Linux 3.12.28
[linux/fpc-iii.git] / net / sctp / socket.c
blobe00a041129c21a631759a225c3e25f252c78bef3
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)
22 * any later version.
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <linux-sctp@vger.kernel.org>
39 * Written or modified by:
40 * La Monte H.P. Yarroll <piggy@acm.org>
41 * Narasimha Budihal <narsi@refcode.org>
42 * Karl Knutson <karl@athena.chicago.il.us>
43 * Jon Grimm <jgrimm@us.ibm.com>
44 * Xingang Guo <xingang.guo@intel.com>
45 * Daisy Chang <daisyc@us.ibm.com>
46 * Sridhar Samudrala <samudrala@us.ibm.com>
47 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
48 * Ardelle Fan <ardelle.fan@intel.com>
49 * Ryan Layer <rmlayer@us.ibm.com>
50 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
51 * Kevin Gao <kevin.gao@intel.com>
54 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
60 #include <linux/ip.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/crypto.h>
66 #include <linux/slab.h>
67 #include <linux/file.h>
68 #include <linux/compat.h>
70 #include <net/ip.h>
71 #include <net/icmp.h>
72 #include <net/route.h>
73 #include <net/ipv6.h>
74 #include <net/inet_common.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
78 #include <net/sock.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,
86 size_t msg_len);
87 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
105 extern struct kmem_cache *sctp_bucket_cachep;
106 extern long sysctl_sctp_mem[3];
107 extern int sysctl_sctp_rmem[3];
108 extern int sysctl_sctp_wmem[3];
110 static int sctp_memory_pressure;
111 static atomic_long_t sctp_memory_allocated;
112 struct percpu_counter sctp_sockets_allocated;
114 static void sctp_enter_memory_pressure(struct sock *sk)
116 sctp_memory_pressure = 1;
120 /* Get the sndbuf space available at the time on the association. */
121 static inline int sctp_wspace(struct sctp_association *asoc)
123 int amt;
125 if (asoc->ep->sndbuf_policy)
126 amt = asoc->sndbuf_used;
127 else
128 amt = sk_wmem_alloc_get(asoc->base.sk);
130 if (amt >= asoc->base.sk->sk_sndbuf) {
131 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
132 amt = 0;
133 else {
134 amt = sk_stream_wspace(asoc->base.sk);
135 if (amt < 0)
136 amt = 0;
138 } else {
139 amt = asoc->base.sk->sk_sndbuf - amt;
141 return amt;
144 /* Increment the used sndbuf space count of the corresponding association by
145 * the size of the outgoing data chunk.
146 * Also, set the skb destructor for sndbuf accounting later.
148 * Since it is always 1-1 between chunk and skb, and also a new skb is always
149 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
150 * destructor in the data chunk skb for the purpose of the sndbuf space
151 * tracking.
153 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
155 struct sctp_association *asoc = chunk->asoc;
156 struct sock *sk = asoc->base.sk;
158 /* The sndbuf space is tracked per association. */
159 sctp_association_hold(asoc);
161 skb_set_owner_w(chunk->skb, sk);
163 chunk->skb->destructor = sctp_wfree;
164 /* Save the chunk pointer in skb for sctp_wfree to use later. */
165 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
167 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
168 sizeof(struct sk_buff) +
169 sizeof(struct sctp_chunk);
171 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
172 sk->sk_wmem_queued += chunk->skb->truesize;
173 sk_mem_charge(sk, chunk->skb->truesize);
176 /* Verify that this is a valid address. */
177 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
178 int len)
180 struct sctp_af *af;
182 /* Verify basic sockaddr. */
183 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
184 if (!af)
185 return -EINVAL;
187 /* Is this a valid SCTP address? */
188 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
189 return -EINVAL;
191 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
192 return -EINVAL;
194 return 0;
197 /* Look up the association by its id. If this is not a UDP-style
198 * socket, the ID field is always ignored.
200 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
202 struct sctp_association *asoc = NULL;
204 /* If this is not a UDP-style socket, assoc id should be ignored. */
205 if (!sctp_style(sk, UDP)) {
206 /* Return NULL if the socket state is not ESTABLISHED. It
207 * could be a TCP-style listening socket or a socket which
208 * hasn't yet called connect() to establish an association.
210 if (!sctp_sstate(sk, ESTABLISHED))
211 return NULL;
213 /* Get the first and the only association from the list. */
214 if (!list_empty(&sctp_sk(sk)->ep->asocs))
215 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
216 struct sctp_association, asocs);
217 return asoc;
220 /* Otherwise this is a UDP-style socket. */
221 if (!id || (id == (sctp_assoc_t)-1))
222 return NULL;
224 spin_lock_bh(&sctp_assocs_id_lock);
225 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
226 spin_unlock_bh(&sctp_assocs_id_lock);
228 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
229 return NULL;
231 return asoc;
234 /* Look up the transport from an address and an assoc id. If both address and
235 * id are specified, the associations matching the address and the id should be
236 * the same.
238 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
239 struct sockaddr_storage *addr,
240 sctp_assoc_t id)
242 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
243 struct sctp_transport *transport;
244 union sctp_addr *laddr = (union sctp_addr *)addr;
246 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
247 laddr,
248 &transport);
250 if (!addr_asoc)
251 return NULL;
253 id_asoc = sctp_id2assoc(sk, id);
254 if (id_asoc && (id_asoc != addr_asoc))
255 return NULL;
257 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
258 (union sctp_addr *)addr);
260 return transport;
263 /* API 3.1.2 bind() - UDP Style Syntax
264 * The syntax of bind() is,
266 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
268 * sd - the socket descriptor returned by socket().
269 * addr - the address structure (struct sockaddr_in or struct
270 * sockaddr_in6 [RFC 2553]),
271 * addr_len - the size of the address structure.
273 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
275 int retval = 0;
277 sctp_lock_sock(sk);
279 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
280 addr, addr_len);
282 /* Disallow binding twice. */
283 if (!sctp_sk(sk)->ep->base.bind_addr.port)
284 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
285 addr_len);
286 else
287 retval = -EINVAL;
289 sctp_release_sock(sk);
291 return retval;
294 static long sctp_get_port_local(struct sock *, union sctp_addr *);
296 /* Verify this is a valid sockaddr. */
297 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
298 union sctp_addr *addr, int len)
300 struct sctp_af *af;
302 /* Check minimum size. */
303 if (len < sizeof (struct sockaddr))
304 return NULL;
306 /* V4 mapped address are really of AF_INET family */
307 if (addr->sa.sa_family == AF_INET6 &&
308 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
309 if (!opt->pf->af_supported(AF_INET, opt))
310 return NULL;
311 } else {
312 /* Does this PF support this AF? */
313 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
314 return NULL;
317 /* If we get this far, af is valid. */
318 af = sctp_get_af_specific(addr->sa.sa_family);
320 if (len < af->sockaddr_len)
321 return NULL;
323 return af;
326 /* Bind a local address either to an endpoint or to an association. */
327 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
329 struct net *net = sock_net(sk);
330 struct sctp_sock *sp = sctp_sk(sk);
331 struct sctp_endpoint *ep = sp->ep;
332 struct sctp_bind_addr *bp = &ep->base.bind_addr;
333 struct sctp_af *af;
334 unsigned short snum;
335 int ret = 0;
337 /* Common sockaddr verification. */
338 af = sctp_sockaddr_af(sp, addr, len);
339 if (!af) {
340 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
341 __func__, sk, addr, len);
342 return -EINVAL;
345 snum = ntohs(addr->v4.sin_port);
347 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
348 __func__, sk, &addr->sa, bp->port, snum, len);
350 /* PF specific bind() address verification. */
351 if (!sp->pf->bind_verify(sp, addr))
352 return -EADDRNOTAVAIL;
354 /* We must either be unbound, or bind to the same port.
355 * It's OK to allow 0 ports if we are already bound.
356 * We'll just inhert an already bound port in this case
358 if (bp->port) {
359 if (!snum)
360 snum = bp->port;
361 else if (snum != bp->port) {
362 pr_debug("%s: new port %d doesn't match existing port "
363 "%d\n", __func__, snum, bp->port);
364 return -EINVAL;
368 if (snum && snum < PROT_SOCK &&
369 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
370 return -EACCES;
372 /* See if the address matches any of the addresses we may have
373 * already bound before checking against other endpoints.
375 if (sctp_bind_addr_match(bp, addr, sp))
376 return -EINVAL;
378 /* Make sure we are allowed to bind here.
379 * The function sctp_get_port_local() does duplicate address
380 * detection.
382 addr->v4.sin_port = htons(snum);
383 if ((ret = sctp_get_port_local(sk, addr))) {
384 return -EADDRINUSE;
387 /* Refresh ephemeral port. */
388 if (!bp->port)
389 bp->port = inet_sk(sk)->inet_num;
391 /* Add the address to the bind address list.
392 * Use GFP_ATOMIC since BHs will be disabled.
394 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
396 /* Copy back into socket for getsockname() use. */
397 if (!ret) {
398 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
399 af->to_sk_saddr(addr, sk);
402 return ret;
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);
419 int retval = 0;
421 /* If there is an outstanding ASCONF chunk, queue it for later
422 * transmission.
424 if (asoc->addip_last_asconf) {
425 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
426 goto out;
429 /* Hold the chunk until an ASCONF_ACK is received. */
430 sctp_chunk_hold(chunk);
431 retval = sctp_primitive_ASCONF(net, asoc, chunk);
432 if (retval)
433 sctp_chunk_free(chunk);
434 else
435 asoc->addip_last_asconf = chunk;
437 out:
438 return retval;
441 /* Add a list of addresses as bind addresses to local endpoint or
442 * association.
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)
455 int cnt;
456 int retval = 0;
457 void *addr_buf;
458 struct sockaddr *sa_addr;
459 struct sctp_af *af;
461 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
462 addrs, addrcnt);
464 addr_buf = addrs;
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.
469 sa_addr = addr_buf;
470 af = sctp_get_af_specific(sa_addr->sa_family);
471 if (!af) {
472 retval = -EINVAL;
473 goto err_bindx_add;
476 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
477 af->sockaddr_len);
479 addr_buf += af->sockaddr_len;
481 err_bindx_add:
482 if (retval < 0) {
483 /* Failed. Cleanup the ones that have been added */
484 if (cnt > 0)
485 sctp_bindx_rem(sk, addrs, cnt);
486 return retval;
490 return retval;
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,
505 int addrcnt)
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;
516 void *addr_buf;
517 struct sctp_af *af;
518 struct list_head *p;
519 int i;
520 int retval = 0;
522 if (!net->sctp.addip_enable)
523 return retval;
525 sp = sctp_sk(sk);
526 ep = sp->ep;
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)
533 continue;
535 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
536 continue;
538 if (!sctp_state(asoc, ESTABLISHED))
539 continue;
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.
546 addr_buf = addrs;
547 for (i = 0; i < addrcnt; i++) {
548 addr = addr_buf;
549 af = sctp_get_af_specific(addr->v4.sin_family);
550 if (!af) {
551 retval = -EINVAL;
552 goto out;
555 if (sctp_assoc_lookup_laddr(asoc, addr))
556 break;
558 addr_buf += af->sockaddr_len;
560 if (i < addrcnt)
561 continue;
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);
571 if (!chunk) {
572 retval = -ENOMEM;
573 goto out;
576 /* Add the new addresses to the bind address list with
577 * use_as_src set to 0.
579 addr_buf = addrs;
580 for (i = 0; i < addrcnt; i++) {
581 addr = addr_buf;
582 af = sctp_get_af_specific(addr->v4.sin_family);
583 memcpy(&saveaddr, addr, af->sockaddr_len);
584 retval = sctp_add_bind_addr(bp, &saveaddr,
585 SCTP_ADDR_NEW, GFP_ATOMIC);
586 addr_buf += af->sockaddr_len;
588 if (asoc->src_out_of_asoc_ok) {
589 struct sctp_transport *trans;
591 list_for_each_entry(trans,
592 &asoc->peer.transport_addr_list, transports) {
593 /* Clear the source and route cache */
594 dst_release(trans->dst);
595 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
596 2*asoc->pathmtu, 4380));
597 trans->ssthresh = asoc->peer.i.a_rwnd;
598 trans->rto = asoc->rto_initial;
599 sctp_max_rto(asoc, trans);
600 trans->rtt = trans->srtt = trans->rttvar = 0;
601 sctp_transport_route(trans, NULL,
602 sctp_sk(asoc->base.sk));
605 retval = sctp_send_asconf(asoc, chunk);
608 out:
609 return retval;
612 /* Remove a list of addresses from bind addresses list. Do not remove the
613 * last address.
615 * Basically run through each address specified in the addrs/addrcnt
616 * array/length pair, determine if it is IPv6 or IPv4 and call
617 * sctp_del_bind() on it.
619 * If any of them fails, then the operation will be reversed and the
620 * ones that were removed will be added back.
622 * At least one address has to be left; if only one address is
623 * available, the operation will return -EBUSY.
625 * Only sctp_setsockopt_bindx() is supposed to call this function.
627 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
629 struct sctp_sock *sp = sctp_sk(sk);
630 struct sctp_endpoint *ep = sp->ep;
631 int cnt;
632 struct sctp_bind_addr *bp = &ep->base.bind_addr;
633 int retval = 0;
634 void *addr_buf;
635 union sctp_addr *sa_addr;
636 struct sctp_af *af;
638 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
639 __func__, sk, addrs, addrcnt);
641 addr_buf = addrs;
642 for (cnt = 0; cnt < addrcnt; cnt++) {
643 /* If the bind address list is empty or if there is only one
644 * bind address, there is nothing more to be removed (we need
645 * at least one address here).
647 if (list_empty(&bp->address_list) ||
648 (sctp_list_single_entry(&bp->address_list))) {
649 retval = -EBUSY;
650 goto err_bindx_rem;
653 sa_addr = addr_buf;
654 af = sctp_get_af_specific(sa_addr->sa.sa_family);
655 if (!af) {
656 retval = -EINVAL;
657 goto err_bindx_rem;
660 if (!af->addr_valid(sa_addr, sp, NULL)) {
661 retval = -EADDRNOTAVAIL;
662 goto err_bindx_rem;
665 if (sa_addr->v4.sin_port &&
666 sa_addr->v4.sin_port != htons(bp->port)) {
667 retval = -EINVAL;
668 goto err_bindx_rem;
671 if (!sa_addr->v4.sin_port)
672 sa_addr->v4.sin_port = htons(bp->port);
674 /* FIXME - There is probably a need to check if sk->sk_saddr and
675 * sk->sk_rcv_addr are currently set to one of the addresses to
676 * be removed. This is something which needs to be looked into
677 * when we are fixing the outstanding issues with multi-homing
678 * socket routing and failover schemes. Refer to comments in
679 * sctp_do_bind(). -daisy
681 retval = sctp_del_bind_addr(bp, sa_addr);
683 addr_buf += af->sockaddr_len;
684 err_bindx_rem:
685 if (retval < 0) {
686 /* Failed. Add the ones that has been removed back */
687 if (cnt > 0)
688 sctp_bindx_add(sk, addrs, cnt);
689 return retval;
693 return retval;
696 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
697 * the associations that are part of the endpoint indicating that a list of
698 * local addresses are removed from the endpoint.
700 * If any of the addresses is already in the bind address list of the
701 * association, we do not send the chunk for that association. But it will not
702 * affect other associations.
704 * Only sctp_setsockopt_bindx() is supposed to call this function.
706 static int sctp_send_asconf_del_ip(struct sock *sk,
707 struct sockaddr *addrs,
708 int addrcnt)
710 struct net *net = sock_net(sk);
711 struct sctp_sock *sp;
712 struct sctp_endpoint *ep;
713 struct sctp_association *asoc;
714 struct sctp_transport *transport;
715 struct sctp_bind_addr *bp;
716 struct sctp_chunk *chunk;
717 union sctp_addr *laddr;
718 void *addr_buf;
719 struct sctp_af *af;
720 struct sctp_sockaddr_entry *saddr;
721 int i;
722 int retval = 0;
723 int stored = 0;
725 chunk = NULL;
726 if (!net->sctp.addip_enable)
727 return retval;
729 sp = sctp_sk(sk);
730 ep = sp->ep;
732 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
733 __func__, sk, addrs, addrcnt);
735 list_for_each_entry(asoc, &ep->asocs, asocs) {
737 if (!asoc->peer.asconf_capable)
738 continue;
740 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
741 continue;
743 if (!sctp_state(asoc, ESTABLISHED))
744 continue;
746 /* Check if any address in the packed array of addresses is
747 * not present in the bind address list of the association.
748 * If so, do not send the asconf chunk to its peer, but
749 * continue with other associations.
751 addr_buf = addrs;
752 for (i = 0; i < addrcnt; i++) {
753 laddr = addr_buf;
754 af = sctp_get_af_specific(laddr->v4.sin_family);
755 if (!af) {
756 retval = -EINVAL;
757 goto out;
760 if (!sctp_assoc_lookup_laddr(asoc, laddr))
761 break;
763 addr_buf += af->sockaddr_len;
765 if (i < addrcnt)
766 continue;
768 /* Find one address in the association's bind address list
769 * that is not in the packed array of addresses. This is to
770 * make sure that we do not delete all the addresses in the
771 * association.
773 bp = &asoc->base.bind_addr;
774 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
775 addrcnt, sp);
776 if ((laddr == NULL) && (addrcnt == 1)) {
777 if (asoc->asconf_addr_del_pending)
778 continue;
779 asoc->asconf_addr_del_pending =
780 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
781 if (asoc->asconf_addr_del_pending == NULL) {
782 retval = -ENOMEM;
783 goto out;
785 asoc->asconf_addr_del_pending->sa.sa_family =
786 addrs->sa_family;
787 asoc->asconf_addr_del_pending->v4.sin_port =
788 htons(bp->port);
789 if (addrs->sa_family == AF_INET) {
790 struct sockaddr_in *sin;
792 sin = (struct sockaddr_in *)addrs;
793 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
794 } else if (addrs->sa_family == AF_INET6) {
795 struct sockaddr_in6 *sin6;
797 sin6 = (struct sockaddr_in6 *)addrs;
798 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
801 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
802 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
803 asoc->asconf_addr_del_pending);
805 asoc->src_out_of_asoc_ok = 1;
806 stored = 1;
807 goto skip_mkasconf;
810 if (laddr == NULL)
811 return -EINVAL;
813 /* We do not need RCU protection throughout this loop
814 * because this is done under a socket lock from the
815 * setsockopt call.
817 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
818 SCTP_PARAM_DEL_IP);
819 if (!chunk) {
820 retval = -ENOMEM;
821 goto out;
824 skip_mkasconf:
825 /* Reset use_as_src flag for the addresses in the bind address
826 * list that are to be deleted.
828 addr_buf = addrs;
829 for (i = 0; i < addrcnt; i++) {
830 laddr = addr_buf;
831 af = sctp_get_af_specific(laddr->v4.sin_family);
832 list_for_each_entry(saddr, &bp->address_list, list) {
833 if (sctp_cmp_addr_exact(&saddr->a, laddr))
834 saddr->state = SCTP_ADDR_DEL;
836 addr_buf += af->sockaddr_len;
839 /* Update the route and saddr entries for all the transports
840 * as some of the addresses in the bind address list are
841 * about to be deleted and cannot be used as source addresses.
843 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
844 transports) {
845 dst_release(transport->dst);
846 sctp_transport_route(transport, NULL,
847 sctp_sk(asoc->base.sk));
850 if (stored)
851 /* We don't need to transmit ASCONF */
852 continue;
853 retval = sctp_send_asconf(asoc, chunk);
855 out:
856 return retval;
859 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
860 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
862 struct sock *sk = sctp_opt2sk(sp);
863 union sctp_addr *addr;
864 struct sctp_af *af;
866 /* It is safe to write port space in caller. */
867 addr = &addrw->a;
868 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
869 af = sctp_get_af_specific(addr->sa.sa_family);
870 if (!af)
871 return -EINVAL;
872 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
873 return -EINVAL;
875 if (addrw->state == SCTP_ADDR_NEW)
876 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
877 else
878 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
881 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
883 * API 8.1
884 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
885 * int flags);
887 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
888 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
889 * or IPv6 addresses.
891 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
892 * Section 3.1.2 for this usage.
894 * addrs is a pointer to an array of one or more socket addresses. Each
895 * address is contained in its appropriate structure (i.e. struct
896 * sockaddr_in or struct sockaddr_in6) the family of the address type
897 * must be used to distinguish the address length (note that this
898 * representation is termed a "packed array" of addresses). The caller
899 * specifies the number of addresses in the array with addrcnt.
901 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
902 * -1, and sets errno to the appropriate error code.
904 * For SCTP, the port given in each socket address must be the same, or
905 * sctp_bindx() will fail, setting errno to EINVAL.
907 * The flags parameter is formed from the bitwise OR of zero or more of
908 * the following currently defined flags:
910 * SCTP_BINDX_ADD_ADDR
912 * SCTP_BINDX_REM_ADDR
914 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
915 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
916 * addresses from the association. The two flags are mutually exclusive;
917 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
918 * not remove all addresses from an association; sctp_bindx() will
919 * reject such an attempt with EINVAL.
921 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
922 * additional addresses with an endpoint after calling bind(). Or use
923 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
924 * socket is associated with so that no new association accepted will be
925 * associated with those addresses. If the endpoint supports dynamic
926 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
927 * endpoint to send the appropriate message to the peer to change the
928 * peers address lists.
930 * Adding and removing addresses from a connected association is
931 * optional functionality. Implementations that do not support this
932 * functionality should return EOPNOTSUPP.
934 * Basically do nothing but copying the addresses from user to kernel
935 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
936 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
937 * from userspace.
939 * We don't use copy_from_user() for optimization: we first do the
940 * sanity checks (buffer size -fast- and access check-healthy
941 * pointer); if all of those succeed, then we can alloc the memory
942 * (expensive operation) needed to copy the data to kernel. Then we do
943 * the copying without checking the user space area
944 * (__copy_from_user()).
946 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
947 * it.
949 * sk The sk of the socket
950 * addrs The pointer to the addresses in user land
951 * addrssize Size of the addrs buffer
952 * op Operation to perform (add or remove, see the flags of
953 * sctp_bindx)
955 * Returns 0 if ok, <0 errno code on error.
957 static int sctp_setsockopt_bindx(struct sock* sk,
958 struct sockaddr __user *addrs,
959 int addrs_size, int op)
961 struct sockaddr *kaddrs;
962 int err;
963 int addrcnt = 0;
964 int walk_size = 0;
965 struct sockaddr *sa_addr;
966 void *addr_buf;
967 struct sctp_af *af;
969 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
970 __func__, sk, addrs, addrs_size, op);
972 if (unlikely(addrs_size <= 0))
973 return -EINVAL;
975 /* Check the user passed a healthy pointer. */
976 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
977 return -EFAULT;
979 /* Alloc space for the address array in kernel memory. */
980 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
981 if (unlikely(!kaddrs))
982 return -ENOMEM;
984 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
985 kfree(kaddrs);
986 return -EFAULT;
989 /* Walk through the addrs buffer and count the number of addresses. */
990 addr_buf = kaddrs;
991 while (walk_size < addrs_size) {
992 if (walk_size + sizeof(sa_family_t) > addrs_size) {
993 kfree(kaddrs);
994 return -EINVAL;
997 sa_addr = addr_buf;
998 af = sctp_get_af_specific(sa_addr->sa_family);
1000 /* If the address family is not supported or if this address
1001 * causes the address buffer to overflow return EINVAL.
1003 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1004 kfree(kaddrs);
1005 return -EINVAL;
1007 addrcnt++;
1008 addr_buf += af->sockaddr_len;
1009 walk_size += af->sockaddr_len;
1012 /* Do the work. */
1013 switch (op) {
1014 case SCTP_BINDX_ADD_ADDR:
1015 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1016 if (err)
1017 goto out;
1018 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1019 break;
1021 case SCTP_BINDX_REM_ADDR:
1022 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1023 if (err)
1024 goto out;
1025 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1026 break;
1028 default:
1029 err = -EINVAL;
1030 break;
1033 out:
1034 kfree(kaddrs);
1036 return err;
1039 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1041 * Common routine for handling connect() and sctp_connectx().
1042 * Connect will come in with just a single address.
1044 static int __sctp_connect(struct sock* sk,
1045 struct sockaddr *kaddrs,
1046 int addrs_size,
1047 sctp_assoc_t *assoc_id)
1049 struct net *net = sock_net(sk);
1050 struct sctp_sock *sp;
1051 struct sctp_endpoint *ep;
1052 struct sctp_association *asoc = NULL;
1053 struct sctp_association *asoc2;
1054 struct sctp_transport *transport;
1055 union sctp_addr to;
1056 struct sctp_af *af;
1057 sctp_scope_t scope;
1058 long timeo;
1059 int err = 0;
1060 int addrcnt = 0;
1061 int walk_size = 0;
1062 union sctp_addr *sa_addr = NULL;
1063 void *addr_buf;
1064 unsigned short port;
1065 unsigned int f_flags = 0;
1067 sp = sctp_sk(sk);
1068 ep = sp->ep;
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) ||
1076 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1077 err = -EISCONN;
1078 goto out_free;
1081 /* Walk through the addrs buffer and count the number of addresses. */
1082 addr_buf = kaddrs;
1083 while (walk_size < addrs_size) {
1084 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1085 err = -EINVAL;
1086 goto out_free;
1089 sa_addr = addr_buf;
1090 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1092 /* If the address family is not supported or if this address
1093 * causes the address buffer to overflow return EINVAL.
1095 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1096 err = -EINVAL;
1097 goto out_free;
1100 port = ntohs(sa_addr->v4.sin_port);
1102 /* Save current address so we can work with it */
1103 memcpy(&to, sa_addr, af->sockaddr_len);
1105 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1106 if (err)
1107 goto out_free;
1109 /* Make sure the destination port is correctly set
1110 * in all addresses.
1112 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1113 err = -EINVAL;
1114 goto out_free;
1117 /* Check if there already is a matching association on the
1118 * endpoint (other than the one created here).
1120 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1121 if (asoc2 && asoc2 != asoc) {
1122 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1123 err = -EISCONN;
1124 else
1125 err = -EALREADY;
1126 goto out_free;
1129 /* If we could not find a matching association on the endpoint,
1130 * make sure that there is no peeled-off association matching
1131 * the peer address even on another socket.
1133 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1134 err = -EADDRNOTAVAIL;
1135 goto out_free;
1138 if (!asoc) {
1139 /* If a bind() or sctp_bindx() is not called prior to
1140 * an sctp_connectx() call, the system picks an
1141 * ephemeral port and will choose an address set
1142 * equivalent to binding with a wildcard address.
1144 if (!ep->base.bind_addr.port) {
1145 if (sctp_autobind(sk)) {
1146 err = -EAGAIN;
1147 goto out_free;
1149 } else {
1151 * If an unprivileged user inherits a 1-many
1152 * style socket with open associations on a
1153 * privileged port, it MAY be permitted to
1154 * accept new associations, but it SHOULD NOT
1155 * be permitted to open new associations.
1157 if (ep->base.bind_addr.port < PROT_SOCK &&
1158 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1159 err = -EACCES;
1160 goto out_free;
1164 scope = sctp_scope(&to);
1165 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1166 if (!asoc) {
1167 err = -ENOMEM;
1168 goto out_free;
1171 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1172 GFP_KERNEL);
1173 if (err < 0) {
1174 goto out_free;
1179 /* Prime the peer's transport structures. */
1180 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1181 SCTP_UNKNOWN);
1182 if (!transport) {
1183 err = -ENOMEM;
1184 goto out_free;
1187 addrcnt++;
1188 addr_buf += af->sockaddr_len;
1189 walk_size += af->sockaddr_len;
1192 /* In case the user of sctp_connectx() wants an association
1193 * id back, assign one now.
1195 if (assoc_id) {
1196 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1197 if (err < 0)
1198 goto out_free;
1201 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1202 if (err < 0) {
1203 goto out_free;
1206 /* Initialize sk's dport and daddr for getpeername() */
1207 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1208 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1209 af->to_sk_daddr(sa_addr, sk);
1210 sk->sk_err = 0;
1212 /* in-kernel sockets don't generally have a file allocated to them
1213 * if all they do is call sock_create_kern().
1215 if (sk->sk_socket->file)
1216 f_flags = sk->sk_socket->file->f_flags;
1218 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1220 err = sctp_wait_for_connect(asoc, &timeo);
1221 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1222 *assoc_id = asoc->assoc_id;
1224 /* Don't free association on exit. */
1225 asoc = NULL;
1227 out_free:
1228 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1229 __func__, asoc, kaddrs, err);
1231 if (asoc) {
1232 /* sctp_primitive_ASSOCIATE may have added this association
1233 * To the hash table, try to unhash it, just in case, its a noop
1234 * if it wasn't hashed so we're safe
1236 sctp_unhash_established(asoc);
1237 sctp_association_free(asoc);
1239 return err;
1242 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1244 * API 8.9
1245 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1246 * sctp_assoc_t *asoc);
1248 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1249 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1250 * or IPv6 addresses.
1252 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1253 * Section 3.1.2 for this usage.
1255 * addrs is a pointer to an array of one or more socket addresses. Each
1256 * address is contained in its appropriate structure (i.e. struct
1257 * sockaddr_in or struct sockaddr_in6) the family of the address type
1258 * must be used to distengish the address length (note that this
1259 * representation is termed a "packed array" of addresses). The caller
1260 * specifies the number of addresses in the array with addrcnt.
1262 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1263 * the association id of the new association. On failure, sctp_connectx()
1264 * returns -1, and sets errno to the appropriate error code. The assoc_id
1265 * is not touched by the kernel.
1267 * For SCTP, the port given in each socket address must be the same, or
1268 * sctp_connectx() will fail, setting errno to EINVAL.
1270 * An application can use sctp_connectx to initiate an association with
1271 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1272 * allows a caller to specify multiple addresses at which a peer can be
1273 * reached. The way the SCTP stack uses the list of addresses to set up
1274 * the association is implementation dependent. This function only
1275 * specifies that the stack will try to make use of all the addresses in
1276 * the list when needed.
1278 * Note that the list of addresses passed in is only used for setting up
1279 * the association. It does not necessarily equal the set of addresses
1280 * the peer uses for the resulting association. If the caller wants to
1281 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1282 * retrieve them after the association has been set up.
1284 * Basically do nothing but copying the addresses from user to kernel
1285 * land and invoking either sctp_connectx(). This is used for tunneling
1286 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1288 * We don't use copy_from_user() for optimization: we first do the
1289 * sanity checks (buffer size -fast- and access check-healthy
1290 * pointer); if all of those succeed, then we can alloc the memory
1291 * (expensive operation) needed to copy the data to kernel. Then we do
1292 * the copying without checking the user space area
1293 * (__copy_from_user()).
1295 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1296 * it.
1298 * sk The sk of the socket
1299 * addrs The pointer to the addresses in user land
1300 * addrssize Size of the addrs buffer
1302 * Returns >=0 if ok, <0 errno code on error.
1304 static int __sctp_setsockopt_connectx(struct sock* sk,
1305 struct sockaddr __user *addrs,
1306 int addrs_size,
1307 sctp_assoc_t *assoc_id)
1309 int err = 0;
1310 struct sockaddr *kaddrs;
1312 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1313 __func__, sk, addrs, addrs_size);
1315 if (unlikely(addrs_size <= 0))
1316 return -EINVAL;
1318 /* Check the user passed a healthy pointer. */
1319 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1320 return -EFAULT;
1322 /* Alloc space for the address array in kernel memory. */
1323 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1324 if (unlikely(!kaddrs))
1325 return -ENOMEM;
1327 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1328 err = -EFAULT;
1329 } else {
1330 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1333 kfree(kaddrs);
1335 return err;
1339 * This is an older interface. It's kept for backward compatibility
1340 * to the option that doesn't provide association id.
1342 static int sctp_setsockopt_connectx_old(struct sock* sk,
1343 struct sockaddr __user *addrs,
1344 int addrs_size)
1346 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1350 * New interface for the API. The since the API is done with a socket
1351 * option, to make it simple we feed back the association id is as a return
1352 * indication to the call. Error is always negative and association id is
1353 * always positive.
1355 static int sctp_setsockopt_connectx(struct sock* sk,
1356 struct sockaddr __user *addrs,
1357 int addrs_size)
1359 sctp_assoc_t assoc_id = 0;
1360 int err = 0;
1362 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1364 if (err)
1365 return err;
1366 else
1367 return assoc_id;
1371 * New (hopefully final) interface for the API.
1372 * We use the sctp_getaddrs_old structure so that use-space library
1373 * can avoid any unnecessary allocations. The only different part
1374 * is that we store the actual length of the address buffer into the
1375 * addrs_num structure member. That way we can re-use the existing
1376 * code.
1378 #ifdef CONFIG_COMPAT
1379 struct compat_sctp_getaddrs_old {
1380 sctp_assoc_t assoc_id;
1381 s32 addr_num;
1382 compat_uptr_t addrs; /* struct sockaddr * */
1384 #endif
1386 static int sctp_getsockopt_connectx3(struct sock* sk, int len,
1387 char __user *optval,
1388 int __user *optlen)
1390 struct sctp_getaddrs_old param;
1391 sctp_assoc_t assoc_id = 0;
1392 int err = 0;
1394 #ifdef CONFIG_COMPAT
1395 if (is_compat_task()) {
1396 struct compat_sctp_getaddrs_old param32;
1398 if (len < sizeof(param32))
1399 return -EINVAL;
1400 if (copy_from_user(&param32, optval, sizeof(param32)))
1401 return -EFAULT;
1403 param.assoc_id = param32.assoc_id;
1404 param.addr_num = param32.addr_num;
1405 param.addrs = compat_ptr(param32.addrs);
1406 } else
1407 #endif
1409 if (len < sizeof(param))
1410 return -EINVAL;
1411 if (copy_from_user(&param, optval, sizeof(param)))
1412 return -EFAULT;
1415 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1416 param.addrs, param.addr_num,
1417 &assoc_id);
1418 if (err == 0 || err == -EINPROGRESS) {
1419 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1420 return -EFAULT;
1421 if (put_user(sizeof(assoc_id), optlen))
1422 return -EFAULT;
1425 return err;
1428 /* API 3.1.4 close() - UDP Style Syntax
1429 * Applications use close() to perform graceful shutdown (as described in
1430 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1431 * by a UDP-style socket.
1433 * The syntax is
1435 * ret = close(int sd);
1437 * sd - the socket descriptor of the associations to be closed.
1439 * To gracefully shutdown a specific association represented by the
1440 * UDP-style socket, an application should use the sendmsg() call,
1441 * passing no user data, but including the appropriate flag in the
1442 * ancillary data (see Section xxxx).
1444 * If sd in the close() call is a branched-off socket representing only
1445 * one association, the shutdown is performed on that association only.
1447 * 4.1.6 close() - TCP Style Syntax
1449 * Applications use close() to gracefully close down an association.
1451 * The syntax is:
1453 * int close(int sd);
1455 * sd - the socket descriptor of the association to be closed.
1457 * After an application calls close() on a socket descriptor, no further
1458 * socket operations will succeed on that descriptor.
1460 * API 7.1.4 SO_LINGER
1462 * An application using the TCP-style socket can use this option to
1463 * perform the SCTP ABORT primitive. The linger option structure is:
1465 * struct linger {
1466 * int l_onoff; // option on/off
1467 * int l_linger; // linger time
1468 * };
1470 * To enable the option, set l_onoff to 1. If the l_linger value is set
1471 * to 0, calling close() is the same as the ABORT primitive. If the
1472 * value is set to a negative value, the setsockopt() call will return
1473 * an error. If the value is set to a positive value linger_time, the
1474 * close() can be blocked for at most linger_time ms. If the graceful
1475 * shutdown phase does not finish during this period, close() will
1476 * return but the graceful shutdown phase continues in the system.
1478 static void sctp_close(struct sock *sk, long timeout)
1480 struct net *net = sock_net(sk);
1481 struct sctp_endpoint *ep;
1482 struct sctp_association *asoc;
1483 struct list_head *pos, *temp;
1484 unsigned int data_was_unread;
1486 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1488 sctp_lock_sock(sk);
1489 sk->sk_shutdown = SHUTDOWN_MASK;
1490 sk->sk_state = SCTP_SS_CLOSING;
1492 ep = sctp_sk(sk)->ep;
1494 /* Clean up any skbs sitting on the receive queue. */
1495 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1496 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1498 /* Walk all associations on an endpoint. */
1499 list_for_each_safe(pos, temp, &ep->asocs) {
1500 asoc = list_entry(pos, struct sctp_association, asocs);
1502 if (sctp_style(sk, TCP)) {
1503 /* A closed association can still be in the list if
1504 * it belongs to a TCP-style listening socket that is
1505 * not yet accepted. If so, free it. If not, send an
1506 * ABORT or SHUTDOWN based on the linger options.
1508 if (sctp_state(asoc, CLOSED)) {
1509 sctp_unhash_established(asoc);
1510 sctp_association_free(asoc);
1511 continue;
1515 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1516 !skb_queue_empty(&asoc->ulpq.reasm) ||
1517 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1518 struct sctp_chunk *chunk;
1520 chunk = sctp_make_abort_user(asoc, NULL, 0);
1521 if (chunk)
1522 sctp_primitive_ABORT(net, asoc, chunk);
1523 } else
1524 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1527 /* On a TCP-style socket, block for at most linger_time if set. */
1528 if (sctp_style(sk, TCP) && timeout)
1529 sctp_wait_for_close(sk, timeout);
1531 /* This will run the backlog queue. */
1532 sctp_release_sock(sk);
1534 /* Supposedly, no process has access to the socket, but
1535 * the net layers still may.
1537 sctp_local_bh_disable();
1538 sctp_bh_lock_sock(sk);
1540 /* Hold the sock, since sk_common_release() will put sock_put()
1541 * and we have just a little more cleanup.
1543 sock_hold(sk);
1544 sk_common_release(sk);
1546 sctp_bh_unlock_sock(sk);
1547 sctp_local_bh_enable();
1549 sock_put(sk);
1551 SCTP_DBG_OBJCNT_DEC(sock);
1554 /* Handle EPIPE error. */
1555 static int sctp_error(struct sock *sk, int flags, int err)
1557 if (err == -EPIPE)
1558 err = sock_error(sk) ? : -EPIPE;
1559 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1560 send_sig(SIGPIPE, current, 0);
1561 return err;
1564 /* API 3.1.3 sendmsg() - UDP Style Syntax
1566 * An application uses sendmsg() and recvmsg() calls to transmit data to
1567 * and receive data from its peer.
1569 * ssize_t sendmsg(int socket, const struct msghdr *message,
1570 * int flags);
1572 * socket - the socket descriptor of the endpoint.
1573 * message - pointer to the msghdr structure which contains a single
1574 * user message and possibly some ancillary data.
1576 * See Section 5 for complete description of the data
1577 * structures.
1579 * flags - flags sent or received with the user message, see Section
1580 * 5 for complete description of the flags.
1582 * Note: This function could use a rewrite especially when explicit
1583 * connect support comes in.
1585 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1587 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1589 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1590 struct msghdr *msg, size_t msg_len)
1592 struct net *net = sock_net(sk);
1593 struct sctp_sock *sp;
1594 struct sctp_endpoint *ep;
1595 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1596 struct sctp_transport *transport, *chunk_tp;
1597 struct sctp_chunk *chunk;
1598 union sctp_addr to;
1599 struct sockaddr *msg_name = NULL;
1600 struct sctp_sndrcvinfo default_sinfo;
1601 struct sctp_sndrcvinfo *sinfo;
1602 struct sctp_initmsg *sinit;
1603 sctp_assoc_t associd = 0;
1604 sctp_cmsgs_t cmsgs = { NULL };
1605 int err;
1606 sctp_scope_t scope;
1607 long timeo;
1608 __u16 sinfo_flags = 0;
1609 struct sctp_datamsg *datamsg;
1610 int msg_flags = msg->msg_flags;
1612 err = 0;
1613 sp = sctp_sk(sk);
1614 ep = sp->ep;
1616 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1617 msg, msg_len, ep);
1619 /* We cannot send a message over a TCP-style listening socket. */
1620 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1621 err = -EPIPE;
1622 goto out_nounlock;
1625 /* Parse out the SCTP CMSGs. */
1626 err = sctp_msghdr_parse(msg, &cmsgs);
1627 if (err) {
1628 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1629 goto out_nounlock;
1632 /* Fetch the destination address for this packet. This
1633 * address only selects the association--it is not necessarily
1634 * the address we will send to.
1635 * For a peeled-off socket, msg_name is ignored.
1637 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1638 int msg_namelen = msg->msg_namelen;
1640 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1641 msg_namelen);
1642 if (err)
1643 return err;
1645 if (msg_namelen > sizeof(to))
1646 msg_namelen = sizeof(to);
1647 memcpy(&to, msg->msg_name, msg_namelen);
1648 msg_name = msg->msg_name;
1651 sinfo = cmsgs.info;
1652 sinit = cmsgs.init;
1654 /* Did the user specify SNDRCVINFO? */
1655 if (sinfo) {
1656 sinfo_flags = sinfo->sinfo_flags;
1657 associd = sinfo->sinfo_assoc_id;
1660 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1661 msg_len, sinfo_flags);
1663 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1664 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1665 err = -EINVAL;
1666 goto out_nounlock;
1669 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1670 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1671 * If SCTP_ABORT is set, the message length could be non zero with
1672 * the msg_iov set to the user abort reason.
1674 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1675 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1676 err = -EINVAL;
1677 goto out_nounlock;
1680 /* If SCTP_ADDR_OVER is set, there must be an address
1681 * specified in msg_name.
1683 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1684 err = -EINVAL;
1685 goto out_nounlock;
1688 transport = NULL;
1690 pr_debug("%s: about to look up association\n", __func__);
1692 sctp_lock_sock(sk);
1694 /* If a msg_name has been specified, assume this is to be used. */
1695 if (msg_name) {
1696 /* Look for a matching association on the endpoint. */
1697 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1698 if (!asoc) {
1699 /* If we could not find a matching association on the
1700 * endpoint, make sure that it is not a TCP-style
1701 * socket that already has an association or there is
1702 * no peeled-off association on another socket.
1704 if ((sctp_style(sk, TCP) &&
1705 sctp_sstate(sk, ESTABLISHED)) ||
1706 sctp_endpoint_is_peeled_off(ep, &to)) {
1707 err = -EADDRNOTAVAIL;
1708 goto out_unlock;
1711 } else {
1712 asoc = sctp_id2assoc(sk, associd);
1713 if (!asoc) {
1714 err = -EPIPE;
1715 goto out_unlock;
1719 if (asoc) {
1720 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1722 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1723 * socket that has an association in CLOSED state. This can
1724 * happen when an accepted socket has an association that is
1725 * already CLOSED.
1727 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1728 err = -EPIPE;
1729 goto out_unlock;
1732 if (sinfo_flags & SCTP_EOF) {
1733 pr_debug("%s: shutting down association:%p\n",
1734 __func__, asoc);
1736 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1737 err = 0;
1738 goto out_unlock;
1740 if (sinfo_flags & SCTP_ABORT) {
1742 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1743 if (!chunk) {
1744 err = -ENOMEM;
1745 goto out_unlock;
1748 pr_debug("%s: aborting association:%p\n",
1749 __func__, asoc);
1751 sctp_primitive_ABORT(net, asoc, chunk);
1752 err = 0;
1753 goto out_unlock;
1757 /* Do we need to create the association? */
1758 if (!asoc) {
1759 pr_debug("%s: there is no association yet\n", __func__);
1761 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1762 err = -EINVAL;
1763 goto out_unlock;
1766 /* Check for invalid stream against the stream counts,
1767 * either the default or the user specified stream counts.
1769 if (sinfo) {
1770 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1771 /* Check against the defaults. */
1772 if (sinfo->sinfo_stream >=
1773 sp->initmsg.sinit_num_ostreams) {
1774 err = -EINVAL;
1775 goto out_unlock;
1777 } else {
1778 /* Check against the requested. */
1779 if (sinfo->sinfo_stream >=
1780 sinit->sinit_num_ostreams) {
1781 err = -EINVAL;
1782 goto out_unlock;
1788 * API 3.1.2 bind() - UDP Style Syntax
1789 * If a bind() or sctp_bindx() is not called prior to a
1790 * sendmsg() call that initiates a new association, the
1791 * system picks an ephemeral port and will choose an address
1792 * set equivalent to binding with a wildcard address.
1794 if (!ep->base.bind_addr.port) {
1795 if (sctp_autobind(sk)) {
1796 err = -EAGAIN;
1797 goto out_unlock;
1799 } else {
1801 * If an unprivileged user inherits a one-to-many
1802 * style socket with open associations on a privileged
1803 * port, it MAY be permitted to accept new associations,
1804 * but it SHOULD NOT be permitted to open new
1805 * associations.
1807 if (ep->base.bind_addr.port < PROT_SOCK &&
1808 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1809 err = -EACCES;
1810 goto out_unlock;
1814 scope = sctp_scope(&to);
1815 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1816 if (!new_asoc) {
1817 err = -ENOMEM;
1818 goto out_unlock;
1820 asoc = new_asoc;
1821 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1822 if (err < 0) {
1823 err = -ENOMEM;
1824 goto out_free;
1827 /* If the SCTP_INIT ancillary data is specified, set all
1828 * the association init values accordingly.
1830 if (sinit) {
1831 if (sinit->sinit_num_ostreams) {
1832 asoc->c.sinit_num_ostreams =
1833 sinit->sinit_num_ostreams;
1835 if (sinit->sinit_max_instreams) {
1836 asoc->c.sinit_max_instreams =
1837 sinit->sinit_max_instreams;
1839 if (sinit->sinit_max_attempts) {
1840 asoc->max_init_attempts
1841 = sinit->sinit_max_attempts;
1843 if (sinit->sinit_max_init_timeo) {
1844 asoc->max_init_timeo =
1845 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1849 /* Prime the peer's transport structures. */
1850 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1851 if (!transport) {
1852 err = -ENOMEM;
1853 goto out_free;
1857 /* ASSERT: we have a valid association at this point. */
1858 pr_debug("%s: we have a valid association\n", __func__);
1860 if (!sinfo) {
1861 /* If the user didn't specify SNDRCVINFO, make up one with
1862 * some defaults.
1864 memset(&default_sinfo, 0, sizeof(default_sinfo));
1865 default_sinfo.sinfo_stream = asoc->default_stream;
1866 default_sinfo.sinfo_flags = asoc->default_flags;
1867 default_sinfo.sinfo_ppid = asoc->default_ppid;
1868 default_sinfo.sinfo_context = asoc->default_context;
1869 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1870 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1871 sinfo = &default_sinfo;
1874 /* API 7.1.7, the sndbuf size per association bounds the
1875 * maximum size of data that can be sent in a single send call.
1877 if (msg_len > sk->sk_sndbuf) {
1878 err = -EMSGSIZE;
1879 goto out_free;
1882 if (asoc->pmtu_pending)
1883 sctp_assoc_pending_pmtu(sk, asoc);
1885 /* If fragmentation is disabled and the message length exceeds the
1886 * association fragmentation point, return EMSGSIZE. The I-D
1887 * does not specify what this error is, but this looks like
1888 * a great fit.
1890 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1891 err = -EMSGSIZE;
1892 goto out_free;
1895 /* Check for invalid stream. */
1896 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1897 err = -EINVAL;
1898 goto out_free;
1901 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1902 if (!sctp_wspace(asoc)) {
1903 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1904 if (err)
1905 goto out_free;
1908 /* If an address is passed with the sendto/sendmsg call, it is used
1909 * to override the primary destination address in the TCP model, or
1910 * when SCTP_ADDR_OVER flag is set in the UDP model.
1912 if ((sctp_style(sk, TCP) && msg_name) ||
1913 (sinfo_flags & SCTP_ADDR_OVER)) {
1914 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1915 if (!chunk_tp) {
1916 err = -EINVAL;
1917 goto out_free;
1919 } else
1920 chunk_tp = NULL;
1922 /* Auto-connect, if we aren't connected already. */
1923 if (sctp_state(asoc, CLOSED)) {
1924 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1925 if (err < 0)
1926 goto out_free;
1928 pr_debug("%s: we associated primitively\n", __func__);
1931 /* Break the message into multiple chunks of maximum size. */
1932 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1933 if (IS_ERR(datamsg)) {
1934 err = PTR_ERR(datamsg);
1935 goto out_free;
1938 /* Now send the (possibly) fragmented message. */
1939 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1940 sctp_chunk_hold(chunk);
1942 /* Do accounting for the write space. */
1943 sctp_set_owner_w(chunk);
1945 chunk->transport = chunk_tp;
1948 /* Send it to the lower layers. Note: all chunks
1949 * must either fail or succeed. The lower layer
1950 * works that way today. Keep it that way or this
1951 * breaks.
1953 err = sctp_primitive_SEND(net, asoc, datamsg);
1954 /* Did the lower layer accept the chunk? */
1955 if (err) {
1956 sctp_datamsg_free(datamsg);
1957 goto out_free;
1960 pr_debug("%s: we sent primitively\n", __func__);
1962 sctp_datamsg_put(datamsg);
1963 err = msg_len;
1965 /* If we are already past ASSOCIATE, the lower
1966 * layers are responsible for association cleanup.
1968 goto out_unlock;
1970 out_free:
1971 if (new_asoc) {
1972 sctp_unhash_established(asoc);
1973 sctp_association_free(asoc);
1975 out_unlock:
1976 sctp_release_sock(sk);
1978 out_nounlock:
1979 return sctp_error(sk, msg_flags, err);
1981 #if 0
1982 do_sock_err:
1983 if (msg_len)
1984 err = msg_len;
1985 else
1986 err = sock_error(sk);
1987 goto out;
1989 do_interrupted:
1990 if (msg_len)
1991 err = msg_len;
1992 goto out;
1993 #endif /* 0 */
1996 /* This is an extended version of skb_pull() that removes the data from the
1997 * start of a skb even when data is spread across the list of skb's in the
1998 * frag_list. len specifies the total amount of data that needs to be removed.
1999 * when 'len' bytes could be removed from the skb, it returns 0.
2000 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2001 * could not be removed.
2003 static int sctp_skb_pull(struct sk_buff *skb, int len)
2005 struct sk_buff *list;
2006 int skb_len = skb_headlen(skb);
2007 int rlen;
2009 if (len <= skb_len) {
2010 __skb_pull(skb, len);
2011 return 0;
2013 len -= skb_len;
2014 __skb_pull(skb, skb_len);
2016 skb_walk_frags(skb, list) {
2017 rlen = sctp_skb_pull(list, len);
2018 skb->len -= (len-rlen);
2019 skb->data_len -= (len-rlen);
2021 if (!rlen)
2022 return 0;
2024 len = rlen;
2027 return len;
2030 /* API 3.1.3 recvmsg() - UDP Style Syntax
2032 * ssize_t recvmsg(int socket, struct msghdr *message,
2033 * int flags);
2035 * socket - the socket descriptor of the endpoint.
2036 * message - pointer to the msghdr structure which contains a single
2037 * user message and possibly some ancillary data.
2039 * See Section 5 for complete description of the data
2040 * structures.
2042 * flags - flags sent or received with the user message, see Section
2043 * 5 for complete description of the flags.
2045 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2047 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2048 struct msghdr *msg, size_t len, int noblock,
2049 int flags, int *addr_len)
2051 struct sctp_ulpevent *event = NULL;
2052 struct sctp_sock *sp = sctp_sk(sk);
2053 struct sk_buff *skb;
2054 int copied;
2055 int err = 0;
2056 int skb_len;
2058 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2059 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2060 addr_len);
2062 sctp_lock_sock(sk);
2064 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2065 err = -ENOTCONN;
2066 goto out;
2069 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2070 if (!skb)
2071 goto out;
2073 /* Get the total length of the skb including any skb's in the
2074 * frag_list.
2076 skb_len = skb->len;
2078 copied = skb_len;
2079 if (copied > len)
2080 copied = len;
2082 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2084 event = sctp_skb2event(skb);
2086 if (err)
2087 goto out_free;
2089 sock_recv_ts_and_drops(msg, sk, skb);
2090 if (sctp_ulpevent_is_notification(event)) {
2091 msg->msg_flags |= MSG_NOTIFICATION;
2092 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2093 } else {
2094 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2097 /* Check if we allow SCTP_SNDRCVINFO. */
2098 if (sp->subscribe.sctp_data_io_event)
2099 sctp_ulpevent_read_sndrcvinfo(event, msg);
2100 #if 0
2101 /* FIXME: we should be calling IP/IPv6 layers. */
2102 if (sk->sk_protinfo.af_inet.cmsg_flags)
2103 ip_cmsg_recv(msg, skb);
2104 #endif
2106 err = copied;
2108 /* If skb's length exceeds the user's buffer, update the skb and
2109 * push it back to the receive_queue so that the next call to
2110 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2112 if (skb_len > copied) {
2113 msg->msg_flags &= ~MSG_EOR;
2114 if (flags & MSG_PEEK)
2115 goto out_free;
2116 sctp_skb_pull(skb, copied);
2117 skb_queue_head(&sk->sk_receive_queue, skb);
2119 /* When only partial message is copied to the user, increase
2120 * rwnd by that amount. If all the data in the skb is read,
2121 * rwnd is updated when the event is freed.
2123 if (!sctp_ulpevent_is_notification(event))
2124 sctp_assoc_rwnd_increase(event->asoc, copied);
2125 goto out;
2126 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2127 (event->msg_flags & MSG_EOR))
2128 msg->msg_flags |= MSG_EOR;
2129 else
2130 msg->msg_flags &= ~MSG_EOR;
2132 out_free:
2133 if (flags & MSG_PEEK) {
2134 /* Release the skb reference acquired after peeking the skb in
2135 * sctp_skb_recv_datagram().
2137 kfree_skb(skb);
2138 } else {
2139 /* Free the event which includes releasing the reference to
2140 * the owner of the skb, freeing the skb and updating the
2141 * rwnd.
2143 sctp_ulpevent_free(event);
2145 out:
2146 sctp_release_sock(sk);
2147 return err;
2150 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2152 * This option is a on/off flag. If enabled no SCTP message
2153 * fragmentation will be performed. Instead if a message being sent
2154 * exceeds the current PMTU size, the message will NOT be sent and
2155 * instead a error will be indicated to the user.
2157 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2158 char __user *optval,
2159 unsigned int optlen)
2161 int val;
2163 if (optlen < sizeof(int))
2164 return -EINVAL;
2166 if (get_user(val, (int __user *)optval))
2167 return -EFAULT;
2169 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2171 return 0;
2174 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2175 unsigned int optlen)
2177 struct sctp_association *asoc;
2178 struct sctp_ulpevent *event;
2180 if (optlen > sizeof(struct sctp_event_subscribe))
2181 return -EINVAL;
2182 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2183 return -EFAULT;
2186 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2187 * if there is no data to be sent or retransmit, the stack will
2188 * immediately send up this notification.
2190 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2191 &sctp_sk(sk)->subscribe)) {
2192 asoc = sctp_id2assoc(sk, 0);
2194 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2195 event = sctp_ulpevent_make_sender_dry_event(asoc,
2196 GFP_ATOMIC);
2197 if (!event)
2198 return -ENOMEM;
2200 sctp_ulpq_tail_event(&asoc->ulpq, event);
2204 return 0;
2207 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2209 * This socket option is applicable to the UDP-style socket only. When
2210 * set it will cause associations that are idle for more than the
2211 * specified number of seconds to automatically close. An association
2212 * being idle is defined an association that has NOT sent or received
2213 * user data. The special value of '0' indicates that no automatic
2214 * close of any associations should be performed. The option expects an
2215 * integer defining the number of seconds of idle time before an
2216 * association is closed.
2218 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2219 unsigned int optlen)
2221 struct sctp_sock *sp = sctp_sk(sk);
2223 /* Applicable to UDP-style socket only */
2224 if (sctp_style(sk, TCP))
2225 return -EOPNOTSUPP;
2226 if (optlen != sizeof(int))
2227 return -EINVAL;
2228 if (copy_from_user(&sp->autoclose, optval, optlen))
2229 return -EFAULT;
2231 return 0;
2234 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2236 * Applications can enable or disable heartbeats for any peer address of
2237 * an association, modify an address's heartbeat interval, force a
2238 * heartbeat to be sent immediately, and adjust the address's maximum
2239 * number of retransmissions sent before an address is considered
2240 * unreachable. The following structure is used to access and modify an
2241 * address's parameters:
2243 * struct sctp_paddrparams {
2244 * sctp_assoc_t spp_assoc_id;
2245 * struct sockaddr_storage spp_address;
2246 * uint32_t spp_hbinterval;
2247 * uint16_t spp_pathmaxrxt;
2248 * uint32_t spp_pathmtu;
2249 * uint32_t spp_sackdelay;
2250 * uint32_t spp_flags;
2251 * };
2253 * spp_assoc_id - (one-to-many style socket) This is filled in the
2254 * application, and identifies the association for
2255 * this query.
2256 * spp_address - This specifies which address is of interest.
2257 * spp_hbinterval - This contains the value of the heartbeat interval,
2258 * in milliseconds. If a value of zero
2259 * is present in this field then no changes are to
2260 * be made to this parameter.
2261 * spp_pathmaxrxt - This contains the maximum number of
2262 * retransmissions before this address shall be
2263 * considered unreachable. If a value of zero
2264 * is present in this field then no changes are to
2265 * be made to this parameter.
2266 * spp_pathmtu - When Path MTU discovery is disabled the value
2267 * specified here will be the "fixed" path mtu.
2268 * Note that if the spp_address field is empty
2269 * then all associations on this address will
2270 * have this fixed path mtu set upon them.
2272 * spp_sackdelay - When delayed sack is enabled, this value specifies
2273 * the number of milliseconds that sacks will be delayed
2274 * for. This value will apply to all addresses of an
2275 * association if the spp_address field is empty. Note
2276 * also, that if delayed sack is enabled and this
2277 * value is set to 0, no change is made to the last
2278 * recorded delayed sack timer value.
2280 * spp_flags - These flags are used to control various features
2281 * on an association. The flag field may contain
2282 * zero or more of the following options.
2284 * SPP_HB_ENABLE - Enable heartbeats on the
2285 * specified address. Note that if the address
2286 * field is empty all addresses for the association
2287 * have heartbeats enabled upon them.
2289 * SPP_HB_DISABLE - Disable heartbeats on the
2290 * speicifed address. Note that if the address
2291 * field is empty all addresses for the association
2292 * will have their heartbeats disabled. Note also
2293 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2294 * mutually exclusive, only one of these two should
2295 * be specified. Enabling both fields will have
2296 * undetermined results.
2298 * SPP_HB_DEMAND - Request a user initiated heartbeat
2299 * to be made immediately.
2301 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2302 * heartbeat delayis to be set to the value of 0
2303 * milliseconds.
2305 * SPP_PMTUD_ENABLE - This field will enable PMTU
2306 * discovery upon the specified address. Note that
2307 * if the address feild is empty then all addresses
2308 * on the association are effected.
2310 * SPP_PMTUD_DISABLE - This field will disable PMTU
2311 * discovery upon the specified address. Note that
2312 * if the address feild is empty then all addresses
2313 * on the association are effected. Not also that
2314 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2315 * exclusive. Enabling both will have undetermined
2316 * results.
2318 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2319 * on delayed sack. The time specified in spp_sackdelay
2320 * is used to specify the sack delay for this address. Note
2321 * that if spp_address is empty then all addresses will
2322 * enable delayed sack and take on the sack delay
2323 * value specified in spp_sackdelay.
2324 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2325 * off delayed sack. If the spp_address field is blank then
2326 * delayed sack is disabled for the entire association. Note
2327 * also that this field is mutually exclusive to
2328 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2329 * results.
2331 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2332 struct sctp_transport *trans,
2333 struct sctp_association *asoc,
2334 struct sctp_sock *sp,
2335 int hb_change,
2336 int pmtud_change,
2337 int sackdelay_change)
2339 int error;
2341 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2342 struct net *net = sock_net(trans->asoc->base.sk);
2344 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2345 if (error)
2346 return error;
2349 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2350 * this field is ignored. Note also that a value of zero indicates
2351 * the current setting should be left unchanged.
2353 if (params->spp_flags & SPP_HB_ENABLE) {
2355 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2356 * set. This lets us use 0 value when this flag
2357 * is set.
2359 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2360 params->spp_hbinterval = 0;
2362 if (params->spp_hbinterval ||
2363 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2364 if (trans) {
2365 trans->hbinterval =
2366 msecs_to_jiffies(params->spp_hbinterval);
2367 } else if (asoc) {
2368 asoc->hbinterval =
2369 msecs_to_jiffies(params->spp_hbinterval);
2370 } else {
2371 sp->hbinterval = params->spp_hbinterval;
2376 if (hb_change) {
2377 if (trans) {
2378 trans->param_flags =
2379 (trans->param_flags & ~SPP_HB) | hb_change;
2380 } else if (asoc) {
2381 asoc->param_flags =
2382 (asoc->param_flags & ~SPP_HB) | hb_change;
2383 } else {
2384 sp->param_flags =
2385 (sp->param_flags & ~SPP_HB) | hb_change;
2389 /* When Path MTU discovery is disabled the value specified here will
2390 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2391 * include the flag SPP_PMTUD_DISABLE for this field to have any
2392 * effect).
2394 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2395 if (trans) {
2396 trans->pathmtu = params->spp_pathmtu;
2397 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2398 } else if (asoc) {
2399 asoc->pathmtu = params->spp_pathmtu;
2400 sctp_frag_point(asoc, params->spp_pathmtu);
2401 } else {
2402 sp->pathmtu = params->spp_pathmtu;
2406 if (pmtud_change) {
2407 if (trans) {
2408 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2409 (params->spp_flags & SPP_PMTUD_ENABLE);
2410 trans->param_flags =
2411 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2412 if (update) {
2413 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2414 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2416 } else if (asoc) {
2417 asoc->param_flags =
2418 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2419 } else {
2420 sp->param_flags =
2421 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2425 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2426 * value of this field is ignored. Note also that a value of zero
2427 * indicates the current setting should be left unchanged.
2429 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2430 if (trans) {
2431 trans->sackdelay =
2432 msecs_to_jiffies(params->spp_sackdelay);
2433 } else if (asoc) {
2434 asoc->sackdelay =
2435 msecs_to_jiffies(params->spp_sackdelay);
2436 } else {
2437 sp->sackdelay = params->spp_sackdelay;
2441 if (sackdelay_change) {
2442 if (trans) {
2443 trans->param_flags =
2444 (trans->param_flags & ~SPP_SACKDELAY) |
2445 sackdelay_change;
2446 } else if (asoc) {
2447 asoc->param_flags =
2448 (asoc->param_flags & ~SPP_SACKDELAY) |
2449 sackdelay_change;
2450 } else {
2451 sp->param_flags =
2452 (sp->param_flags & ~SPP_SACKDELAY) |
2453 sackdelay_change;
2457 /* Note that a value of zero indicates the current setting should be
2458 left unchanged.
2460 if (params->spp_pathmaxrxt) {
2461 if (trans) {
2462 trans->pathmaxrxt = params->spp_pathmaxrxt;
2463 } else if (asoc) {
2464 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2465 } else {
2466 sp->pathmaxrxt = params->spp_pathmaxrxt;
2470 return 0;
2473 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2474 char __user *optval,
2475 unsigned int optlen)
2477 struct sctp_paddrparams params;
2478 struct sctp_transport *trans = NULL;
2479 struct sctp_association *asoc = NULL;
2480 struct sctp_sock *sp = sctp_sk(sk);
2481 int error;
2482 int hb_change, pmtud_change, sackdelay_change;
2484 if (optlen != sizeof(struct sctp_paddrparams))
2485 return - EINVAL;
2487 if (copy_from_user(&params, optval, optlen))
2488 return -EFAULT;
2490 /* Validate flags and value parameters. */
2491 hb_change = params.spp_flags & SPP_HB;
2492 pmtud_change = params.spp_flags & SPP_PMTUD;
2493 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2495 if (hb_change == SPP_HB ||
2496 pmtud_change == SPP_PMTUD ||
2497 sackdelay_change == SPP_SACKDELAY ||
2498 params.spp_sackdelay > 500 ||
2499 (params.spp_pathmtu &&
2500 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2501 return -EINVAL;
2503 /* If an address other than INADDR_ANY is specified, and
2504 * no transport is found, then the request is invalid.
2506 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2507 trans = sctp_addr_id2transport(sk, &params.spp_address,
2508 params.spp_assoc_id);
2509 if (!trans)
2510 return -EINVAL;
2513 /* Get association, if assoc_id != 0 and the socket is a one
2514 * to many style socket, and an association was not found, then
2515 * the id was invalid.
2517 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2518 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2519 return -EINVAL;
2521 /* Heartbeat demand can only be sent on a transport or
2522 * association, but not a socket.
2524 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2525 return -EINVAL;
2527 /* Process parameters. */
2528 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2529 hb_change, pmtud_change,
2530 sackdelay_change);
2532 if (error)
2533 return error;
2535 /* If changes are for association, also apply parameters to each
2536 * transport.
2538 if (!trans && asoc) {
2539 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2540 transports) {
2541 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2542 hb_change, pmtud_change,
2543 sackdelay_change);
2547 return 0;
2551 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2553 * This option will effect the way delayed acks are performed. This
2554 * option allows you to get or set the delayed ack time, in
2555 * milliseconds. It also allows changing the delayed ack frequency.
2556 * Changing the frequency to 1 disables the delayed sack algorithm. If
2557 * the assoc_id is 0, then this sets or gets the endpoints default
2558 * values. If the assoc_id field is non-zero, then the set or get
2559 * effects the specified association for the one to many model (the
2560 * assoc_id field is ignored by the one to one model). Note that if
2561 * sack_delay or sack_freq are 0 when setting this option, then the
2562 * current values will remain unchanged.
2564 * struct sctp_sack_info {
2565 * sctp_assoc_t sack_assoc_id;
2566 * uint32_t sack_delay;
2567 * uint32_t sack_freq;
2568 * };
2570 * sack_assoc_id - This parameter, indicates which association the user
2571 * is performing an action upon. Note that if this field's value is
2572 * zero then the endpoints default value is changed (effecting future
2573 * associations only).
2575 * sack_delay - This parameter contains the number of milliseconds that
2576 * the user is requesting the delayed ACK timer be set to. Note that
2577 * this value is defined in the standard to be between 200 and 500
2578 * milliseconds.
2580 * sack_freq - This parameter contains the number of packets that must
2581 * be received before a sack is sent without waiting for the delay
2582 * timer to expire. The default value for this is 2, setting this
2583 * value to 1 will disable the delayed sack algorithm.
2586 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2587 char __user *optval, unsigned int optlen)
2589 struct sctp_sack_info params;
2590 struct sctp_transport *trans = NULL;
2591 struct sctp_association *asoc = NULL;
2592 struct sctp_sock *sp = sctp_sk(sk);
2594 if (optlen == sizeof(struct sctp_sack_info)) {
2595 if (copy_from_user(&params, optval, optlen))
2596 return -EFAULT;
2598 if (params.sack_delay == 0 && params.sack_freq == 0)
2599 return 0;
2600 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2601 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2602 pr_warn("Use struct sctp_sack_info instead\n");
2603 if (copy_from_user(&params, optval, optlen))
2604 return -EFAULT;
2606 if (params.sack_delay == 0)
2607 params.sack_freq = 1;
2608 else
2609 params.sack_freq = 0;
2610 } else
2611 return - EINVAL;
2613 /* Validate value parameter. */
2614 if (params.sack_delay > 500)
2615 return -EINVAL;
2617 /* Get association, if sack_assoc_id != 0 and the socket is a one
2618 * to many style socket, and an association was not found, then
2619 * the id was invalid.
2621 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2622 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2623 return -EINVAL;
2625 if (params.sack_delay) {
2626 if (asoc) {
2627 asoc->sackdelay =
2628 msecs_to_jiffies(params.sack_delay);
2629 asoc->param_flags =
2630 (asoc->param_flags & ~SPP_SACKDELAY) |
2631 SPP_SACKDELAY_ENABLE;
2632 } else {
2633 sp->sackdelay = params.sack_delay;
2634 sp->param_flags =
2635 (sp->param_flags & ~SPP_SACKDELAY) |
2636 SPP_SACKDELAY_ENABLE;
2640 if (params.sack_freq == 1) {
2641 if (asoc) {
2642 asoc->param_flags =
2643 (asoc->param_flags & ~SPP_SACKDELAY) |
2644 SPP_SACKDELAY_DISABLE;
2645 } else {
2646 sp->param_flags =
2647 (sp->param_flags & ~SPP_SACKDELAY) |
2648 SPP_SACKDELAY_DISABLE;
2650 } else if (params.sack_freq > 1) {
2651 if (asoc) {
2652 asoc->sackfreq = params.sack_freq;
2653 asoc->param_flags =
2654 (asoc->param_flags & ~SPP_SACKDELAY) |
2655 SPP_SACKDELAY_ENABLE;
2656 } else {
2657 sp->sackfreq = params.sack_freq;
2658 sp->param_flags =
2659 (sp->param_flags & ~SPP_SACKDELAY) |
2660 SPP_SACKDELAY_ENABLE;
2664 /* If change is for association, also apply to each transport. */
2665 if (asoc) {
2666 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2667 transports) {
2668 if (params.sack_delay) {
2669 trans->sackdelay =
2670 msecs_to_jiffies(params.sack_delay);
2671 trans->param_flags =
2672 (trans->param_flags & ~SPP_SACKDELAY) |
2673 SPP_SACKDELAY_ENABLE;
2675 if (params.sack_freq == 1) {
2676 trans->param_flags =
2677 (trans->param_flags & ~SPP_SACKDELAY) |
2678 SPP_SACKDELAY_DISABLE;
2679 } else if (params.sack_freq > 1) {
2680 trans->sackfreq = params.sack_freq;
2681 trans->param_flags =
2682 (trans->param_flags & ~SPP_SACKDELAY) |
2683 SPP_SACKDELAY_ENABLE;
2688 return 0;
2691 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2693 * Applications can specify protocol parameters for the default association
2694 * initialization. The option name argument to setsockopt() and getsockopt()
2695 * is SCTP_INITMSG.
2697 * Setting initialization parameters is effective only on an unconnected
2698 * socket (for UDP-style sockets only future associations are effected
2699 * by the change). With TCP-style sockets, this option is inherited by
2700 * sockets derived from a listener socket.
2702 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2704 struct sctp_initmsg sinit;
2705 struct sctp_sock *sp = sctp_sk(sk);
2707 if (optlen != sizeof(struct sctp_initmsg))
2708 return -EINVAL;
2709 if (copy_from_user(&sinit, optval, optlen))
2710 return -EFAULT;
2712 if (sinit.sinit_num_ostreams)
2713 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2714 if (sinit.sinit_max_instreams)
2715 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2716 if (sinit.sinit_max_attempts)
2717 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2718 if (sinit.sinit_max_init_timeo)
2719 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2721 return 0;
2725 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2727 * Applications that wish to use the sendto() system call may wish to
2728 * specify a default set of parameters that would normally be supplied
2729 * through the inclusion of ancillary data. This socket option allows
2730 * such an application to set the default sctp_sndrcvinfo structure.
2731 * The application that wishes to use this socket option simply passes
2732 * in to this call the sctp_sndrcvinfo structure defined in Section
2733 * 5.2.2) The input parameters accepted by this call include
2734 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2735 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2736 * to this call if the caller is using the UDP model.
2738 static int sctp_setsockopt_default_send_param(struct sock *sk,
2739 char __user *optval,
2740 unsigned int optlen)
2742 struct sctp_sndrcvinfo info;
2743 struct sctp_association *asoc;
2744 struct sctp_sock *sp = sctp_sk(sk);
2746 if (optlen != sizeof(struct sctp_sndrcvinfo))
2747 return -EINVAL;
2748 if (copy_from_user(&info, optval, optlen))
2749 return -EFAULT;
2751 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2752 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2753 return -EINVAL;
2755 if (asoc) {
2756 asoc->default_stream = info.sinfo_stream;
2757 asoc->default_flags = info.sinfo_flags;
2758 asoc->default_ppid = info.sinfo_ppid;
2759 asoc->default_context = info.sinfo_context;
2760 asoc->default_timetolive = info.sinfo_timetolive;
2761 } else {
2762 sp->default_stream = info.sinfo_stream;
2763 sp->default_flags = info.sinfo_flags;
2764 sp->default_ppid = info.sinfo_ppid;
2765 sp->default_context = info.sinfo_context;
2766 sp->default_timetolive = info.sinfo_timetolive;
2769 return 0;
2772 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2774 * Requests that the local SCTP stack use the enclosed peer address as
2775 * the association primary. The enclosed address must be one of the
2776 * association peer's addresses.
2778 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2779 unsigned int optlen)
2781 struct sctp_prim prim;
2782 struct sctp_transport *trans;
2784 if (optlen != sizeof(struct sctp_prim))
2785 return -EINVAL;
2787 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2788 return -EFAULT;
2790 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2791 if (!trans)
2792 return -EINVAL;
2794 sctp_assoc_set_primary(trans->asoc, trans);
2796 return 0;
2800 * 7.1.5 SCTP_NODELAY
2802 * Turn on/off any Nagle-like algorithm. This means that packets are
2803 * generally sent as soon as possible and no unnecessary delays are
2804 * introduced, at the cost of more packets in the network. Expects an
2805 * integer boolean flag.
2807 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2808 unsigned int optlen)
2810 int val;
2812 if (optlen < sizeof(int))
2813 return -EINVAL;
2814 if (get_user(val, (int __user *)optval))
2815 return -EFAULT;
2817 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2818 return 0;
2823 * 7.1.1 SCTP_RTOINFO
2825 * The protocol parameters used to initialize and bound retransmission
2826 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2827 * and modify these parameters.
2828 * All parameters are time values, in milliseconds. A value of 0, when
2829 * modifying the parameters, indicates that the current value should not
2830 * be changed.
2833 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2835 struct sctp_rtoinfo rtoinfo;
2836 struct sctp_association *asoc;
2838 if (optlen != sizeof (struct sctp_rtoinfo))
2839 return -EINVAL;
2841 if (copy_from_user(&rtoinfo, optval, optlen))
2842 return -EFAULT;
2844 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2846 /* Set the values to the specific association */
2847 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2848 return -EINVAL;
2850 if (asoc) {
2851 if (rtoinfo.srto_initial != 0)
2852 asoc->rto_initial =
2853 msecs_to_jiffies(rtoinfo.srto_initial);
2854 if (rtoinfo.srto_max != 0)
2855 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2856 if (rtoinfo.srto_min != 0)
2857 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2858 } else {
2859 /* If there is no association or the association-id = 0
2860 * set the values to the endpoint.
2862 struct sctp_sock *sp = sctp_sk(sk);
2864 if (rtoinfo.srto_initial != 0)
2865 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2866 if (rtoinfo.srto_max != 0)
2867 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2868 if (rtoinfo.srto_min != 0)
2869 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2872 return 0;
2877 * 7.1.2 SCTP_ASSOCINFO
2879 * This option is used to tune the maximum retransmission attempts
2880 * of the association.
2881 * Returns an error if the new association retransmission value is
2882 * greater than the sum of the retransmission value of the peer.
2883 * See [SCTP] for more information.
2886 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2889 struct sctp_assocparams assocparams;
2890 struct sctp_association *asoc;
2892 if (optlen != sizeof(struct sctp_assocparams))
2893 return -EINVAL;
2894 if (copy_from_user(&assocparams, optval, optlen))
2895 return -EFAULT;
2897 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2899 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2900 return -EINVAL;
2902 /* Set the values to the specific association */
2903 if (asoc) {
2904 if (assocparams.sasoc_asocmaxrxt != 0) {
2905 __u32 path_sum = 0;
2906 int paths = 0;
2907 struct sctp_transport *peer_addr;
2909 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2910 transports) {
2911 path_sum += peer_addr->pathmaxrxt;
2912 paths++;
2915 /* Only validate asocmaxrxt if we have more than
2916 * one path/transport. We do this because path
2917 * retransmissions are only counted when we have more
2918 * then one path.
2920 if (paths > 1 &&
2921 assocparams.sasoc_asocmaxrxt > path_sum)
2922 return -EINVAL;
2924 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2927 if (assocparams.sasoc_cookie_life != 0)
2928 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2929 } else {
2930 /* Set the values to the endpoint */
2931 struct sctp_sock *sp = sctp_sk(sk);
2933 if (assocparams.sasoc_asocmaxrxt != 0)
2934 sp->assocparams.sasoc_asocmaxrxt =
2935 assocparams.sasoc_asocmaxrxt;
2936 if (assocparams.sasoc_cookie_life != 0)
2937 sp->assocparams.sasoc_cookie_life =
2938 assocparams.sasoc_cookie_life;
2940 return 0;
2944 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2946 * This socket option is a boolean flag which turns on or off mapped V4
2947 * addresses. If this option is turned on and the socket is type
2948 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2949 * If this option is turned off, then no mapping will be done of V4
2950 * addresses and a user will receive both PF_INET6 and PF_INET type
2951 * addresses on the socket.
2953 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2955 int val;
2956 struct sctp_sock *sp = sctp_sk(sk);
2958 if (optlen < sizeof(int))
2959 return -EINVAL;
2960 if (get_user(val, (int __user *)optval))
2961 return -EFAULT;
2962 if (val)
2963 sp->v4mapped = 1;
2964 else
2965 sp->v4mapped = 0;
2967 return 0;
2971 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2972 * This option will get or set the maximum size to put in any outgoing
2973 * SCTP DATA chunk. If a message is larger than this size it will be
2974 * fragmented by SCTP into the specified size. Note that the underlying
2975 * SCTP implementation may fragment into smaller sized chunks when the
2976 * PMTU of the underlying association is smaller than the value set by
2977 * the user. The default value for this option is '0' which indicates
2978 * the user is NOT limiting fragmentation and only the PMTU will effect
2979 * SCTP's choice of DATA chunk size. Note also that values set larger
2980 * than the maximum size of an IP datagram will effectively let SCTP
2981 * control fragmentation (i.e. the same as setting this option to 0).
2983 * The following structure is used to access and modify this parameter:
2985 * struct sctp_assoc_value {
2986 * sctp_assoc_t assoc_id;
2987 * uint32_t assoc_value;
2988 * };
2990 * assoc_id: This parameter is ignored for one-to-one style sockets.
2991 * For one-to-many style sockets this parameter indicates which
2992 * association the user is performing an action upon. Note that if
2993 * this field's value is zero then the endpoints default value is
2994 * changed (effecting future associations only).
2995 * assoc_value: This parameter specifies the maximum size in bytes.
2997 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2999 struct sctp_assoc_value params;
3000 struct sctp_association *asoc;
3001 struct sctp_sock *sp = sctp_sk(sk);
3002 int val;
3004 if (optlen == sizeof(int)) {
3005 pr_warn("Use of int in maxseg socket option deprecated\n");
3006 pr_warn("Use struct sctp_assoc_value instead\n");
3007 if (copy_from_user(&val, optval, optlen))
3008 return -EFAULT;
3009 params.assoc_id = 0;
3010 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3011 if (copy_from_user(&params, optval, optlen))
3012 return -EFAULT;
3013 val = params.assoc_value;
3014 } else
3015 return -EINVAL;
3017 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3018 return -EINVAL;
3020 asoc = sctp_id2assoc(sk, params.assoc_id);
3021 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3022 return -EINVAL;
3024 if (asoc) {
3025 if (val == 0) {
3026 val = asoc->pathmtu;
3027 val -= sp->pf->af->net_header_len;
3028 val -= sizeof(struct sctphdr) +
3029 sizeof(struct sctp_data_chunk);
3031 asoc->user_frag = val;
3032 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3033 } else {
3034 sp->user_frag = val;
3037 return 0;
3042 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3044 * Requests that the peer mark the enclosed address as the association
3045 * primary. The enclosed address must be one of the association's
3046 * locally bound addresses. The following structure is used to make a
3047 * set primary request:
3049 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3050 unsigned int optlen)
3052 struct net *net = sock_net(sk);
3053 struct sctp_sock *sp;
3054 struct sctp_association *asoc = NULL;
3055 struct sctp_setpeerprim prim;
3056 struct sctp_chunk *chunk;
3057 struct sctp_af *af;
3058 int err;
3060 sp = sctp_sk(sk);
3062 if (!net->sctp.addip_enable)
3063 return -EPERM;
3065 if (optlen != sizeof(struct sctp_setpeerprim))
3066 return -EINVAL;
3068 if (copy_from_user(&prim, optval, optlen))
3069 return -EFAULT;
3071 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3072 if (!asoc)
3073 return -EINVAL;
3075 if (!asoc->peer.asconf_capable)
3076 return -EPERM;
3078 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3079 return -EPERM;
3081 if (!sctp_state(asoc, ESTABLISHED))
3082 return -ENOTCONN;
3084 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3085 if (!af)
3086 return -EINVAL;
3088 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3089 return -EADDRNOTAVAIL;
3091 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3092 return -EADDRNOTAVAIL;
3094 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3095 chunk = sctp_make_asconf_set_prim(asoc,
3096 (union sctp_addr *)&prim.sspp_addr);
3097 if (!chunk)
3098 return -ENOMEM;
3100 err = sctp_send_asconf(asoc, chunk);
3102 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3104 return err;
3107 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3108 unsigned int optlen)
3110 struct sctp_setadaptation adaptation;
3112 if (optlen != sizeof(struct sctp_setadaptation))
3113 return -EINVAL;
3114 if (copy_from_user(&adaptation, optval, optlen))
3115 return -EFAULT;
3117 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3119 return 0;
3123 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3125 * The context field in the sctp_sndrcvinfo structure is normally only
3126 * used when a failed message is retrieved holding the value that was
3127 * sent down on the actual send call. This option allows the setting of
3128 * a default context on an association basis that will be received on
3129 * reading messages from the peer. This is especially helpful in the
3130 * one-2-many model for an application to keep some reference to an
3131 * internal state machine that is processing messages on the
3132 * association. Note that the setting of this value only effects
3133 * received messages from the peer and does not effect the value that is
3134 * saved with outbound messages.
3136 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3137 unsigned int optlen)
3139 struct sctp_assoc_value params;
3140 struct sctp_sock *sp;
3141 struct sctp_association *asoc;
3143 if (optlen != sizeof(struct sctp_assoc_value))
3144 return -EINVAL;
3145 if (copy_from_user(&params, optval, optlen))
3146 return -EFAULT;
3148 sp = sctp_sk(sk);
3150 if (params.assoc_id != 0) {
3151 asoc = sctp_id2assoc(sk, params.assoc_id);
3152 if (!asoc)
3153 return -EINVAL;
3154 asoc->default_rcv_context = params.assoc_value;
3155 } else {
3156 sp->default_rcv_context = params.assoc_value;
3159 return 0;
3163 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3165 * This options will at a minimum specify if the implementation is doing
3166 * fragmented interleave. Fragmented interleave, for a one to many
3167 * socket, is when subsequent calls to receive a message may return
3168 * parts of messages from different associations. Some implementations
3169 * may allow you to turn this value on or off. If so, when turned off,
3170 * no fragment interleave will occur (which will cause a head of line
3171 * blocking amongst multiple associations sharing the same one to many
3172 * socket). When this option is turned on, then each receive call may
3173 * come from a different association (thus the user must receive data
3174 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3175 * association each receive belongs to.
3177 * This option takes a boolean value. A non-zero value indicates that
3178 * fragmented interleave is on. A value of zero indicates that
3179 * fragmented interleave is off.
3181 * Note that it is important that an implementation that allows this
3182 * option to be turned on, have it off by default. Otherwise an unaware
3183 * application using the one to many model may become confused and act
3184 * incorrectly.
3186 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3187 char __user *optval,
3188 unsigned int optlen)
3190 int val;
3192 if (optlen != sizeof(int))
3193 return -EINVAL;
3194 if (get_user(val, (int __user *)optval))
3195 return -EFAULT;
3197 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3199 return 0;
3203 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3204 * (SCTP_PARTIAL_DELIVERY_POINT)
3206 * This option will set or get the SCTP partial delivery point. This
3207 * point is the size of a message where the partial delivery API will be
3208 * invoked to help free up rwnd space for the peer. Setting this to a
3209 * lower value will cause partial deliveries to happen more often. The
3210 * calls argument is an integer that sets or gets the partial delivery
3211 * point. Note also that the call will fail if the user attempts to set
3212 * this value larger than the socket receive buffer size.
3214 * Note that any single message having a length smaller than or equal to
3215 * the SCTP partial delivery point will be delivered in one single read
3216 * call as long as the user provided buffer is large enough to hold the
3217 * message.
3219 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3220 char __user *optval,
3221 unsigned int optlen)
3223 u32 val;
3225 if (optlen != sizeof(u32))
3226 return -EINVAL;
3227 if (get_user(val, (int __user *)optval))
3228 return -EFAULT;
3230 /* Note: We double the receive buffer from what the user sets
3231 * it to be, also initial rwnd is based on rcvbuf/2.
3233 if (val > (sk->sk_rcvbuf >> 1))
3234 return -EINVAL;
3236 sctp_sk(sk)->pd_point = val;
3238 return 0; /* is this the right error code? */
3242 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3244 * This option will allow a user to change the maximum burst of packets
3245 * that can be emitted by this association. Note that the default value
3246 * is 4, and some implementations may restrict this setting so that it
3247 * can only be lowered.
3249 * NOTE: This text doesn't seem right. Do this on a socket basis with
3250 * future associations inheriting the socket value.
3252 static int sctp_setsockopt_maxburst(struct sock *sk,
3253 char __user *optval,
3254 unsigned int optlen)
3256 struct sctp_assoc_value params;
3257 struct sctp_sock *sp;
3258 struct sctp_association *asoc;
3259 int val;
3260 int assoc_id = 0;
3262 if (optlen == sizeof(int)) {
3263 pr_warn("Use of int in max_burst socket option deprecated\n");
3264 pr_warn("Use struct sctp_assoc_value instead\n");
3265 if (copy_from_user(&val, optval, optlen))
3266 return -EFAULT;
3267 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3268 if (copy_from_user(&params, optval, optlen))
3269 return -EFAULT;
3270 val = params.assoc_value;
3271 assoc_id = params.assoc_id;
3272 } else
3273 return -EINVAL;
3275 sp = sctp_sk(sk);
3277 if (assoc_id != 0) {
3278 asoc = sctp_id2assoc(sk, assoc_id);
3279 if (!asoc)
3280 return -EINVAL;
3281 asoc->max_burst = val;
3282 } else
3283 sp->max_burst = val;
3285 return 0;
3289 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3291 * This set option adds a chunk type that the user is requesting to be
3292 * received only in an authenticated way. Changes to the list of chunks
3293 * will only effect future associations on the socket.
3295 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3296 char __user *optval,
3297 unsigned int optlen)
3299 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3300 struct sctp_authchunk val;
3302 if (!ep->auth_enable)
3303 return -EACCES;
3305 if (optlen != sizeof(struct sctp_authchunk))
3306 return -EINVAL;
3307 if (copy_from_user(&val, optval, optlen))
3308 return -EFAULT;
3310 switch (val.sauth_chunk) {
3311 case SCTP_CID_INIT:
3312 case SCTP_CID_INIT_ACK:
3313 case SCTP_CID_SHUTDOWN_COMPLETE:
3314 case SCTP_CID_AUTH:
3315 return -EINVAL;
3318 /* add this chunk id to the endpoint */
3319 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3323 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3325 * This option gets or sets the list of HMAC algorithms that the local
3326 * endpoint requires the peer to use.
3328 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3329 char __user *optval,
3330 unsigned int optlen)
3332 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3333 struct sctp_hmacalgo *hmacs;
3334 u32 idents;
3335 int err;
3337 if (!ep->auth_enable)
3338 return -EACCES;
3340 if (optlen < sizeof(struct sctp_hmacalgo))
3341 return -EINVAL;
3343 hmacs= memdup_user(optval, optlen);
3344 if (IS_ERR(hmacs))
3345 return PTR_ERR(hmacs);
3347 idents = hmacs->shmac_num_idents;
3348 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3349 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3350 err = -EINVAL;
3351 goto out;
3354 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3355 out:
3356 kfree(hmacs);
3357 return err;
3361 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3363 * This option will set a shared secret key which is used to build an
3364 * association shared key.
3366 static int sctp_setsockopt_auth_key(struct sock *sk,
3367 char __user *optval,
3368 unsigned int optlen)
3370 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3371 struct sctp_authkey *authkey;
3372 struct sctp_association *asoc;
3373 int ret;
3375 if (!ep->auth_enable)
3376 return -EACCES;
3378 if (optlen <= sizeof(struct sctp_authkey))
3379 return -EINVAL;
3381 authkey= memdup_user(optval, optlen);
3382 if (IS_ERR(authkey))
3383 return PTR_ERR(authkey);
3385 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3386 ret = -EINVAL;
3387 goto out;
3390 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3391 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3392 ret = -EINVAL;
3393 goto out;
3396 ret = sctp_auth_set_key(ep, asoc, authkey);
3397 out:
3398 kzfree(authkey);
3399 return ret;
3403 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3405 * This option will get or set the active shared key to be used to build
3406 * the association shared key.
3408 static int sctp_setsockopt_active_key(struct sock *sk,
3409 char __user *optval,
3410 unsigned int optlen)
3412 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3413 struct sctp_authkeyid val;
3414 struct sctp_association *asoc;
3416 if (!ep->auth_enable)
3417 return -EACCES;
3419 if (optlen != sizeof(struct sctp_authkeyid))
3420 return -EINVAL;
3421 if (copy_from_user(&val, optval, optlen))
3422 return -EFAULT;
3424 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3425 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3426 return -EINVAL;
3428 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3432 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3434 * This set option will delete a shared secret key from use.
3436 static int sctp_setsockopt_del_key(struct sock *sk,
3437 char __user *optval,
3438 unsigned int optlen)
3440 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3441 struct sctp_authkeyid val;
3442 struct sctp_association *asoc;
3444 if (!ep->auth_enable)
3445 return -EACCES;
3447 if (optlen != sizeof(struct sctp_authkeyid))
3448 return -EINVAL;
3449 if (copy_from_user(&val, optval, optlen))
3450 return -EFAULT;
3452 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3453 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3454 return -EINVAL;
3456 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3461 * 8.1.23 SCTP_AUTO_ASCONF
3463 * This option will enable or disable the use of the automatic generation of
3464 * ASCONF chunks to add and delete addresses to an existing association. Note
3465 * that this option has two caveats namely: a) it only affects sockets that
3466 * are bound to all addresses available to the SCTP stack, and b) the system
3467 * administrator may have an overriding control that turns the ASCONF feature
3468 * off no matter what setting the socket option may have.
3469 * This option expects an integer boolean flag, where a non-zero value turns on
3470 * the option, and a zero value turns off the option.
3471 * Note. In this implementation, socket operation overrides default parameter
3472 * being set by sysctl as well as FreeBSD implementation
3474 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3475 unsigned int optlen)
3477 int val;
3478 struct sctp_sock *sp = sctp_sk(sk);
3480 if (optlen < sizeof(int))
3481 return -EINVAL;
3482 if (get_user(val, (int __user *)optval))
3483 return -EFAULT;
3484 if (!sctp_is_ep_boundall(sk) && val)
3485 return -EINVAL;
3486 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3487 return 0;
3489 if (val == 0 && sp->do_auto_asconf) {
3490 list_del(&sp->auto_asconf_list);
3491 sp->do_auto_asconf = 0;
3492 } else if (val && !sp->do_auto_asconf) {
3493 list_add_tail(&sp->auto_asconf_list,
3494 &sock_net(sk)->sctp.auto_asconf_splist);
3495 sp->do_auto_asconf = 1;
3497 return 0;
3502 * SCTP_PEER_ADDR_THLDS
3504 * This option allows us to alter the partially failed threshold for one or all
3505 * transports in an association. See Section 6.1 of:
3506 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3508 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3509 char __user *optval,
3510 unsigned int optlen)
3512 struct sctp_paddrthlds val;
3513 struct sctp_transport *trans;
3514 struct sctp_association *asoc;
3516 if (optlen < sizeof(struct sctp_paddrthlds))
3517 return -EINVAL;
3518 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3519 sizeof(struct sctp_paddrthlds)))
3520 return -EFAULT;
3523 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3524 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3525 if (!asoc)
3526 return -ENOENT;
3527 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3528 transports) {
3529 if (val.spt_pathmaxrxt)
3530 trans->pathmaxrxt = val.spt_pathmaxrxt;
3531 trans->pf_retrans = val.spt_pathpfthld;
3534 if (val.spt_pathmaxrxt)
3535 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3536 asoc->pf_retrans = val.spt_pathpfthld;
3537 } else {
3538 trans = sctp_addr_id2transport(sk, &val.spt_address,
3539 val.spt_assoc_id);
3540 if (!trans)
3541 return -ENOENT;
3543 if (val.spt_pathmaxrxt)
3544 trans->pathmaxrxt = val.spt_pathmaxrxt;
3545 trans->pf_retrans = val.spt_pathpfthld;
3548 return 0;
3551 /* API 6.2 setsockopt(), getsockopt()
3553 * Applications use setsockopt() and getsockopt() to set or retrieve
3554 * socket options. Socket options are used to change the default
3555 * behavior of sockets calls. They are described in Section 7.
3557 * The syntax is:
3559 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3560 * int __user *optlen);
3561 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3562 * int optlen);
3564 * sd - the socket descript.
3565 * level - set to IPPROTO_SCTP for all SCTP options.
3566 * optname - the option name.
3567 * optval - the buffer to store the value of the option.
3568 * optlen - the size of the buffer.
3570 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3571 char __user *optval, unsigned int optlen)
3573 int retval = 0;
3575 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3577 /* I can hardly begin to describe how wrong this is. This is
3578 * so broken as to be worse than useless. The API draft
3579 * REALLY is NOT helpful here... I am not convinced that the
3580 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3581 * are at all well-founded.
3583 if (level != SOL_SCTP) {
3584 struct sctp_af *af = sctp_sk(sk)->pf->af;
3585 retval = af->setsockopt(sk, level, optname, optval, optlen);
3586 goto out_nounlock;
3589 sctp_lock_sock(sk);
3591 switch (optname) {
3592 case SCTP_SOCKOPT_BINDX_ADD:
3593 /* 'optlen' is the size of the addresses buffer. */
3594 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3595 optlen, SCTP_BINDX_ADD_ADDR);
3596 break;
3598 case SCTP_SOCKOPT_BINDX_REM:
3599 /* 'optlen' is the size of the addresses buffer. */
3600 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3601 optlen, SCTP_BINDX_REM_ADDR);
3602 break;
3604 case SCTP_SOCKOPT_CONNECTX_OLD:
3605 /* 'optlen' is the size of the addresses buffer. */
3606 retval = sctp_setsockopt_connectx_old(sk,
3607 (struct sockaddr __user *)optval,
3608 optlen);
3609 break;
3611 case SCTP_SOCKOPT_CONNECTX:
3612 /* 'optlen' is the size of the addresses buffer. */
3613 retval = sctp_setsockopt_connectx(sk,
3614 (struct sockaddr __user *)optval,
3615 optlen);
3616 break;
3618 case SCTP_DISABLE_FRAGMENTS:
3619 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3620 break;
3622 case SCTP_EVENTS:
3623 retval = sctp_setsockopt_events(sk, optval, optlen);
3624 break;
3626 case SCTP_AUTOCLOSE:
3627 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3628 break;
3630 case SCTP_PEER_ADDR_PARAMS:
3631 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3632 break;
3634 case SCTP_DELAYED_SACK:
3635 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3636 break;
3637 case SCTP_PARTIAL_DELIVERY_POINT:
3638 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3639 break;
3641 case SCTP_INITMSG:
3642 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3643 break;
3644 case SCTP_DEFAULT_SEND_PARAM:
3645 retval = sctp_setsockopt_default_send_param(sk, optval,
3646 optlen);
3647 break;
3648 case SCTP_PRIMARY_ADDR:
3649 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3650 break;
3651 case SCTP_SET_PEER_PRIMARY_ADDR:
3652 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3653 break;
3654 case SCTP_NODELAY:
3655 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3656 break;
3657 case SCTP_RTOINFO:
3658 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3659 break;
3660 case SCTP_ASSOCINFO:
3661 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3662 break;
3663 case SCTP_I_WANT_MAPPED_V4_ADDR:
3664 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3665 break;
3666 case SCTP_MAXSEG:
3667 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3668 break;
3669 case SCTP_ADAPTATION_LAYER:
3670 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3671 break;
3672 case SCTP_CONTEXT:
3673 retval = sctp_setsockopt_context(sk, optval, optlen);
3674 break;
3675 case SCTP_FRAGMENT_INTERLEAVE:
3676 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3677 break;
3678 case SCTP_MAX_BURST:
3679 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3680 break;
3681 case SCTP_AUTH_CHUNK:
3682 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3683 break;
3684 case SCTP_HMAC_IDENT:
3685 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3686 break;
3687 case SCTP_AUTH_KEY:
3688 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3689 break;
3690 case SCTP_AUTH_ACTIVE_KEY:
3691 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3692 break;
3693 case SCTP_AUTH_DELETE_KEY:
3694 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3695 break;
3696 case SCTP_AUTO_ASCONF:
3697 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3698 break;
3699 case SCTP_PEER_ADDR_THLDS:
3700 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3701 break;
3702 default:
3703 retval = -ENOPROTOOPT;
3704 break;
3707 sctp_release_sock(sk);
3709 out_nounlock:
3710 return retval;
3713 /* API 3.1.6 connect() - UDP Style Syntax
3715 * An application may use the connect() call in the UDP model to initiate an
3716 * association without sending data.
3718 * The syntax is:
3720 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3722 * sd: the socket descriptor to have a new association added to.
3724 * nam: the address structure (either struct sockaddr_in or struct
3725 * sockaddr_in6 defined in RFC2553 [7]).
3727 * len: the size of the address.
3729 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3730 int addr_len)
3732 int err = 0;
3733 struct sctp_af *af;
3735 sctp_lock_sock(sk);
3737 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3738 addr, addr_len);
3740 /* Validate addr_len before calling common connect/connectx routine. */
3741 af = sctp_get_af_specific(addr->sa_family);
3742 if (!af || addr_len < af->sockaddr_len) {
3743 err = -EINVAL;
3744 } else {
3745 /* Pass correct addr len to common routine (so it knows there
3746 * is only one address being passed.
3748 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3751 sctp_release_sock(sk);
3752 return err;
3755 /* FIXME: Write comments. */
3756 static int sctp_disconnect(struct sock *sk, int flags)
3758 return -EOPNOTSUPP; /* STUB */
3761 /* 4.1.4 accept() - TCP Style Syntax
3763 * Applications use accept() call to remove an established SCTP
3764 * association from the accept queue of the endpoint. A new socket
3765 * descriptor will be returned from accept() to represent the newly
3766 * formed association.
3768 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3770 struct sctp_sock *sp;
3771 struct sctp_endpoint *ep;
3772 struct sock *newsk = NULL;
3773 struct sctp_association *asoc;
3774 long timeo;
3775 int error = 0;
3777 sctp_lock_sock(sk);
3779 sp = sctp_sk(sk);
3780 ep = sp->ep;
3782 if (!sctp_style(sk, TCP)) {
3783 error = -EOPNOTSUPP;
3784 goto out;
3787 if (!sctp_sstate(sk, LISTENING)) {
3788 error = -EINVAL;
3789 goto out;
3792 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3794 error = sctp_wait_for_accept(sk, timeo);
3795 if (error)
3796 goto out;
3798 /* We treat the list of associations on the endpoint as the accept
3799 * queue and pick the first association on the list.
3801 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3803 newsk = sp->pf->create_accept_sk(sk, asoc);
3804 if (!newsk) {
3805 error = -ENOMEM;
3806 goto out;
3809 /* Populate the fields of the newsk from the oldsk and migrate the
3810 * asoc to the newsk.
3812 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3814 out:
3815 sctp_release_sock(sk);
3816 *err = error;
3817 return newsk;
3820 /* The SCTP ioctl handler. */
3821 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3823 int rc = -ENOTCONN;
3825 sctp_lock_sock(sk);
3828 * SEQPACKET-style sockets in LISTENING state are valid, for
3829 * SCTP, so only discard TCP-style sockets in LISTENING state.
3831 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3832 goto out;
3834 switch (cmd) {
3835 case SIOCINQ: {
3836 struct sk_buff *skb;
3837 unsigned int amount = 0;
3839 skb = skb_peek(&sk->sk_receive_queue);
3840 if (skb != NULL) {
3842 * We will only return the amount of this packet since
3843 * that is all that will be read.
3845 amount = skb->len;
3847 rc = put_user(amount, (int __user *)arg);
3848 break;
3850 default:
3851 rc = -ENOIOCTLCMD;
3852 break;
3854 out:
3855 sctp_release_sock(sk);
3856 return rc;
3859 /* This is the function which gets called during socket creation to
3860 * initialized the SCTP-specific portion of the sock.
3861 * The sock structure should already be zero-filled memory.
3863 static int sctp_init_sock(struct sock *sk)
3865 struct net *net = sock_net(sk);
3866 struct sctp_sock *sp;
3868 pr_debug("%s: sk:%p\n", __func__, sk);
3870 sp = sctp_sk(sk);
3872 /* Initialize the SCTP per socket area. */
3873 switch (sk->sk_type) {
3874 case SOCK_SEQPACKET:
3875 sp->type = SCTP_SOCKET_UDP;
3876 break;
3877 case SOCK_STREAM:
3878 sp->type = SCTP_SOCKET_TCP;
3879 break;
3880 default:
3881 return -ESOCKTNOSUPPORT;
3884 /* Initialize default send parameters. These parameters can be
3885 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3887 sp->default_stream = 0;
3888 sp->default_ppid = 0;
3889 sp->default_flags = 0;
3890 sp->default_context = 0;
3891 sp->default_timetolive = 0;
3893 sp->default_rcv_context = 0;
3894 sp->max_burst = net->sctp.max_burst;
3896 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3898 /* Initialize default setup parameters. These parameters
3899 * can be modified with the SCTP_INITMSG socket option or
3900 * overridden by the SCTP_INIT CMSG.
3902 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3903 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3904 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3905 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3907 /* Initialize default RTO related parameters. These parameters can
3908 * be modified for with the SCTP_RTOINFO socket option.
3910 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3911 sp->rtoinfo.srto_max = net->sctp.rto_max;
3912 sp->rtoinfo.srto_min = net->sctp.rto_min;
3914 /* Initialize default association related parameters. These parameters
3915 * can be modified with the SCTP_ASSOCINFO socket option.
3917 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3918 sp->assocparams.sasoc_number_peer_destinations = 0;
3919 sp->assocparams.sasoc_peer_rwnd = 0;
3920 sp->assocparams.sasoc_local_rwnd = 0;
3921 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3923 /* Initialize default event subscriptions. By default, all the
3924 * options are off.
3926 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3928 /* Default Peer Address Parameters. These defaults can
3929 * be modified via SCTP_PEER_ADDR_PARAMS
3931 sp->hbinterval = net->sctp.hb_interval;
3932 sp->pathmaxrxt = net->sctp.max_retrans_path;
3933 sp->pathmtu = 0; // allow default discovery
3934 sp->sackdelay = net->sctp.sack_timeout;
3935 sp->sackfreq = 2;
3936 sp->param_flags = SPP_HB_ENABLE |
3937 SPP_PMTUD_ENABLE |
3938 SPP_SACKDELAY_ENABLE;
3940 /* If enabled no SCTP message fragmentation will be performed.
3941 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3943 sp->disable_fragments = 0;
3945 /* Enable Nagle algorithm by default. */
3946 sp->nodelay = 0;
3948 /* Enable by default. */
3949 sp->v4mapped = 1;
3951 /* Auto-close idle associations after the configured
3952 * number of seconds. A value of 0 disables this
3953 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3954 * for UDP-style sockets only.
3956 sp->autoclose = 0;
3958 /* User specified fragmentation limit. */
3959 sp->user_frag = 0;
3961 sp->adaptation_ind = 0;
3963 sp->pf = sctp_get_pf_specific(sk->sk_family);
3965 /* Control variables for partial data delivery. */
3966 atomic_set(&sp->pd_mode, 0);
3967 skb_queue_head_init(&sp->pd_lobby);
3968 sp->frag_interleave = 0;
3970 /* Create a per socket endpoint structure. Even if we
3971 * change the data structure relationships, this may still
3972 * be useful for storing pre-connect address information.
3974 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3975 if (!sp->ep)
3976 return -ENOMEM;
3978 sp->hmac = NULL;
3980 sk->sk_destruct = sctp_destruct_sock;
3982 SCTP_DBG_OBJCNT_INC(sock);
3984 local_bh_disable();
3985 percpu_counter_inc(&sctp_sockets_allocated);
3986 sock_prot_inuse_add(net, sk->sk_prot, 1);
3987 if (net->sctp.default_auto_asconf) {
3988 list_add_tail(&sp->auto_asconf_list,
3989 &net->sctp.auto_asconf_splist);
3990 sp->do_auto_asconf = 1;
3991 } else
3992 sp->do_auto_asconf = 0;
3993 local_bh_enable();
3995 return 0;
3998 /* Cleanup any SCTP per socket resources. */
3999 static void sctp_destroy_sock(struct sock *sk)
4001 struct sctp_sock *sp;
4003 pr_debug("%s: sk:%p\n", __func__, sk);
4005 /* Release our hold on the endpoint. */
4006 sp = sctp_sk(sk);
4007 /* This could happen during socket init, thus we bail out
4008 * early, since the rest of the below is not setup either.
4010 if (sp->ep == NULL)
4011 return;
4013 if (sp->do_auto_asconf) {
4014 sp->do_auto_asconf = 0;
4015 list_del(&sp->auto_asconf_list);
4017 sctp_endpoint_free(sp->ep);
4018 local_bh_disable();
4019 percpu_counter_dec(&sctp_sockets_allocated);
4020 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4021 local_bh_enable();
4024 /* Triggered when there are no references on the socket anymore */
4025 static void sctp_destruct_sock(struct sock *sk)
4027 struct sctp_sock *sp = sctp_sk(sk);
4029 /* Free up the HMAC transform. */
4030 crypto_free_hash(sp->hmac);
4032 inet_sock_destruct(sk);
4035 /* API 4.1.7 shutdown() - TCP Style Syntax
4036 * int shutdown(int socket, int how);
4038 * sd - the socket descriptor of the association to be closed.
4039 * how - Specifies the type of shutdown. The values are
4040 * as follows:
4041 * SHUT_RD
4042 * Disables further receive operations. No SCTP
4043 * protocol action is taken.
4044 * SHUT_WR
4045 * Disables further send operations, and initiates
4046 * the SCTP shutdown sequence.
4047 * SHUT_RDWR
4048 * Disables further send and receive operations
4049 * and initiates the SCTP shutdown sequence.
4051 static void sctp_shutdown(struct sock *sk, int how)
4053 struct net *net = sock_net(sk);
4054 struct sctp_endpoint *ep;
4055 struct sctp_association *asoc;
4057 if (!sctp_style(sk, TCP))
4058 return;
4060 if (how & SEND_SHUTDOWN) {
4061 ep = sctp_sk(sk)->ep;
4062 if (!list_empty(&ep->asocs)) {
4063 asoc = list_entry(ep->asocs.next,
4064 struct sctp_association, asocs);
4065 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4070 /* 7.2.1 Association Status (SCTP_STATUS)
4072 * Applications can retrieve current status information about an
4073 * association, including association state, peer receiver window size,
4074 * number of unacked data chunks, and number of data chunks pending
4075 * receipt. This information is read-only.
4077 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4078 char __user *optval,
4079 int __user *optlen)
4081 struct sctp_status status;
4082 struct sctp_association *asoc = NULL;
4083 struct sctp_transport *transport;
4084 sctp_assoc_t associd;
4085 int retval = 0;
4087 if (len < sizeof(status)) {
4088 retval = -EINVAL;
4089 goto out;
4092 len = sizeof(status);
4093 if (copy_from_user(&status, optval, len)) {
4094 retval = -EFAULT;
4095 goto out;
4098 associd = status.sstat_assoc_id;
4099 asoc = sctp_id2assoc(sk, associd);
4100 if (!asoc) {
4101 retval = -EINVAL;
4102 goto out;
4105 transport = asoc->peer.primary_path;
4107 status.sstat_assoc_id = sctp_assoc2id(asoc);
4108 status.sstat_state = asoc->state;
4109 status.sstat_rwnd = asoc->peer.rwnd;
4110 status.sstat_unackdata = asoc->unack_data;
4112 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4113 status.sstat_instrms = asoc->c.sinit_max_instreams;
4114 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4115 status.sstat_fragmentation_point = asoc->frag_point;
4116 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4117 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4118 transport->af_specific->sockaddr_len);
4119 /* Map ipv4 address into v4-mapped-on-v6 address. */
4120 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4121 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4122 status.sstat_primary.spinfo_state = transport->state;
4123 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4124 status.sstat_primary.spinfo_srtt = transport->srtt;
4125 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4126 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4128 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4129 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4131 if (put_user(len, optlen)) {
4132 retval = -EFAULT;
4133 goto out;
4136 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4137 __func__, len, status.sstat_state, status.sstat_rwnd,
4138 status.sstat_assoc_id);
4140 if (copy_to_user(optval, &status, len)) {
4141 retval = -EFAULT;
4142 goto out;
4145 out:
4146 return retval;
4150 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4152 * Applications can retrieve information about a specific peer address
4153 * of an association, including its reachability state, congestion
4154 * window, and retransmission timer values. This information is
4155 * read-only.
4157 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4158 char __user *optval,
4159 int __user *optlen)
4161 struct sctp_paddrinfo pinfo;
4162 struct sctp_transport *transport;
4163 int retval = 0;
4165 if (len < sizeof(pinfo)) {
4166 retval = -EINVAL;
4167 goto out;
4170 len = sizeof(pinfo);
4171 if (copy_from_user(&pinfo, optval, len)) {
4172 retval = -EFAULT;
4173 goto out;
4176 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4177 pinfo.spinfo_assoc_id);
4178 if (!transport)
4179 return -EINVAL;
4181 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4182 pinfo.spinfo_state = transport->state;
4183 pinfo.spinfo_cwnd = transport->cwnd;
4184 pinfo.spinfo_srtt = transport->srtt;
4185 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4186 pinfo.spinfo_mtu = transport->pathmtu;
4188 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4189 pinfo.spinfo_state = SCTP_ACTIVE;
4191 if (put_user(len, optlen)) {
4192 retval = -EFAULT;
4193 goto out;
4196 if (copy_to_user(optval, &pinfo, len)) {
4197 retval = -EFAULT;
4198 goto out;
4201 out:
4202 return retval;
4205 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4207 * This option is a on/off flag. If enabled no SCTP message
4208 * fragmentation will be performed. Instead if a message being sent
4209 * exceeds the current PMTU size, the message will NOT be sent and
4210 * instead a error will be indicated to the user.
4212 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4213 char __user *optval, int __user *optlen)
4215 int val;
4217 if (len < sizeof(int))
4218 return -EINVAL;
4220 len = sizeof(int);
4221 val = (sctp_sk(sk)->disable_fragments == 1);
4222 if (put_user(len, optlen))
4223 return -EFAULT;
4224 if (copy_to_user(optval, &val, len))
4225 return -EFAULT;
4226 return 0;
4229 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4231 * This socket option is used to specify various notifications and
4232 * ancillary data the user wishes to receive.
4234 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4235 int __user *optlen)
4237 if (len <= 0)
4238 return -EINVAL;
4239 if (len > sizeof(struct sctp_event_subscribe))
4240 len = sizeof(struct sctp_event_subscribe);
4241 if (put_user(len, optlen))
4242 return -EFAULT;
4243 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4244 return -EFAULT;
4245 return 0;
4248 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4250 * This socket option is applicable to the UDP-style socket only. When
4251 * set it will cause associations that are idle for more than the
4252 * specified number of seconds to automatically close. An association
4253 * being idle is defined an association that has NOT sent or received
4254 * user data. The special value of '0' indicates that no automatic
4255 * close of any associations should be performed. The option expects an
4256 * integer defining the number of seconds of idle time before an
4257 * association is closed.
4259 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4261 /* Applicable to UDP-style socket only */
4262 if (sctp_style(sk, TCP))
4263 return -EOPNOTSUPP;
4264 if (len < sizeof(int))
4265 return -EINVAL;
4266 len = sizeof(int);
4267 if (put_user(len, optlen))
4268 return -EFAULT;
4269 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4270 return -EFAULT;
4271 return 0;
4274 /* Helper routine to branch off an association to a new socket. */
4275 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4277 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4278 struct socket *sock;
4279 struct sctp_af *af;
4280 int err = 0;
4282 if (!asoc)
4283 return -EINVAL;
4285 /* An association cannot be branched off from an already peeled-off
4286 * socket, nor is this supported for tcp style sockets.
4288 if (!sctp_style(sk, UDP))
4289 return -EINVAL;
4291 /* Create a new socket. */
4292 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4293 if (err < 0)
4294 return err;
4296 sctp_copy_sock(sock->sk, sk, asoc);
4298 /* Make peeled-off sockets more like 1-1 accepted sockets.
4299 * Set the daddr and initialize id to something more random
4301 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4302 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4304 /* Populate the fields of the newsk from the oldsk and migrate the
4305 * asoc to the newsk.
4307 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4309 *sockp = sock;
4311 return err;
4313 EXPORT_SYMBOL(sctp_do_peeloff);
4315 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4317 sctp_peeloff_arg_t peeloff;
4318 struct socket *newsock;
4319 struct file *newfile;
4320 int retval = 0;
4322 if (len < sizeof(sctp_peeloff_arg_t))
4323 return -EINVAL;
4324 len = sizeof(sctp_peeloff_arg_t);
4325 if (copy_from_user(&peeloff, optval, len))
4326 return -EFAULT;
4328 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4329 if (retval < 0)
4330 goto out;
4332 /* Map the socket to an unused fd that can be returned to the user. */
4333 retval = get_unused_fd_flags(0);
4334 if (retval < 0) {
4335 sock_release(newsock);
4336 goto out;
4339 newfile = sock_alloc_file(newsock, 0, NULL);
4340 if (unlikely(IS_ERR(newfile))) {
4341 put_unused_fd(retval);
4342 sock_release(newsock);
4343 return PTR_ERR(newfile);
4346 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4347 retval);
4349 /* Return the fd mapped to the new socket. */
4350 if (put_user(len, optlen)) {
4351 fput(newfile);
4352 put_unused_fd(retval);
4353 return -EFAULT;
4355 peeloff.sd = retval;
4356 if (copy_to_user(optval, &peeloff, len)) {
4357 fput(newfile);
4358 put_unused_fd(retval);
4359 return -EFAULT;
4361 fd_install(retval, newfile);
4362 out:
4363 return retval;
4366 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4368 * Applications can enable or disable heartbeats for any peer address of
4369 * an association, modify an address's heartbeat interval, force a
4370 * heartbeat to be sent immediately, and adjust the address's maximum
4371 * number of retransmissions sent before an address is considered
4372 * unreachable. The following structure is used to access and modify an
4373 * address's parameters:
4375 * struct sctp_paddrparams {
4376 * sctp_assoc_t spp_assoc_id;
4377 * struct sockaddr_storage spp_address;
4378 * uint32_t spp_hbinterval;
4379 * uint16_t spp_pathmaxrxt;
4380 * uint32_t spp_pathmtu;
4381 * uint32_t spp_sackdelay;
4382 * uint32_t spp_flags;
4383 * };
4385 * spp_assoc_id - (one-to-many style socket) This is filled in the
4386 * application, and identifies the association for
4387 * this query.
4388 * spp_address - This specifies which address is of interest.
4389 * spp_hbinterval - This contains the value of the heartbeat interval,
4390 * in milliseconds. If a value of zero
4391 * is present in this field then no changes are to
4392 * be made to this parameter.
4393 * spp_pathmaxrxt - This contains the maximum number of
4394 * retransmissions before this address shall be
4395 * considered unreachable. If a value of zero
4396 * is present in this field then no changes are to
4397 * be made to this parameter.
4398 * spp_pathmtu - When Path MTU discovery is disabled the value
4399 * specified here will be the "fixed" path mtu.
4400 * Note that if the spp_address field is empty
4401 * then all associations on this address will
4402 * have this fixed path mtu set upon them.
4404 * spp_sackdelay - When delayed sack is enabled, this value specifies
4405 * the number of milliseconds that sacks will be delayed
4406 * for. This value will apply to all addresses of an
4407 * association if the spp_address field is empty. Note
4408 * also, that if delayed sack is enabled and this
4409 * value is set to 0, no change is made to the last
4410 * recorded delayed sack timer value.
4412 * spp_flags - These flags are used to control various features
4413 * on an association. The flag field may contain
4414 * zero or more of the following options.
4416 * SPP_HB_ENABLE - Enable heartbeats on the
4417 * specified address. Note that if the address
4418 * field is empty all addresses for the association
4419 * have heartbeats enabled upon them.
4421 * SPP_HB_DISABLE - Disable heartbeats on the
4422 * speicifed address. Note that if the address
4423 * field is empty all addresses for the association
4424 * will have their heartbeats disabled. Note also
4425 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4426 * mutually exclusive, only one of these two should
4427 * be specified. Enabling both fields will have
4428 * undetermined results.
4430 * SPP_HB_DEMAND - Request a user initiated heartbeat
4431 * to be made immediately.
4433 * SPP_PMTUD_ENABLE - This field will enable PMTU
4434 * discovery upon the specified address. Note that
4435 * if the address feild is empty then all addresses
4436 * on the association are effected.
4438 * SPP_PMTUD_DISABLE - This field will disable PMTU
4439 * discovery upon the specified address. Note that
4440 * if the address feild is empty then all addresses
4441 * on the association are effected. Not also that
4442 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4443 * exclusive. Enabling both will have undetermined
4444 * results.
4446 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4447 * on delayed sack. The time specified in spp_sackdelay
4448 * is used to specify the sack delay for this address. Note
4449 * that if spp_address is empty then all addresses will
4450 * enable delayed sack and take on the sack delay
4451 * value specified in spp_sackdelay.
4452 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4453 * off delayed sack. If the spp_address field is blank then
4454 * delayed sack is disabled for the entire association. Note
4455 * also that this field is mutually exclusive to
4456 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4457 * results.
4459 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4460 char __user *optval, int __user *optlen)
4462 struct sctp_paddrparams params;
4463 struct sctp_transport *trans = NULL;
4464 struct sctp_association *asoc = NULL;
4465 struct sctp_sock *sp = sctp_sk(sk);
4467 if (len < sizeof(struct sctp_paddrparams))
4468 return -EINVAL;
4469 len = sizeof(struct sctp_paddrparams);
4470 if (copy_from_user(&params, optval, len))
4471 return -EFAULT;
4473 /* If an address other than INADDR_ANY is specified, and
4474 * no transport is found, then the request is invalid.
4476 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4477 trans = sctp_addr_id2transport(sk, &params.spp_address,
4478 params.spp_assoc_id);
4479 if (!trans) {
4480 pr_debug("%s: failed no transport\n", __func__);
4481 return -EINVAL;
4485 /* Get association, if assoc_id != 0 and the socket is a one
4486 * to many style socket, and an association was not found, then
4487 * the id was invalid.
4489 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4490 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4491 pr_debug("%s: failed no association\n", __func__);
4492 return -EINVAL;
4495 if (trans) {
4496 /* Fetch transport values. */
4497 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4498 params.spp_pathmtu = trans->pathmtu;
4499 params.spp_pathmaxrxt = trans->pathmaxrxt;
4500 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4502 /*draft-11 doesn't say what to return in spp_flags*/
4503 params.spp_flags = trans->param_flags;
4504 } else if (asoc) {
4505 /* Fetch association values. */
4506 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4507 params.spp_pathmtu = asoc->pathmtu;
4508 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4509 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4511 /*draft-11 doesn't say what to return in spp_flags*/
4512 params.spp_flags = asoc->param_flags;
4513 } else {
4514 /* Fetch socket values. */
4515 params.spp_hbinterval = sp->hbinterval;
4516 params.spp_pathmtu = sp->pathmtu;
4517 params.spp_sackdelay = sp->sackdelay;
4518 params.spp_pathmaxrxt = sp->pathmaxrxt;
4520 /*draft-11 doesn't say what to return in spp_flags*/
4521 params.spp_flags = sp->param_flags;
4524 if (copy_to_user(optval, &params, len))
4525 return -EFAULT;
4527 if (put_user(len, optlen))
4528 return -EFAULT;
4530 return 0;
4534 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4536 * This option will effect the way delayed acks are performed. This
4537 * option allows you to get or set the delayed ack time, in
4538 * milliseconds. It also allows changing the delayed ack frequency.
4539 * Changing the frequency to 1 disables the delayed sack algorithm. If
4540 * the assoc_id is 0, then this sets or gets the endpoints default
4541 * values. If the assoc_id field is non-zero, then the set or get
4542 * effects the specified association for the one to many model (the
4543 * assoc_id field is ignored by the one to one model). Note that if
4544 * sack_delay or sack_freq are 0 when setting this option, then the
4545 * current values will remain unchanged.
4547 * struct sctp_sack_info {
4548 * sctp_assoc_t sack_assoc_id;
4549 * uint32_t sack_delay;
4550 * uint32_t sack_freq;
4551 * };
4553 * sack_assoc_id - This parameter, indicates which association the user
4554 * is performing an action upon. Note that if this field's value is
4555 * zero then the endpoints default value is changed (effecting future
4556 * associations only).
4558 * sack_delay - This parameter contains the number of milliseconds that
4559 * the user is requesting the delayed ACK timer be set to. Note that
4560 * this value is defined in the standard to be between 200 and 500
4561 * milliseconds.
4563 * sack_freq - This parameter contains the number of packets that must
4564 * be received before a sack is sent without waiting for the delay
4565 * timer to expire. The default value for this is 2, setting this
4566 * value to 1 will disable the delayed sack algorithm.
4568 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4569 char __user *optval,
4570 int __user *optlen)
4572 struct sctp_sack_info params;
4573 struct sctp_association *asoc = NULL;
4574 struct sctp_sock *sp = sctp_sk(sk);
4576 if (len >= sizeof(struct sctp_sack_info)) {
4577 len = sizeof(struct sctp_sack_info);
4579 if (copy_from_user(&params, optval, len))
4580 return -EFAULT;
4581 } else if (len == sizeof(struct sctp_assoc_value)) {
4582 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4583 pr_warn("Use struct sctp_sack_info instead\n");
4584 if (copy_from_user(&params, optval, len))
4585 return -EFAULT;
4586 } else
4587 return - EINVAL;
4589 /* Get association, if sack_assoc_id != 0 and the socket is a one
4590 * to many style socket, and an association was not found, then
4591 * the id was invalid.
4593 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4594 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4595 return -EINVAL;
4597 if (asoc) {
4598 /* Fetch association values. */
4599 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4600 params.sack_delay = jiffies_to_msecs(
4601 asoc->sackdelay);
4602 params.sack_freq = asoc->sackfreq;
4604 } else {
4605 params.sack_delay = 0;
4606 params.sack_freq = 1;
4608 } else {
4609 /* Fetch socket values. */
4610 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4611 params.sack_delay = sp->sackdelay;
4612 params.sack_freq = sp->sackfreq;
4613 } else {
4614 params.sack_delay = 0;
4615 params.sack_freq = 1;
4619 if (copy_to_user(optval, &params, len))
4620 return -EFAULT;
4622 if (put_user(len, optlen))
4623 return -EFAULT;
4625 return 0;
4628 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4630 * Applications can specify protocol parameters for the default association
4631 * initialization. The option name argument to setsockopt() and getsockopt()
4632 * is SCTP_INITMSG.
4634 * Setting initialization parameters is effective only on an unconnected
4635 * socket (for UDP-style sockets only future associations are effected
4636 * by the change). With TCP-style sockets, this option is inherited by
4637 * sockets derived from a listener socket.
4639 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4641 if (len < sizeof(struct sctp_initmsg))
4642 return -EINVAL;
4643 len = sizeof(struct sctp_initmsg);
4644 if (put_user(len, optlen))
4645 return -EFAULT;
4646 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4647 return -EFAULT;
4648 return 0;
4652 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4653 char __user *optval, int __user *optlen)
4655 struct sctp_association *asoc;
4656 int cnt = 0;
4657 struct sctp_getaddrs getaddrs;
4658 struct sctp_transport *from;
4659 void __user *to;
4660 union sctp_addr temp;
4661 struct sctp_sock *sp = sctp_sk(sk);
4662 int addrlen;
4663 size_t space_left;
4664 int bytes_copied;
4666 if (len < sizeof(struct sctp_getaddrs))
4667 return -EINVAL;
4669 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4670 return -EFAULT;
4672 /* For UDP-style sockets, id specifies the association to query. */
4673 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4674 if (!asoc)
4675 return -EINVAL;
4677 to = optval + offsetof(struct sctp_getaddrs,addrs);
4678 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4680 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4681 transports) {
4682 memcpy(&temp, &from->ipaddr, sizeof(temp));
4683 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4684 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4685 if (space_left < addrlen)
4686 return -ENOMEM;
4687 if (copy_to_user(to, &temp, addrlen))
4688 return -EFAULT;
4689 to += addrlen;
4690 cnt++;
4691 space_left -= addrlen;
4694 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4695 return -EFAULT;
4696 bytes_copied = ((char __user *)to) - optval;
4697 if (put_user(bytes_copied, optlen))
4698 return -EFAULT;
4700 return 0;
4703 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4704 size_t space_left, int *bytes_copied)
4706 struct sctp_sockaddr_entry *addr;
4707 union sctp_addr temp;
4708 int cnt = 0;
4709 int addrlen;
4710 struct net *net = sock_net(sk);
4712 rcu_read_lock();
4713 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4714 if (!addr->valid)
4715 continue;
4717 if ((PF_INET == sk->sk_family) &&
4718 (AF_INET6 == addr->a.sa.sa_family))
4719 continue;
4720 if ((PF_INET6 == sk->sk_family) &&
4721 inet_v6_ipv6only(sk) &&
4722 (AF_INET == addr->a.sa.sa_family))
4723 continue;
4724 memcpy(&temp, &addr->a, sizeof(temp));
4725 if (!temp.v4.sin_port)
4726 temp.v4.sin_port = htons(port);
4728 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4729 &temp);
4730 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4731 if (space_left < addrlen) {
4732 cnt = -ENOMEM;
4733 break;
4735 memcpy(to, &temp, addrlen);
4737 to += addrlen;
4738 cnt ++;
4739 space_left -= addrlen;
4740 *bytes_copied += addrlen;
4742 rcu_read_unlock();
4744 return cnt;
4748 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4749 char __user *optval, int __user *optlen)
4751 struct sctp_bind_addr *bp;
4752 struct sctp_association *asoc;
4753 int cnt = 0;
4754 struct sctp_getaddrs getaddrs;
4755 struct sctp_sockaddr_entry *addr;
4756 void __user *to;
4757 union sctp_addr temp;
4758 struct sctp_sock *sp = sctp_sk(sk);
4759 int addrlen;
4760 int err = 0;
4761 size_t space_left;
4762 int bytes_copied = 0;
4763 void *addrs;
4764 void *buf;
4766 if (len < sizeof(struct sctp_getaddrs))
4767 return -EINVAL;
4769 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4770 return -EFAULT;
4773 * For UDP-style sockets, id specifies the association to query.
4774 * If the id field is set to the value '0' then the locally bound
4775 * addresses are returned without regard to any particular
4776 * association.
4778 if (0 == getaddrs.assoc_id) {
4779 bp = &sctp_sk(sk)->ep->base.bind_addr;
4780 } else {
4781 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4782 if (!asoc)
4783 return -EINVAL;
4784 bp = &asoc->base.bind_addr;
4787 to = optval + offsetof(struct sctp_getaddrs,addrs);
4788 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4790 addrs = kmalloc(space_left, GFP_KERNEL);
4791 if (!addrs)
4792 return -ENOMEM;
4794 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4795 * addresses from the global local address list.
4797 if (sctp_list_single_entry(&bp->address_list)) {
4798 addr = list_entry(bp->address_list.next,
4799 struct sctp_sockaddr_entry, list);
4800 if (sctp_is_any(sk, &addr->a)) {
4801 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4802 space_left, &bytes_copied);
4803 if (cnt < 0) {
4804 err = cnt;
4805 goto out;
4807 goto copy_getaddrs;
4811 buf = addrs;
4812 /* Protection on the bound address list is not needed since
4813 * in the socket option context we hold a socket lock and
4814 * thus the bound address list can't change.
4816 list_for_each_entry(addr, &bp->address_list, list) {
4817 memcpy(&temp, &addr->a, sizeof(temp));
4818 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4819 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4820 if (space_left < addrlen) {
4821 err = -ENOMEM; /*fixme: right error?*/
4822 goto out;
4824 memcpy(buf, &temp, addrlen);
4825 buf += addrlen;
4826 bytes_copied += addrlen;
4827 cnt ++;
4828 space_left -= addrlen;
4831 copy_getaddrs:
4832 if (copy_to_user(to, addrs, bytes_copied)) {
4833 err = -EFAULT;
4834 goto out;
4836 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4837 err = -EFAULT;
4838 goto out;
4840 if (put_user(bytes_copied, optlen))
4841 err = -EFAULT;
4842 out:
4843 kfree(addrs);
4844 return err;
4847 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4849 * Requests that the local SCTP stack use the enclosed peer address as
4850 * the association primary. The enclosed address must be one of the
4851 * association peer's addresses.
4853 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4854 char __user *optval, int __user *optlen)
4856 struct sctp_prim prim;
4857 struct sctp_association *asoc;
4858 struct sctp_sock *sp = sctp_sk(sk);
4860 if (len < sizeof(struct sctp_prim))
4861 return -EINVAL;
4863 len = sizeof(struct sctp_prim);
4865 if (copy_from_user(&prim, optval, len))
4866 return -EFAULT;
4868 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4869 if (!asoc)
4870 return -EINVAL;
4872 if (!asoc->peer.primary_path)
4873 return -ENOTCONN;
4875 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4876 asoc->peer.primary_path->af_specific->sockaddr_len);
4878 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4879 (union sctp_addr *)&prim.ssp_addr);
4881 if (put_user(len, optlen))
4882 return -EFAULT;
4883 if (copy_to_user(optval, &prim, len))
4884 return -EFAULT;
4886 return 0;
4890 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4892 * Requests that the local endpoint set the specified Adaptation Layer
4893 * Indication parameter for all future INIT and INIT-ACK exchanges.
4895 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4896 char __user *optval, int __user *optlen)
4898 struct sctp_setadaptation adaptation;
4900 if (len < sizeof(struct sctp_setadaptation))
4901 return -EINVAL;
4903 len = sizeof(struct sctp_setadaptation);
4905 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4907 if (put_user(len, optlen))
4908 return -EFAULT;
4909 if (copy_to_user(optval, &adaptation, len))
4910 return -EFAULT;
4912 return 0;
4917 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4919 * Applications that wish to use the sendto() system call may wish to
4920 * specify a default set of parameters that would normally be supplied
4921 * through the inclusion of ancillary data. This socket option allows
4922 * such an application to set the default sctp_sndrcvinfo structure.
4925 * The application that wishes to use this socket option simply passes
4926 * in to this call the sctp_sndrcvinfo structure defined in Section
4927 * 5.2.2) The input parameters accepted by this call include
4928 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4929 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4930 * to this call if the caller is using the UDP model.
4932 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4934 static int sctp_getsockopt_default_send_param(struct sock *sk,
4935 int len, char __user *optval,
4936 int __user *optlen)
4938 struct sctp_sndrcvinfo info;
4939 struct sctp_association *asoc;
4940 struct sctp_sock *sp = sctp_sk(sk);
4942 if (len < sizeof(struct sctp_sndrcvinfo))
4943 return -EINVAL;
4945 len = sizeof(struct sctp_sndrcvinfo);
4947 if (copy_from_user(&info, optval, len))
4948 return -EFAULT;
4950 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4951 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4952 return -EINVAL;
4954 if (asoc) {
4955 info.sinfo_stream = asoc->default_stream;
4956 info.sinfo_flags = asoc->default_flags;
4957 info.sinfo_ppid = asoc->default_ppid;
4958 info.sinfo_context = asoc->default_context;
4959 info.sinfo_timetolive = asoc->default_timetolive;
4960 } else {
4961 info.sinfo_stream = sp->default_stream;
4962 info.sinfo_flags = sp->default_flags;
4963 info.sinfo_ppid = sp->default_ppid;
4964 info.sinfo_context = sp->default_context;
4965 info.sinfo_timetolive = sp->default_timetolive;
4968 if (put_user(len, optlen))
4969 return -EFAULT;
4970 if (copy_to_user(optval, &info, len))
4971 return -EFAULT;
4973 return 0;
4978 * 7.1.5 SCTP_NODELAY
4980 * Turn on/off any Nagle-like algorithm. This means that packets are
4981 * generally sent as soon as possible and no unnecessary delays are
4982 * introduced, at the cost of more packets in the network. Expects an
4983 * integer boolean flag.
4986 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4987 char __user *optval, int __user *optlen)
4989 int val;
4991 if (len < sizeof(int))
4992 return -EINVAL;
4994 len = sizeof(int);
4995 val = (sctp_sk(sk)->nodelay == 1);
4996 if (put_user(len, optlen))
4997 return -EFAULT;
4998 if (copy_to_user(optval, &val, len))
4999 return -EFAULT;
5000 return 0;
5005 * 7.1.1 SCTP_RTOINFO
5007 * The protocol parameters used to initialize and bound retransmission
5008 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5009 * and modify these parameters.
5010 * All parameters are time values, in milliseconds. A value of 0, when
5011 * modifying the parameters, indicates that the current value should not
5012 * be changed.
5015 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5016 char __user *optval,
5017 int __user *optlen) {
5018 struct sctp_rtoinfo rtoinfo;
5019 struct sctp_association *asoc;
5021 if (len < sizeof (struct sctp_rtoinfo))
5022 return -EINVAL;
5024 len = sizeof(struct sctp_rtoinfo);
5026 if (copy_from_user(&rtoinfo, optval, len))
5027 return -EFAULT;
5029 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5031 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5032 return -EINVAL;
5034 /* Values corresponding to the specific association. */
5035 if (asoc) {
5036 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5037 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5038 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5039 } else {
5040 /* Values corresponding to the endpoint. */
5041 struct sctp_sock *sp = sctp_sk(sk);
5043 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5044 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5045 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5048 if (put_user(len, optlen))
5049 return -EFAULT;
5051 if (copy_to_user(optval, &rtoinfo, len))
5052 return -EFAULT;
5054 return 0;
5059 * 7.1.2 SCTP_ASSOCINFO
5061 * This option is used to tune the maximum retransmission attempts
5062 * of the association.
5063 * Returns an error if the new association retransmission value is
5064 * greater than the sum of the retransmission value of the peer.
5065 * See [SCTP] for more information.
5068 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5069 char __user *optval,
5070 int __user *optlen)
5073 struct sctp_assocparams assocparams;
5074 struct sctp_association *asoc;
5075 struct list_head *pos;
5076 int cnt = 0;
5078 if (len < sizeof (struct sctp_assocparams))
5079 return -EINVAL;
5081 len = sizeof(struct sctp_assocparams);
5083 if (copy_from_user(&assocparams, optval, len))
5084 return -EFAULT;
5086 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5088 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5089 return -EINVAL;
5091 /* Values correspoinding to the specific association */
5092 if (asoc) {
5093 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5094 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5095 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5096 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5098 list_for_each(pos, &asoc->peer.transport_addr_list) {
5099 cnt ++;
5102 assocparams.sasoc_number_peer_destinations = cnt;
5103 } else {
5104 /* Values corresponding to the endpoint */
5105 struct sctp_sock *sp = sctp_sk(sk);
5107 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5108 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5109 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5110 assocparams.sasoc_cookie_life =
5111 sp->assocparams.sasoc_cookie_life;
5112 assocparams.sasoc_number_peer_destinations =
5113 sp->assocparams.
5114 sasoc_number_peer_destinations;
5117 if (put_user(len, optlen))
5118 return -EFAULT;
5120 if (copy_to_user(optval, &assocparams, len))
5121 return -EFAULT;
5123 return 0;
5127 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5129 * This socket option is a boolean flag which turns on or off mapped V4
5130 * addresses. If this option is turned on and the socket is type
5131 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5132 * If this option is turned off, then no mapping will be done of V4
5133 * addresses and a user will receive both PF_INET6 and PF_INET type
5134 * addresses on the socket.
5136 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5137 char __user *optval, int __user *optlen)
5139 int val;
5140 struct sctp_sock *sp = sctp_sk(sk);
5142 if (len < sizeof(int))
5143 return -EINVAL;
5145 len = sizeof(int);
5146 val = sp->v4mapped;
5147 if (put_user(len, optlen))
5148 return -EFAULT;
5149 if (copy_to_user(optval, &val, len))
5150 return -EFAULT;
5152 return 0;
5156 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5157 * (chapter and verse is quoted at sctp_setsockopt_context())
5159 static int sctp_getsockopt_context(struct sock *sk, int len,
5160 char __user *optval, int __user *optlen)
5162 struct sctp_assoc_value params;
5163 struct sctp_sock *sp;
5164 struct sctp_association *asoc;
5166 if (len < sizeof(struct sctp_assoc_value))
5167 return -EINVAL;
5169 len = sizeof(struct sctp_assoc_value);
5171 if (copy_from_user(&params, optval, len))
5172 return -EFAULT;
5174 sp = sctp_sk(sk);
5176 if (params.assoc_id != 0) {
5177 asoc = sctp_id2assoc(sk, params.assoc_id);
5178 if (!asoc)
5179 return -EINVAL;
5180 params.assoc_value = asoc->default_rcv_context;
5181 } else {
5182 params.assoc_value = sp->default_rcv_context;
5185 if (put_user(len, optlen))
5186 return -EFAULT;
5187 if (copy_to_user(optval, &params, len))
5188 return -EFAULT;
5190 return 0;
5194 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5195 * This option will get or set the maximum size to put in any outgoing
5196 * SCTP DATA chunk. If a message is larger than this size it will be
5197 * fragmented by SCTP into the specified size. Note that the underlying
5198 * SCTP implementation may fragment into smaller sized chunks when the
5199 * PMTU of the underlying association is smaller than the value set by
5200 * the user. The default value for this option is '0' which indicates
5201 * the user is NOT limiting fragmentation and only the PMTU will effect
5202 * SCTP's choice of DATA chunk size. Note also that values set larger
5203 * than the maximum size of an IP datagram will effectively let SCTP
5204 * control fragmentation (i.e. the same as setting this option to 0).
5206 * The following structure is used to access and modify this parameter:
5208 * struct sctp_assoc_value {
5209 * sctp_assoc_t assoc_id;
5210 * uint32_t assoc_value;
5211 * };
5213 * assoc_id: This parameter is ignored for one-to-one style sockets.
5214 * For one-to-many style sockets this parameter indicates which
5215 * association the user is performing an action upon. Note that if
5216 * this field's value is zero then the endpoints default value is
5217 * changed (effecting future associations only).
5218 * assoc_value: This parameter specifies the maximum size in bytes.
5220 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5221 char __user *optval, int __user *optlen)
5223 struct sctp_assoc_value params;
5224 struct sctp_association *asoc;
5226 if (len == sizeof(int)) {
5227 pr_warn("Use of int in maxseg socket option deprecated\n");
5228 pr_warn("Use struct sctp_assoc_value instead\n");
5229 params.assoc_id = 0;
5230 } else if (len >= sizeof(struct sctp_assoc_value)) {
5231 len = sizeof(struct sctp_assoc_value);
5232 if (copy_from_user(&params, optval, sizeof(params)))
5233 return -EFAULT;
5234 } else
5235 return -EINVAL;
5237 asoc = sctp_id2assoc(sk, params.assoc_id);
5238 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5239 return -EINVAL;
5241 if (asoc)
5242 params.assoc_value = asoc->frag_point;
5243 else
5244 params.assoc_value = sctp_sk(sk)->user_frag;
5246 if (put_user(len, optlen))
5247 return -EFAULT;
5248 if (len == sizeof(int)) {
5249 if (copy_to_user(optval, &params.assoc_value, len))
5250 return -EFAULT;
5251 } else {
5252 if (copy_to_user(optval, &params, len))
5253 return -EFAULT;
5256 return 0;
5260 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5261 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5263 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5264 char __user *optval, int __user *optlen)
5266 int val;
5268 if (len < sizeof(int))
5269 return -EINVAL;
5271 len = sizeof(int);
5273 val = sctp_sk(sk)->frag_interleave;
5274 if (put_user(len, optlen))
5275 return -EFAULT;
5276 if (copy_to_user(optval, &val, len))
5277 return -EFAULT;
5279 return 0;
5283 * 7.1.25. Set or Get the sctp partial delivery point
5284 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5286 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5287 char __user *optval,
5288 int __user *optlen)
5290 u32 val;
5292 if (len < sizeof(u32))
5293 return -EINVAL;
5295 len = sizeof(u32);
5297 val = sctp_sk(sk)->pd_point;
5298 if (put_user(len, optlen))
5299 return -EFAULT;
5300 if (copy_to_user(optval, &val, len))
5301 return -EFAULT;
5303 return 0;
5307 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5308 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5310 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5311 char __user *optval,
5312 int __user *optlen)
5314 struct sctp_assoc_value params;
5315 struct sctp_sock *sp;
5316 struct sctp_association *asoc;
5318 if (len == sizeof(int)) {
5319 pr_warn("Use of int in max_burst socket option deprecated\n");
5320 pr_warn("Use struct sctp_assoc_value instead\n");
5321 params.assoc_id = 0;
5322 } else if (len >= sizeof(struct sctp_assoc_value)) {
5323 len = sizeof(struct sctp_assoc_value);
5324 if (copy_from_user(&params, optval, len))
5325 return -EFAULT;
5326 } else
5327 return -EINVAL;
5329 sp = sctp_sk(sk);
5331 if (params.assoc_id != 0) {
5332 asoc = sctp_id2assoc(sk, params.assoc_id);
5333 if (!asoc)
5334 return -EINVAL;
5335 params.assoc_value = asoc->max_burst;
5336 } else
5337 params.assoc_value = sp->max_burst;
5339 if (len == sizeof(int)) {
5340 if (copy_to_user(optval, &params.assoc_value, len))
5341 return -EFAULT;
5342 } else {
5343 if (copy_to_user(optval, &params, len))
5344 return -EFAULT;
5347 return 0;
5351 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5352 char __user *optval, int __user *optlen)
5354 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5355 struct sctp_hmacalgo __user *p = (void __user *)optval;
5356 struct sctp_hmac_algo_param *hmacs;
5357 __u16 data_len = 0;
5358 u32 num_idents;
5360 if (!ep->auth_enable)
5361 return -EACCES;
5363 hmacs = ep->auth_hmacs_list;
5364 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5366 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5367 return -EINVAL;
5369 len = sizeof(struct sctp_hmacalgo) + data_len;
5370 num_idents = data_len / sizeof(u16);
5372 if (put_user(len, optlen))
5373 return -EFAULT;
5374 if (put_user(num_idents, &p->shmac_num_idents))
5375 return -EFAULT;
5376 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5377 return -EFAULT;
5378 return 0;
5381 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5382 char __user *optval, int __user *optlen)
5384 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5385 struct sctp_authkeyid val;
5386 struct sctp_association *asoc;
5388 if (!ep->auth_enable)
5389 return -EACCES;
5391 if (len < sizeof(struct sctp_authkeyid))
5392 return -EINVAL;
5393 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5394 return -EFAULT;
5396 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5397 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5398 return -EINVAL;
5400 if (asoc)
5401 val.scact_keynumber = asoc->active_key_id;
5402 else
5403 val.scact_keynumber = ep->active_key_id;
5405 len = sizeof(struct sctp_authkeyid);
5406 if (put_user(len, optlen))
5407 return -EFAULT;
5408 if (copy_to_user(optval, &val, len))
5409 return -EFAULT;
5411 return 0;
5414 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5415 char __user *optval, int __user *optlen)
5417 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5418 struct sctp_authchunks __user *p = (void __user *)optval;
5419 struct sctp_authchunks val;
5420 struct sctp_association *asoc;
5421 struct sctp_chunks_param *ch;
5422 u32 num_chunks = 0;
5423 char __user *to;
5425 if (!ep->auth_enable)
5426 return -EACCES;
5428 if (len < sizeof(struct sctp_authchunks))
5429 return -EINVAL;
5431 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5432 return -EFAULT;
5434 to = p->gauth_chunks;
5435 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5436 if (!asoc)
5437 return -EINVAL;
5439 ch = asoc->peer.peer_chunks;
5440 if (!ch)
5441 goto num;
5443 /* See if the user provided enough room for all the data */
5444 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5445 if (len < num_chunks)
5446 return -EINVAL;
5448 if (copy_to_user(to, ch->chunks, num_chunks))
5449 return -EFAULT;
5450 num:
5451 len = sizeof(struct sctp_authchunks) + num_chunks;
5452 if (put_user(len, optlen)) return -EFAULT;
5453 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5454 return -EFAULT;
5455 return 0;
5458 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5459 char __user *optval, int __user *optlen)
5461 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5462 struct sctp_authchunks __user *p = (void __user *)optval;
5463 struct sctp_authchunks val;
5464 struct sctp_association *asoc;
5465 struct sctp_chunks_param *ch;
5466 u32 num_chunks = 0;
5467 char __user *to;
5469 if (!ep->auth_enable)
5470 return -EACCES;
5472 if (len < sizeof(struct sctp_authchunks))
5473 return -EINVAL;
5475 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5476 return -EFAULT;
5478 to = p->gauth_chunks;
5479 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5480 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5481 return -EINVAL;
5483 if (asoc)
5484 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5485 else
5486 ch = ep->auth_chunk_list;
5488 if (!ch)
5489 goto num;
5491 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5492 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5493 return -EINVAL;
5495 if (copy_to_user(to, ch->chunks, num_chunks))
5496 return -EFAULT;
5497 num:
5498 len = sizeof(struct sctp_authchunks) + num_chunks;
5499 if (put_user(len, optlen))
5500 return -EFAULT;
5501 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5502 return -EFAULT;
5504 return 0;
5508 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5509 * This option gets the current number of associations that are attached
5510 * to a one-to-many style socket. The option value is an uint32_t.
5512 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5513 char __user *optval, int __user *optlen)
5515 struct sctp_sock *sp = sctp_sk(sk);
5516 struct sctp_association *asoc;
5517 u32 val = 0;
5519 if (sctp_style(sk, TCP))
5520 return -EOPNOTSUPP;
5522 if (len < sizeof(u32))
5523 return -EINVAL;
5525 len = sizeof(u32);
5527 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5528 val++;
5531 if (put_user(len, optlen))
5532 return -EFAULT;
5533 if (copy_to_user(optval, &val, len))
5534 return -EFAULT;
5536 return 0;
5540 * 8.1.23 SCTP_AUTO_ASCONF
5541 * See the corresponding setsockopt entry as description
5543 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5544 char __user *optval, int __user *optlen)
5546 int val = 0;
5548 if (len < sizeof(int))
5549 return -EINVAL;
5551 len = sizeof(int);
5552 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5553 val = 1;
5554 if (put_user(len, optlen))
5555 return -EFAULT;
5556 if (copy_to_user(optval, &val, len))
5557 return -EFAULT;
5558 return 0;
5562 * 8.2.6. Get the Current Identifiers of Associations
5563 * (SCTP_GET_ASSOC_ID_LIST)
5565 * This option gets the current list of SCTP association identifiers of
5566 * the SCTP associations handled by a one-to-many style socket.
5568 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5569 char __user *optval, int __user *optlen)
5571 struct sctp_sock *sp = sctp_sk(sk);
5572 struct sctp_association *asoc;
5573 struct sctp_assoc_ids *ids;
5574 u32 num = 0;
5576 if (sctp_style(sk, TCP))
5577 return -EOPNOTSUPP;
5579 if (len < sizeof(struct sctp_assoc_ids))
5580 return -EINVAL;
5582 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5583 num++;
5586 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5587 return -EINVAL;
5589 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5591 ids = kmalloc(len, GFP_KERNEL);
5592 if (unlikely(!ids))
5593 return -ENOMEM;
5595 ids->gaids_number_of_ids = num;
5596 num = 0;
5597 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5598 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5601 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5602 kfree(ids);
5603 return -EFAULT;
5606 kfree(ids);
5607 return 0;
5611 * SCTP_PEER_ADDR_THLDS
5613 * This option allows us to fetch the partially failed threshold for one or all
5614 * transports in an association. See Section 6.1 of:
5615 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5617 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5618 char __user *optval,
5619 int len,
5620 int __user *optlen)
5622 struct sctp_paddrthlds val;
5623 struct sctp_transport *trans;
5624 struct sctp_association *asoc;
5626 if (len < sizeof(struct sctp_paddrthlds))
5627 return -EINVAL;
5628 len = sizeof(struct sctp_paddrthlds);
5629 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5630 return -EFAULT;
5632 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5633 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5634 if (!asoc)
5635 return -ENOENT;
5637 val.spt_pathpfthld = asoc->pf_retrans;
5638 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5639 } else {
5640 trans = sctp_addr_id2transport(sk, &val.spt_address,
5641 val.spt_assoc_id);
5642 if (!trans)
5643 return -ENOENT;
5645 val.spt_pathmaxrxt = trans->pathmaxrxt;
5646 val.spt_pathpfthld = trans->pf_retrans;
5649 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5650 return -EFAULT;
5652 return 0;
5656 * SCTP_GET_ASSOC_STATS
5658 * This option retrieves local per endpoint statistics. It is modeled
5659 * after OpenSolaris' implementation
5661 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5662 char __user *optval,
5663 int __user *optlen)
5665 struct sctp_assoc_stats sas;
5666 struct sctp_association *asoc = NULL;
5668 /* User must provide at least the assoc id */
5669 if (len < sizeof(sctp_assoc_t))
5670 return -EINVAL;
5672 /* Allow the struct to grow and fill in as much as possible */
5673 len = min_t(size_t, len, sizeof(sas));
5675 if (copy_from_user(&sas, optval, len))
5676 return -EFAULT;
5678 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5679 if (!asoc)
5680 return -EINVAL;
5682 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5683 sas.sas_gapcnt = asoc->stats.gapcnt;
5684 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5685 sas.sas_osacks = asoc->stats.osacks;
5686 sas.sas_isacks = asoc->stats.isacks;
5687 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5688 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5689 sas.sas_oodchunks = asoc->stats.oodchunks;
5690 sas.sas_iodchunks = asoc->stats.iodchunks;
5691 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5692 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5693 sas.sas_idupchunks = asoc->stats.idupchunks;
5694 sas.sas_opackets = asoc->stats.opackets;
5695 sas.sas_ipackets = asoc->stats.ipackets;
5697 /* New high max rto observed, will return 0 if not a single
5698 * RTO update took place. obs_rto_ipaddr will be bogus
5699 * in such a case
5701 sas.sas_maxrto = asoc->stats.max_obs_rto;
5702 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5703 sizeof(struct sockaddr_storage));
5705 /* Mark beginning of a new observation period */
5706 asoc->stats.max_obs_rto = asoc->rto_min;
5708 if (put_user(len, optlen))
5709 return -EFAULT;
5711 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5713 if (copy_to_user(optval, &sas, len))
5714 return -EFAULT;
5716 return 0;
5719 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5720 char __user *optval, int __user *optlen)
5722 int retval = 0;
5723 int len;
5725 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5727 /* I can hardly begin to describe how wrong this is. This is
5728 * so broken as to be worse than useless. The API draft
5729 * REALLY is NOT helpful here... I am not convinced that the
5730 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5731 * are at all well-founded.
5733 if (level != SOL_SCTP) {
5734 struct sctp_af *af = sctp_sk(sk)->pf->af;
5736 retval = af->getsockopt(sk, level, optname, optval, optlen);
5737 return retval;
5740 if (get_user(len, optlen))
5741 return -EFAULT;
5743 sctp_lock_sock(sk);
5745 switch (optname) {
5746 case SCTP_STATUS:
5747 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5748 break;
5749 case SCTP_DISABLE_FRAGMENTS:
5750 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5751 optlen);
5752 break;
5753 case SCTP_EVENTS:
5754 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5755 break;
5756 case SCTP_AUTOCLOSE:
5757 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5758 break;
5759 case SCTP_SOCKOPT_PEELOFF:
5760 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5761 break;
5762 case SCTP_PEER_ADDR_PARAMS:
5763 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5764 optlen);
5765 break;
5766 case SCTP_DELAYED_SACK:
5767 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5768 optlen);
5769 break;
5770 case SCTP_INITMSG:
5771 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5772 break;
5773 case SCTP_GET_PEER_ADDRS:
5774 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5775 optlen);
5776 break;
5777 case SCTP_GET_LOCAL_ADDRS:
5778 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5779 optlen);
5780 break;
5781 case SCTP_SOCKOPT_CONNECTX3:
5782 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5783 break;
5784 case SCTP_DEFAULT_SEND_PARAM:
5785 retval = sctp_getsockopt_default_send_param(sk, len,
5786 optval, optlen);
5787 break;
5788 case SCTP_PRIMARY_ADDR:
5789 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5790 break;
5791 case SCTP_NODELAY:
5792 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5793 break;
5794 case SCTP_RTOINFO:
5795 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5796 break;
5797 case SCTP_ASSOCINFO:
5798 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5799 break;
5800 case SCTP_I_WANT_MAPPED_V4_ADDR:
5801 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5802 break;
5803 case SCTP_MAXSEG:
5804 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5805 break;
5806 case SCTP_GET_PEER_ADDR_INFO:
5807 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5808 optlen);
5809 break;
5810 case SCTP_ADAPTATION_LAYER:
5811 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5812 optlen);
5813 break;
5814 case SCTP_CONTEXT:
5815 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5816 break;
5817 case SCTP_FRAGMENT_INTERLEAVE:
5818 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5819 optlen);
5820 break;
5821 case SCTP_PARTIAL_DELIVERY_POINT:
5822 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5823 optlen);
5824 break;
5825 case SCTP_MAX_BURST:
5826 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5827 break;
5828 case SCTP_AUTH_KEY:
5829 case SCTP_AUTH_CHUNK:
5830 case SCTP_AUTH_DELETE_KEY:
5831 retval = -EOPNOTSUPP;
5832 break;
5833 case SCTP_HMAC_IDENT:
5834 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5835 break;
5836 case SCTP_AUTH_ACTIVE_KEY:
5837 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5838 break;
5839 case SCTP_PEER_AUTH_CHUNKS:
5840 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5841 optlen);
5842 break;
5843 case SCTP_LOCAL_AUTH_CHUNKS:
5844 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5845 optlen);
5846 break;
5847 case SCTP_GET_ASSOC_NUMBER:
5848 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5849 break;
5850 case SCTP_GET_ASSOC_ID_LIST:
5851 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5852 break;
5853 case SCTP_AUTO_ASCONF:
5854 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5855 break;
5856 case SCTP_PEER_ADDR_THLDS:
5857 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5858 break;
5859 case SCTP_GET_ASSOC_STATS:
5860 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5861 break;
5862 default:
5863 retval = -ENOPROTOOPT;
5864 break;
5867 sctp_release_sock(sk);
5868 return retval;
5871 static void sctp_hash(struct sock *sk)
5873 /* STUB */
5876 static void sctp_unhash(struct sock *sk)
5878 /* STUB */
5881 /* Check if port is acceptable. Possibly find first available port.
5883 * The port hash table (contained in the 'global' SCTP protocol storage
5884 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5885 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5886 * list (the list number is the port number hashed out, so as you
5887 * would expect from a hash function, all the ports in a given list have
5888 * such a number that hashes out to the same list number; you were
5889 * expecting that, right?); so each list has a set of ports, with a
5890 * link to the socket (struct sock) that uses it, the port number and
5891 * a fastreuse flag (FIXME: NPI ipg).
5893 static struct sctp_bind_bucket *sctp_bucket_create(
5894 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5896 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5898 struct sctp_bind_hashbucket *head; /* hash list */
5899 struct sctp_bind_bucket *pp;
5900 unsigned short snum;
5901 int ret;
5903 snum = ntohs(addr->v4.sin_port);
5905 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5907 sctp_local_bh_disable();
5909 if (snum == 0) {
5910 /* Search for an available port. */
5911 int low, high, remaining, index;
5912 unsigned int rover;
5914 inet_get_local_port_range(&low, &high);
5915 remaining = (high - low) + 1;
5916 rover = net_random() % remaining + low;
5918 do {
5919 rover++;
5920 if ((rover < low) || (rover > high))
5921 rover = low;
5922 if (inet_is_reserved_local_port(rover))
5923 continue;
5924 index = sctp_phashfn(sock_net(sk), rover);
5925 head = &sctp_port_hashtable[index];
5926 sctp_spin_lock(&head->lock);
5927 sctp_for_each_hentry(pp, &head->chain)
5928 if ((pp->port == rover) &&
5929 net_eq(sock_net(sk), pp->net))
5930 goto next;
5931 break;
5932 next:
5933 sctp_spin_unlock(&head->lock);
5934 } while (--remaining > 0);
5936 /* Exhausted local port range during search? */
5937 ret = 1;
5938 if (remaining <= 0)
5939 goto fail;
5941 /* OK, here is the one we will use. HEAD (the port
5942 * hash table list entry) is non-NULL and we hold it's
5943 * mutex.
5945 snum = rover;
5946 } else {
5947 /* We are given an specific port number; we verify
5948 * that it is not being used. If it is used, we will
5949 * exahust the search in the hash list corresponding
5950 * to the port number (snum) - we detect that with the
5951 * port iterator, pp being NULL.
5953 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5954 sctp_spin_lock(&head->lock);
5955 sctp_for_each_hentry(pp, &head->chain) {
5956 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5957 goto pp_found;
5960 pp = NULL;
5961 goto pp_not_found;
5962 pp_found:
5963 if (!hlist_empty(&pp->owner)) {
5964 /* We had a port hash table hit - there is an
5965 * available port (pp != NULL) and it is being
5966 * used by other socket (pp->owner not empty); that other
5967 * socket is going to be sk2.
5969 int reuse = sk->sk_reuse;
5970 struct sock *sk2;
5972 pr_debug("%s: found a possible match\n", __func__);
5974 if (pp->fastreuse && sk->sk_reuse &&
5975 sk->sk_state != SCTP_SS_LISTENING)
5976 goto success;
5978 /* Run through the list of sockets bound to the port
5979 * (pp->port) [via the pointers bind_next and
5980 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5981 * we get the endpoint they describe and run through
5982 * the endpoint's list of IP (v4 or v6) addresses,
5983 * comparing each of the addresses with the address of
5984 * the socket sk. If we find a match, then that means
5985 * that this port/socket (sk) combination are already
5986 * in an endpoint.
5988 sk_for_each_bound(sk2, &pp->owner) {
5989 struct sctp_endpoint *ep2;
5990 ep2 = sctp_sk(sk2)->ep;
5992 if (sk == sk2 ||
5993 (reuse && sk2->sk_reuse &&
5994 sk2->sk_state != SCTP_SS_LISTENING))
5995 continue;
5997 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5998 sctp_sk(sk2), sctp_sk(sk))) {
5999 ret = (long)sk2;
6000 goto fail_unlock;
6004 pr_debug("%s: found a match\n", __func__);
6006 pp_not_found:
6007 /* If there was a hash table miss, create a new port. */
6008 ret = 1;
6009 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6010 goto fail_unlock;
6012 /* In either case (hit or miss), make sure fastreuse is 1 only
6013 * if sk->sk_reuse is too (that is, if the caller requested
6014 * SO_REUSEADDR on this socket -sk-).
6016 if (hlist_empty(&pp->owner)) {
6017 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6018 pp->fastreuse = 1;
6019 else
6020 pp->fastreuse = 0;
6021 } else if (pp->fastreuse &&
6022 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6023 pp->fastreuse = 0;
6025 /* We are set, so fill up all the data in the hash table
6026 * entry, tie the socket list information with the rest of the
6027 * sockets FIXME: Blurry, NPI (ipg).
6029 success:
6030 if (!sctp_sk(sk)->bind_hash) {
6031 inet_sk(sk)->inet_num = snum;
6032 sk_add_bind_node(sk, &pp->owner);
6033 sctp_sk(sk)->bind_hash = pp;
6035 ret = 0;
6037 fail_unlock:
6038 sctp_spin_unlock(&head->lock);
6040 fail:
6041 sctp_local_bh_enable();
6042 return ret;
6045 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6046 * port is requested.
6048 static int sctp_get_port(struct sock *sk, unsigned short snum)
6050 union sctp_addr addr;
6051 struct sctp_af *af = sctp_sk(sk)->pf->af;
6053 /* Set up a dummy address struct from the sk. */
6054 af->from_sk(&addr, sk);
6055 addr.v4.sin_port = htons(snum);
6057 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6058 return !!sctp_get_port_local(sk, &addr);
6062 * Move a socket to LISTENING state.
6064 static int sctp_listen_start(struct sock *sk, int backlog)
6066 struct sctp_sock *sp = sctp_sk(sk);
6067 struct sctp_endpoint *ep = sp->ep;
6068 struct crypto_hash *tfm = NULL;
6069 char alg[32];
6071 /* Allocate HMAC for generating cookie. */
6072 if (!sp->hmac && sp->sctp_hmac_alg) {
6073 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6074 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6075 if (IS_ERR(tfm)) {
6076 net_info_ratelimited("failed to load transform for %s: %ld\n",
6077 sp->sctp_hmac_alg, PTR_ERR(tfm));
6078 return -ENOSYS;
6080 sctp_sk(sk)->hmac = tfm;
6084 * If a bind() or sctp_bindx() is not called prior to a listen()
6085 * call that allows new associations to be accepted, the system
6086 * picks an ephemeral port and will choose an address set equivalent
6087 * to binding with a wildcard address.
6089 * This is not currently spelled out in the SCTP sockets
6090 * extensions draft, but follows the practice as seen in TCP
6091 * sockets.
6094 sk->sk_state = SCTP_SS_LISTENING;
6095 if (!ep->base.bind_addr.port) {
6096 if (sctp_autobind(sk))
6097 return -EAGAIN;
6098 } else {
6099 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6100 sk->sk_state = SCTP_SS_CLOSED;
6101 return -EADDRINUSE;
6105 sk->sk_max_ack_backlog = backlog;
6106 sctp_hash_endpoint(ep);
6107 return 0;
6111 * 4.1.3 / 5.1.3 listen()
6113 * By default, new associations are not accepted for UDP style sockets.
6114 * An application uses listen() to mark a socket as being able to
6115 * accept new associations.
6117 * On TCP style sockets, applications use listen() to ready the SCTP
6118 * endpoint for accepting inbound associations.
6120 * On both types of endpoints a backlog of '0' disables listening.
6122 * Move a socket to LISTENING state.
6124 int sctp_inet_listen(struct socket *sock, int backlog)
6126 struct sock *sk = sock->sk;
6127 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6128 int err = -EINVAL;
6130 if (unlikely(backlog < 0))
6131 return err;
6133 sctp_lock_sock(sk);
6135 /* Peeled-off sockets are not allowed to listen(). */
6136 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6137 goto out;
6139 if (sock->state != SS_UNCONNECTED)
6140 goto out;
6142 /* If backlog is zero, disable listening. */
6143 if (!backlog) {
6144 if (sctp_sstate(sk, CLOSED))
6145 goto out;
6147 err = 0;
6148 sctp_unhash_endpoint(ep);
6149 sk->sk_state = SCTP_SS_CLOSED;
6150 if (sk->sk_reuse)
6151 sctp_sk(sk)->bind_hash->fastreuse = 1;
6152 goto out;
6155 /* If we are already listening, just update the backlog */
6156 if (sctp_sstate(sk, LISTENING))
6157 sk->sk_max_ack_backlog = backlog;
6158 else {
6159 err = sctp_listen_start(sk, backlog);
6160 if (err)
6161 goto out;
6164 err = 0;
6165 out:
6166 sctp_release_sock(sk);
6167 return err;
6171 * This function is done by modeling the current datagram_poll() and the
6172 * tcp_poll(). Note that, based on these implementations, we don't
6173 * lock the socket in this function, even though it seems that,
6174 * ideally, locking or some other mechanisms can be used to ensure
6175 * the integrity of the counters (sndbuf and wmem_alloc) used
6176 * in this place. We assume that we don't need locks either until proven
6177 * otherwise.
6179 * Another thing to note is that we include the Async I/O support
6180 * here, again, by modeling the current TCP/UDP code. We don't have
6181 * a good way to test with it yet.
6183 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6185 struct sock *sk = sock->sk;
6186 struct sctp_sock *sp = sctp_sk(sk);
6187 unsigned int mask;
6189 poll_wait(file, sk_sleep(sk), wait);
6191 /* A TCP-style listening socket becomes readable when the accept queue
6192 * is not empty.
6194 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6195 return (!list_empty(&sp->ep->asocs)) ?
6196 (POLLIN | POLLRDNORM) : 0;
6198 mask = 0;
6200 /* Is there any exceptional events? */
6201 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6202 mask |= POLLERR |
6203 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6204 if (sk->sk_shutdown & RCV_SHUTDOWN)
6205 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6206 if (sk->sk_shutdown == SHUTDOWN_MASK)
6207 mask |= POLLHUP;
6209 /* Is it readable? Reconsider this code with TCP-style support. */
6210 if (!skb_queue_empty(&sk->sk_receive_queue))
6211 mask |= POLLIN | POLLRDNORM;
6213 /* The association is either gone or not ready. */
6214 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6215 return mask;
6217 /* Is it writable? */
6218 if (sctp_writeable(sk)) {
6219 mask |= POLLOUT | POLLWRNORM;
6220 } else {
6221 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6223 * Since the socket is not locked, the buffer
6224 * might be made available after the writeable check and
6225 * before the bit is set. This could cause a lost I/O
6226 * signal. tcp_poll() has a race breaker for this race
6227 * condition. Based on their implementation, we put
6228 * in the following code to cover it as well.
6230 if (sctp_writeable(sk))
6231 mask |= POLLOUT | POLLWRNORM;
6233 return mask;
6236 /********************************************************************
6237 * 2nd Level Abstractions
6238 ********************************************************************/
6240 static struct sctp_bind_bucket *sctp_bucket_create(
6241 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6243 struct sctp_bind_bucket *pp;
6245 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6246 if (pp) {
6247 SCTP_DBG_OBJCNT_INC(bind_bucket);
6248 pp->port = snum;
6249 pp->fastreuse = 0;
6250 INIT_HLIST_HEAD(&pp->owner);
6251 pp->net = net;
6252 hlist_add_head(&pp->node, &head->chain);
6254 return pp;
6257 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6258 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6260 if (pp && hlist_empty(&pp->owner)) {
6261 __hlist_del(&pp->node);
6262 kmem_cache_free(sctp_bucket_cachep, pp);
6263 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6267 /* Release this socket's reference to a local port. */
6268 static inline void __sctp_put_port(struct sock *sk)
6270 struct sctp_bind_hashbucket *head =
6271 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6272 inet_sk(sk)->inet_num)];
6273 struct sctp_bind_bucket *pp;
6275 sctp_spin_lock(&head->lock);
6276 pp = sctp_sk(sk)->bind_hash;
6277 __sk_del_bind_node(sk);
6278 sctp_sk(sk)->bind_hash = NULL;
6279 inet_sk(sk)->inet_num = 0;
6280 sctp_bucket_destroy(pp);
6281 sctp_spin_unlock(&head->lock);
6284 void sctp_put_port(struct sock *sk)
6286 sctp_local_bh_disable();
6287 __sctp_put_port(sk);
6288 sctp_local_bh_enable();
6292 * The system picks an ephemeral port and choose an address set equivalent
6293 * to binding with a wildcard address.
6294 * One of those addresses will be the primary address for the association.
6295 * This automatically enables the multihoming capability of SCTP.
6297 static int sctp_autobind(struct sock *sk)
6299 union sctp_addr autoaddr;
6300 struct sctp_af *af;
6301 __be16 port;
6303 /* Initialize a local sockaddr structure to INADDR_ANY. */
6304 af = sctp_sk(sk)->pf->af;
6306 port = htons(inet_sk(sk)->inet_num);
6307 af->inaddr_any(&autoaddr, port);
6309 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6312 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6314 * From RFC 2292
6315 * 4.2 The cmsghdr Structure *
6317 * When ancillary data is sent or received, any number of ancillary data
6318 * objects can be specified by the msg_control and msg_controllen members of
6319 * the msghdr structure, because each object is preceded by
6320 * a cmsghdr structure defining the object's length (the cmsg_len member).
6321 * Historically Berkeley-derived implementations have passed only one object
6322 * at a time, but this API allows multiple objects to be
6323 * passed in a single call to sendmsg() or recvmsg(). The following example
6324 * shows two ancillary data objects in a control buffer.
6326 * |<--------------------------- msg_controllen -------------------------->|
6327 * | |
6329 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6331 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6332 * | | |
6334 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6336 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6337 * | | | | |
6339 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6340 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6342 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6344 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6348 * msg_control
6349 * points here
6351 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6353 struct cmsghdr *cmsg;
6354 struct msghdr *my_msg = (struct msghdr *)msg;
6356 for (cmsg = CMSG_FIRSTHDR(msg);
6357 cmsg != NULL;
6358 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6359 if (!CMSG_OK(my_msg, cmsg))
6360 return -EINVAL;
6362 /* Should we parse this header or ignore? */
6363 if (cmsg->cmsg_level != IPPROTO_SCTP)
6364 continue;
6366 /* Strictly check lengths following example in SCM code. */
6367 switch (cmsg->cmsg_type) {
6368 case SCTP_INIT:
6369 /* SCTP Socket API Extension
6370 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6372 * This cmsghdr structure provides information for
6373 * initializing new SCTP associations with sendmsg().
6374 * The SCTP_INITMSG socket option uses this same data
6375 * structure. This structure is not used for
6376 * recvmsg().
6378 * cmsg_level cmsg_type cmsg_data[]
6379 * ------------ ------------ ----------------------
6380 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6382 if (cmsg->cmsg_len !=
6383 CMSG_LEN(sizeof(struct sctp_initmsg)))
6384 return -EINVAL;
6385 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6386 break;
6388 case SCTP_SNDRCV:
6389 /* SCTP Socket API Extension
6390 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6392 * This cmsghdr structure specifies SCTP options for
6393 * sendmsg() and describes SCTP header information
6394 * about a received message through recvmsg().
6396 * cmsg_level cmsg_type cmsg_data[]
6397 * ------------ ------------ ----------------------
6398 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6400 if (cmsg->cmsg_len !=
6401 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6402 return -EINVAL;
6404 cmsgs->info =
6405 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6407 /* Minimally, validate the sinfo_flags. */
6408 if (cmsgs->info->sinfo_flags &
6409 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6410 SCTP_ABORT | SCTP_EOF))
6411 return -EINVAL;
6412 break;
6414 default:
6415 return -EINVAL;
6418 return 0;
6422 * Wait for a packet..
6423 * Note: This function is the same function as in core/datagram.c
6424 * with a few modifications to make lksctp work.
6426 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6428 int error;
6429 DEFINE_WAIT(wait);
6431 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6433 /* Socket errors? */
6434 error = sock_error(sk);
6435 if (error)
6436 goto out;
6438 if (!skb_queue_empty(&sk->sk_receive_queue))
6439 goto ready;
6441 /* Socket shut down? */
6442 if (sk->sk_shutdown & RCV_SHUTDOWN)
6443 goto out;
6445 /* Sequenced packets can come disconnected. If so we report the
6446 * problem.
6448 error = -ENOTCONN;
6450 /* Is there a good reason to think that we may receive some data? */
6451 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6452 goto out;
6454 /* Handle signals. */
6455 if (signal_pending(current))
6456 goto interrupted;
6458 /* Let another process have a go. Since we are going to sleep
6459 * anyway. Note: This may cause odd behaviors if the message
6460 * does not fit in the user's buffer, but this seems to be the
6461 * only way to honor MSG_DONTWAIT realistically.
6463 sctp_release_sock(sk);
6464 *timeo_p = schedule_timeout(*timeo_p);
6465 sctp_lock_sock(sk);
6467 ready:
6468 finish_wait(sk_sleep(sk), &wait);
6469 return 0;
6471 interrupted:
6472 error = sock_intr_errno(*timeo_p);
6474 out:
6475 finish_wait(sk_sleep(sk), &wait);
6476 *err = error;
6477 return error;
6480 /* Receive a datagram.
6481 * Note: This is pretty much the same routine as in core/datagram.c
6482 * with a few changes to make lksctp work.
6484 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6485 int noblock, int *err)
6487 int error;
6488 struct sk_buff *skb;
6489 long timeo;
6491 timeo = sock_rcvtimeo(sk, noblock);
6493 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6494 MAX_SCHEDULE_TIMEOUT);
6496 do {
6497 /* Again only user level code calls this function,
6498 * so nothing interrupt level
6499 * will suddenly eat the receive_queue.
6501 * Look at current nfs client by the way...
6502 * However, this function was correct in any case. 8)
6504 if (flags & MSG_PEEK) {
6505 spin_lock_bh(&sk->sk_receive_queue.lock);
6506 skb = skb_peek(&sk->sk_receive_queue);
6507 if (skb)
6508 atomic_inc(&skb->users);
6509 spin_unlock_bh(&sk->sk_receive_queue.lock);
6510 } else {
6511 skb = skb_dequeue(&sk->sk_receive_queue);
6514 if (skb)
6515 return skb;
6517 /* Caller is allowed not to check sk->sk_err before calling. */
6518 error = sock_error(sk);
6519 if (error)
6520 goto no_packet;
6522 if (sk->sk_shutdown & RCV_SHUTDOWN)
6523 break;
6525 /* User doesn't want to wait. */
6526 error = -EAGAIN;
6527 if (!timeo)
6528 goto no_packet;
6529 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6531 return NULL;
6533 no_packet:
6534 *err = error;
6535 return NULL;
6538 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6539 static void __sctp_write_space(struct sctp_association *asoc)
6541 struct sock *sk = asoc->base.sk;
6542 struct socket *sock = sk->sk_socket;
6544 if ((sctp_wspace(asoc) > 0) && sock) {
6545 if (waitqueue_active(&asoc->wait))
6546 wake_up_interruptible(&asoc->wait);
6548 if (sctp_writeable(sk)) {
6549 wait_queue_head_t *wq = sk_sleep(sk);
6551 if (wq && waitqueue_active(wq))
6552 wake_up_interruptible(wq);
6554 /* Note that we try to include the Async I/O support
6555 * here by modeling from the current TCP/UDP code.
6556 * We have not tested with it yet.
6558 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6559 sock_wake_async(sock,
6560 SOCK_WAKE_SPACE, POLL_OUT);
6565 static void sctp_wake_up_waiters(struct sock *sk,
6566 struct sctp_association *asoc)
6568 struct sctp_association *tmp = asoc;
6570 /* We do accounting for the sndbuf space per association,
6571 * so we only need to wake our own association.
6573 if (asoc->ep->sndbuf_policy)
6574 return __sctp_write_space(asoc);
6576 /* If association goes down and is just flushing its
6577 * outq, then just normally notify others.
6579 if (asoc->base.dead)
6580 return sctp_write_space(sk);
6582 /* Accounting for the sndbuf space is per socket, so we
6583 * need to wake up others, try to be fair and in case of
6584 * other associations, let them have a go first instead
6585 * of just doing a sctp_write_space() call.
6587 * Note that we reach sctp_wake_up_waiters() only when
6588 * associations free up queued chunks, thus we are under
6589 * lock and the list of associations on a socket is
6590 * guaranteed not to change.
6592 for (tmp = list_next_entry(tmp, asocs); 1;
6593 tmp = list_next_entry(tmp, asocs)) {
6594 /* Manually skip the head element. */
6595 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6596 continue;
6597 /* Wake up association. */
6598 __sctp_write_space(tmp);
6599 /* We've reached the end. */
6600 if (tmp == asoc)
6601 break;
6605 /* Do accounting for the sndbuf space.
6606 * Decrement the used sndbuf space of the corresponding association by the
6607 * data size which was just transmitted(freed).
6609 static void sctp_wfree(struct sk_buff *skb)
6611 struct sctp_association *asoc;
6612 struct sctp_chunk *chunk;
6613 struct sock *sk;
6615 /* Get the saved chunk pointer. */
6616 chunk = *((struct sctp_chunk **)(skb->cb));
6617 asoc = chunk->asoc;
6618 sk = asoc->base.sk;
6619 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6620 sizeof(struct sk_buff) +
6621 sizeof(struct sctp_chunk);
6623 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6626 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6628 sk->sk_wmem_queued -= skb->truesize;
6629 sk_mem_uncharge(sk, skb->truesize);
6631 sock_wfree(skb);
6632 sctp_wake_up_waiters(sk, asoc);
6634 sctp_association_put(asoc);
6637 /* Do accounting for the receive space on the socket.
6638 * Accounting for the association is done in ulpevent.c
6639 * We set this as a destructor for the cloned data skbs so that
6640 * accounting is done at the correct time.
6642 void sctp_sock_rfree(struct sk_buff *skb)
6644 struct sock *sk = skb->sk;
6645 struct sctp_ulpevent *event = sctp_skb2event(skb);
6647 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6650 * Mimic the behavior of sock_rfree
6652 sk_mem_uncharge(sk, event->rmem_len);
6656 /* Helper function to wait for space in the sndbuf. */
6657 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6658 size_t msg_len)
6660 struct sock *sk = asoc->base.sk;
6661 int err = 0;
6662 long current_timeo = *timeo_p;
6663 DEFINE_WAIT(wait);
6665 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6666 *timeo_p, msg_len);
6668 /* Increment the association's refcnt. */
6669 sctp_association_hold(asoc);
6671 /* Wait on the association specific sndbuf space. */
6672 for (;;) {
6673 prepare_to_wait_exclusive(&asoc->wait, &wait,
6674 TASK_INTERRUPTIBLE);
6675 if (!*timeo_p)
6676 goto do_nonblock;
6677 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6678 asoc->base.dead)
6679 goto do_error;
6680 if (signal_pending(current))
6681 goto do_interrupted;
6682 if (msg_len <= sctp_wspace(asoc))
6683 break;
6685 /* Let another process have a go. Since we are going
6686 * to sleep anyway.
6688 sctp_release_sock(sk);
6689 current_timeo = schedule_timeout(current_timeo);
6690 BUG_ON(sk != asoc->base.sk);
6691 sctp_lock_sock(sk);
6693 *timeo_p = current_timeo;
6696 out:
6697 finish_wait(&asoc->wait, &wait);
6699 /* Release the association's refcnt. */
6700 sctp_association_put(asoc);
6702 return err;
6704 do_error:
6705 err = -EPIPE;
6706 goto out;
6708 do_interrupted:
6709 err = sock_intr_errno(*timeo_p);
6710 goto out;
6712 do_nonblock:
6713 err = -EAGAIN;
6714 goto out;
6717 void sctp_data_ready(struct sock *sk, int len)
6719 struct socket_wq *wq;
6721 rcu_read_lock();
6722 wq = rcu_dereference(sk->sk_wq);
6723 if (wq_has_sleeper(wq))
6724 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6725 POLLRDNORM | POLLRDBAND);
6726 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6727 rcu_read_unlock();
6730 /* If socket sndbuf has changed, wake up all per association waiters. */
6731 void sctp_write_space(struct sock *sk)
6733 struct sctp_association *asoc;
6735 /* Wake up the tasks in each wait queue. */
6736 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6737 __sctp_write_space(asoc);
6741 /* Is there any sndbuf space available on the socket?
6743 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6744 * associations on the same socket. For a UDP-style socket with
6745 * multiple associations, it is possible for it to be "unwriteable"
6746 * prematurely. I assume that this is acceptable because
6747 * a premature "unwriteable" is better than an accidental "writeable" which
6748 * would cause an unwanted block under certain circumstances. For the 1-1
6749 * UDP-style sockets or TCP-style sockets, this code should work.
6750 * - Daisy
6752 static int sctp_writeable(struct sock *sk)
6754 int amt = 0;
6756 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6757 if (amt < 0)
6758 amt = 0;
6759 return amt;
6762 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6763 * returns immediately with EINPROGRESS.
6765 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6767 struct sock *sk = asoc->base.sk;
6768 int err = 0;
6769 long current_timeo = *timeo_p;
6770 DEFINE_WAIT(wait);
6772 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6774 /* Increment the association's refcnt. */
6775 sctp_association_hold(asoc);
6777 for (;;) {
6778 prepare_to_wait_exclusive(&asoc->wait, &wait,
6779 TASK_INTERRUPTIBLE);
6780 if (!*timeo_p)
6781 goto do_nonblock;
6782 if (sk->sk_shutdown & RCV_SHUTDOWN)
6783 break;
6784 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6785 asoc->base.dead)
6786 goto do_error;
6787 if (signal_pending(current))
6788 goto do_interrupted;
6790 if (sctp_state(asoc, ESTABLISHED))
6791 break;
6793 /* Let another process have a go. Since we are going
6794 * to sleep anyway.
6796 sctp_release_sock(sk);
6797 current_timeo = schedule_timeout(current_timeo);
6798 sctp_lock_sock(sk);
6800 *timeo_p = current_timeo;
6803 out:
6804 finish_wait(&asoc->wait, &wait);
6806 /* Release the association's refcnt. */
6807 sctp_association_put(asoc);
6809 return err;
6811 do_error:
6812 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6813 err = -ETIMEDOUT;
6814 else
6815 err = -ECONNREFUSED;
6816 goto out;
6818 do_interrupted:
6819 err = sock_intr_errno(*timeo_p);
6820 goto out;
6822 do_nonblock:
6823 err = -EINPROGRESS;
6824 goto out;
6827 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6829 struct sctp_endpoint *ep;
6830 int err = 0;
6831 DEFINE_WAIT(wait);
6833 ep = sctp_sk(sk)->ep;
6836 for (;;) {
6837 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6838 TASK_INTERRUPTIBLE);
6840 if (list_empty(&ep->asocs)) {
6841 sctp_release_sock(sk);
6842 timeo = schedule_timeout(timeo);
6843 sctp_lock_sock(sk);
6846 err = -EINVAL;
6847 if (!sctp_sstate(sk, LISTENING))
6848 break;
6850 err = 0;
6851 if (!list_empty(&ep->asocs))
6852 break;
6854 err = sock_intr_errno(timeo);
6855 if (signal_pending(current))
6856 break;
6858 err = -EAGAIN;
6859 if (!timeo)
6860 break;
6863 finish_wait(sk_sleep(sk), &wait);
6865 return err;
6868 static void sctp_wait_for_close(struct sock *sk, long timeout)
6870 DEFINE_WAIT(wait);
6872 do {
6873 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6874 if (list_empty(&sctp_sk(sk)->ep->asocs))
6875 break;
6876 sctp_release_sock(sk);
6877 timeout = schedule_timeout(timeout);
6878 sctp_lock_sock(sk);
6879 } while (!signal_pending(current) && timeout);
6881 finish_wait(sk_sleep(sk), &wait);
6884 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6886 struct sk_buff *frag;
6888 if (!skb->data_len)
6889 goto done;
6891 /* Don't forget the fragments. */
6892 skb_walk_frags(skb, frag)
6893 sctp_skb_set_owner_r_frag(frag, sk);
6895 done:
6896 sctp_skb_set_owner_r(skb, sk);
6899 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6900 struct sctp_association *asoc)
6902 struct inet_sock *inet = inet_sk(sk);
6903 struct inet_sock *newinet;
6905 newsk->sk_type = sk->sk_type;
6906 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6907 newsk->sk_flags = sk->sk_flags;
6908 newsk->sk_no_check = sk->sk_no_check;
6909 newsk->sk_reuse = sk->sk_reuse;
6911 newsk->sk_shutdown = sk->sk_shutdown;
6912 newsk->sk_destruct = sctp_destruct_sock;
6913 newsk->sk_family = sk->sk_family;
6914 newsk->sk_protocol = IPPROTO_SCTP;
6915 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6916 newsk->sk_sndbuf = sk->sk_sndbuf;
6917 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6918 newsk->sk_lingertime = sk->sk_lingertime;
6919 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6920 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6922 newinet = inet_sk(newsk);
6924 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6925 * getsockname() and getpeername()
6927 newinet->inet_sport = inet->inet_sport;
6928 newinet->inet_saddr = inet->inet_saddr;
6929 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6930 newinet->inet_dport = htons(asoc->peer.port);
6931 newinet->pmtudisc = inet->pmtudisc;
6932 newinet->inet_id = asoc->next_tsn ^ jiffies;
6934 newinet->uc_ttl = inet->uc_ttl;
6935 newinet->mc_loop = 1;
6936 newinet->mc_ttl = 1;
6937 newinet->mc_index = 0;
6938 newinet->mc_list = NULL;
6941 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6942 * and its messages to the newsk.
6944 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6945 struct sctp_association *assoc,
6946 sctp_socket_type_t type)
6948 struct sctp_sock *oldsp = sctp_sk(oldsk);
6949 struct sctp_sock *newsp = sctp_sk(newsk);
6950 struct sctp_bind_bucket *pp; /* hash list port iterator */
6951 struct sctp_endpoint *newep = newsp->ep;
6952 struct sk_buff *skb, *tmp;
6953 struct sctp_ulpevent *event;
6954 struct sctp_bind_hashbucket *head;
6955 struct list_head tmplist;
6957 /* Migrate socket buffer sizes and all the socket level options to the
6958 * new socket.
6960 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6961 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6962 /* Brute force copy old sctp opt. */
6963 if (oldsp->do_auto_asconf) {
6964 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6965 inet_sk_copy_descendant(newsk, oldsk);
6966 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6967 } else
6968 inet_sk_copy_descendant(newsk, oldsk);
6970 /* Restore the ep value that was overwritten with the above structure
6971 * copy.
6973 newsp->ep = newep;
6974 newsp->hmac = NULL;
6976 /* Hook this new socket in to the bind_hash list. */
6977 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6978 inet_sk(oldsk)->inet_num)];
6979 sctp_local_bh_disable();
6980 sctp_spin_lock(&head->lock);
6981 pp = sctp_sk(oldsk)->bind_hash;
6982 sk_add_bind_node(newsk, &pp->owner);
6983 sctp_sk(newsk)->bind_hash = pp;
6984 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6985 sctp_spin_unlock(&head->lock);
6986 sctp_local_bh_enable();
6988 /* Copy the bind_addr list from the original endpoint to the new
6989 * endpoint so that we can handle restarts properly
6991 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6992 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6994 /* Move any messages in the old socket's receive queue that are for the
6995 * peeled off association to the new socket's receive queue.
6997 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6998 event = sctp_skb2event(skb);
6999 if (event->asoc == assoc) {
7000 __skb_unlink(skb, &oldsk->sk_receive_queue);
7001 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7002 sctp_skb_set_owner_r_frag(skb, newsk);
7006 /* Clean up any messages pending delivery due to partial
7007 * delivery. Three cases:
7008 * 1) No partial deliver; no work.
7009 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7010 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7012 skb_queue_head_init(&newsp->pd_lobby);
7013 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7015 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7016 struct sk_buff_head *queue;
7018 /* Decide which queue to move pd_lobby skbs to. */
7019 if (assoc->ulpq.pd_mode) {
7020 queue = &newsp->pd_lobby;
7021 } else
7022 queue = &newsk->sk_receive_queue;
7024 /* Walk through the pd_lobby, looking for skbs that
7025 * need moved to the new socket.
7027 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7028 event = sctp_skb2event(skb);
7029 if (event->asoc == assoc) {
7030 __skb_unlink(skb, &oldsp->pd_lobby);
7031 __skb_queue_tail(queue, skb);
7032 sctp_skb_set_owner_r_frag(skb, newsk);
7036 /* Clear up any skbs waiting for the partial
7037 * delivery to finish.
7039 if (assoc->ulpq.pd_mode)
7040 sctp_clear_pd(oldsk, NULL);
7044 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7045 sctp_skb_set_owner_r_frag(skb, newsk);
7047 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7048 sctp_skb_set_owner_r_frag(skb, newsk);
7050 /* Set the type of socket to indicate that it is peeled off from the
7051 * original UDP-style socket or created with the accept() call on a
7052 * TCP-style socket..
7054 newsp->type = type;
7056 /* Mark the new socket "in-use" by the user so that any packets
7057 * that may arrive on the association after we've moved it are
7058 * queued to the backlog. This prevents a potential race between
7059 * backlog processing on the old socket and new-packet processing
7060 * on the new socket.
7062 * The caller has just allocated newsk so we can guarantee that other
7063 * paths won't try to lock it and then oldsk.
7065 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7066 sctp_assoc_migrate(assoc, newsk);
7068 /* If the association on the newsk is already closed before accept()
7069 * is called, set RCV_SHUTDOWN flag.
7071 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7072 newsk->sk_shutdown |= RCV_SHUTDOWN;
7074 newsk->sk_state = SCTP_SS_ESTABLISHED;
7075 sctp_release_sock(newsk);
7079 /* This proto struct describes the ULP interface for SCTP. */
7080 struct proto sctp_prot = {
7081 .name = "SCTP",
7082 .owner = THIS_MODULE,
7083 .close = sctp_close,
7084 .connect = sctp_connect,
7085 .disconnect = sctp_disconnect,
7086 .accept = sctp_accept,
7087 .ioctl = sctp_ioctl,
7088 .init = sctp_init_sock,
7089 .destroy = sctp_destroy_sock,
7090 .shutdown = sctp_shutdown,
7091 .setsockopt = sctp_setsockopt,
7092 .getsockopt = sctp_getsockopt,
7093 .sendmsg = sctp_sendmsg,
7094 .recvmsg = sctp_recvmsg,
7095 .bind = sctp_bind,
7096 .backlog_rcv = sctp_backlog_rcv,
7097 .hash = sctp_hash,
7098 .unhash = sctp_unhash,
7099 .get_port = sctp_get_port,
7100 .obj_size = sizeof(struct sctp_sock),
7101 .sysctl_mem = sysctl_sctp_mem,
7102 .sysctl_rmem = sysctl_sctp_rmem,
7103 .sysctl_wmem = sysctl_sctp_wmem,
7104 .memory_pressure = &sctp_memory_pressure,
7105 .enter_memory_pressure = sctp_enter_memory_pressure,
7106 .memory_allocated = &sctp_memory_allocated,
7107 .sockets_allocated = &sctp_sockets_allocated,
7110 #if IS_ENABLED(CONFIG_IPV6)
7112 struct proto sctpv6_prot = {
7113 .name = "SCTPv6",
7114 .owner = THIS_MODULE,
7115 .close = sctp_close,
7116 .connect = sctp_connect,
7117 .disconnect = sctp_disconnect,
7118 .accept = sctp_accept,
7119 .ioctl = sctp_ioctl,
7120 .init = sctp_init_sock,
7121 .destroy = sctp_destroy_sock,
7122 .shutdown = sctp_shutdown,
7123 .setsockopt = sctp_setsockopt,
7124 .getsockopt = sctp_getsockopt,
7125 .sendmsg = sctp_sendmsg,
7126 .recvmsg = sctp_recvmsg,
7127 .bind = sctp_bind,
7128 .backlog_rcv = sctp_backlog_rcv,
7129 .hash = sctp_hash,
7130 .unhash = sctp_unhash,
7131 .get_port = sctp_get_port,
7132 .obj_size = sizeof(struct sctp6_sock),
7133 .sysctl_mem = sysctl_sctp_mem,
7134 .sysctl_rmem = sysctl_sctp_rmem,
7135 .sysctl_wmem = sysctl_sctp_wmem,
7136 .memory_pressure = &sctp_memory_pressure,
7137 .enter_memory_pressure = sctp_enter_memory_pressure,
7138 .memory_allocated = &sctp_memory_allocated,
7139 .sockets_allocated = &sctp_sockets_allocated,
7141 #endif /* IS_ENABLED(CONFIG_IPV6) */