sizeof(params
);
6369 if (copy_from_user(¶ms
, optval
, len
)) {
6374 policy
= params
.sprstat_policy
;
6375 if (policy
& ~SCTP_PR_SCTP_MASK
)
6378 asoc
= sctp_id2assoc(sk
, params
.sprstat_assoc_id
);
6382 if (policy
== SCTP_PR_SCTP_NONE
) {
6383 params
.sprstat_abandoned_unsent
= 0;
6384 params
.sprstat_abandoned_sent
= 0;
6385 for (policy
= 0; policy
<= SCTP_PR_INDEX(MAX
); policy
++) {
6386 params
.sprstat_abandoned_unsent
+=
6387 asoc
->abandoned_unsent
[policy
];
6388 params
.sprstat_abandoned_sent
+=
6389 asoc
->abandoned_sent
[policy
];
6392 params
.sprstat_abandoned_unsent
=
6393 asoc
->abandoned_unsent
[__SCTP_PR_INDEX(policy
)];
6394 params
.sprstat_abandoned_sent
=
6395 asoc
->abandoned_sent
[__SCTP_PR_INDEX(policy
)];
6398 if (put_user(len
, optlen
)) {
6403 if (copy_to_user(optval
, ¶ms
, len
)) {
6414 static int sctp_getsockopt(struct sock
*sk
, int level
, int optname
,
6415 char __user
*optval
, int __user
*optlen
)
6420 pr_debug("%s: sk:%p, optname:%d\n", __func__
, sk
, optname
);
6422 /* I can hardly begin to describe how wrong this is. This is
6423 * so broken as to be worse than useless. The API draft
6424 * REALLY is NOT helpful here... I am not convinced that the
6425 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6426 * are at all well-founded.
6428 if (level
!= SOL_SCTP
) {
6429 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6431 retval
= af
->getsockopt(sk
, level
, optname
, optval
, optlen
);
6435 if (get_user(len
, optlen
))
6445 retval
= sctp_getsockopt_sctp_status(sk
, len
, optval
, optlen
);
6447 case SCTP_DISABLE_FRAGMENTS
:
6448 retval
= sctp_getsockopt_disable_fragments(sk
, len
, optval
,
6452 retval
= sctp_getsockopt_events(sk
, len
, optval
, optlen
);
6454 case SCTP_AUTOCLOSE
:
6455 retval
= sctp_getsockopt_autoclose(sk
, len
, optval
, optlen
);
6457 case SCTP_SOCKOPT_PEELOFF
:
6458 retval
= sctp_getsockopt_peeloff(sk
, len
, optval
, optlen
);
6460 case SCTP_PEER_ADDR_PARAMS
:
6461 retval
= sctp_getsockopt_peer_addr_params(sk
, len
, optval
,
6464 case SCTP_DELAYED_SACK
:
6465 retval
= sctp_getsockopt_delayed_ack(sk
, len
, optval
,
6469 retval
= sctp_getsockopt_initmsg(sk
, len
, optval
, optlen
);
6471 case SCTP_GET_PEER_ADDRS
:
6472 retval
= sctp_getsockopt_peer_addrs(sk
, len
, optval
,
6475 case SCTP_GET_LOCAL_ADDRS
:
6476 retval
= sctp_getsockopt_local_addrs(sk
, len
, optval
,
6479 case SCTP_SOCKOPT_CONNECTX3
:
6480 retval
= sctp_getsockopt_connectx3(sk
, len
, optval
, optlen
);
6482 case SCTP_DEFAULT_SEND_PARAM
:
6483 retval
= sctp_getsockopt_default_send_param(sk
, len
,
6486 case SCTP_DEFAULT_SNDINFO
:
6487 retval
= sctp_getsockopt_default_sndinfo(sk
, len
,
6490 case SCTP_PRIMARY_ADDR
:
6491 retval
= sctp_getsockopt_primary_addr(sk
, len
, optval
, optlen
);
6494 retval
= sctp_getsockopt_nodelay(sk
, len
, optval
, optlen
);
6497 retval
= sctp_getsockopt_rtoinfo(sk
, len
, optval
, optlen
);
6499 case SCTP_ASSOCINFO
:
6500 retval
= sctp_getsockopt_associnfo(sk
, len
, optval
, optlen
);
6502 case SCTP_I_WANT_MAPPED_V4_ADDR
:
6503 retval
= sctp_getsockopt_mappedv4(sk
, len
, optval
, optlen
);
6506 retval
= sctp_getsockopt_maxseg(sk
, len
, optval
, optlen
);
6508 case SCTP_GET_PEER_ADDR_INFO
:
6509 retval
= sctp_getsockopt_peer_addr_info(sk
, len
, optval
,
6512 case SCTP_ADAPTATION_LAYER
:
6513 retval
= sctp_getsockopt_adaptation_layer(sk
, len
, optval
,
6517 retval
= sctp_getsockopt_context(sk
, len
, optval
, optlen
);
6519 case SCTP_FRAGMENT_INTERLEAVE
:
6520 retval
= sctp_getsockopt_fragment_interleave(sk
, len
, optval
,
6523 case SCTP_PARTIAL_DELIVERY_POINT
:
6524 retval
= sctp_getsockopt_partial_delivery_point(sk
, len
, optval
,
6527 case SCTP_MAX_BURST
:
6528 retval
= sctp_getsockopt_maxburst(sk
, len
, optval
, optlen
);
6531 case SCTP_AUTH_CHUNK
:
6532 case SCTP_AUTH_DELETE_KEY
:
6533 retval
= -EOPNOTSUPP
;
6535 case SCTP_HMAC_IDENT
:
6536 retval
= sctp_getsockopt_hmac_ident(sk
, len
, optval
, optlen
);
6538 case SCTP_AUTH_ACTIVE_KEY
:
6539 retval
= sctp_getsockopt_active_key(sk
, len
, optval
, optlen
);
6541 case SCTP_PEER_AUTH_CHUNKS
:
6542 retval
= sctp_getsockopt_peer_auth_chunks(sk
, len
, optval
,
6545 case SCTP_LOCAL_AUTH_CHUNKS
:
6546 retval
= sctp_getsockopt_local_auth_chunks(sk
, len
, optval
,
6549 case SCTP_GET_ASSOC_NUMBER
:
6550 retval
= sctp_getsockopt_assoc_number(sk
, len
, optval
, optlen
);
6552 case SCTP_GET_ASSOC_ID_LIST
:
6553 retval
= sctp_getsockopt_assoc_ids(sk
, len
, optval
, optlen
);
6555 case SCTP_AUTO_ASCONF
:
6556 retval
= sctp_getsockopt_auto_asconf(sk
, len
, optval
, optlen
);
6558 case SCTP_PEER_ADDR_THLDS
:
6559 retval
= sctp_getsockopt_paddr_thresholds(sk
, optval
, len
, optlen
);
6561 case SCTP_GET_ASSOC_STATS
:
6562 retval
= sctp_getsockopt_assoc_stats(sk
, len
, optval
, optlen
);
6564 case SCTP_RECVRCVINFO
:
6565 retval
= sctp_getsockopt_recvrcvinfo(sk
, len
, optval
, optlen
);
6567 case SCTP_RECVNXTINFO
:
6568 retval
= sctp_getsockopt_recvnxtinfo(sk
, len
, optval
, optlen
);
6570 case SCTP_PR_SUPPORTED
:
6571 retval
= sctp_getsockopt_pr_supported(sk
, len
, optval
, optlen
);
6573 case SCTP_DEFAULT_PRINFO
:
6574 retval
= sctp_getsockopt_default_prinfo(sk
, len
, optval
,
6577 case SCTP_PR_ASSOC_STATUS
:
6578 retval
= sctp_getsockopt_pr_assocstatus(sk
, len
, optval
,
6582 retval
= -ENOPROTOOPT
;
6590 static int sctp_hash(struct sock
*sk
)
6596 static void sctp_unhash(struct sock
*sk
)
6601 /* Check if port is acceptable. Possibly find first available port.
6603 * The port hash table (contained in the 'global' SCTP protocol storage
6604 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6605 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6606 * list (the list number is the port number hashed out, so as you
6607 * would expect from a hash function, all the ports in a given list have
6608 * such a number that hashes out to the same list number; you were
6609 * expecting that, right?); so each list has a set of ports, with a
6610 * link to the socket (struct sock) that uses it, the port number and
6611 * a fastreuse flag (FIXME: NPI ipg).
6613 static struct sctp_bind_bucket
*sctp_bucket_create(
6614 struct sctp_bind_hashbucket
*head
, struct net
*, unsigned short snum
);
6616 static long sctp_get_port_local(struct sock
*sk
, union sctp_addr
*addr
)
6618 struct sctp_bind_hashbucket
*head
; /* hash list */
6619 struct sctp_bind_bucket
*pp
;
6620 unsigned short snum
;
6623 snum
= ntohs(addr
->v4
.sin_port
);
6625 pr_debug("%s: begins, snum:%d\n", __func__
, snum
);
6630 /* Search for an available port. */
6631 int low
, high
, remaining
, index
;
6633 struct net
*net
= sock_net(sk
);
6635 inet_get_local_port_range(net
, &low
, &high
);
6636 remaining
= (high
- low
) + 1;
6637 rover
= prandom_u32() % remaining
+ low
;
6641 if ((rover
< low
) || (rover
> high
))
6643 if (inet_is_local_reserved_port(net
, rover
))
6645 index
= sctp_phashfn(sock_net(sk
), rover
);
6646 head
= &sctp_port_hashtable
[index
];
6647 spin_lock(&head
->lock
);
6648 sctp_for_each_hentry(pp
, &head
->chain
)
6649 if ((pp
->port
== rover
) &&
6650 net_eq(sock_net(sk
), pp
->net
))
6654 spin_unlock(&head
->lock
);
6655 } while (--remaining
> 0);
6657 /* Exhausted local port range during search? */
6662 /* OK, here is the one we will use. HEAD (the port
6663 * hash table list entry) is non-NULL and we hold it's
6668 /* We are given an specific port number; we verify
6669 * that it is not being used. If it is used, we will
6670 * exahust the search in the hash list corresponding
6671 * to the port number (snum) - we detect that with the
6672 * port iterator, pp being NULL.
6674 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
), snum
)];
6675 spin_lock(&head
->lock
);
6676 sctp_for_each_hentry(pp
, &head
->chain
) {
6677 if ((pp
->port
== snum
) && net_eq(pp
->net
, sock_net(sk
)))
6684 if (!hlist_empty(&pp
->owner
)) {
6685 /* We had a port hash table hit - there is an
6686 * available port (pp != NULL) and it is being
6687 * used by other socket (pp->owner not empty); that other
6688 * socket is going to be sk2.
6690 int reuse
= sk
->sk_reuse
;
6693 pr_debug("%s: found a possible match\n", __func__
);
6695 if (pp
->fastreuse
&& sk
->sk_reuse
&&
6696 sk
->sk_state
!= SCTP_SS_LISTENING
)
6699 /* Run through the list of sockets bound to the port
6700 * (pp->port) [via the pointers bind_next and
6701 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6702 * we get the endpoint they describe and run through
6703 * the endpoint's list of IP (v4 or v6) addresses,
6704 * comparing each of the addresses with the address of
6705 * the socket sk. If we find a match, then that means
6706 * that this port/socket (sk) combination are already
6709 sk_for_each_bound(sk2
, &pp
->owner
) {
6710 struct sctp_endpoint
*ep2
;
6711 ep2
= sctp_sk(sk2
)->ep
;
6714 (reuse
&& sk2
->sk_reuse
&&
6715 sk2
->sk_state
!= SCTP_SS_LISTENING
))
6718 if (sctp_bind_addr_conflict(&ep2
->base
.bind_addr
, addr
,
6719 sctp_sk(sk2
), sctp_sk(sk
))) {
6725 pr_debug("%s: found a match\n", __func__
);
6728 /* If there was a hash table miss, create a new port. */
6730 if (!pp
&& !(pp
= sctp_bucket_create(head
, sock_net(sk
), snum
)))
6733 /* In either case (hit or miss), make sure fastreuse is 1 only
6734 * if sk->sk_reuse is too (that is, if the caller requested
6735 * SO_REUSEADDR on this socket -sk-).
6737 if (hlist_empty(&pp
->owner
)) {
6738 if (sk
->sk_reuse
&& sk
->sk_state
!= SCTP_SS_LISTENING
)
6742 } else if (pp
->fastreuse
&&
6743 (!sk
->sk_reuse
|| sk
->sk_state
== SCTP_SS_LISTENING
))
6746 /* We are set, so fill up all the data in the hash table
6747 * entry, tie the socket list information with the rest of the
6748 * sockets FIXME: Blurry, NPI (ipg).
6751 if (!sctp_sk(sk
)->bind_hash
) {
6752 inet_sk(sk
)->inet_num
= snum
;
6753 sk_add_bind_node(sk
, &pp
->owner
);
6754 sctp_sk(sk
)->bind_hash
= pp
;
6759 spin_unlock(&head
->lock
);
6766 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6767 * port is requested.
6769 static int sctp_get_port(struct sock
*sk
, unsigned short snum
)
6771 union sctp_addr addr
;
6772 struct sctp_af
*af
= sctp_sk(sk
)->pf
->af
;
6774 /* Set up a dummy address struct from the sk. */
6775 af
->from_sk(&addr
, sk
);
6776 addr
.v4
.sin_port
= htons(snum
);
6778 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6779 return !!sctp_get_port_local(sk
, &addr
);
6783 * Move a socket to LISTENING state.
6785 static int sctp_listen_start(struct sock
*sk
, int backlog
)
6787 struct sctp_sock
*sp
= sctp_sk(sk
);
6788 struct sctp_endpoint
*ep
= sp
->ep
;
6789 struct crypto_shash
*tfm
= NULL
;
6792 /* Allocate HMAC for generating cookie. */
6793 if (!sp
->hmac
&& sp
->sctp_hmac_alg
) {
6794 sprintf(alg
, "hmac(%s)", sp
->sctp_hmac_alg
);
6795 tfm
= crypto_alloc_shash(alg
, 0, 0);
6797 net_info_ratelimited("failed to load transform for %s: %ld\n",
6798 sp
->sctp_hmac_alg
, PTR_ERR(tfm
));
6801 sctp_sk(sk
)->hmac
= tfm
;
6805 * If a bind() or sctp_bindx() is not called prior to a listen()
6806 * call that allows new associations to be accepted, the system
6807 * picks an ephemeral port and will choose an address set equivalent
6808 * to binding with a wildcard address.
6810 * This is not currently spelled out in the SCTP sockets
6811 * extensions draft, but follows the practice as seen in TCP
6815 sk
->sk_state
= SCTP_SS_LISTENING
;
6816 if (!ep
->base
.bind_addr
.port
) {
6817 if (sctp_autobind(sk
))
6820 if (sctp_get_port(sk
, inet_sk(sk
)->inet_num
)) {
6821 sk
->sk_state
= SCTP_SS_CLOSED
;
6826 sk
->sk_max_ack_backlog
= backlog
;
6827 sctp_hash_endpoint(ep
);
6832 * 4.1.3 / 5.1.3 listen()
6834 * By default, new associations are not accepted for UDP style sockets.
6835 * An application uses listen() to mark a socket as being able to
6836 * accept new associations.
6838 * On TCP style sockets, applications use listen() to ready the SCTP
6839 * endpoint for accepting inbound associations.
6841 * On both types of endpoints a backlog of '0' disables listening.
6843 * Move a socket to LISTENING state.
6845 int sctp_inet_listen(struct socket
*sock
, int backlog
)
6847 struct sock
*sk
= sock
->sk
;
6848 struct sctp_endpoint
*ep
= sctp_sk(sk
)->ep
;
6851 if (unlikely(backlog
< 0))
6856 /* Peeled-off sockets are not allowed to listen(). */
6857 if (sctp_style(sk
, UDP_HIGH_BANDWIDTH
))
6860 if (sock
->state
!= SS_UNCONNECTED
)
6863 if (!sctp_sstate(sk
, LISTENING
) && !sctp_sstate(sk
, CLOSED
))
6866 /* If backlog is zero, disable listening. */
6868 if (sctp_sstate(sk
, CLOSED
))
6872 sctp_unhash_endpoint(ep
);
6873 sk
->sk_state
= SCTP_SS_CLOSED
;
6875 sctp_sk(sk
)->bind_hash
->fastreuse
= 1;
6879 /* If we are already listening, just update the backlog */
6880 if (sctp_sstate(sk
, LISTENING
))
6881 sk
->sk_max_ack_backlog
= backlog
;
6883 err
= sctp_listen_start(sk
, backlog
);
6895 * This function is done by modeling the current datagram_poll() and the
6896 * tcp_poll(). Note that, based on these implementations, we don't
6897 * lock the socket in this function, even though it seems that,
6898 * ideally, locking or some other mechanisms can be used to ensure
6899 * the integrity of the counters (sndbuf and wmem_alloc) used
6900 * in this place. We assume that we don't need locks either until proven
6903 * Another thing to note is that we include the Async I/O support
6904 * here, again, by modeling the current TCP/UDP code. We don't have
6905 * a good way to test with it yet.
6907 unsigned int sctp_poll(struct file
*file
, struct socket
*sock
, poll_table
*wait
)
6909 struct sock
*sk
= sock
->sk
;
6910 struct sctp_sock
*sp
= sctp_sk(sk
);
6913 poll_wait(file
, sk_sleep(sk
), wait
);
6915 sock_rps_record_flow(sk
);
6917 /* A TCP-style listening socket becomes readable when the accept queue
6920 if (sctp_style(sk
, TCP
) && sctp_sstate(sk
, LISTENING
))
6921 return (!list_empty(&sp
->ep
->asocs
)) ?
6922 (POLLIN
| POLLRDNORM
) : 0;
6926 /* Is there any exceptional events? */
6927 if (sk
->sk_err
|| !skb_queue_empty(&sk
->sk_error_queue
))
6929 (sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
) ? POLLPRI
: 0);
6930 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
6931 mask
|= POLLRDHUP
| POLLIN
| POLLRDNORM
;
6932 if (sk
->sk_shutdown
== SHUTDOWN_MASK
)
6935 /* Is it readable? Reconsider this code with TCP-style support. */
6936 if (!skb_queue_empty(&sk
->sk_receive_queue
))
6937 mask
|= POLLIN
| POLLRDNORM
;
6939 /* The association is either gone or not ready. */
6940 if (!sctp_style(sk
, UDP
) && sctp_sstate(sk
, CLOSED
))
6943 /* Is it writable? */
6944 if (sctp_writeable(sk
)) {
6945 mask
|= POLLOUT
| POLLWRNORM
;
6947 sk_set_bit(SOCKWQ_ASYNC_NOSPACE
, sk
);
6949 * Since the socket is not locked, the buffer
6950 * might be made available after the writeable check and
6951 * before the bit is set. This could cause a lost I/O
6952 * signal. tcp_poll() has a race breaker for this race
6953 * condition. Based on their implementation, we put
6954 * in the following code to cover it as well.
6956 if (sctp_writeable(sk
))
6957 mask
|= POLLOUT
| POLLWRNORM
;
6962 /********************************************************************
6963 * 2nd Level Abstractions
6964 ********************************************************************/
6966 static struct sctp_bind_bucket
*sctp_bucket_create(
6967 struct sctp_bind_hashbucket
*head
, struct net
*net
, unsigned short snum
)
6969 struct sctp_bind_bucket
*pp
;
6971 pp
= kmem_cache_alloc(sctp_bucket_cachep
, GFP_ATOMIC
);
6973 SCTP_DBG_OBJCNT_INC(bind_bucket
);
6976 INIT_HLIST_HEAD(&pp
->owner
);
6978 hlist_add_head(&pp
->node
, &head
->chain
);
6983 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6984 static void sctp_bucket_destroy(struct sctp_bind_bucket
*pp
)
6986 if (pp
&& hlist_empty(&pp
->owner
)) {
6987 __hlist_del(&pp
->node
);
6988 kmem_cache_free(sctp_bucket_cachep
, pp
);
6989 SCTP_DBG_OBJCNT_DEC(bind_bucket
);
6993 /* Release this socket's reference to a local port. */
6994 static inline void __sctp_put_port(struct sock
*sk
)
6996 struct sctp_bind_hashbucket
*head
=
6997 &sctp_port_hashtable
[sctp_phashfn(sock_net(sk
),
6998 inet_sk(sk
)->inet_num
)];
6999 struct sctp_bind_bucket
*pp
;
7001 spin_lock(&head
->lock
);
7002 pp
= sctp_sk(sk
)->bind_hash
;
7003 __sk_del_bind_node(sk
);
7004 sctp_sk(sk
)->bind_hash
= NULL
;
7005 inet_sk(sk
)->inet_num
= 0;
7006 sctp_bucket_destroy(pp
);
7007 spin_unlock(&head
->lock
);
7010 void sctp_put_port(struct sock
*sk
)
7013 __sctp_put_port(sk
);
7018 * The system picks an ephemeral port and choose an address set equivalent
7019 * to binding with a wildcard address.
7020 * One of those addresses will be the primary address for the association.
7021 * This automatically enables the multihoming capability of SCTP.
7023 static int sctp_autobind(struct sock
*sk
)
7025 union sctp_addr autoaddr
;
7029 /* Initialize a local sockaddr structure to INADDR_ANY. */
7030 af
= sctp_sk(sk
)->pf
->af
;
7032 port
= htons(inet_sk(sk
)->inet_num
);
7033 af
->inaddr_any(&autoaddr
, port
);
7035 return sctp_do_bind(sk
, &autoaddr
, af
->sockaddr_len
);
7038 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7041 * 4.2 The cmsghdr Structure *
7043 * When ancillary data is sent or received, any number of ancillary data
7044 * objects can be specified by the msg_control and msg_controllen members of
7045 * the msghdr structure, because each object is preceded by
7046 * a cmsghdr structure defining the object's length (the cmsg_len member).
7047 * Historically Berkeley-derived implementations have passed only one object
7048 * at a time, but this API allows multiple objects to be
7049 * passed in a single call to sendmsg() or recvmsg(). The following example
7050 * shows two ancillary data objects in a control buffer.
7052 * |<--------------------------- msg_controllen -------------------------->|
7055 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7057 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7060 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7062 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7065 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7066 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7068 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7070 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7077 static int sctp_msghdr_parse(const struct msghdr
*msg
, sctp_cmsgs_t
*cmsgs
)
7079 struct cmsghdr
*cmsg
;
7080 struct msghdr
*my_msg
= (struct msghdr
*)msg
;
7082 for_each_cmsghdr(cmsg
, my_msg
) {
7083 if (!CMSG_OK(my_msg
, cmsg
))
7086 /* Should we parse this header or ignore? */
7087 if (cmsg
->cmsg_level
!= IPPROTO_SCTP
)
7090 /* Strictly check lengths following example in SCM code. */
7091 switch (cmsg
->cmsg_type
) {
7093 /* SCTP Socket API Extension
7094 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7096 * This cmsghdr structure provides information for
7097 * initializing new SCTP associations with sendmsg().
7098 * The SCTP_INITMSG socket option uses this same data
7099 * structure. This structure is not used for
7102 * cmsg_level cmsg_type cmsg_data[]
7103 * ------------ ------------ ----------------------
7104 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7106 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_initmsg
)))
7109 cmsgs
->init
= CMSG_DATA(cmsg
);
7113 /* SCTP Socket API Extension
7114 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7116 * This cmsghdr structure specifies SCTP options for
7117 * sendmsg() and describes SCTP header information
7118 * about a received message through recvmsg().
7120 * cmsg_level cmsg_type cmsg_data[]
7121 * ------------ ------------ ----------------------
7122 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7124 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndrcvinfo
)))
7127 cmsgs
->srinfo
= CMSG_DATA(cmsg
);
7129 if (cmsgs
->srinfo
->sinfo_flags
&
7130 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7131 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7132 SCTP_ABORT
| SCTP_EOF
))
7137 /* SCTP Socket API Extension
7138 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7140 * This cmsghdr structure specifies SCTP options for
7141 * sendmsg(). This structure and SCTP_RCVINFO replaces
7142 * SCTP_SNDRCV which has been deprecated.
7144 * cmsg_level cmsg_type cmsg_data[]
7145 * ------------ ------------ ---------------------
7146 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7148 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(struct sctp_sndinfo
)))
7151 cmsgs
->sinfo
= CMSG_DATA(cmsg
);
7153 if (cmsgs
->sinfo
->snd_flags
&
7154 ~(SCTP_UNORDERED
| SCTP_ADDR_OVER
|
7155 SCTP_SACK_IMMEDIATELY
| SCTP_PR_SCTP_MASK
|
7156 SCTP_ABORT
| SCTP_EOF
))
7168 * Wait for a packet..
7169 * Note: This function is the same function as in core/datagram.c
7170 * with a few modifications to make lksctp work.
7172 static int sctp_wait_for_packet(struct sock
*sk
, int *err
, long *timeo_p
)
7177 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7179 /* Socket errors? */
7180 error
= sock_error(sk
);
7184 if (!skb_queue_empty(&sk
->sk_receive_queue
))
7187 /* Socket shut down? */
7188 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7191 /* Sequenced packets can come disconnected. If so we report the
7196 /* Is there a good reason to think that we may receive some data? */
7197 if (list_empty(&sctp_sk(sk
)->ep
->asocs
) && !sctp_sstate(sk
, LISTENING
))
7200 /* Handle signals. */
7201 if (signal_pending(current
))
7204 /* Let another process have a go. Since we are going to sleep
7205 * anyway. Note: This may cause odd behaviors if the message
7206 * does not fit in the user's buffer, but this seems to be the
7207 * only way to honor MSG_DONTWAIT realistically.
7210 *timeo_p
= schedule_timeout(*timeo_p
);
7214 finish_wait(sk_sleep(sk
), &wait
);
7218 error
= sock_intr_errno(*timeo_p
);
7221 finish_wait(sk_sleep(sk
), &wait
);
7226 /* Receive a datagram.
7227 * Note: This is pretty much the same routine as in core/datagram.c
7228 * with a few changes to make lksctp work.
7230 struct sk_buff
*sctp_skb_recv_datagram(struct sock
*sk
, int flags
,
7231 int noblock
, int *err
)
7234 struct sk_buff
*skb
;
7237 timeo
= sock_rcvtimeo(sk
, noblock
);
7239 pr_debug("%s: timeo:%ld, max:%ld\n", __func__
, timeo
,
7240 MAX_SCHEDULE_TIMEOUT
);
7243 /* Again only user level code calls this function,
7244 * so nothing interrupt level
7245 * will suddenly eat the receive_queue.
7247 * Look at current nfs client by the way...
7248 * However, this function was correct in any case. 8)
7250 if (flags
& MSG_PEEK
) {
7251 skb
= skb_peek(&sk
->sk_receive_queue
);
7253 atomic_inc(&skb
->users
);
7255 skb
= __skb_dequeue(&sk
->sk_receive_queue
);
7261 /* Caller is allowed not to check sk->sk_err before calling. */
7262 error
= sock_error(sk
);
7266 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7269 if (sk_can_busy_loop(sk
) &&
7270 sk_busy_loop(sk
, noblock
))
7273 /* User doesn't want to wait. */
7277 } while (sctp_wait_for_packet(sk
, err
, &timeo
) == 0);
7286 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7287 static void __sctp_write_space(struct sctp_association
*asoc
)
7289 struct sock
*sk
= asoc
->base
.sk
;
7291 if (sctp_wspace(asoc
) <= 0)
7294 if (waitqueue_active(&asoc
->wait
))
7295 wake_up_interruptible(&asoc
->wait
);
7297 if (sctp_writeable(sk
)) {
7298 struct socket_wq
*wq
;
7301 wq
= rcu_dereference(sk
->sk_wq
);
7303 if (waitqueue_active(&wq
->wait
))
7304 wake_up_interruptible(&wq
->wait
);
7306 /* Note that we try to include the Async I/O support
7307 * here by modeling from the current TCP/UDP code.
7308 * We have not tested with it yet.
7310 if (!(sk
->sk_shutdown
& SEND_SHUTDOWN
))
7311 sock_wake_async(wq
, SOCK_WAKE_SPACE
, POLL_OUT
);
7317 static void sctp_wake_up_waiters(struct sock
*sk
,
7318 struct sctp_association
*asoc
)
7320 struct sctp_association
*tmp
= asoc
;
7322 /* We do accounting for the sndbuf space per association,
7323 * so we only need to wake our own association.
7325 if (asoc
->ep
->sndbuf_policy
)
7326 return __sctp_write_space(asoc
);
7328 /* If association goes down and is just flushing its
7329 * outq, then just normally notify others.
7331 if (asoc
->base
.dead
)
7332 return sctp_write_space(sk
);
7334 /* Accounting for the sndbuf space is per socket, so we
7335 * need to wake up others, try to be fair and in case of
7336 * other associations, let them have a go first instead
7337 * of just doing a sctp_write_space() call.
7339 * Note that we reach sctp_wake_up_waiters() only when
7340 * associations free up queued chunks, thus we are under
7341 * lock and the list of associations on a socket is
7342 * guaranteed not to change.
7344 for (tmp
= list_next_entry(tmp
, asocs
); 1;
7345 tmp
= list_next_entry(tmp
, asocs
)) {
7346 /* Manually skip the head element. */
7347 if (&tmp
->asocs
== &((sctp_sk(sk
))->ep
->asocs
))
7349 /* Wake up association. */
7350 __sctp_write_space(tmp
);
7351 /* We've reached the end. */
7357 /* Do accounting for the sndbuf space.
7358 * Decrement the used sndbuf space of the corresponding association by the
7359 * data size which was just transmitted(freed).
7361 static void sctp_wfree(struct sk_buff
*skb
)
7363 struct sctp_chunk
*chunk
= skb_shinfo(skb
)->destructor_arg
;
7364 struct sctp_association
*asoc
= chunk
->asoc
;
7365 struct sock
*sk
= asoc
->base
.sk
;
7367 asoc
->sndbuf_used
-= SCTP_DATA_SNDSIZE(chunk
) +
7368 sizeof(struct sk_buff
) +
7369 sizeof(struct sctp_chunk
);
7371 atomic_sub(sizeof(struct sctp_chunk
), &sk
->sk_wmem_alloc
);
7374 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7376 sk
->sk_wmem_queued
-= skb
->truesize
;
7377 sk_mem_uncharge(sk
, skb
->truesize
);
7380 sctp_wake_up_waiters(sk
, asoc
);
7382 sctp_association_put(asoc
);
7385 /* Do accounting for the receive space on the socket.
7386 * Accounting for the association is done in ulpevent.c
7387 * We set this as a destructor for the cloned data skbs so that
7388 * accounting is done at the correct time.
7390 void sctp_sock_rfree(struct sk_buff
*skb
)
7392 struct sock
*sk
= skb
->sk
;
7393 struct sctp_ulpevent
*event
= sctp_skb2event(skb
);
7395 atomic_sub(event
->rmem_len
, &sk
->sk_rmem_alloc
);
7398 * Mimic the behavior of sock_rfree
7400 sk_mem_uncharge(sk
, event
->rmem_len
);
7404 /* Helper function to wait for space in the sndbuf. */
7405 static int sctp_wait_for_sndbuf(struct sctp_association
*asoc
, long *timeo_p
,
7408 struct sock
*sk
= asoc
->base
.sk
;
7410 long current_timeo
= *timeo_p
;
7413 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__
, asoc
,
7416 /* Increment the association's refcnt. */
7417 sctp_association_hold(asoc
);
7419 /* Wait on the association specific sndbuf space. */
7421 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7422 TASK_INTERRUPTIBLE
);
7425 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7428 if (signal_pending(current
))
7429 goto do_interrupted
;
7430 if (msg_len
<= sctp_wspace(asoc
))
7433 /* Let another process have a go. Since we are going
7437 current_timeo
= schedule_timeout(current_timeo
);
7440 *timeo_p
= current_timeo
;
7444 finish_wait(&asoc
->wait
, &wait
);
7446 /* Release the association's refcnt. */
7447 sctp_association_put(asoc
);
7456 err
= sock_intr_errno(*timeo_p
);
7464 void sctp_data_ready(struct sock
*sk
)
7466 struct socket_wq
*wq
;
7469 wq
= rcu_dereference(sk
->sk_wq
);
7470 if (skwq_has_sleeper(wq
))
7471 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
|
7472 POLLRDNORM
| POLLRDBAND
);
7473 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
7477 /* If socket sndbuf has changed, wake up all per association waiters. */
7478 void sctp_write_space(struct sock
*sk
)
7480 struct sctp_association
*asoc
;
7482 /* Wake up the tasks in each wait queue. */
7483 list_for_each_entry(asoc
, &((sctp_sk(sk
))->ep
->asocs
), asocs
) {
7484 __sctp_write_space(asoc
);
7488 /* Is there any sndbuf space available on the socket?
7490 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7491 * associations on the same socket. For a UDP-style socket with
7492 * multiple associations, it is possible for it to be "unwriteable"
7493 * prematurely. I assume that this is acceptable because
7494 * a premature "unwriteable" is better than an accidental "writeable" which
7495 * would cause an unwanted block under certain circumstances. For the 1-1
7496 * UDP-style sockets or TCP-style sockets, this code should work.
7499 static int sctp_writeable(struct sock
*sk
)
7503 amt
= sk
->sk_sndbuf
- sk_wmem_alloc_get(sk
);
7509 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7510 * returns immediately with EINPROGRESS.
7512 static int sctp_wait_for_connect(struct sctp_association
*asoc
, long *timeo_p
)
7514 struct sock
*sk
= asoc
->base
.sk
;
7516 long current_timeo
= *timeo_p
;
7519 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__
, asoc
, *timeo_p
);
7521 /* Increment the association's refcnt. */
7522 sctp_association_hold(asoc
);
7525 prepare_to_wait_exclusive(&asoc
->wait
, &wait
,
7526 TASK_INTERRUPTIBLE
);
7529 if (sk
->sk_shutdown
& RCV_SHUTDOWN
)
7531 if (sk
->sk_err
|| asoc
->state
>= SCTP_STATE_SHUTDOWN_PENDING
||
7534 if (signal_pending(current
))
7535 goto do_interrupted
;
7537 if (sctp_state(asoc
, ESTABLISHED
))
7540 /* Let another process have a go. Since we are going
7544 current_timeo
= schedule_timeout(current_timeo
);
7547 *timeo_p
= current_timeo
;
7551 finish_wait(&asoc
->wait
, &wait
);
7553 /* Release the association's refcnt. */
7554 sctp_association_put(asoc
);
7559 if (asoc
->init_err_counter
+ 1 > asoc
->max_init_attempts
)
7562 err
= -ECONNREFUSED
;
7566 err
= sock_intr_errno(*timeo_p
);
7574 static int sctp_wait_for_accept(struct sock
*sk
, long timeo
)
7576 struct sctp_endpoint
*ep
;
7580 ep
= sctp_sk(sk
)->ep
;
7584 prepare_to_wait_exclusive(sk_sleep(sk
), &wait
,
7585 TASK_INTERRUPTIBLE
);
7587 if (list_empty(&ep
->asocs
)) {
7589 timeo
= schedule_timeout(timeo
);
7594 if (!sctp_sstate(sk
, LISTENING
))
7598 if (!list_empty(&ep
->asocs
))
7601 err
= sock_intr_errno(timeo
);
7602 if (signal_pending(current
))
7610 finish_wait(sk_sleep(sk
), &wait
);
7615 static void sctp_wait_for_close(struct sock
*sk
, long timeout
)
7620 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
7621 if (list_empty(&sctp_sk(sk
)->ep
->asocs
))
7624 timeout
= schedule_timeout(timeout
);
7626 } while (!signal_pending(current
) && timeout
);
7628 finish_wait(sk_sleep(sk
), &wait
);
7631 static void sctp_skb_set_owner_r_frag(struct sk_buff
*skb
, struct sock
*sk
)
7633 struct sk_buff
*frag
;
7638 /* Don't forget the fragments. */
7639 skb_walk_frags(skb
, frag
)
7640 sctp_skb_set_owner_r_frag(frag
, sk
);
7643 sctp_skb_set_owner_r(skb
, sk
);
7646 void sctp_copy_sock(struct sock
*newsk
, struct sock
*sk
,
7647 struct sctp_association
*asoc
)
7649 struct inet_sock
*inet
= inet_sk(sk
);
7650 struct inet_sock
*newinet
;
7652 newsk
->sk_type
= sk
->sk_type
;
7653 newsk
->sk_bound_dev_if
= sk
->sk_bound_dev_if
;
7654 newsk
->sk_flags
= sk
->sk_flags
;
7655 newsk
->sk_tsflags
= sk
->sk_tsflags
;
7656 newsk
->sk_no_check_tx
= sk
->sk_no_check_tx
;
7657 newsk
->sk_no_check_rx
= sk
->sk_no_check_rx
;
7658 newsk
->sk_reuse
= sk
->sk_reuse
;
7660 newsk
->sk_shutdown
= sk
->sk_shutdown
;
7661 newsk
->sk_destruct
= sctp_destruct_sock
;
7662 newsk
->sk_family
= sk
->sk_family
;
7663 newsk
->sk_protocol
= IPPROTO_SCTP
;
7664 newsk
->sk_backlog_rcv
= sk
->sk_prot
->backlog_rcv
;
7665 newsk
->sk_sndbuf
= sk
->sk_sndbuf
;
7666 newsk
->sk_rcvbuf
= sk
->sk_rcvbuf
;
7667 newsk
->sk_lingertime
= sk
->sk_lingertime
;
7668 newsk
->sk_rcvtimeo
= sk
->sk_rcvtimeo
;
7669 newsk
->sk_sndtimeo
= sk
->sk_sndtimeo
;
7670 newsk
->sk_rxhash
= sk
->sk_rxhash
;
7672 newinet
= inet_sk(newsk
);
7674 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7675 * getsockname() and getpeername()
7677 newinet
->inet_sport
= inet
->inet_sport
;
7678 newinet
->inet_saddr
= inet
->inet_saddr
;
7679 newinet
->inet_rcv_saddr
= inet
->inet_rcv_saddr
;
7680 newinet
->inet_dport
= htons(asoc
->peer
.port
);
7681 newinet
->pmtudisc
= inet
->pmtudisc
;
7682 newinet
->inet_id
= asoc
->next_tsn
^ jiffies
;
7684 newinet
->uc_ttl
= inet
->uc_ttl
;
7685 newinet
->mc_loop
= 1;
7686 newinet
->mc_ttl
= 1;
7687 newinet
->mc_index
= 0;
7688 newinet
->mc_list
= NULL
;
7690 if (newsk
->sk_flags
& SK_FLAGS_TIMESTAMP
)
7691 net_enable_timestamp();
7693 security_sk_clone(sk
, newsk
);
7696 static inline void sctp_copy_descendant(struct sock
*sk_to
,
7697 const struct sock
*sk_from
)
7699 int ancestor_size
= sizeof(struct inet_sock
) +
7700 sizeof(struct sctp_sock
) -
7701 offsetof(struct sctp_sock
, auto_asconf_list
);
7703 if (sk_from
->sk_family
== PF_INET6
)
7704 ancestor_size
+= sizeof(struct ipv6_pinfo
);
7706 __inet_sk_copy_descendant(sk_to
, sk_from
, ancestor_size
);
7709 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7710 * and its messages to the newsk.
7712 static void sctp_sock_migrate(struct sock
*oldsk
, struct sock
*newsk
,
7713 struct sctp_association
*assoc
,
7714 sctp_socket_type_t type
)
7716 struct sctp_sock
*oldsp
= sctp_sk(oldsk
);
7717 struct sctp_sock
*newsp
= sctp_sk(newsk
);
7718 struct sctp_bind_bucket
*pp
; /* hash list port iterator */
7719 struct sctp_endpoint
*newep
= newsp
->ep
;
7720 struct sk_buff
*skb
, *tmp
;
7721 struct sctp_ulpevent
*event
;
7722 struct sctp_bind_hashbucket
*head
;
7724 /* Migrate socket buffer sizes and all the socket level options to the
7727 newsk
->sk_sndbuf
= oldsk
->sk_sndbuf
;
7728 newsk
->sk_rcvbuf
= oldsk
->sk_rcvbuf
;
7729 /* Brute force copy old sctp opt. */
7730 sctp_copy_descendant(newsk
, oldsk
);
7732 /* Restore the ep value that was overwritten with the above structure
7738 /* Hook this new socket in to the bind_hash list. */
7739 head
= &sctp_port_hashtable
[sctp_phashfn(sock_net(oldsk
),
7740 inet_sk(oldsk
)->inet_num
)];
7741 spin_lock_bh(&head
->lock
);
7742 pp
= sctp_sk(oldsk
)->bind_hash
;
7743 sk_add_bind_node(newsk
, &pp
->owner
);
7744 sctp_sk(newsk
)->bind_hash
= pp
;
7745 inet_sk(newsk
)->inet_num
= inet_sk(oldsk
)->inet_num
;
7746 spin_unlock_bh(&head
->lock
);
7748 /* Copy the bind_addr list from the original endpoint to the new
7749 * endpoint so that we can handle restarts properly
7751 sctp_bind_addr_dup(&newsp
->ep
->base
.bind_addr
,
7752 &oldsp
->ep
->base
.bind_addr
, GFP_KERNEL
);
7754 /* Move any messages in the old socket's receive queue that are for the
7755 * peeled off association to the new socket's receive queue.
7757 sctp_skb_for_each(skb
, &oldsk
->sk_receive_queue
, tmp
) {
7758 event
= sctp_skb2event(skb
);
7759 if (event
->asoc
== assoc
) {
7760 __skb_unlink(skb
, &oldsk
->sk_receive_queue
);
7761 __skb_queue_tail(&newsk
->sk_receive_queue
, skb
);
7762 sctp_skb_set_owner_r_frag(skb
, newsk
);
7766 /* Clean up any messages pending delivery due to partial
7767 * delivery. Three cases:
7768 * 1) No partial deliver; no work.
7769 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7770 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7772 skb_queue_head_init(&newsp
->pd_lobby
);
7773 atomic_set(&sctp_sk(newsk
)->pd_mode
, assoc
->ulpq
.pd_mode
);
7775 if (atomic_read(&sctp_sk(oldsk
)->pd_mode
)) {
7776 struct sk_buff_head
*queue
;
7778 /* Decide which queue to move pd_lobby skbs to. */
7779 if (assoc
->ulpq
.pd_mode
) {
7780 queue
= &newsp
->pd_lobby
;
7782 queue
= &newsk
->sk_receive_queue
;
7784 /* Walk through the pd_lobby, looking for skbs that
7785 * need moved to the new socket.
7787 sctp_skb_for_each(skb
, &oldsp
->pd_lobby
, tmp
) {
7788 event
= sctp_skb2event(skb
);
7789 if (event
->asoc
== assoc
) {
7790 __skb_unlink(skb
, &oldsp
->pd_lobby
);
7791 __skb_queue_tail(queue
, skb
);
7792 sctp_skb_set_owner_r_frag(skb
, newsk
);
7796 /* Clear up any skbs waiting for the partial
7797 * delivery to finish.
7799 if (assoc
->ulpq
.pd_mode
)
7800 sctp_clear_pd(oldsk
, NULL
);
7804 sctp_skb_for_each(skb
, &assoc
->ulpq
.reasm
, tmp
)
7805 sctp_skb_set_owner_r_frag(skb
, newsk
);
7807 sctp_skb_for_each(skb
, &assoc
->ulpq
.lobby
, tmp
)
7808 sctp_skb_set_owner_r_frag(skb
, newsk
);
7810 /* Set the type of socket to indicate that it is peeled off from the
7811 * original UDP-style socket or created with the accept() call on a
7812 * TCP-style socket..
7816 /* Mark the new socket "in-use" by the user so that any packets
7817 * that may arrive on the association after we've moved it are
7818 * queued to the backlog. This prevents a potential race between
7819 * backlog processing on the old socket and new-packet processing
7820 * on the new socket.
7822 * The caller has just allocated newsk so we can guarantee that other
7823 * paths won't try to lock it and then oldsk.
7825 lock_sock_nested(newsk
, SINGLE_DEPTH_NESTING
);
7826 sctp_assoc_migrate(assoc
, newsk
);
7828 /* If the association on the newsk is already closed before accept()
7829 * is called, set RCV_SHUTDOWN flag.
7831 if (sctp_state(assoc
, CLOSED
) && sctp_style(newsk
, TCP
)) {
7832 newsk
->sk_state
= SCTP_SS_CLOSED
;
7833 newsk
->sk_shutdown
|= RCV_SHUTDOWN
;
7835 newsk
->sk_state
= SCTP_SS_ESTABLISHED
;
7838 release_sock(newsk
);
7842 /* This proto struct describes the ULP interface for SCTP. */
7843 struct proto sctp_prot
= {
7845 .owner
= THIS_MODULE
,
7846 .close
= sctp_close
,
7847 .connect
= sctp_connect
,
7848 .disconnect
= sctp_disconnect
,
7849 .accept
= sctp_accept
,
7850 .ioctl
= sctp_ioctl
,
7851 .init
= sctp_init_sock
,
7852 .destroy
= sctp_destroy_sock
,
7853 .shutdown
= sctp_shutdown
,
7854 .setsockopt
= sctp_setsockopt
,
7855 .getsockopt
= sctp_getsockopt
,
7856 .sendmsg
= sctp_sendmsg
,
7857 .recvmsg
= sctp_recvmsg
,
7859 .backlog_rcv
= sctp_backlog_rcv
,
7861 .unhash
= sctp_unhash
,
7862 .get_port
= sctp_get_port
,
7863 .obj_size
= sizeof(struct sctp_sock
),
7864 .sysctl_mem
= sysctl_sctp_mem
,
7865 .sysctl_rmem
= sysctl_sctp_rmem
,
7866 .sysctl_wmem
= sysctl_sctp_wmem
,
7867 .memory_pressure
= &sctp_memory_pressure
,
7868 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7869 .memory_allocated
= &sctp_memory_allocated
,
7870 .sockets_allocated
= &sctp_sockets_allocated
,
7873 #if IS_ENABLED(CONFIG_IPV6)
7875 #include <net/transp_v6.h>
7876 static void sctp_v6_destroy_sock(struct sock
*sk
)
7878 sctp_destroy_sock(sk
);
7879 inet6_destroy_sock(sk
);
7882 struct proto sctpv6_prot
= {
7884 .owner
= THIS_MODULE
,
7885 .close
= sctp_close
,
7886 .connect
= sctp_connect
,
7887 .disconnect
= sctp_disconnect
,
7888 .accept
= sctp_accept
,
7889 .ioctl
= sctp_ioctl
,
7890 .init
= sctp_init_sock
,
7891 .destroy
= sctp_v6_destroy_sock
,
7892 .shutdown
= sctp_shutdown
,
7893 .setsockopt
= sctp_setsockopt
,
7894 .getsockopt
= sctp_getsockopt
,
7895 .sendmsg
= sctp_sendmsg
,
7896 .recvmsg
= sctp_recvmsg
,
7898 .backlog_rcv
= sctp_backlog_rcv
,
7900 .unhash
= sctp_unhash
,
7901 .get_port
= sctp_get_port
,
7902 .obj_size
= sizeof(struct sctp6_sock
),
7903 .sysctl_mem
= sysctl_sctp_mem
,
7904 .sysctl_rmem
= sysctl_sctp_rmem
,
7905 .sysctl_wmem
= sysctl_sctp_wmem
,
7906 .memory_pressure
= &sctp_memory_pressure
,
7907 .enter_memory_pressure
= sctp_enter_memory_pressure
,
7908 .memory_allocated
= &sctp_memory_allocated
,
7909 .sockets_allocated
= &sctp_sockets_allocated
,
7911 #endif /* IS_ENABLED(CONFIG_IPV6) */