Merge branch 'for-3.11' of git://linux-nfs.org/~bfields/linux
[linux-2.6.git] / net / sctp / socket.c
blobc6670d2e3f8d1bfa7c7fd0cef617de4944ca95e0
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 <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
66 #include <linux/ip.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
73 #include <linux/file.h>
75 #include <net/ip.h>
76 #include <net/icmp.h>
77 #include <net/route.h>
78 #include <net/ipv6.h>
79 #include <net/inet_common.h>
81 #include <linux/socket.h> /* for sa_family_t */
82 #include <linux/export.h>
83 #include <net/sock.h>
84 #include <net/sctp/sctp.h>
85 #include <net/sctp/sm.h>
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91 size_t msg_len);
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static void sctp_destruct_sock(struct sock *sk);
97 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
98 union sctp_addr *addr, int len);
99 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
100 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
103 static int sctp_send_asconf(struct sctp_association *asoc,
104 struct sctp_chunk *chunk);
105 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
106 static int sctp_autobind(struct sock *sk);
107 static void sctp_sock_migrate(struct sock *, struct sock *,
108 struct sctp_association *, sctp_socket_type_t);
110 extern struct kmem_cache *sctp_bucket_cachep;
111 extern long sysctl_sctp_mem[3];
112 extern int sysctl_sctp_rmem[3];
113 extern int sysctl_sctp_wmem[3];
115 static int sctp_memory_pressure;
116 static atomic_long_t sctp_memory_allocated;
117 struct percpu_counter sctp_sockets_allocated;
119 static void sctp_enter_memory_pressure(struct sock *sk)
121 sctp_memory_pressure = 1;
125 /* Get the sndbuf space available at the time on the association. */
126 static inline int sctp_wspace(struct sctp_association *asoc)
128 int amt;
130 if (asoc->ep->sndbuf_policy)
131 amt = asoc->sndbuf_used;
132 else
133 amt = sk_wmem_alloc_get(asoc->base.sk);
135 if (amt >= asoc->base.sk->sk_sndbuf) {
136 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
137 amt = 0;
138 else {
139 amt = sk_stream_wspace(asoc->base.sk);
140 if (amt < 0)
141 amt = 0;
143 } else {
144 amt = asoc->base.sk->sk_sndbuf - amt;
146 return amt;
149 /* Increment the used sndbuf space count of the corresponding association by
150 * the size of the outgoing data chunk.
151 * Also, set the skb destructor for sndbuf accounting later.
153 * Since it is always 1-1 between chunk and skb, and also a new skb is always
154 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
155 * destructor in the data chunk skb for the purpose of the sndbuf space
156 * tracking.
158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
160 struct sctp_association *asoc = chunk->asoc;
161 struct sock *sk = asoc->base.sk;
163 /* The sndbuf space is tracked per association. */
164 sctp_association_hold(asoc);
166 skb_set_owner_w(chunk->skb, sk);
168 chunk->skb->destructor = sctp_wfree;
169 /* Save the chunk pointer in skb for sctp_wfree to use later. */
170 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
172 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
173 sizeof(struct sk_buff) +
174 sizeof(struct sctp_chunk);
176 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
177 sk->sk_wmem_queued += chunk->skb->truesize;
178 sk_mem_charge(sk, chunk->skb->truesize);
181 /* Verify that this is a valid address. */
182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
183 int len)
185 struct sctp_af *af;
187 /* Verify basic sockaddr. */
188 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
189 if (!af)
190 return -EINVAL;
192 /* Is this a valid SCTP address? */
193 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
194 return -EINVAL;
196 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
197 return -EINVAL;
199 return 0;
202 /* Look up the association by its id. If this is not a UDP-style
203 * socket, the ID field is always ignored.
205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
207 struct sctp_association *asoc = NULL;
209 /* If this is not a UDP-style socket, assoc id should be ignored. */
210 if (!sctp_style(sk, UDP)) {
211 /* Return NULL if the socket state is not ESTABLISHED. It
212 * could be a TCP-style listening socket or a socket which
213 * hasn't yet called connect() to establish an association.
215 if (!sctp_sstate(sk, ESTABLISHED))
216 return NULL;
218 /* Get the first and the only association from the list. */
219 if (!list_empty(&sctp_sk(sk)->ep->asocs))
220 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
221 struct sctp_association, asocs);
222 return asoc;
225 /* Otherwise this is a UDP-style socket. */
226 if (!id || (id == (sctp_assoc_t)-1))
227 return NULL;
229 spin_lock_bh(&sctp_assocs_id_lock);
230 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
231 spin_unlock_bh(&sctp_assocs_id_lock);
233 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
234 return NULL;
236 return asoc;
239 /* Look up the transport from an address and an assoc id. If both address and
240 * id are specified, the associations matching the address and the id should be
241 * the same.
243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
244 struct sockaddr_storage *addr,
245 sctp_assoc_t id)
247 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
248 struct sctp_transport *transport;
249 union sctp_addr *laddr = (union sctp_addr *)addr;
251 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
252 laddr,
253 &transport);
255 if (!addr_asoc)
256 return NULL;
258 id_asoc = sctp_id2assoc(sk, id);
259 if (id_asoc && (id_asoc != addr_asoc))
260 return NULL;
262 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
263 (union sctp_addr *)addr);
265 return transport;
268 /* API 3.1.2 bind() - UDP Style Syntax
269 * The syntax of bind() is,
271 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
273 * sd - the socket descriptor returned by socket().
274 * addr - the address structure (struct sockaddr_in or struct
275 * sockaddr_in6 [RFC 2553]),
276 * addr_len - the size of the address structure.
278 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
280 int retval = 0;
282 sctp_lock_sock(sk);
284 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
285 addr, addr_len);
287 /* Disallow binding twice. */
288 if (!sctp_sk(sk)->ep->base.bind_addr.port)
289 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
290 addr_len);
291 else
292 retval = -EINVAL;
294 sctp_release_sock(sk);
296 return retval;
299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
301 /* Verify this is a valid sockaddr. */
302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
303 union sctp_addr *addr, int len)
305 struct sctp_af *af;
307 /* Check minimum size. */
308 if (len < sizeof (struct sockaddr))
309 return NULL;
311 /* V4 mapped address are really of AF_INET family */
312 if (addr->sa.sa_family == AF_INET6 &&
313 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
314 if (!opt->pf->af_supported(AF_INET, opt))
315 return NULL;
316 } else {
317 /* Does this PF support this AF? */
318 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
319 return NULL;
322 /* If we get this far, af is valid. */
323 af = sctp_get_af_specific(addr->sa.sa_family);
325 if (len < af->sockaddr_len)
326 return NULL;
328 return af;
331 /* Bind a local address either to an endpoint or to an association. */
332 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
334 struct net *net = sock_net(sk);
335 struct sctp_sock *sp = sctp_sk(sk);
336 struct sctp_endpoint *ep = sp->ep;
337 struct sctp_bind_addr *bp = &ep->base.bind_addr;
338 struct sctp_af *af;
339 unsigned short snum;
340 int ret = 0;
342 /* Common sockaddr verification. */
343 af = sctp_sockaddr_af(sp, addr, len);
344 if (!af) {
345 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
346 __func__, sk, addr, len);
347 return -EINVAL;
350 snum = ntohs(addr->v4.sin_port);
352 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
353 __func__, sk, &addr->sa, bp->port, snum, len);
355 /* PF specific bind() address verification. */
356 if (!sp->pf->bind_verify(sp, addr))
357 return -EADDRNOTAVAIL;
359 /* We must either be unbound, or bind to the same port.
360 * It's OK to allow 0 ports if we are already bound.
361 * We'll just inhert an already bound port in this case
363 if (bp->port) {
364 if (!snum)
365 snum = bp->port;
366 else if (snum != bp->port) {
367 pr_debug("%s: new port %d doesn't match existing port "
368 "%d\n", __func__, snum, bp->port);
369 return -EINVAL;
373 if (snum && snum < PROT_SOCK &&
374 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
375 return -EACCES;
377 /* See if the address matches any of the addresses we may have
378 * already bound before checking against other endpoints.
380 if (sctp_bind_addr_match(bp, addr, sp))
381 return -EINVAL;
383 /* Make sure we are allowed to bind here.
384 * The function sctp_get_port_local() does duplicate address
385 * detection.
387 addr->v4.sin_port = htons(snum);
388 if ((ret = sctp_get_port_local(sk, addr))) {
389 return -EADDRINUSE;
392 /* Refresh ephemeral port. */
393 if (!bp->port)
394 bp->port = inet_sk(sk)->inet_num;
396 /* Add the address to the bind address list.
397 * Use GFP_ATOMIC since BHs will be disabled.
399 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
401 /* Copy back into socket for getsockname() use. */
402 if (!ret) {
403 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
404 af->to_sk_saddr(addr, sk);
407 return ret;
410 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
412 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
413 * at any one time. If a sender, after sending an ASCONF chunk, decides
414 * it needs to transfer another ASCONF Chunk, it MUST wait until the
415 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
416 * subsequent ASCONF. Note this restriction binds each side, so at any
417 * time two ASCONF may be in-transit on any given association (one sent
418 * from each endpoint).
420 static int sctp_send_asconf(struct sctp_association *asoc,
421 struct sctp_chunk *chunk)
423 struct net *net = sock_net(asoc->base.sk);
424 int retval = 0;
426 /* If there is an outstanding ASCONF chunk, queue it for later
427 * transmission.
429 if (asoc->addip_last_asconf) {
430 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
431 goto out;
434 /* Hold the chunk until an ASCONF_ACK is received. */
435 sctp_chunk_hold(chunk);
436 retval = sctp_primitive_ASCONF(net, asoc, chunk);
437 if (retval)
438 sctp_chunk_free(chunk);
439 else
440 asoc->addip_last_asconf = chunk;
442 out:
443 return retval;
446 /* Add a list of addresses as bind addresses to local endpoint or
447 * association.
449 * Basically run through each address specified in the addrs/addrcnt
450 * array/length pair, determine if it is IPv6 or IPv4 and call
451 * sctp_do_bind() on it.
453 * If any of them fails, then the operation will be reversed and the
454 * ones that were added will be removed.
456 * Only sctp_setsockopt_bindx() is supposed to call this function.
458 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
460 int cnt;
461 int retval = 0;
462 void *addr_buf;
463 struct sockaddr *sa_addr;
464 struct sctp_af *af;
466 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
467 addrs, addrcnt);
469 addr_buf = addrs;
470 for (cnt = 0; cnt < addrcnt; cnt++) {
471 /* The list may contain either IPv4 or IPv6 address;
472 * determine the address length for walking thru the list.
474 sa_addr = addr_buf;
475 af = sctp_get_af_specific(sa_addr->sa_family);
476 if (!af) {
477 retval = -EINVAL;
478 goto err_bindx_add;
481 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
482 af->sockaddr_len);
484 addr_buf += af->sockaddr_len;
486 err_bindx_add:
487 if (retval < 0) {
488 /* Failed. Cleanup the ones that have been added */
489 if (cnt > 0)
490 sctp_bindx_rem(sk, addrs, cnt);
491 return retval;
495 return retval;
498 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
499 * associations that are part of the endpoint indicating that a list of local
500 * addresses are added to the endpoint.
502 * If any of the addresses is already in the bind address list of the
503 * association, we do not send the chunk for that association. But it will not
504 * affect other associations.
506 * Only sctp_setsockopt_bindx() is supposed to call this function.
508 static int sctp_send_asconf_add_ip(struct sock *sk,
509 struct sockaddr *addrs,
510 int addrcnt)
512 struct net *net = sock_net(sk);
513 struct sctp_sock *sp;
514 struct sctp_endpoint *ep;
515 struct sctp_association *asoc;
516 struct sctp_bind_addr *bp;
517 struct sctp_chunk *chunk;
518 struct sctp_sockaddr_entry *laddr;
519 union sctp_addr *addr;
520 union sctp_addr saveaddr;
521 void *addr_buf;
522 struct sctp_af *af;
523 struct list_head *p;
524 int i;
525 int retval = 0;
527 if (!net->sctp.addip_enable)
528 return retval;
530 sp = sctp_sk(sk);
531 ep = sp->ep;
533 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
534 __func__, sk, addrs, addrcnt);
536 list_for_each_entry(asoc, &ep->asocs, asocs) {
537 if (!asoc->peer.asconf_capable)
538 continue;
540 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
541 continue;
543 if (!sctp_state(asoc, ESTABLISHED))
544 continue;
546 /* Check if any address in the packed array of addresses is
547 * in the bind address list of the association. If so,
548 * do not send the asconf chunk to its peer, but continue with
549 * other associations.
551 addr_buf = addrs;
552 for (i = 0; i < addrcnt; i++) {
553 addr = addr_buf;
554 af = sctp_get_af_specific(addr->v4.sin_family);
555 if (!af) {
556 retval = -EINVAL;
557 goto out;
560 if (sctp_assoc_lookup_laddr(asoc, addr))
561 break;
563 addr_buf += af->sockaddr_len;
565 if (i < addrcnt)
566 continue;
568 /* Use the first valid address in bind addr list of
569 * association as Address Parameter of ASCONF CHUNK.
571 bp = &asoc->base.bind_addr;
572 p = bp->address_list.next;
573 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
574 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
575 addrcnt, SCTP_PARAM_ADD_IP);
576 if (!chunk) {
577 retval = -ENOMEM;
578 goto out;
581 /* Add the new addresses to the bind address list with
582 * use_as_src set to 0.
584 addr_buf = addrs;
585 for (i = 0; i < addrcnt; i++) {
586 addr = addr_buf;
587 af = sctp_get_af_specific(addr->v4.sin_family);
588 memcpy(&saveaddr, addr, af->sockaddr_len);
589 retval = sctp_add_bind_addr(bp, &saveaddr,
590 SCTP_ADDR_NEW, GFP_ATOMIC);
591 addr_buf += af->sockaddr_len;
593 if (asoc->src_out_of_asoc_ok) {
594 struct sctp_transport *trans;
596 list_for_each_entry(trans,
597 &asoc->peer.transport_addr_list, transports) {
598 /* Clear the source and route cache */
599 dst_release(trans->dst);
600 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
601 2*asoc->pathmtu, 4380));
602 trans->ssthresh = asoc->peer.i.a_rwnd;
603 trans->rto = asoc->rto_initial;
604 sctp_max_rto(asoc, trans);
605 trans->rtt = trans->srtt = trans->rttvar = 0;
606 sctp_transport_route(trans, NULL,
607 sctp_sk(asoc->base.sk));
610 retval = sctp_send_asconf(asoc, chunk);
613 out:
614 return retval;
617 /* Remove a list of addresses from bind addresses list. Do not remove the
618 * last address.
620 * Basically run through each address specified in the addrs/addrcnt
621 * array/length pair, determine if it is IPv6 or IPv4 and call
622 * sctp_del_bind() on it.
624 * If any of them fails, then the operation will be reversed and the
625 * ones that were removed will be added back.
627 * At least one address has to be left; if only one address is
628 * available, the operation will return -EBUSY.
630 * Only sctp_setsockopt_bindx() is supposed to call this function.
632 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
634 struct sctp_sock *sp = sctp_sk(sk);
635 struct sctp_endpoint *ep = sp->ep;
636 int cnt;
637 struct sctp_bind_addr *bp = &ep->base.bind_addr;
638 int retval = 0;
639 void *addr_buf;
640 union sctp_addr *sa_addr;
641 struct sctp_af *af;
643 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
644 __func__, sk, addrs, addrcnt);
646 addr_buf = addrs;
647 for (cnt = 0; cnt < addrcnt; cnt++) {
648 /* If the bind address list is empty or if there is only one
649 * bind address, there is nothing more to be removed (we need
650 * at least one address here).
652 if (list_empty(&bp->address_list) ||
653 (sctp_list_single_entry(&bp->address_list))) {
654 retval = -EBUSY;
655 goto err_bindx_rem;
658 sa_addr = addr_buf;
659 af = sctp_get_af_specific(sa_addr->sa.sa_family);
660 if (!af) {
661 retval = -EINVAL;
662 goto err_bindx_rem;
665 if (!af->addr_valid(sa_addr, sp, NULL)) {
666 retval = -EADDRNOTAVAIL;
667 goto err_bindx_rem;
670 if (sa_addr->v4.sin_port &&
671 sa_addr->v4.sin_port != htons(bp->port)) {
672 retval = -EINVAL;
673 goto err_bindx_rem;
676 if (!sa_addr->v4.sin_port)
677 sa_addr->v4.sin_port = htons(bp->port);
679 /* FIXME - There is probably a need to check if sk->sk_saddr and
680 * sk->sk_rcv_addr are currently set to one of the addresses to
681 * be removed. This is something which needs to be looked into
682 * when we are fixing the outstanding issues with multi-homing
683 * socket routing and failover schemes. Refer to comments in
684 * sctp_do_bind(). -daisy
686 retval = sctp_del_bind_addr(bp, sa_addr);
688 addr_buf += af->sockaddr_len;
689 err_bindx_rem:
690 if (retval < 0) {
691 /* Failed. Add the ones that has been removed back */
692 if (cnt > 0)
693 sctp_bindx_add(sk, addrs, cnt);
694 return retval;
698 return retval;
701 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
702 * the associations that are part of the endpoint indicating that a list of
703 * local addresses are removed from the endpoint.
705 * If any of the addresses is already in the bind address list of the
706 * association, we do not send the chunk for that association. But it will not
707 * affect other associations.
709 * Only sctp_setsockopt_bindx() is supposed to call this function.
711 static int sctp_send_asconf_del_ip(struct sock *sk,
712 struct sockaddr *addrs,
713 int addrcnt)
715 struct net *net = sock_net(sk);
716 struct sctp_sock *sp;
717 struct sctp_endpoint *ep;
718 struct sctp_association *asoc;
719 struct sctp_transport *transport;
720 struct sctp_bind_addr *bp;
721 struct sctp_chunk *chunk;
722 union sctp_addr *laddr;
723 void *addr_buf;
724 struct sctp_af *af;
725 struct sctp_sockaddr_entry *saddr;
726 int i;
727 int retval = 0;
728 int stored = 0;
730 chunk = NULL;
731 if (!net->sctp.addip_enable)
732 return retval;
734 sp = sctp_sk(sk);
735 ep = sp->ep;
737 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
738 __func__, sk, addrs, addrcnt);
740 list_for_each_entry(asoc, &ep->asocs, asocs) {
742 if (!asoc->peer.asconf_capable)
743 continue;
745 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
746 continue;
748 if (!sctp_state(asoc, ESTABLISHED))
749 continue;
751 /* Check if any address in the packed array of addresses is
752 * not present in the bind address list of the association.
753 * If so, do not send the asconf chunk to its peer, but
754 * continue with other associations.
756 addr_buf = addrs;
757 for (i = 0; i < addrcnt; i++) {
758 laddr = addr_buf;
759 af = sctp_get_af_specific(laddr->v4.sin_family);
760 if (!af) {
761 retval = -EINVAL;
762 goto out;
765 if (!sctp_assoc_lookup_laddr(asoc, laddr))
766 break;
768 addr_buf += af->sockaddr_len;
770 if (i < addrcnt)
771 continue;
773 /* Find one address in the association's bind address list
774 * that is not in the packed array of addresses. This is to
775 * make sure that we do not delete all the addresses in the
776 * association.
778 bp = &asoc->base.bind_addr;
779 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
780 addrcnt, sp);
781 if ((laddr == NULL) && (addrcnt == 1)) {
782 if (asoc->asconf_addr_del_pending)
783 continue;
784 asoc->asconf_addr_del_pending =
785 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
786 if (asoc->asconf_addr_del_pending == NULL) {
787 retval = -ENOMEM;
788 goto out;
790 asoc->asconf_addr_del_pending->sa.sa_family =
791 addrs->sa_family;
792 asoc->asconf_addr_del_pending->v4.sin_port =
793 htons(bp->port);
794 if (addrs->sa_family == AF_INET) {
795 struct sockaddr_in *sin;
797 sin = (struct sockaddr_in *)addrs;
798 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
799 } else if (addrs->sa_family == AF_INET6) {
800 struct sockaddr_in6 *sin6;
802 sin6 = (struct sockaddr_in6 *)addrs;
803 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
806 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
807 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
808 asoc->asconf_addr_del_pending);
810 asoc->src_out_of_asoc_ok = 1;
811 stored = 1;
812 goto skip_mkasconf;
815 /* We do not need RCU protection throughout this loop
816 * because this is done under a socket lock from the
817 * setsockopt call.
819 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
820 SCTP_PARAM_DEL_IP);
821 if (!chunk) {
822 retval = -ENOMEM;
823 goto out;
826 skip_mkasconf:
827 /* Reset use_as_src flag for the addresses in the bind address
828 * list that are to be deleted.
830 addr_buf = addrs;
831 for (i = 0; i < addrcnt; i++) {
832 laddr = addr_buf;
833 af = sctp_get_af_specific(laddr->v4.sin_family);
834 list_for_each_entry(saddr, &bp->address_list, list) {
835 if (sctp_cmp_addr_exact(&saddr->a, laddr))
836 saddr->state = SCTP_ADDR_DEL;
838 addr_buf += af->sockaddr_len;
841 /* Update the route and saddr entries for all the transports
842 * as some of the addresses in the bind address list are
843 * about to be deleted and cannot be used as source addresses.
845 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
846 transports) {
847 dst_release(transport->dst);
848 sctp_transport_route(transport, NULL,
849 sctp_sk(asoc->base.sk));
852 if (stored)
853 /* We don't need to transmit ASCONF */
854 continue;
855 retval = sctp_send_asconf(asoc, chunk);
857 out:
858 return retval;
861 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
862 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
864 struct sock *sk = sctp_opt2sk(sp);
865 union sctp_addr *addr;
866 struct sctp_af *af;
868 /* It is safe to write port space in caller. */
869 addr = &addrw->a;
870 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
871 af = sctp_get_af_specific(addr->sa.sa_family);
872 if (!af)
873 return -EINVAL;
874 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
875 return -EINVAL;
877 if (addrw->state == SCTP_ADDR_NEW)
878 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
879 else
880 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
883 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
885 * API 8.1
886 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
887 * int flags);
889 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
890 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
891 * or IPv6 addresses.
893 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
894 * Section 3.1.2 for this usage.
896 * addrs is a pointer to an array of one or more socket addresses. Each
897 * address is contained in its appropriate structure (i.e. struct
898 * sockaddr_in or struct sockaddr_in6) the family of the address type
899 * must be used to distinguish the address length (note that this
900 * representation is termed a "packed array" of addresses). The caller
901 * specifies the number of addresses in the array with addrcnt.
903 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
904 * -1, and sets errno to the appropriate error code.
906 * For SCTP, the port given in each socket address must be the same, or
907 * sctp_bindx() will fail, setting errno to EINVAL.
909 * The flags parameter is formed from the bitwise OR of zero or more of
910 * the following currently defined flags:
912 * SCTP_BINDX_ADD_ADDR
914 * SCTP_BINDX_REM_ADDR
916 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
917 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
918 * addresses from the association. The two flags are mutually exclusive;
919 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
920 * not remove all addresses from an association; sctp_bindx() will
921 * reject such an attempt with EINVAL.
923 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
924 * additional addresses with an endpoint after calling bind(). Or use
925 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
926 * socket is associated with so that no new association accepted will be
927 * associated with those addresses. If the endpoint supports dynamic
928 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
929 * endpoint to send the appropriate message to the peer to change the
930 * peers address lists.
932 * Adding and removing addresses from a connected association is
933 * optional functionality. Implementations that do not support this
934 * functionality should return EOPNOTSUPP.
936 * Basically do nothing but copying the addresses from user to kernel
937 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
938 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
939 * from userspace.
941 * We don't use copy_from_user() for optimization: we first do the
942 * sanity checks (buffer size -fast- and access check-healthy
943 * pointer); if all of those succeed, then we can alloc the memory
944 * (expensive operation) needed to copy the data to kernel. Then we do
945 * the copying without checking the user space area
946 * (__copy_from_user()).
948 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
949 * it.
951 * sk The sk of the socket
952 * addrs The pointer to the addresses in user land
953 * addrssize Size of the addrs buffer
954 * op Operation to perform (add or remove, see the flags of
955 * sctp_bindx)
957 * Returns 0 if ok, <0 errno code on error.
959 static int sctp_setsockopt_bindx(struct sock* sk,
960 struct sockaddr __user *addrs,
961 int addrs_size, int op)
963 struct sockaddr *kaddrs;
964 int err;
965 int addrcnt = 0;
966 int walk_size = 0;
967 struct sockaddr *sa_addr;
968 void *addr_buf;
969 struct sctp_af *af;
971 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
972 __func__, sk, addrs, addrs_size, op);
974 if (unlikely(addrs_size <= 0))
975 return -EINVAL;
977 /* Check the user passed a healthy pointer. */
978 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
979 return -EFAULT;
981 /* Alloc space for the address array in kernel memory. */
982 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
983 if (unlikely(!kaddrs))
984 return -ENOMEM;
986 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
987 kfree(kaddrs);
988 return -EFAULT;
991 /* Walk through the addrs buffer and count the number of addresses. */
992 addr_buf = kaddrs;
993 while (walk_size < addrs_size) {
994 if (walk_size + sizeof(sa_family_t) > addrs_size) {
995 kfree(kaddrs);
996 return -EINVAL;
999 sa_addr = addr_buf;
1000 af = sctp_get_af_specific(sa_addr->sa_family);
1002 /* If the address family is not supported or if this address
1003 * causes the address buffer to overflow return EINVAL.
1005 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1006 kfree(kaddrs);
1007 return -EINVAL;
1009 addrcnt++;
1010 addr_buf += af->sockaddr_len;
1011 walk_size += af->sockaddr_len;
1014 /* Do the work. */
1015 switch (op) {
1016 case SCTP_BINDX_ADD_ADDR:
1017 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1018 if (err)
1019 goto out;
1020 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1021 break;
1023 case SCTP_BINDX_REM_ADDR:
1024 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1025 if (err)
1026 goto out;
1027 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1028 break;
1030 default:
1031 err = -EINVAL;
1032 break;
1035 out:
1036 kfree(kaddrs);
1038 return err;
1041 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1043 * Common routine for handling connect() and sctp_connectx().
1044 * Connect will come in with just a single address.
1046 static int __sctp_connect(struct sock* sk,
1047 struct sockaddr *kaddrs,
1048 int addrs_size,
1049 sctp_assoc_t *assoc_id)
1051 struct net *net = sock_net(sk);
1052 struct sctp_sock *sp;
1053 struct sctp_endpoint *ep;
1054 struct sctp_association *asoc = NULL;
1055 struct sctp_association *asoc2;
1056 struct sctp_transport *transport;
1057 union sctp_addr to;
1058 struct sctp_af *af;
1059 sctp_scope_t scope;
1060 long timeo;
1061 int err = 0;
1062 int addrcnt = 0;
1063 int walk_size = 0;
1064 union sctp_addr *sa_addr = NULL;
1065 void *addr_buf;
1066 unsigned short port;
1067 unsigned int f_flags = 0;
1069 sp = sctp_sk(sk);
1070 ep = sp->ep;
1072 /* connect() cannot be done on a socket that is already in ESTABLISHED
1073 * state - UDP-style peeled off socket or a TCP-style socket that
1074 * is already connected.
1075 * It cannot be done even on a TCP-style listening socket.
1077 if (sctp_sstate(sk, ESTABLISHED) ||
1078 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1079 err = -EISCONN;
1080 goto out_free;
1083 /* Walk through the addrs buffer and count the number of addresses. */
1084 addr_buf = kaddrs;
1085 while (walk_size < addrs_size) {
1086 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1087 err = -EINVAL;
1088 goto out_free;
1091 sa_addr = addr_buf;
1092 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1094 /* If the address family is not supported or if this address
1095 * causes the address buffer to overflow return EINVAL.
1097 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1098 err = -EINVAL;
1099 goto out_free;
1102 port = ntohs(sa_addr->v4.sin_port);
1104 /* Save current address so we can work with it */
1105 memcpy(&to, sa_addr, af->sockaddr_len);
1107 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1108 if (err)
1109 goto out_free;
1111 /* Make sure the destination port is correctly set
1112 * in all addresses.
1114 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1115 err = -EINVAL;
1116 goto out_free;
1119 /* Check if there already is a matching association on the
1120 * endpoint (other than the one created here).
1122 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1123 if (asoc2 && asoc2 != asoc) {
1124 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1125 err = -EISCONN;
1126 else
1127 err = -EALREADY;
1128 goto out_free;
1131 /* If we could not find a matching association on the endpoint,
1132 * make sure that there is no peeled-off association matching
1133 * the peer address even on another socket.
1135 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1136 err = -EADDRNOTAVAIL;
1137 goto out_free;
1140 if (!asoc) {
1141 /* If a bind() or sctp_bindx() is not called prior to
1142 * an sctp_connectx() call, the system picks an
1143 * ephemeral port and will choose an address set
1144 * equivalent to binding with a wildcard address.
1146 if (!ep->base.bind_addr.port) {
1147 if (sctp_autobind(sk)) {
1148 err = -EAGAIN;
1149 goto out_free;
1151 } else {
1153 * If an unprivileged user inherits a 1-many
1154 * style socket with open associations on a
1155 * privileged port, it MAY be permitted to
1156 * accept new associations, but it SHOULD NOT
1157 * be permitted to open new associations.
1159 if (ep->base.bind_addr.port < PROT_SOCK &&
1160 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1161 err = -EACCES;
1162 goto out_free;
1166 scope = sctp_scope(&to);
1167 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1168 if (!asoc) {
1169 err = -ENOMEM;
1170 goto out_free;
1173 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1174 GFP_KERNEL);
1175 if (err < 0) {
1176 goto out_free;
1181 /* Prime the peer's transport structures. */
1182 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1183 SCTP_UNKNOWN);
1184 if (!transport) {
1185 err = -ENOMEM;
1186 goto out_free;
1189 addrcnt++;
1190 addr_buf += af->sockaddr_len;
1191 walk_size += af->sockaddr_len;
1194 /* In case the user of sctp_connectx() wants an association
1195 * id back, assign one now.
1197 if (assoc_id) {
1198 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1199 if (err < 0)
1200 goto out_free;
1203 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1204 if (err < 0) {
1205 goto out_free;
1208 /* Initialize sk's dport and daddr for getpeername() */
1209 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1210 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1211 af->to_sk_daddr(sa_addr, sk);
1212 sk->sk_err = 0;
1214 /* in-kernel sockets don't generally have a file allocated to them
1215 * if all they do is call sock_create_kern().
1217 if (sk->sk_socket->file)
1218 f_flags = sk->sk_socket->file->f_flags;
1220 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1222 err = sctp_wait_for_connect(asoc, &timeo);
1223 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1224 *assoc_id = asoc->assoc_id;
1226 /* Don't free association on exit. */
1227 asoc = NULL;
1229 out_free:
1230 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1231 __func__, asoc, kaddrs, err);
1233 if (asoc) {
1234 /* sctp_primitive_ASSOCIATE may have added this association
1235 * To the hash table, try to unhash it, just in case, its a noop
1236 * if it wasn't hashed so we're safe
1238 sctp_unhash_established(asoc);
1239 sctp_association_free(asoc);
1241 return err;
1244 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1246 * API 8.9
1247 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1248 * sctp_assoc_t *asoc);
1250 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1251 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1252 * or IPv6 addresses.
1254 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1255 * Section 3.1.2 for this usage.
1257 * addrs is a pointer to an array of one or more socket addresses. Each
1258 * address is contained in its appropriate structure (i.e. struct
1259 * sockaddr_in or struct sockaddr_in6) the family of the address type
1260 * must be used to distengish the address length (note that this
1261 * representation is termed a "packed array" of addresses). The caller
1262 * specifies the number of addresses in the array with addrcnt.
1264 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1265 * the association id of the new association. On failure, sctp_connectx()
1266 * returns -1, and sets errno to the appropriate error code. The assoc_id
1267 * is not touched by the kernel.
1269 * For SCTP, the port given in each socket address must be the same, or
1270 * sctp_connectx() will fail, setting errno to EINVAL.
1272 * An application can use sctp_connectx to initiate an association with
1273 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1274 * allows a caller to specify multiple addresses at which a peer can be
1275 * reached. The way the SCTP stack uses the list of addresses to set up
1276 * the association is implementation dependent. This function only
1277 * specifies that the stack will try to make use of all the addresses in
1278 * the list when needed.
1280 * Note that the list of addresses passed in is only used for setting up
1281 * the association. It does not necessarily equal the set of addresses
1282 * the peer uses for the resulting association. If the caller wants to
1283 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1284 * retrieve them after the association has been set up.
1286 * Basically do nothing but copying the addresses from user to kernel
1287 * land and invoking either sctp_connectx(). This is used for tunneling
1288 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1290 * We don't use copy_from_user() for optimization: we first do the
1291 * sanity checks (buffer size -fast- and access check-healthy
1292 * pointer); if all of those succeed, then we can alloc the memory
1293 * (expensive operation) needed to copy the data to kernel. Then we do
1294 * the copying without checking the user space area
1295 * (__copy_from_user()).
1297 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1298 * it.
1300 * sk The sk of the socket
1301 * addrs The pointer to the addresses in user land
1302 * addrssize Size of the addrs buffer
1304 * Returns >=0 if ok, <0 errno code on error.
1306 static int __sctp_setsockopt_connectx(struct sock* sk,
1307 struct sockaddr __user *addrs,
1308 int addrs_size,
1309 sctp_assoc_t *assoc_id)
1311 int err = 0;
1312 struct sockaddr *kaddrs;
1314 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1315 __func__, sk, addrs, addrs_size);
1317 if (unlikely(addrs_size <= 0))
1318 return -EINVAL;
1320 /* Check the user passed a healthy pointer. */
1321 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1322 return -EFAULT;
1324 /* Alloc space for the address array in kernel memory. */
1325 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1326 if (unlikely(!kaddrs))
1327 return -ENOMEM;
1329 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1330 err = -EFAULT;
1331 } else {
1332 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1335 kfree(kaddrs);
1337 return err;
1341 * This is an older interface. It's kept for backward compatibility
1342 * to the option that doesn't provide association id.
1344 static int sctp_setsockopt_connectx_old(struct sock* sk,
1345 struct sockaddr __user *addrs,
1346 int addrs_size)
1348 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1352 * New interface for the API. The since the API is done with a socket
1353 * option, to make it simple we feed back the association id is as a return
1354 * indication to the call. Error is always negative and association id is
1355 * always positive.
1357 static int sctp_setsockopt_connectx(struct sock* sk,
1358 struct sockaddr __user *addrs,
1359 int addrs_size)
1361 sctp_assoc_t assoc_id = 0;
1362 int err = 0;
1364 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1366 if (err)
1367 return err;
1368 else
1369 return assoc_id;
1373 * New (hopefully final) interface for the API.
1374 * We use the sctp_getaddrs_old structure so that use-space library
1375 * can avoid any unnecessary allocations. The only defferent part
1376 * is that we store the actual length of the address buffer into the
1377 * addrs_num structure member. That way we can re-use the existing
1378 * code.
1380 static int sctp_getsockopt_connectx3(struct sock* sk, int len,
1381 char __user *optval,
1382 int __user *optlen)
1384 struct sctp_getaddrs_old param;
1385 sctp_assoc_t assoc_id = 0;
1386 int err = 0;
1388 if (len < sizeof(param))
1389 return -EINVAL;
1391 if (copy_from_user(&param, optval, sizeof(param)))
1392 return -EFAULT;
1394 err = __sctp_setsockopt_connectx(sk,
1395 (struct sockaddr __user *)param.addrs,
1396 param.addr_num, &assoc_id);
1398 if (err == 0 || err == -EINPROGRESS) {
1399 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1400 return -EFAULT;
1401 if (put_user(sizeof(assoc_id), optlen))
1402 return -EFAULT;
1405 return err;
1408 /* API 3.1.4 close() - UDP Style Syntax
1409 * Applications use close() to perform graceful shutdown (as described in
1410 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1411 * by a UDP-style socket.
1413 * The syntax is
1415 * ret = close(int sd);
1417 * sd - the socket descriptor of the associations to be closed.
1419 * To gracefully shutdown a specific association represented by the
1420 * UDP-style socket, an application should use the sendmsg() call,
1421 * passing no user data, but including the appropriate flag in the
1422 * ancillary data (see Section xxxx).
1424 * If sd in the close() call is a branched-off socket representing only
1425 * one association, the shutdown is performed on that association only.
1427 * 4.1.6 close() - TCP Style Syntax
1429 * Applications use close() to gracefully close down an association.
1431 * The syntax is:
1433 * int close(int sd);
1435 * sd - the socket descriptor of the association to be closed.
1437 * After an application calls close() on a socket descriptor, no further
1438 * socket operations will succeed on that descriptor.
1440 * API 7.1.4 SO_LINGER
1442 * An application using the TCP-style socket can use this option to
1443 * perform the SCTP ABORT primitive. The linger option structure is:
1445 * struct linger {
1446 * int l_onoff; // option on/off
1447 * int l_linger; // linger time
1448 * };
1450 * To enable the option, set l_onoff to 1. If the l_linger value is set
1451 * to 0, calling close() is the same as the ABORT primitive. If the
1452 * value is set to a negative value, the setsockopt() call will return
1453 * an error. If the value is set to a positive value linger_time, the
1454 * close() can be blocked for at most linger_time ms. If the graceful
1455 * shutdown phase does not finish during this period, close() will
1456 * return but the graceful shutdown phase continues in the system.
1458 static void sctp_close(struct sock *sk, long timeout)
1460 struct net *net = sock_net(sk);
1461 struct sctp_endpoint *ep;
1462 struct sctp_association *asoc;
1463 struct list_head *pos, *temp;
1464 unsigned int data_was_unread;
1466 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1468 sctp_lock_sock(sk);
1469 sk->sk_shutdown = SHUTDOWN_MASK;
1470 sk->sk_state = SCTP_SS_CLOSING;
1472 ep = sctp_sk(sk)->ep;
1474 /* Clean up any skbs sitting on the receive queue. */
1475 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1476 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1478 /* Walk all associations on an endpoint. */
1479 list_for_each_safe(pos, temp, &ep->asocs) {
1480 asoc = list_entry(pos, struct sctp_association, asocs);
1482 if (sctp_style(sk, TCP)) {
1483 /* A closed association can still be in the list if
1484 * it belongs to a TCP-style listening socket that is
1485 * not yet accepted. If so, free it. If not, send an
1486 * ABORT or SHUTDOWN based on the linger options.
1488 if (sctp_state(asoc, CLOSED)) {
1489 sctp_unhash_established(asoc);
1490 sctp_association_free(asoc);
1491 continue;
1495 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1496 !skb_queue_empty(&asoc->ulpq.reasm) ||
1497 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1498 struct sctp_chunk *chunk;
1500 chunk = sctp_make_abort_user(asoc, NULL, 0);
1501 if (chunk)
1502 sctp_primitive_ABORT(net, asoc, chunk);
1503 } else
1504 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1507 /* On a TCP-style socket, block for at most linger_time if set. */
1508 if (sctp_style(sk, TCP) && timeout)
1509 sctp_wait_for_close(sk, timeout);
1511 /* This will run the backlog queue. */
1512 sctp_release_sock(sk);
1514 /* Supposedly, no process has access to the socket, but
1515 * the net layers still may.
1517 sctp_local_bh_disable();
1518 sctp_bh_lock_sock(sk);
1520 /* Hold the sock, since sk_common_release() will put sock_put()
1521 * and we have just a little more cleanup.
1523 sock_hold(sk);
1524 sk_common_release(sk);
1526 sctp_bh_unlock_sock(sk);
1527 sctp_local_bh_enable();
1529 sock_put(sk);
1531 SCTP_DBG_OBJCNT_DEC(sock);
1534 /* Handle EPIPE error. */
1535 static int sctp_error(struct sock *sk, int flags, int err)
1537 if (err == -EPIPE)
1538 err = sock_error(sk) ? : -EPIPE;
1539 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1540 send_sig(SIGPIPE, current, 0);
1541 return err;
1544 /* API 3.1.3 sendmsg() - UDP Style Syntax
1546 * An application uses sendmsg() and recvmsg() calls to transmit data to
1547 * and receive data from its peer.
1549 * ssize_t sendmsg(int socket, const struct msghdr *message,
1550 * int flags);
1552 * socket - the socket descriptor of the endpoint.
1553 * message - pointer to the msghdr structure which contains a single
1554 * user message and possibly some ancillary data.
1556 * See Section 5 for complete description of the data
1557 * structures.
1559 * flags - flags sent or received with the user message, see Section
1560 * 5 for complete description of the flags.
1562 * Note: This function could use a rewrite especially when explicit
1563 * connect support comes in.
1565 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1567 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1569 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1570 struct msghdr *msg, size_t msg_len)
1572 struct net *net = sock_net(sk);
1573 struct sctp_sock *sp;
1574 struct sctp_endpoint *ep;
1575 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1576 struct sctp_transport *transport, *chunk_tp;
1577 struct sctp_chunk *chunk;
1578 union sctp_addr to;
1579 struct sockaddr *msg_name = NULL;
1580 struct sctp_sndrcvinfo default_sinfo;
1581 struct sctp_sndrcvinfo *sinfo;
1582 struct sctp_initmsg *sinit;
1583 sctp_assoc_t associd = 0;
1584 sctp_cmsgs_t cmsgs = { NULL };
1585 int err;
1586 sctp_scope_t scope;
1587 long timeo;
1588 __u16 sinfo_flags = 0;
1589 struct sctp_datamsg *datamsg;
1590 int msg_flags = msg->msg_flags;
1592 err = 0;
1593 sp = sctp_sk(sk);
1594 ep = sp->ep;
1596 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1597 msg, msg_len, ep);
1599 /* We cannot send a message over a TCP-style listening socket. */
1600 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1601 err = -EPIPE;
1602 goto out_nounlock;
1605 /* Parse out the SCTP CMSGs. */
1606 err = sctp_msghdr_parse(msg, &cmsgs);
1607 if (err) {
1608 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1609 goto out_nounlock;
1612 /* Fetch the destination address for this packet. This
1613 * address only selects the association--it is not necessarily
1614 * the address we will send to.
1615 * For a peeled-off socket, msg_name is ignored.
1617 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1618 int msg_namelen = msg->msg_namelen;
1620 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1621 msg_namelen);
1622 if (err)
1623 return err;
1625 if (msg_namelen > sizeof(to))
1626 msg_namelen = sizeof(to);
1627 memcpy(&to, msg->msg_name, msg_namelen);
1628 msg_name = msg->msg_name;
1631 sinfo = cmsgs.info;
1632 sinit = cmsgs.init;
1634 /* Did the user specify SNDRCVINFO? */
1635 if (sinfo) {
1636 sinfo_flags = sinfo->sinfo_flags;
1637 associd = sinfo->sinfo_assoc_id;
1640 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1641 msg_len, sinfo_flags);
1643 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1644 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1645 err = -EINVAL;
1646 goto out_nounlock;
1649 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1650 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1651 * If SCTP_ABORT is set, the message length could be non zero with
1652 * the msg_iov set to the user abort reason.
1654 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1655 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1656 err = -EINVAL;
1657 goto out_nounlock;
1660 /* If SCTP_ADDR_OVER is set, there must be an address
1661 * specified in msg_name.
1663 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1664 err = -EINVAL;
1665 goto out_nounlock;
1668 transport = NULL;
1670 pr_debug("%s: about to look up association\n", __func__);
1672 sctp_lock_sock(sk);
1674 /* If a msg_name has been specified, assume this is to be used. */
1675 if (msg_name) {
1676 /* Look for a matching association on the endpoint. */
1677 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1678 if (!asoc) {
1679 /* If we could not find a matching association on the
1680 * endpoint, make sure that it is not a TCP-style
1681 * socket that already has an association or there is
1682 * no peeled-off association on another socket.
1684 if ((sctp_style(sk, TCP) &&
1685 sctp_sstate(sk, ESTABLISHED)) ||
1686 sctp_endpoint_is_peeled_off(ep, &to)) {
1687 err = -EADDRNOTAVAIL;
1688 goto out_unlock;
1691 } else {
1692 asoc = sctp_id2assoc(sk, associd);
1693 if (!asoc) {
1694 err = -EPIPE;
1695 goto out_unlock;
1699 if (asoc) {
1700 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1702 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1703 * socket that has an association in CLOSED state. This can
1704 * happen when an accepted socket has an association that is
1705 * already CLOSED.
1707 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1708 err = -EPIPE;
1709 goto out_unlock;
1712 if (sinfo_flags & SCTP_EOF) {
1713 pr_debug("%s: shutting down association:%p\n",
1714 __func__, asoc);
1716 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1717 err = 0;
1718 goto out_unlock;
1720 if (sinfo_flags & SCTP_ABORT) {
1722 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1723 if (!chunk) {
1724 err = -ENOMEM;
1725 goto out_unlock;
1728 pr_debug("%s: aborting association:%p\n",
1729 __func__, asoc);
1731 sctp_primitive_ABORT(net, asoc, chunk);
1732 err = 0;
1733 goto out_unlock;
1737 /* Do we need to create the association? */
1738 if (!asoc) {
1739 pr_debug("%s: there is no association yet\n", __func__);
1741 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1742 err = -EINVAL;
1743 goto out_unlock;
1746 /* Check for invalid stream against the stream counts,
1747 * either the default or the user specified stream counts.
1749 if (sinfo) {
1750 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1751 /* Check against the defaults. */
1752 if (sinfo->sinfo_stream >=
1753 sp->initmsg.sinit_num_ostreams) {
1754 err = -EINVAL;
1755 goto out_unlock;
1757 } else {
1758 /* Check against the requested. */
1759 if (sinfo->sinfo_stream >=
1760 sinit->sinit_num_ostreams) {
1761 err = -EINVAL;
1762 goto out_unlock;
1768 * API 3.1.2 bind() - UDP Style Syntax
1769 * If a bind() or sctp_bindx() is not called prior to a
1770 * sendmsg() call that initiates a new association, the
1771 * system picks an ephemeral port and will choose an address
1772 * set equivalent to binding with a wildcard address.
1774 if (!ep->base.bind_addr.port) {
1775 if (sctp_autobind(sk)) {
1776 err = -EAGAIN;
1777 goto out_unlock;
1779 } else {
1781 * If an unprivileged user inherits a one-to-many
1782 * style socket with open associations on a privileged
1783 * port, it MAY be permitted to accept new associations,
1784 * but it SHOULD NOT be permitted to open new
1785 * associations.
1787 if (ep->base.bind_addr.port < PROT_SOCK &&
1788 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1789 err = -EACCES;
1790 goto out_unlock;
1794 scope = sctp_scope(&to);
1795 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1796 if (!new_asoc) {
1797 err = -ENOMEM;
1798 goto out_unlock;
1800 asoc = new_asoc;
1801 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1802 if (err < 0) {
1803 err = -ENOMEM;
1804 goto out_free;
1807 /* If the SCTP_INIT ancillary data is specified, set all
1808 * the association init values accordingly.
1810 if (sinit) {
1811 if (sinit->sinit_num_ostreams) {
1812 asoc->c.sinit_num_ostreams =
1813 sinit->sinit_num_ostreams;
1815 if (sinit->sinit_max_instreams) {
1816 asoc->c.sinit_max_instreams =
1817 sinit->sinit_max_instreams;
1819 if (sinit->sinit_max_attempts) {
1820 asoc->max_init_attempts
1821 = sinit->sinit_max_attempts;
1823 if (sinit->sinit_max_init_timeo) {
1824 asoc->max_init_timeo =
1825 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1829 /* Prime the peer's transport structures. */
1830 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1831 if (!transport) {
1832 err = -ENOMEM;
1833 goto out_free;
1837 /* ASSERT: we have a valid association at this point. */
1838 pr_debug("%s: we have a valid association\n", __func__);
1840 if (!sinfo) {
1841 /* If the user didn't specify SNDRCVINFO, make up one with
1842 * some defaults.
1844 memset(&default_sinfo, 0, sizeof(default_sinfo));
1845 default_sinfo.sinfo_stream = asoc->default_stream;
1846 default_sinfo.sinfo_flags = asoc->default_flags;
1847 default_sinfo.sinfo_ppid = asoc->default_ppid;
1848 default_sinfo.sinfo_context = asoc->default_context;
1849 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1850 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1851 sinfo = &default_sinfo;
1854 /* API 7.1.7, the sndbuf size per association bounds the
1855 * maximum size of data that can be sent in a single send call.
1857 if (msg_len > sk->sk_sndbuf) {
1858 err = -EMSGSIZE;
1859 goto out_free;
1862 if (asoc->pmtu_pending)
1863 sctp_assoc_pending_pmtu(sk, asoc);
1865 /* If fragmentation is disabled and the message length exceeds the
1866 * association fragmentation point, return EMSGSIZE. The I-D
1867 * does not specify what this error is, but this looks like
1868 * a great fit.
1870 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1871 err = -EMSGSIZE;
1872 goto out_free;
1875 /* Check for invalid stream. */
1876 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1877 err = -EINVAL;
1878 goto out_free;
1881 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1882 if (!sctp_wspace(asoc)) {
1883 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1884 if (err)
1885 goto out_free;
1888 /* If an address is passed with the sendto/sendmsg call, it is used
1889 * to override the primary destination address in the TCP model, or
1890 * when SCTP_ADDR_OVER flag is set in the UDP model.
1892 if ((sctp_style(sk, TCP) && msg_name) ||
1893 (sinfo_flags & SCTP_ADDR_OVER)) {
1894 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1895 if (!chunk_tp) {
1896 err = -EINVAL;
1897 goto out_free;
1899 } else
1900 chunk_tp = NULL;
1902 /* Auto-connect, if we aren't connected already. */
1903 if (sctp_state(asoc, CLOSED)) {
1904 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1905 if (err < 0)
1906 goto out_free;
1908 pr_debug("%s: we associated primitively\n", __func__);
1911 /* Break the message into multiple chunks of maximum size. */
1912 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1913 if (IS_ERR(datamsg)) {
1914 err = PTR_ERR(datamsg);
1915 goto out_free;
1918 /* Now send the (possibly) fragmented message. */
1919 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1920 sctp_chunk_hold(chunk);
1922 /* Do accounting for the write space. */
1923 sctp_set_owner_w(chunk);
1925 chunk->transport = chunk_tp;
1928 /* Send it to the lower layers. Note: all chunks
1929 * must either fail or succeed. The lower layer
1930 * works that way today. Keep it that way or this
1931 * breaks.
1933 err = sctp_primitive_SEND(net, asoc, datamsg);
1934 /* Did the lower layer accept the chunk? */
1935 if (err) {
1936 sctp_datamsg_free(datamsg);
1937 goto out_free;
1940 pr_debug("%s: we sent primitively\n", __func__);
1942 sctp_datamsg_put(datamsg);
1943 err = msg_len;
1945 /* If we are already past ASSOCIATE, the lower
1946 * layers are responsible for association cleanup.
1948 goto out_unlock;
1950 out_free:
1951 if (new_asoc) {
1952 sctp_unhash_established(asoc);
1953 sctp_association_free(asoc);
1955 out_unlock:
1956 sctp_release_sock(sk);
1958 out_nounlock:
1959 return sctp_error(sk, msg_flags, err);
1961 #if 0
1962 do_sock_err:
1963 if (msg_len)
1964 err = msg_len;
1965 else
1966 err = sock_error(sk);
1967 goto out;
1969 do_interrupted:
1970 if (msg_len)
1971 err = msg_len;
1972 goto out;
1973 #endif /* 0 */
1976 /* This is an extended version of skb_pull() that removes the data from the
1977 * start of a skb even when data is spread across the list of skb's in the
1978 * frag_list. len specifies the total amount of data that needs to be removed.
1979 * when 'len' bytes could be removed from the skb, it returns 0.
1980 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1981 * could not be removed.
1983 static int sctp_skb_pull(struct sk_buff *skb, int len)
1985 struct sk_buff *list;
1986 int skb_len = skb_headlen(skb);
1987 int rlen;
1989 if (len <= skb_len) {
1990 __skb_pull(skb, len);
1991 return 0;
1993 len -= skb_len;
1994 __skb_pull(skb, skb_len);
1996 skb_walk_frags(skb, list) {
1997 rlen = sctp_skb_pull(list, len);
1998 skb->len -= (len-rlen);
1999 skb->data_len -= (len-rlen);
2001 if (!rlen)
2002 return 0;
2004 len = rlen;
2007 return len;
2010 /* API 3.1.3 recvmsg() - UDP Style Syntax
2012 * ssize_t recvmsg(int socket, struct msghdr *message,
2013 * int flags);
2015 * socket - the socket descriptor of the endpoint.
2016 * message - pointer to the msghdr structure which contains a single
2017 * user message and possibly some ancillary data.
2019 * See Section 5 for complete description of the data
2020 * structures.
2022 * flags - flags sent or received with the user message, see Section
2023 * 5 for complete description of the flags.
2025 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2027 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2028 struct msghdr *msg, size_t len, int noblock,
2029 int flags, int *addr_len)
2031 struct sctp_ulpevent *event = NULL;
2032 struct sctp_sock *sp = sctp_sk(sk);
2033 struct sk_buff *skb;
2034 int copied;
2035 int err = 0;
2036 int skb_len;
2038 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2039 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2040 addr_len);
2042 sctp_lock_sock(sk);
2044 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2045 err = -ENOTCONN;
2046 goto out;
2049 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2050 if (!skb)
2051 goto out;
2053 /* Get the total length of the skb including any skb's in the
2054 * frag_list.
2056 skb_len = skb->len;
2058 copied = skb_len;
2059 if (copied > len)
2060 copied = len;
2062 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2064 event = sctp_skb2event(skb);
2066 if (err)
2067 goto out_free;
2069 sock_recv_ts_and_drops(msg, sk, skb);
2070 if (sctp_ulpevent_is_notification(event)) {
2071 msg->msg_flags |= MSG_NOTIFICATION;
2072 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2073 } else {
2074 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2077 /* Check if we allow SCTP_SNDRCVINFO. */
2078 if (sp->subscribe.sctp_data_io_event)
2079 sctp_ulpevent_read_sndrcvinfo(event, msg);
2080 #if 0
2081 /* FIXME: we should be calling IP/IPv6 layers. */
2082 if (sk->sk_protinfo.af_inet.cmsg_flags)
2083 ip_cmsg_recv(msg, skb);
2084 #endif
2086 err = copied;
2088 /* If skb's length exceeds the user's buffer, update the skb and
2089 * push it back to the receive_queue so that the next call to
2090 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2092 if (skb_len > copied) {
2093 msg->msg_flags &= ~MSG_EOR;
2094 if (flags & MSG_PEEK)
2095 goto out_free;
2096 sctp_skb_pull(skb, copied);
2097 skb_queue_head(&sk->sk_receive_queue, skb);
2099 /* When only partial message is copied to the user, increase
2100 * rwnd by that amount. If all the data in the skb is read,
2101 * rwnd is updated when the event is freed.
2103 if (!sctp_ulpevent_is_notification(event))
2104 sctp_assoc_rwnd_increase(event->asoc, copied);
2105 goto out;
2106 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2107 (event->msg_flags & MSG_EOR))
2108 msg->msg_flags |= MSG_EOR;
2109 else
2110 msg->msg_flags &= ~MSG_EOR;
2112 out_free:
2113 if (flags & MSG_PEEK) {
2114 /* Release the skb reference acquired after peeking the skb in
2115 * sctp_skb_recv_datagram().
2117 kfree_skb(skb);
2118 } else {
2119 /* Free the event which includes releasing the reference to
2120 * the owner of the skb, freeing the skb and updating the
2121 * rwnd.
2123 sctp_ulpevent_free(event);
2125 out:
2126 sctp_release_sock(sk);
2127 return err;
2130 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2132 * This option is a on/off flag. If enabled no SCTP message
2133 * fragmentation will be performed. Instead if a message being sent
2134 * exceeds the current PMTU size, the message will NOT be sent and
2135 * instead a error will be indicated to the user.
2137 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2138 char __user *optval,
2139 unsigned int optlen)
2141 int val;
2143 if (optlen < sizeof(int))
2144 return -EINVAL;
2146 if (get_user(val, (int __user *)optval))
2147 return -EFAULT;
2149 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2151 return 0;
2154 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2155 unsigned int optlen)
2157 struct sctp_association *asoc;
2158 struct sctp_ulpevent *event;
2160 if (optlen > sizeof(struct sctp_event_subscribe))
2161 return -EINVAL;
2162 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2163 return -EFAULT;
2166 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2167 * if there is no data to be sent or retransmit, the stack will
2168 * immediately send up this notification.
2170 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2171 &sctp_sk(sk)->subscribe)) {
2172 asoc = sctp_id2assoc(sk, 0);
2174 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2175 event = sctp_ulpevent_make_sender_dry_event(asoc,
2176 GFP_ATOMIC);
2177 if (!event)
2178 return -ENOMEM;
2180 sctp_ulpq_tail_event(&asoc->ulpq, event);
2184 return 0;
2187 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2189 * This socket option is applicable to the UDP-style socket only. When
2190 * set it will cause associations that are idle for more than the
2191 * specified number of seconds to automatically close. An association
2192 * being idle is defined an association that has NOT sent or received
2193 * user data. The special value of '0' indicates that no automatic
2194 * close of any associations should be performed. The option expects an
2195 * integer defining the number of seconds of idle time before an
2196 * association is closed.
2198 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2199 unsigned int optlen)
2201 struct sctp_sock *sp = sctp_sk(sk);
2203 /* Applicable to UDP-style socket only */
2204 if (sctp_style(sk, TCP))
2205 return -EOPNOTSUPP;
2206 if (optlen != sizeof(int))
2207 return -EINVAL;
2208 if (copy_from_user(&sp->autoclose, optval, optlen))
2209 return -EFAULT;
2211 return 0;
2214 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2216 * Applications can enable or disable heartbeats for any peer address of
2217 * an association, modify an address's heartbeat interval, force a
2218 * heartbeat to be sent immediately, and adjust the address's maximum
2219 * number of retransmissions sent before an address is considered
2220 * unreachable. The following structure is used to access and modify an
2221 * address's parameters:
2223 * struct sctp_paddrparams {
2224 * sctp_assoc_t spp_assoc_id;
2225 * struct sockaddr_storage spp_address;
2226 * uint32_t spp_hbinterval;
2227 * uint16_t spp_pathmaxrxt;
2228 * uint32_t spp_pathmtu;
2229 * uint32_t spp_sackdelay;
2230 * uint32_t spp_flags;
2231 * };
2233 * spp_assoc_id - (one-to-many style socket) This is filled in the
2234 * application, and identifies the association for
2235 * this query.
2236 * spp_address - This specifies which address is of interest.
2237 * spp_hbinterval - This contains the value of the heartbeat interval,
2238 * in milliseconds. If a value of zero
2239 * is present in this field then no changes are to
2240 * be made to this parameter.
2241 * spp_pathmaxrxt - This contains the maximum number of
2242 * retransmissions before this address shall be
2243 * considered unreachable. If a value of zero
2244 * is present in this field then no changes are to
2245 * be made to this parameter.
2246 * spp_pathmtu - When Path MTU discovery is disabled the value
2247 * specified here will be the "fixed" path mtu.
2248 * Note that if the spp_address field is empty
2249 * then all associations on this address will
2250 * have this fixed path mtu set upon them.
2252 * spp_sackdelay - When delayed sack is enabled, this value specifies
2253 * the number of milliseconds that sacks will be delayed
2254 * for. This value will apply to all addresses of an
2255 * association if the spp_address field is empty. Note
2256 * also, that if delayed sack is enabled and this
2257 * value is set to 0, no change is made to the last
2258 * recorded delayed sack timer value.
2260 * spp_flags - These flags are used to control various features
2261 * on an association. The flag field may contain
2262 * zero or more of the following options.
2264 * SPP_HB_ENABLE - Enable heartbeats on the
2265 * specified address. Note that if the address
2266 * field is empty all addresses for the association
2267 * have heartbeats enabled upon them.
2269 * SPP_HB_DISABLE - Disable heartbeats on the
2270 * speicifed address. Note that if the address
2271 * field is empty all addresses for the association
2272 * will have their heartbeats disabled. Note also
2273 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2274 * mutually exclusive, only one of these two should
2275 * be specified. Enabling both fields will have
2276 * undetermined results.
2278 * SPP_HB_DEMAND - Request a user initiated heartbeat
2279 * to be made immediately.
2281 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2282 * heartbeat delayis to be set to the value of 0
2283 * milliseconds.
2285 * SPP_PMTUD_ENABLE - This field will enable PMTU
2286 * discovery upon the specified address. Note that
2287 * if the address feild is empty then all addresses
2288 * on the association are effected.
2290 * SPP_PMTUD_DISABLE - This field will disable PMTU
2291 * discovery upon the specified address. Note that
2292 * if the address feild is empty then all addresses
2293 * on the association are effected. Not also that
2294 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2295 * exclusive. Enabling both will have undetermined
2296 * results.
2298 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2299 * on delayed sack. The time specified in spp_sackdelay
2300 * is used to specify the sack delay for this address. Note
2301 * that if spp_address is empty then all addresses will
2302 * enable delayed sack and take on the sack delay
2303 * value specified in spp_sackdelay.
2304 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2305 * off delayed sack. If the spp_address field is blank then
2306 * delayed sack is disabled for the entire association. Note
2307 * also that this field is mutually exclusive to
2308 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2309 * results.
2311 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2312 struct sctp_transport *trans,
2313 struct sctp_association *asoc,
2314 struct sctp_sock *sp,
2315 int hb_change,
2316 int pmtud_change,
2317 int sackdelay_change)
2319 int error;
2321 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2322 struct net *net = sock_net(trans->asoc->base.sk);
2324 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2325 if (error)
2326 return error;
2329 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2330 * this field is ignored. Note also that a value of zero indicates
2331 * the current setting should be left unchanged.
2333 if (params->spp_flags & SPP_HB_ENABLE) {
2335 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2336 * set. This lets us use 0 value when this flag
2337 * is set.
2339 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2340 params->spp_hbinterval = 0;
2342 if (params->spp_hbinterval ||
2343 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2344 if (trans) {
2345 trans->hbinterval =
2346 msecs_to_jiffies(params->spp_hbinterval);
2347 } else if (asoc) {
2348 asoc->hbinterval =
2349 msecs_to_jiffies(params->spp_hbinterval);
2350 } else {
2351 sp->hbinterval = params->spp_hbinterval;
2356 if (hb_change) {
2357 if (trans) {
2358 trans->param_flags =
2359 (trans->param_flags & ~SPP_HB) | hb_change;
2360 } else if (asoc) {
2361 asoc->param_flags =
2362 (asoc->param_flags & ~SPP_HB) | hb_change;
2363 } else {
2364 sp->param_flags =
2365 (sp->param_flags & ~SPP_HB) | hb_change;
2369 /* When Path MTU discovery is disabled the value specified here will
2370 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2371 * include the flag SPP_PMTUD_DISABLE for this field to have any
2372 * effect).
2374 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2375 if (trans) {
2376 trans->pathmtu = params->spp_pathmtu;
2377 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2378 } else if (asoc) {
2379 asoc->pathmtu = params->spp_pathmtu;
2380 sctp_frag_point(asoc, params->spp_pathmtu);
2381 } else {
2382 sp->pathmtu = params->spp_pathmtu;
2386 if (pmtud_change) {
2387 if (trans) {
2388 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2389 (params->spp_flags & SPP_PMTUD_ENABLE);
2390 trans->param_flags =
2391 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2392 if (update) {
2393 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2394 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2396 } else if (asoc) {
2397 asoc->param_flags =
2398 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2399 } else {
2400 sp->param_flags =
2401 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2405 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2406 * value of this field is ignored. Note also that a value of zero
2407 * indicates the current setting should be left unchanged.
2409 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2410 if (trans) {
2411 trans->sackdelay =
2412 msecs_to_jiffies(params->spp_sackdelay);
2413 } else if (asoc) {
2414 asoc->sackdelay =
2415 msecs_to_jiffies(params->spp_sackdelay);
2416 } else {
2417 sp->sackdelay = params->spp_sackdelay;
2421 if (sackdelay_change) {
2422 if (trans) {
2423 trans->param_flags =
2424 (trans->param_flags & ~SPP_SACKDELAY) |
2425 sackdelay_change;
2426 } else if (asoc) {
2427 asoc->param_flags =
2428 (asoc->param_flags & ~SPP_SACKDELAY) |
2429 sackdelay_change;
2430 } else {
2431 sp->param_flags =
2432 (sp->param_flags & ~SPP_SACKDELAY) |
2433 sackdelay_change;
2437 /* Note that a value of zero indicates the current setting should be
2438 left unchanged.
2440 if (params->spp_pathmaxrxt) {
2441 if (trans) {
2442 trans->pathmaxrxt = params->spp_pathmaxrxt;
2443 } else if (asoc) {
2444 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2445 } else {
2446 sp->pathmaxrxt = params->spp_pathmaxrxt;
2450 return 0;
2453 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2454 char __user *optval,
2455 unsigned int optlen)
2457 struct sctp_paddrparams params;
2458 struct sctp_transport *trans = NULL;
2459 struct sctp_association *asoc = NULL;
2460 struct sctp_sock *sp = sctp_sk(sk);
2461 int error;
2462 int hb_change, pmtud_change, sackdelay_change;
2464 if (optlen != sizeof(struct sctp_paddrparams))
2465 return - EINVAL;
2467 if (copy_from_user(&params, optval, optlen))
2468 return -EFAULT;
2470 /* Validate flags and value parameters. */
2471 hb_change = params.spp_flags & SPP_HB;
2472 pmtud_change = params.spp_flags & SPP_PMTUD;
2473 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2475 if (hb_change == SPP_HB ||
2476 pmtud_change == SPP_PMTUD ||
2477 sackdelay_change == SPP_SACKDELAY ||
2478 params.spp_sackdelay > 500 ||
2479 (params.spp_pathmtu &&
2480 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2481 return -EINVAL;
2483 /* If an address other than INADDR_ANY is specified, and
2484 * no transport is found, then the request is invalid.
2486 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2487 trans = sctp_addr_id2transport(sk, &params.spp_address,
2488 params.spp_assoc_id);
2489 if (!trans)
2490 return -EINVAL;
2493 /* Get association, if assoc_id != 0 and the socket is a one
2494 * to many style socket, and an association was not found, then
2495 * the id was invalid.
2497 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2498 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2499 return -EINVAL;
2501 /* Heartbeat demand can only be sent on a transport or
2502 * association, but not a socket.
2504 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2505 return -EINVAL;
2507 /* Process parameters. */
2508 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2509 hb_change, pmtud_change,
2510 sackdelay_change);
2512 if (error)
2513 return error;
2515 /* If changes are for association, also apply parameters to each
2516 * transport.
2518 if (!trans && asoc) {
2519 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2520 transports) {
2521 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2522 hb_change, pmtud_change,
2523 sackdelay_change);
2527 return 0;
2531 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2533 * This option will effect the way delayed acks are performed. This
2534 * option allows you to get or set the delayed ack time, in
2535 * milliseconds. It also allows changing the delayed ack frequency.
2536 * Changing the frequency to 1 disables the delayed sack algorithm. If
2537 * the assoc_id is 0, then this sets or gets the endpoints default
2538 * values. If the assoc_id field is non-zero, then the set or get
2539 * effects the specified association for the one to many model (the
2540 * assoc_id field is ignored by the one to one model). Note that if
2541 * sack_delay or sack_freq are 0 when setting this option, then the
2542 * current values will remain unchanged.
2544 * struct sctp_sack_info {
2545 * sctp_assoc_t sack_assoc_id;
2546 * uint32_t sack_delay;
2547 * uint32_t sack_freq;
2548 * };
2550 * sack_assoc_id - This parameter, indicates which association the user
2551 * is performing an action upon. Note that if this field's value is
2552 * zero then the endpoints default value is changed (effecting future
2553 * associations only).
2555 * sack_delay - This parameter contains the number of milliseconds that
2556 * the user is requesting the delayed ACK timer be set to. Note that
2557 * this value is defined in the standard to be between 200 and 500
2558 * milliseconds.
2560 * sack_freq - This parameter contains the number of packets that must
2561 * be received before a sack is sent without waiting for the delay
2562 * timer to expire. The default value for this is 2, setting this
2563 * value to 1 will disable the delayed sack algorithm.
2566 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2567 char __user *optval, unsigned int optlen)
2569 struct sctp_sack_info params;
2570 struct sctp_transport *trans = NULL;
2571 struct sctp_association *asoc = NULL;
2572 struct sctp_sock *sp = sctp_sk(sk);
2574 if (optlen == sizeof(struct sctp_sack_info)) {
2575 if (copy_from_user(&params, optval, optlen))
2576 return -EFAULT;
2578 if (params.sack_delay == 0 && params.sack_freq == 0)
2579 return 0;
2580 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2581 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2582 pr_warn("Use struct sctp_sack_info instead\n");
2583 if (copy_from_user(&params, optval, optlen))
2584 return -EFAULT;
2586 if (params.sack_delay == 0)
2587 params.sack_freq = 1;
2588 else
2589 params.sack_freq = 0;
2590 } else
2591 return - EINVAL;
2593 /* Validate value parameter. */
2594 if (params.sack_delay > 500)
2595 return -EINVAL;
2597 /* Get association, if sack_assoc_id != 0 and the socket is a one
2598 * to many style socket, and an association was not found, then
2599 * the id was invalid.
2601 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2602 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2603 return -EINVAL;
2605 if (params.sack_delay) {
2606 if (asoc) {
2607 asoc->sackdelay =
2608 msecs_to_jiffies(params.sack_delay);
2609 asoc->param_flags =
2610 (asoc->param_flags & ~SPP_SACKDELAY) |
2611 SPP_SACKDELAY_ENABLE;
2612 } else {
2613 sp->sackdelay = params.sack_delay;
2614 sp->param_flags =
2615 (sp->param_flags & ~SPP_SACKDELAY) |
2616 SPP_SACKDELAY_ENABLE;
2620 if (params.sack_freq == 1) {
2621 if (asoc) {
2622 asoc->param_flags =
2623 (asoc->param_flags & ~SPP_SACKDELAY) |
2624 SPP_SACKDELAY_DISABLE;
2625 } else {
2626 sp->param_flags =
2627 (sp->param_flags & ~SPP_SACKDELAY) |
2628 SPP_SACKDELAY_DISABLE;
2630 } else if (params.sack_freq > 1) {
2631 if (asoc) {
2632 asoc->sackfreq = params.sack_freq;
2633 asoc->param_flags =
2634 (asoc->param_flags & ~SPP_SACKDELAY) |
2635 SPP_SACKDELAY_ENABLE;
2636 } else {
2637 sp->sackfreq = params.sack_freq;
2638 sp->param_flags =
2639 (sp->param_flags & ~SPP_SACKDELAY) |
2640 SPP_SACKDELAY_ENABLE;
2644 /* If change is for association, also apply to each transport. */
2645 if (asoc) {
2646 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2647 transports) {
2648 if (params.sack_delay) {
2649 trans->sackdelay =
2650 msecs_to_jiffies(params.sack_delay);
2651 trans->param_flags =
2652 (trans->param_flags & ~SPP_SACKDELAY) |
2653 SPP_SACKDELAY_ENABLE;
2655 if (params.sack_freq == 1) {
2656 trans->param_flags =
2657 (trans->param_flags & ~SPP_SACKDELAY) |
2658 SPP_SACKDELAY_DISABLE;
2659 } else if (params.sack_freq > 1) {
2660 trans->sackfreq = params.sack_freq;
2661 trans->param_flags =
2662 (trans->param_flags & ~SPP_SACKDELAY) |
2663 SPP_SACKDELAY_ENABLE;
2668 return 0;
2671 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2673 * Applications can specify protocol parameters for the default association
2674 * initialization. The option name argument to setsockopt() and getsockopt()
2675 * is SCTP_INITMSG.
2677 * Setting initialization parameters is effective only on an unconnected
2678 * socket (for UDP-style sockets only future associations are effected
2679 * by the change). With TCP-style sockets, this option is inherited by
2680 * sockets derived from a listener socket.
2682 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2684 struct sctp_initmsg sinit;
2685 struct sctp_sock *sp = sctp_sk(sk);
2687 if (optlen != sizeof(struct sctp_initmsg))
2688 return -EINVAL;
2689 if (copy_from_user(&sinit, optval, optlen))
2690 return -EFAULT;
2692 if (sinit.sinit_num_ostreams)
2693 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2694 if (sinit.sinit_max_instreams)
2695 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2696 if (sinit.sinit_max_attempts)
2697 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2698 if (sinit.sinit_max_init_timeo)
2699 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2701 return 0;
2705 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2707 * Applications that wish to use the sendto() system call may wish to
2708 * specify a default set of parameters that would normally be supplied
2709 * through the inclusion of ancillary data. This socket option allows
2710 * such an application to set the default sctp_sndrcvinfo structure.
2711 * The application that wishes to use this socket option simply passes
2712 * in to this call the sctp_sndrcvinfo structure defined in Section
2713 * 5.2.2) The input parameters accepted by this call include
2714 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2715 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2716 * to this call if the caller is using the UDP model.
2718 static int sctp_setsockopt_default_send_param(struct sock *sk,
2719 char __user *optval,
2720 unsigned int optlen)
2722 struct sctp_sndrcvinfo info;
2723 struct sctp_association *asoc;
2724 struct sctp_sock *sp = sctp_sk(sk);
2726 if (optlen != sizeof(struct sctp_sndrcvinfo))
2727 return -EINVAL;
2728 if (copy_from_user(&info, optval, optlen))
2729 return -EFAULT;
2731 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2732 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2733 return -EINVAL;
2735 if (asoc) {
2736 asoc->default_stream = info.sinfo_stream;
2737 asoc->default_flags = info.sinfo_flags;
2738 asoc->default_ppid = info.sinfo_ppid;
2739 asoc->default_context = info.sinfo_context;
2740 asoc->default_timetolive = info.sinfo_timetolive;
2741 } else {
2742 sp->default_stream = info.sinfo_stream;
2743 sp->default_flags = info.sinfo_flags;
2744 sp->default_ppid = info.sinfo_ppid;
2745 sp->default_context = info.sinfo_context;
2746 sp->default_timetolive = info.sinfo_timetolive;
2749 return 0;
2752 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2754 * Requests that the local SCTP stack use the enclosed peer address as
2755 * the association primary. The enclosed address must be one of the
2756 * association peer's addresses.
2758 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2759 unsigned int optlen)
2761 struct sctp_prim prim;
2762 struct sctp_transport *trans;
2764 if (optlen != sizeof(struct sctp_prim))
2765 return -EINVAL;
2767 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2768 return -EFAULT;
2770 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2771 if (!trans)
2772 return -EINVAL;
2774 sctp_assoc_set_primary(trans->asoc, trans);
2776 return 0;
2780 * 7.1.5 SCTP_NODELAY
2782 * Turn on/off any Nagle-like algorithm. This means that packets are
2783 * generally sent as soon as possible and no unnecessary delays are
2784 * introduced, at the cost of more packets in the network. Expects an
2785 * integer boolean flag.
2787 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2788 unsigned int optlen)
2790 int val;
2792 if (optlen < sizeof(int))
2793 return -EINVAL;
2794 if (get_user(val, (int __user *)optval))
2795 return -EFAULT;
2797 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2798 return 0;
2803 * 7.1.1 SCTP_RTOINFO
2805 * The protocol parameters used to initialize and bound retransmission
2806 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2807 * and modify these parameters.
2808 * All parameters are time values, in milliseconds. A value of 0, when
2809 * modifying the parameters, indicates that the current value should not
2810 * be changed.
2813 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2815 struct sctp_rtoinfo rtoinfo;
2816 struct sctp_association *asoc;
2818 if (optlen != sizeof (struct sctp_rtoinfo))
2819 return -EINVAL;
2821 if (copy_from_user(&rtoinfo, optval, optlen))
2822 return -EFAULT;
2824 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2826 /* Set the values to the specific association */
2827 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2828 return -EINVAL;
2830 if (asoc) {
2831 if (rtoinfo.srto_initial != 0)
2832 asoc->rto_initial =
2833 msecs_to_jiffies(rtoinfo.srto_initial);
2834 if (rtoinfo.srto_max != 0)
2835 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2836 if (rtoinfo.srto_min != 0)
2837 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2838 } else {
2839 /* If there is no association or the association-id = 0
2840 * set the values to the endpoint.
2842 struct sctp_sock *sp = sctp_sk(sk);
2844 if (rtoinfo.srto_initial != 0)
2845 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2846 if (rtoinfo.srto_max != 0)
2847 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2848 if (rtoinfo.srto_min != 0)
2849 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2852 return 0;
2857 * 7.1.2 SCTP_ASSOCINFO
2859 * This option is used to tune the maximum retransmission attempts
2860 * of the association.
2861 * Returns an error if the new association retransmission value is
2862 * greater than the sum of the retransmission value of the peer.
2863 * See [SCTP] for more information.
2866 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2869 struct sctp_assocparams assocparams;
2870 struct sctp_association *asoc;
2872 if (optlen != sizeof(struct sctp_assocparams))
2873 return -EINVAL;
2874 if (copy_from_user(&assocparams, optval, optlen))
2875 return -EFAULT;
2877 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2879 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2880 return -EINVAL;
2882 /* Set the values to the specific association */
2883 if (asoc) {
2884 if (assocparams.sasoc_asocmaxrxt != 0) {
2885 __u32 path_sum = 0;
2886 int paths = 0;
2887 struct sctp_transport *peer_addr;
2889 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2890 transports) {
2891 path_sum += peer_addr->pathmaxrxt;
2892 paths++;
2895 /* Only validate asocmaxrxt if we have more than
2896 * one path/transport. We do this because path
2897 * retransmissions are only counted when we have more
2898 * then one path.
2900 if (paths > 1 &&
2901 assocparams.sasoc_asocmaxrxt > path_sum)
2902 return -EINVAL;
2904 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2907 if (assocparams.sasoc_cookie_life != 0)
2908 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2909 } else {
2910 /* Set the values to the endpoint */
2911 struct sctp_sock *sp = sctp_sk(sk);
2913 if (assocparams.sasoc_asocmaxrxt != 0)
2914 sp->assocparams.sasoc_asocmaxrxt =
2915 assocparams.sasoc_asocmaxrxt;
2916 if (assocparams.sasoc_cookie_life != 0)
2917 sp->assocparams.sasoc_cookie_life =
2918 assocparams.sasoc_cookie_life;
2920 return 0;
2924 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2926 * This socket option is a boolean flag which turns on or off mapped V4
2927 * addresses. If this option is turned on and the socket is type
2928 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2929 * If this option is turned off, then no mapping will be done of V4
2930 * addresses and a user will receive both PF_INET6 and PF_INET type
2931 * addresses on the socket.
2933 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2935 int val;
2936 struct sctp_sock *sp = sctp_sk(sk);
2938 if (optlen < sizeof(int))
2939 return -EINVAL;
2940 if (get_user(val, (int __user *)optval))
2941 return -EFAULT;
2942 if (val)
2943 sp->v4mapped = 1;
2944 else
2945 sp->v4mapped = 0;
2947 return 0;
2951 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2952 * This option will get or set the maximum size to put in any outgoing
2953 * SCTP DATA chunk. If a message is larger than this size it will be
2954 * fragmented by SCTP into the specified size. Note that the underlying
2955 * SCTP implementation may fragment into smaller sized chunks when the
2956 * PMTU of the underlying association is smaller than the value set by
2957 * the user. The default value for this option is '0' which indicates
2958 * the user is NOT limiting fragmentation and only the PMTU will effect
2959 * SCTP's choice of DATA chunk size. Note also that values set larger
2960 * than the maximum size of an IP datagram will effectively let SCTP
2961 * control fragmentation (i.e. the same as setting this option to 0).
2963 * The following structure is used to access and modify this parameter:
2965 * struct sctp_assoc_value {
2966 * sctp_assoc_t assoc_id;
2967 * uint32_t assoc_value;
2968 * };
2970 * assoc_id: This parameter is ignored for one-to-one style sockets.
2971 * For one-to-many style sockets this parameter indicates which
2972 * association the user is performing an action upon. Note that if
2973 * this field's value is zero then the endpoints default value is
2974 * changed (effecting future associations only).
2975 * assoc_value: This parameter specifies the maximum size in bytes.
2977 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2979 struct sctp_assoc_value params;
2980 struct sctp_association *asoc;
2981 struct sctp_sock *sp = sctp_sk(sk);
2982 int val;
2984 if (optlen == sizeof(int)) {
2985 pr_warn("Use of int in maxseg socket option deprecated\n");
2986 pr_warn("Use struct sctp_assoc_value instead\n");
2987 if (copy_from_user(&val, optval, optlen))
2988 return -EFAULT;
2989 params.assoc_id = 0;
2990 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2991 if (copy_from_user(&params, optval, optlen))
2992 return -EFAULT;
2993 val = params.assoc_value;
2994 } else
2995 return -EINVAL;
2997 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2998 return -EINVAL;
3000 asoc = sctp_id2assoc(sk, params.assoc_id);
3001 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3002 return -EINVAL;
3004 if (asoc) {
3005 if (val == 0) {
3006 val = asoc->pathmtu;
3007 val -= sp->pf->af->net_header_len;
3008 val -= sizeof(struct sctphdr) +
3009 sizeof(struct sctp_data_chunk);
3011 asoc->user_frag = val;
3012 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3013 } else {
3014 sp->user_frag = val;
3017 return 0;
3022 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3024 * Requests that the peer mark the enclosed address as the association
3025 * primary. The enclosed address must be one of the association's
3026 * locally bound addresses. The following structure is used to make a
3027 * set primary request:
3029 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3030 unsigned int optlen)
3032 struct net *net = sock_net(sk);
3033 struct sctp_sock *sp;
3034 struct sctp_association *asoc = NULL;
3035 struct sctp_setpeerprim prim;
3036 struct sctp_chunk *chunk;
3037 struct sctp_af *af;
3038 int err;
3040 sp = sctp_sk(sk);
3042 if (!net->sctp.addip_enable)
3043 return -EPERM;
3045 if (optlen != sizeof(struct sctp_setpeerprim))
3046 return -EINVAL;
3048 if (copy_from_user(&prim, optval, optlen))
3049 return -EFAULT;
3051 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3052 if (!asoc)
3053 return -EINVAL;
3055 if (!asoc->peer.asconf_capable)
3056 return -EPERM;
3058 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3059 return -EPERM;
3061 if (!sctp_state(asoc, ESTABLISHED))
3062 return -ENOTCONN;
3064 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3065 if (!af)
3066 return -EINVAL;
3068 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3069 return -EADDRNOTAVAIL;
3071 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3072 return -EADDRNOTAVAIL;
3074 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3075 chunk = sctp_make_asconf_set_prim(asoc,
3076 (union sctp_addr *)&prim.sspp_addr);
3077 if (!chunk)
3078 return -ENOMEM;
3080 err = sctp_send_asconf(asoc, chunk);
3082 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3084 return err;
3087 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3088 unsigned int optlen)
3090 struct sctp_setadaptation adaptation;
3092 if (optlen != sizeof(struct sctp_setadaptation))
3093 return -EINVAL;
3094 if (copy_from_user(&adaptation, optval, optlen))
3095 return -EFAULT;
3097 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3099 return 0;
3103 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3105 * The context field in the sctp_sndrcvinfo structure is normally only
3106 * used when a failed message is retrieved holding the value that was
3107 * sent down on the actual send call. This option allows the setting of
3108 * a default context on an association basis that will be received on
3109 * reading messages from the peer. This is especially helpful in the
3110 * one-2-many model for an application to keep some reference to an
3111 * internal state machine that is processing messages on the
3112 * association. Note that the setting of this value only effects
3113 * received messages from the peer and does not effect the value that is
3114 * saved with outbound messages.
3116 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3117 unsigned int optlen)
3119 struct sctp_assoc_value params;
3120 struct sctp_sock *sp;
3121 struct sctp_association *asoc;
3123 if (optlen != sizeof(struct sctp_assoc_value))
3124 return -EINVAL;
3125 if (copy_from_user(&params, optval, optlen))
3126 return -EFAULT;
3128 sp = sctp_sk(sk);
3130 if (params.assoc_id != 0) {
3131 asoc = sctp_id2assoc(sk, params.assoc_id);
3132 if (!asoc)
3133 return -EINVAL;
3134 asoc->default_rcv_context = params.assoc_value;
3135 } else {
3136 sp->default_rcv_context = params.assoc_value;
3139 return 0;
3143 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3145 * This options will at a minimum specify if the implementation is doing
3146 * fragmented interleave. Fragmented interleave, for a one to many
3147 * socket, is when subsequent calls to receive a message may return
3148 * parts of messages from different associations. Some implementations
3149 * may allow you to turn this value on or off. If so, when turned off,
3150 * no fragment interleave will occur (which will cause a head of line
3151 * blocking amongst multiple associations sharing the same one to many
3152 * socket). When this option is turned on, then each receive call may
3153 * come from a different association (thus the user must receive data
3154 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3155 * association each receive belongs to.
3157 * This option takes a boolean value. A non-zero value indicates that
3158 * fragmented interleave is on. A value of zero indicates that
3159 * fragmented interleave is off.
3161 * Note that it is important that an implementation that allows this
3162 * option to be turned on, have it off by default. Otherwise an unaware
3163 * application using the one to many model may become confused and act
3164 * incorrectly.
3166 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3167 char __user *optval,
3168 unsigned int optlen)
3170 int val;
3172 if (optlen != sizeof(int))
3173 return -EINVAL;
3174 if (get_user(val, (int __user *)optval))
3175 return -EFAULT;
3177 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3179 return 0;
3183 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3184 * (SCTP_PARTIAL_DELIVERY_POINT)
3186 * This option will set or get the SCTP partial delivery point. This
3187 * point is the size of a message where the partial delivery API will be
3188 * invoked to help free up rwnd space for the peer. Setting this to a
3189 * lower value will cause partial deliveries to happen more often. The
3190 * calls argument is an integer that sets or gets the partial delivery
3191 * point. Note also that the call will fail if the user attempts to set
3192 * this value larger than the socket receive buffer size.
3194 * Note that any single message having a length smaller than or equal to
3195 * the SCTP partial delivery point will be delivered in one single read
3196 * call as long as the user provided buffer is large enough to hold the
3197 * message.
3199 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3200 char __user *optval,
3201 unsigned int optlen)
3203 u32 val;
3205 if (optlen != sizeof(u32))
3206 return -EINVAL;
3207 if (get_user(val, (int __user *)optval))
3208 return -EFAULT;
3210 /* Note: We double the receive buffer from what the user sets
3211 * it to be, also initial rwnd is based on rcvbuf/2.
3213 if (val > (sk->sk_rcvbuf >> 1))
3214 return -EINVAL;
3216 sctp_sk(sk)->pd_point = val;
3218 return 0; /* is this the right error code? */
3222 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3224 * This option will allow a user to change the maximum burst of packets
3225 * that can be emitted by this association. Note that the default value
3226 * is 4, and some implementations may restrict this setting so that it
3227 * can only be lowered.
3229 * NOTE: This text doesn't seem right. Do this on a socket basis with
3230 * future associations inheriting the socket value.
3232 static int sctp_setsockopt_maxburst(struct sock *sk,
3233 char __user *optval,
3234 unsigned int optlen)
3236 struct sctp_assoc_value params;
3237 struct sctp_sock *sp;
3238 struct sctp_association *asoc;
3239 int val;
3240 int assoc_id = 0;
3242 if (optlen == sizeof(int)) {
3243 pr_warn("Use of int in max_burst socket option deprecated\n");
3244 pr_warn("Use struct sctp_assoc_value instead\n");
3245 if (copy_from_user(&val, optval, optlen))
3246 return -EFAULT;
3247 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3248 if (copy_from_user(&params, optval, optlen))
3249 return -EFAULT;
3250 val = params.assoc_value;
3251 assoc_id = params.assoc_id;
3252 } else
3253 return -EINVAL;
3255 sp = sctp_sk(sk);
3257 if (assoc_id != 0) {
3258 asoc = sctp_id2assoc(sk, assoc_id);
3259 if (!asoc)
3260 return -EINVAL;
3261 asoc->max_burst = val;
3262 } else
3263 sp->max_burst = val;
3265 return 0;
3269 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3271 * This set option adds a chunk type that the user is requesting to be
3272 * received only in an authenticated way. Changes to the list of chunks
3273 * will only effect future associations on the socket.
3275 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3276 char __user *optval,
3277 unsigned int optlen)
3279 struct net *net = sock_net(sk);
3280 struct sctp_authchunk val;
3282 if (!net->sctp.auth_enable)
3283 return -EACCES;
3285 if (optlen != sizeof(struct sctp_authchunk))
3286 return -EINVAL;
3287 if (copy_from_user(&val, optval, optlen))
3288 return -EFAULT;
3290 switch (val.sauth_chunk) {
3291 case SCTP_CID_INIT:
3292 case SCTP_CID_INIT_ACK:
3293 case SCTP_CID_SHUTDOWN_COMPLETE:
3294 case SCTP_CID_AUTH:
3295 return -EINVAL;
3298 /* add this chunk id to the endpoint */
3299 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3303 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3305 * This option gets or sets the list of HMAC algorithms that the local
3306 * endpoint requires the peer to use.
3308 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3309 char __user *optval,
3310 unsigned int optlen)
3312 struct net *net = sock_net(sk);
3313 struct sctp_hmacalgo *hmacs;
3314 u32 idents;
3315 int err;
3317 if (!net->sctp.auth_enable)
3318 return -EACCES;
3320 if (optlen < sizeof(struct sctp_hmacalgo))
3321 return -EINVAL;
3323 hmacs= memdup_user(optval, optlen);
3324 if (IS_ERR(hmacs))
3325 return PTR_ERR(hmacs);
3327 idents = hmacs->shmac_num_idents;
3328 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3329 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3330 err = -EINVAL;
3331 goto out;
3334 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3335 out:
3336 kfree(hmacs);
3337 return err;
3341 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3343 * This option will set a shared secret key which is used to build an
3344 * association shared key.
3346 static int sctp_setsockopt_auth_key(struct sock *sk,
3347 char __user *optval,
3348 unsigned int optlen)
3350 struct net *net = sock_net(sk);
3351 struct sctp_authkey *authkey;
3352 struct sctp_association *asoc;
3353 int ret;
3355 if (!net->sctp.auth_enable)
3356 return -EACCES;
3358 if (optlen <= sizeof(struct sctp_authkey))
3359 return -EINVAL;
3361 authkey= memdup_user(optval, optlen);
3362 if (IS_ERR(authkey))
3363 return PTR_ERR(authkey);
3365 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3366 ret = -EINVAL;
3367 goto out;
3370 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3371 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3372 ret = -EINVAL;
3373 goto out;
3376 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3377 out:
3378 kzfree(authkey);
3379 return ret;
3383 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3385 * This option will get or set the active shared key to be used to build
3386 * the association shared key.
3388 static int sctp_setsockopt_active_key(struct sock *sk,
3389 char __user *optval,
3390 unsigned int optlen)
3392 struct net *net = sock_net(sk);
3393 struct sctp_authkeyid val;
3394 struct sctp_association *asoc;
3396 if (!net->sctp.auth_enable)
3397 return -EACCES;
3399 if (optlen != sizeof(struct sctp_authkeyid))
3400 return -EINVAL;
3401 if (copy_from_user(&val, optval, optlen))
3402 return -EFAULT;
3404 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3405 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3406 return -EINVAL;
3408 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3409 val.scact_keynumber);
3413 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3415 * This set option will delete a shared secret key from use.
3417 static int sctp_setsockopt_del_key(struct sock *sk,
3418 char __user *optval,
3419 unsigned int optlen)
3421 struct net *net = sock_net(sk);
3422 struct sctp_authkeyid val;
3423 struct sctp_association *asoc;
3425 if (!net->sctp.auth_enable)
3426 return -EACCES;
3428 if (optlen != sizeof(struct sctp_authkeyid))
3429 return -EINVAL;
3430 if (copy_from_user(&val, optval, optlen))
3431 return -EFAULT;
3433 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3434 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3435 return -EINVAL;
3437 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3438 val.scact_keynumber);
3443 * 8.1.23 SCTP_AUTO_ASCONF
3445 * This option will enable or disable the use of the automatic generation of
3446 * ASCONF chunks to add and delete addresses to an existing association. Note
3447 * that this option has two caveats namely: a) it only affects sockets that
3448 * are bound to all addresses available to the SCTP stack, and b) the system
3449 * administrator may have an overriding control that turns the ASCONF feature
3450 * off no matter what setting the socket option may have.
3451 * This option expects an integer boolean flag, where a non-zero value turns on
3452 * the option, and a zero value turns off the option.
3453 * Note. In this implementation, socket operation overrides default parameter
3454 * being set by sysctl as well as FreeBSD implementation
3456 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3457 unsigned int optlen)
3459 int val;
3460 struct sctp_sock *sp = sctp_sk(sk);
3462 if (optlen < sizeof(int))
3463 return -EINVAL;
3464 if (get_user(val, (int __user *)optval))
3465 return -EFAULT;
3466 if (!sctp_is_ep_boundall(sk) && val)
3467 return -EINVAL;
3468 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3469 return 0;
3471 if (val == 0 && sp->do_auto_asconf) {
3472 list_del(&sp->auto_asconf_list);
3473 sp->do_auto_asconf = 0;
3474 } else if (val && !sp->do_auto_asconf) {
3475 list_add_tail(&sp->auto_asconf_list,
3476 &sock_net(sk)->sctp.auto_asconf_splist);
3477 sp->do_auto_asconf = 1;
3479 return 0;
3484 * SCTP_PEER_ADDR_THLDS
3486 * This option allows us to alter the partially failed threshold for one or all
3487 * transports in an association. See Section 6.1 of:
3488 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3490 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3491 char __user *optval,
3492 unsigned int optlen)
3494 struct sctp_paddrthlds val;
3495 struct sctp_transport *trans;
3496 struct sctp_association *asoc;
3498 if (optlen < sizeof(struct sctp_paddrthlds))
3499 return -EINVAL;
3500 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3501 sizeof(struct sctp_paddrthlds)))
3502 return -EFAULT;
3505 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3506 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3507 if (!asoc)
3508 return -ENOENT;
3509 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3510 transports) {
3511 if (val.spt_pathmaxrxt)
3512 trans->pathmaxrxt = val.spt_pathmaxrxt;
3513 trans->pf_retrans = val.spt_pathpfthld;
3516 if (val.spt_pathmaxrxt)
3517 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3518 asoc->pf_retrans = val.spt_pathpfthld;
3519 } else {
3520 trans = sctp_addr_id2transport(sk, &val.spt_address,
3521 val.spt_assoc_id);
3522 if (!trans)
3523 return -ENOENT;
3525 if (val.spt_pathmaxrxt)
3526 trans->pathmaxrxt = val.spt_pathmaxrxt;
3527 trans->pf_retrans = val.spt_pathpfthld;
3530 return 0;
3533 /* API 6.2 setsockopt(), getsockopt()
3535 * Applications use setsockopt() and getsockopt() to set or retrieve
3536 * socket options. Socket options are used to change the default
3537 * behavior of sockets calls. They are described in Section 7.
3539 * The syntax is:
3541 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3542 * int __user *optlen);
3543 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3544 * int optlen);
3546 * sd - the socket descript.
3547 * level - set to IPPROTO_SCTP for all SCTP options.
3548 * optname - the option name.
3549 * optval - the buffer to store the value of the option.
3550 * optlen - the size of the buffer.
3552 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3553 char __user *optval, unsigned int optlen)
3555 int retval = 0;
3557 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3559 /* I can hardly begin to describe how wrong this is. This is
3560 * so broken as to be worse than useless. The API draft
3561 * REALLY is NOT helpful here... I am not convinced that the
3562 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3563 * are at all well-founded.
3565 if (level != SOL_SCTP) {
3566 struct sctp_af *af = sctp_sk(sk)->pf->af;
3567 retval = af->setsockopt(sk, level, optname, optval, optlen);
3568 goto out_nounlock;
3571 sctp_lock_sock(sk);
3573 switch (optname) {
3574 case SCTP_SOCKOPT_BINDX_ADD:
3575 /* 'optlen' is the size of the addresses buffer. */
3576 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3577 optlen, SCTP_BINDX_ADD_ADDR);
3578 break;
3580 case SCTP_SOCKOPT_BINDX_REM:
3581 /* 'optlen' is the size of the addresses buffer. */
3582 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3583 optlen, SCTP_BINDX_REM_ADDR);
3584 break;
3586 case SCTP_SOCKOPT_CONNECTX_OLD:
3587 /* 'optlen' is the size of the addresses buffer. */
3588 retval = sctp_setsockopt_connectx_old(sk,
3589 (struct sockaddr __user *)optval,
3590 optlen);
3591 break;
3593 case SCTP_SOCKOPT_CONNECTX:
3594 /* 'optlen' is the size of the addresses buffer. */
3595 retval = sctp_setsockopt_connectx(sk,
3596 (struct sockaddr __user *)optval,
3597 optlen);
3598 break;
3600 case SCTP_DISABLE_FRAGMENTS:
3601 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3602 break;
3604 case SCTP_EVENTS:
3605 retval = sctp_setsockopt_events(sk, optval, optlen);
3606 break;
3608 case SCTP_AUTOCLOSE:
3609 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3610 break;
3612 case SCTP_PEER_ADDR_PARAMS:
3613 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3614 break;
3616 case SCTP_DELAYED_SACK:
3617 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3618 break;
3619 case SCTP_PARTIAL_DELIVERY_POINT:
3620 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3621 break;
3623 case SCTP_INITMSG:
3624 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3625 break;
3626 case SCTP_DEFAULT_SEND_PARAM:
3627 retval = sctp_setsockopt_default_send_param(sk, optval,
3628 optlen);
3629 break;
3630 case SCTP_PRIMARY_ADDR:
3631 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3632 break;
3633 case SCTP_SET_PEER_PRIMARY_ADDR:
3634 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3635 break;
3636 case SCTP_NODELAY:
3637 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3638 break;
3639 case SCTP_RTOINFO:
3640 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3641 break;
3642 case SCTP_ASSOCINFO:
3643 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3644 break;
3645 case SCTP_I_WANT_MAPPED_V4_ADDR:
3646 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3647 break;
3648 case SCTP_MAXSEG:
3649 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3650 break;
3651 case SCTP_ADAPTATION_LAYER:
3652 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3653 break;
3654 case SCTP_CONTEXT:
3655 retval = sctp_setsockopt_context(sk, optval, optlen);
3656 break;
3657 case SCTP_FRAGMENT_INTERLEAVE:
3658 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3659 break;
3660 case SCTP_MAX_BURST:
3661 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3662 break;
3663 case SCTP_AUTH_CHUNK:
3664 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3665 break;
3666 case SCTP_HMAC_IDENT:
3667 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3668 break;
3669 case SCTP_AUTH_KEY:
3670 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3671 break;
3672 case SCTP_AUTH_ACTIVE_KEY:
3673 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3674 break;
3675 case SCTP_AUTH_DELETE_KEY:
3676 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3677 break;
3678 case SCTP_AUTO_ASCONF:
3679 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3680 break;
3681 case SCTP_PEER_ADDR_THLDS:
3682 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3683 break;
3684 default:
3685 retval = -ENOPROTOOPT;
3686 break;
3689 sctp_release_sock(sk);
3691 out_nounlock:
3692 return retval;
3695 /* API 3.1.6 connect() - UDP Style Syntax
3697 * An application may use the connect() call in the UDP model to initiate an
3698 * association without sending data.
3700 * The syntax is:
3702 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3704 * sd: the socket descriptor to have a new association added to.
3706 * nam: the address structure (either struct sockaddr_in or struct
3707 * sockaddr_in6 defined in RFC2553 [7]).
3709 * len: the size of the address.
3711 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3712 int addr_len)
3714 int err = 0;
3715 struct sctp_af *af;
3717 sctp_lock_sock(sk);
3719 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3720 addr, addr_len);
3722 /* Validate addr_len before calling common connect/connectx routine. */
3723 af = sctp_get_af_specific(addr->sa_family);
3724 if (!af || addr_len < af->sockaddr_len) {
3725 err = -EINVAL;
3726 } else {
3727 /* Pass correct addr len to common routine (so it knows there
3728 * is only one address being passed.
3730 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3733 sctp_release_sock(sk);
3734 return err;
3737 /* FIXME: Write comments. */
3738 static int sctp_disconnect(struct sock *sk, int flags)
3740 return -EOPNOTSUPP; /* STUB */
3743 /* 4.1.4 accept() - TCP Style Syntax
3745 * Applications use accept() call to remove an established SCTP
3746 * association from the accept queue of the endpoint. A new socket
3747 * descriptor will be returned from accept() to represent the newly
3748 * formed association.
3750 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3752 struct sctp_sock *sp;
3753 struct sctp_endpoint *ep;
3754 struct sock *newsk = NULL;
3755 struct sctp_association *asoc;
3756 long timeo;
3757 int error = 0;
3759 sctp_lock_sock(sk);
3761 sp = sctp_sk(sk);
3762 ep = sp->ep;
3764 if (!sctp_style(sk, TCP)) {
3765 error = -EOPNOTSUPP;
3766 goto out;
3769 if (!sctp_sstate(sk, LISTENING)) {
3770 error = -EINVAL;
3771 goto out;
3774 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3776 error = sctp_wait_for_accept(sk, timeo);
3777 if (error)
3778 goto out;
3780 /* We treat the list of associations on the endpoint as the accept
3781 * queue and pick the first association on the list.
3783 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3785 newsk = sp->pf->create_accept_sk(sk, asoc);
3786 if (!newsk) {
3787 error = -ENOMEM;
3788 goto out;
3791 /* Populate the fields of the newsk from the oldsk and migrate the
3792 * asoc to the newsk.
3794 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3796 out:
3797 sctp_release_sock(sk);
3798 *err = error;
3799 return newsk;
3802 /* The SCTP ioctl handler. */
3803 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3805 int rc = -ENOTCONN;
3807 sctp_lock_sock(sk);
3810 * SEQPACKET-style sockets in LISTENING state are valid, for
3811 * SCTP, so only discard TCP-style sockets in LISTENING state.
3813 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3814 goto out;
3816 switch (cmd) {
3817 case SIOCINQ: {
3818 struct sk_buff *skb;
3819 unsigned int amount = 0;
3821 skb = skb_peek(&sk->sk_receive_queue);
3822 if (skb != NULL) {
3824 * We will only return the amount of this packet since
3825 * that is all that will be read.
3827 amount = skb->len;
3829 rc = put_user(amount, (int __user *)arg);
3830 break;
3832 default:
3833 rc = -ENOIOCTLCMD;
3834 break;
3836 out:
3837 sctp_release_sock(sk);
3838 return rc;
3841 /* This is the function which gets called during socket creation to
3842 * initialized the SCTP-specific portion of the sock.
3843 * The sock structure should already be zero-filled memory.
3845 static int sctp_init_sock(struct sock *sk)
3847 struct net *net = sock_net(sk);
3848 struct sctp_sock *sp;
3850 pr_debug("%s: sk:%p\n", __func__, sk);
3852 sp = sctp_sk(sk);
3854 /* Initialize the SCTP per socket area. */
3855 switch (sk->sk_type) {
3856 case SOCK_SEQPACKET:
3857 sp->type = SCTP_SOCKET_UDP;
3858 break;
3859 case SOCK_STREAM:
3860 sp->type = SCTP_SOCKET_TCP;
3861 break;
3862 default:
3863 return -ESOCKTNOSUPPORT;
3866 /* Initialize default send parameters. These parameters can be
3867 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3869 sp->default_stream = 0;
3870 sp->default_ppid = 0;
3871 sp->default_flags = 0;
3872 sp->default_context = 0;
3873 sp->default_timetolive = 0;
3875 sp->default_rcv_context = 0;
3876 sp->max_burst = net->sctp.max_burst;
3878 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3880 /* Initialize default setup parameters. These parameters
3881 * can be modified with the SCTP_INITMSG socket option or
3882 * overridden by the SCTP_INIT CMSG.
3884 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3885 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3886 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3887 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3889 /* Initialize default RTO related parameters. These parameters can
3890 * be modified for with the SCTP_RTOINFO socket option.
3892 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3893 sp->rtoinfo.srto_max = net->sctp.rto_max;
3894 sp->rtoinfo.srto_min = net->sctp.rto_min;
3896 /* Initialize default association related parameters. These parameters
3897 * can be modified with the SCTP_ASSOCINFO socket option.
3899 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3900 sp->assocparams.sasoc_number_peer_destinations = 0;
3901 sp->assocparams.sasoc_peer_rwnd = 0;
3902 sp->assocparams.sasoc_local_rwnd = 0;
3903 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3905 /* Initialize default event subscriptions. By default, all the
3906 * options are off.
3908 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3910 /* Default Peer Address Parameters. These defaults can
3911 * be modified via SCTP_PEER_ADDR_PARAMS
3913 sp->hbinterval = net->sctp.hb_interval;
3914 sp->pathmaxrxt = net->sctp.max_retrans_path;
3915 sp->pathmtu = 0; // allow default discovery
3916 sp->sackdelay = net->sctp.sack_timeout;
3917 sp->sackfreq = 2;
3918 sp->param_flags = SPP_HB_ENABLE |
3919 SPP_PMTUD_ENABLE |
3920 SPP_SACKDELAY_ENABLE;
3922 /* If enabled no SCTP message fragmentation will be performed.
3923 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3925 sp->disable_fragments = 0;
3927 /* Enable Nagle algorithm by default. */
3928 sp->nodelay = 0;
3930 /* Enable by default. */
3931 sp->v4mapped = 1;
3933 /* Auto-close idle associations after the configured
3934 * number of seconds. A value of 0 disables this
3935 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3936 * for UDP-style sockets only.
3938 sp->autoclose = 0;
3940 /* User specified fragmentation limit. */
3941 sp->user_frag = 0;
3943 sp->adaptation_ind = 0;
3945 sp->pf = sctp_get_pf_specific(sk->sk_family);
3947 /* Control variables for partial data delivery. */
3948 atomic_set(&sp->pd_mode, 0);
3949 skb_queue_head_init(&sp->pd_lobby);
3950 sp->frag_interleave = 0;
3952 /* Create a per socket endpoint structure. Even if we
3953 * change the data structure relationships, this may still
3954 * be useful for storing pre-connect address information.
3956 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
3957 if (!sp->ep)
3958 return -ENOMEM;
3960 sp->hmac = NULL;
3962 sk->sk_destruct = sctp_destruct_sock;
3964 SCTP_DBG_OBJCNT_INC(sock);
3966 local_bh_disable();
3967 percpu_counter_inc(&sctp_sockets_allocated);
3968 sock_prot_inuse_add(net, sk->sk_prot, 1);
3969 if (net->sctp.default_auto_asconf) {
3970 list_add_tail(&sp->auto_asconf_list,
3971 &net->sctp.auto_asconf_splist);
3972 sp->do_auto_asconf = 1;
3973 } else
3974 sp->do_auto_asconf = 0;
3975 local_bh_enable();
3977 return 0;
3980 /* Cleanup any SCTP per socket resources. */
3981 static void sctp_destroy_sock(struct sock *sk)
3983 struct sctp_sock *sp;
3985 pr_debug("%s: sk:%p\n", __func__, sk);
3987 /* Release our hold on the endpoint. */
3988 sp = sctp_sk(sk);
3989 /* This could happen during socket init, thus we bail out
3990 * early, since the rest of the below is not setup either.
3992 if (sp->ep == NULL)
3993 return;
3995 if (sp->do_auto_asconf) {
3996 sp->do_auto_asconf = 0;
3997 list_del(&sp->auto_asconf_list);
3999 sctp_endpoint_free(sp->ep);
4000 local_bh_disable();
4001 percpu_counter_dec(&sctp_sockets_allocated);
4002 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4003 local_bh_enable();
4006 /* Triggered when there are no references on the socket anymore */
4007 static void sctp_destruct_sock(struct sock *sk)
4009 struct sctp_sock *sp = sctp_sk(sk);
4011 /* Free up the HMAC transform. */
4012 crypto_free_hash(sp->hmac);
4014 inet_sock_destruct(sk);
4017 /* API 4.1.7 shutdown() - TCP Style Syntax
4018 * int shutdown(int socket, int how);
4020 * sd - the socket descriptor of the association to be closed.
4021 * how - Specifies the type of shutdown. The values are
4022 * as follows:
4023 * SHUT_RD
4024 * Disables further receive operations. No SCTP
4025 * protocol action is taken.
4026 * SHUT_WR
4027 * Disables further send operations, and initiates
4028 * the SCTP shutdown sequence.
4029 * SHUT_RDWR
4030 * Disables further send and receive operations
4031 * and initiates the SCTP shutdown sequence.
4033 static void sctp_shutdown(struct sock *sk, int how)
4035 struct net *net = sock_net(sk);
4036 struct sctp_endpoint *ep;
4037 struct sctp_association *asoc;
4039 if (!sctp_style(sk, TCP))
4040 return;
4042 if (how & SEND_SHUTDOWN) {
4043 ep = sctp_sk(sk)->ep;
4044 if (!list_empty(&ep->asocs)) {
4045 asoc = list_entry(ep->asocs.next,
4046 struct sctp_association, asocs);
4047 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4052 /* 7.2.1 Association Status (SCTP_STATUS)
4054 * Applications can retrieve current status information about an
4055 * association, including association state, peer receiver window size,
4056 * number of unacked data chunks, and number of data chunks pending
4057 * receipt. This information is read-only.
4059 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4060 char __user *optval,
4061 int __user *optlen)
4063 struct sctp_status status;
4064 struct sctp_association *asoc = NULL;
4065 struct sctp_transport *transport;
4066 sctp_assoc_t associd;
4067 int retval = 0;
4069 if (len < sizeof(status)) {
4070 retval = -EINVAL;
4071 goto out;
4074 len = sizeof(status);
4075 if (copy_from_user(&status, optval, len)) {
4076 retval = -EFAULT;
4077 goto out;
4080 associd = status.sstat_assoc_id;
4081 asoc = sctp_id2assoc(sk, associd);
4082 if (!asoc) {
4083 retval = -EINVAL;
4084 goto out;
4087 transport = asoc->peer.primary_path;
4089 status.sstat_assoc_id = sctp_assoc2id(asoc);
4090 status.sstat_state = asoc->state;
4091 status.sstat_rwnd = asoc->peer.rwnd;
4092 status.sstat_unackdata = asoc->unack_data;
4094 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4095 status.sstat_instrms = asoc->c.sinit_max_instreams;
4096 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4097 status.sstat_fragmentation_point = asoc->frag_point;
4098 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4099 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4100 transport->af_specific->sockaddr_len);
4101 /* Map ipv4 address into v4-mapped-on-v6 address. */
4102 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4103 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4104 status.sstat_primary.spinfo_state = transport->state;
4105 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4106 status.sstat_primary.spinfo_srtt = transport->srtt;
4107 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4108 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4110 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4111 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4113 if (put_user(len, optlen)) {
4114 retval = -EFAULT;
4115 goto out;
4118 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4119 __func__, len, status.sstat_state, status.sstat_rwnd,
4120 status.sstat_assoc_id);
4122 if (copy_to_user(optval, &status, len)) {
4123 retval = -EFAULT;
4124 goto out;
4127 out:
4128 return retval;
4132 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4134 * Applications can retrieve information about a specific peer address
4135 * of an association, including its reachability state, congestion
4136 * window, and retransmission timer values. This information is
4137 * read-only.
4139 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4140 char __user *optval,
4141 int __user *optlen)
4143 struct sctp_paddrinfo pinfo;
4144 struct sctp_transport *transport;
4145 int retval = 0;
4147 if (len < sizeof(pinfo)) {
4148 retval = -EINVAL;
4149 goto out;
4152 len = sizeof(pinfo);
4153 if (copy_from_user(&pinfo, optval, len)) {
4154 retval = -EFAULT;
4155 goto out;
4158 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4159 pinfo.spinfo_assoc_id);
4160 if (!transport)
4161 return -EINVAL;
4163 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4164 pinfo.spinfo_state = transport->state;
4165 pinfo.spinfo_cwnd = transport->cwnd;
4166 pinfo.spinfo_srtt = transport->srtt;
4167 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4168 pinfo.spinfo_mtu = transport->pathmtu;
4170 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4171 pinfo.spinfo_state = SCTP_ACTIVE;
4173 if (put_user(len, optlen)) {
4174 retval = -EFAULT;
4175 goto out;
4178 if (copy_to_user(optval, &pinfo, len)) {
4179 retval = -EFAULT;
4180 goto out;
4183 out:
4184 return retval;
4187 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4189 * This option is a on/off flag. If enabled no SCTP message
4190 * fragmentation will be performed. Instead if a message being sent
4191 * exceeds the current PMTU size, the message will NOT be sent and
4192 * instead a error will be indicated to the user.
4194 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4195 char __user *optval, int __user *optlen)
4197 int val;
4199 if (len < sizeof(int))
4200 return -EINVAL;
4202 len = sizeof(int);
4203 val = (sctp_sk(sk)->disable_fragments == 1);
4204 if (put_user(len, optlen))
4205 return -EFAULT;
4206 if (copy_to_user(optval, &val, len))
4207 return -EFAULT;
4208 return 0;
4211 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4213 * This socket option is used to specify various notifications and
4214 * ancillary data the user wishes to receive.
4216 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4217 int __user *optlen)
4219 if (len <= 0)
4220 return -EINVAL;
4221 if (len > sizeof(struct sctp_event_subscribe))
4222 len = sizeof(struct sctp_event_subscribe);
4223 if (put_user(len, optlen))
4224 return -EFAULT;
4225 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4226 return -EFAULT;
4227 return 0;
4230 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4232 * This socket option is applicable to the UDP-style socket only. When
4233 * set it will cause associations that are idle for more than the
4234 * specified number of seconds to automatically close. An association
4235 * being idle is defined an association that has NOT sent or received
4236 * user data. The special value of '0' indicates that no automatic
4237 * close of any associations should be performed. The option expects an
4238 * integer defining the number of seconds of idle time before an
4239 * association is closed.
4241 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4243 /* Applicable to UDP-style socket only */
4244 if (sctp_style(sk, TCP))
4245 return -EOPNOTSUPP;
4246 if (len < sizeof(int))
4247 return -EINVAL;
4248 len = sizeof(int);
4249 if (put_user(len, optlen))
4250 return -EFAULT;
4251 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4252 return -EFAULT;
4253 return 0;
4256 /* Helper routine to branch off an association to a new socket. */
4257 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4259 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4260 struct socket *sock;
4261 struct sctp_af *af;
4262 int err = 0;
4264 if (!asoc)
4265 return -EINVAL;
4267 /* An association cannot be branched off from an already peeled-off
4268 * socket, nor is this supported for tcp style sockets.
4270 if (!sctp_style(sk, UDP))
4271 return -EINVAL;
4273 /* Create a new socket. */
4274 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4275 if (err < 0)
4276 return err;
4278 sctp_copy_sock(sock->sk, sk, asoc);
4280 /* Make peeled-off sockets more like 1-1 accepted sockets.
4281 * Set the daddr and initialize id to something more random
4283 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4284 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4286 /* Populate the fields of the newsk from the oldsk and migrate the
4287 * asoc to the newsk.
4289 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4291 *sockp = sock;
4293 return err;
4295 EXPORT_SYMBOL(sctp_do_peeloff);
4297 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4299 sctp_peeloff_arg_t peeloff;
4300 struct socket *newsock;
4301 struct file *newfile;
4302 int retval = 0;
4304 if (len < sizeof(sctp_peeloff_arg_t))
4305 return -EINVAL;
4306 len = sizeof(sctp_peeloff_arg_t);
4307 if (copy_from_user(&peeloff, optval, len))
4308 return -EFAULT;
4310 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4311 if (retval < 0)
4312 goto out;
4314 /* Map the socket to an unused fd that can be returned to the user. */
4315 retval = get_unused_fd_flags(0);
4316 if (retval < 0) {
4317 sock_release(newsock);
4318 goto out;
4321 newfile = sock_alloc_file(newsock, 0, NULL);
4322 if (unlikely(IS_ERR(newfile))) {
4323 put_unused_fd(retval);
4324 sock_release(newsock);
4325 return PTR_ERR(newfile);
4328 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4329 retval);
4331 /* Return the fd mapped to the new socket. */
4332 if (put_user(len, optlen)) {
4333 fput(newfile);
4334 put_unused_fd(retval);
4335 return -EFAULT;
4337 peeloff.sd = retval;
4338 if (copy_to_user(optval, &peeloff, len)) {
4339 fput(newfile);
4340 put_unused_fd(retval);
4341 return -EFAULT;
4343 fd_install(retval, newfile);
4344 out:
4345 return retval;
4348 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4350 * Applications can enable or disable heartbeats for any peer address of
4351 * an association, modify an address's heartbeat interval, force a
4352 * heartbeat to be sent immediately, and adjust the address's maximum
4353 * number of retransmissions sent before an address is considered
4354 * unreachable. The following structure is used to access and modify an
4355 * address's parameters:
4357 * struct sctp_paddrparams {
4358 * sctp_assoc_t spp_assoc_id;
4359 * struct sockaddr_storage spp_address;
4360 * uint32_t spp_hbinterval;
4361 * uint16_t spp_pathmaxrxt;
4362 * uint32_t spp_pathmtu;
4363 * uint32_t spp_sackdelay;
4364 * uint32_t spp_flags;
4365 * };
4367 * spp_assoc_id - (one-to-many style socket) This is filled in the
4368 * application, and identifies the association for
4369 * this query.
4370 * spp_address - This specifies which address is of interest.
4371 * spp_hbinterval - This contains the value of the heartbeat interval,
4372 * in milliseconds. If a value of zero
4373 * is present in this field then no changes are to
4374 * be made to this parameter.
4375 * spp_pathmaxrxt - This contains the maximum number of
4376 * retransmissions before this address shall be
4377 * considered unreachable. If a value of zero
4378 * is present in this field then no changes are to
4379 * be made to this parameter.
4380 * spp_pathmtu - When Path MTU discovery is disabled the value
4381 * specified here will be the "fixed" path mtu.
4382 * Note that if the spp_address field is empty
4383 * then all associations on this address will
4384 * have this fixed path mtu set upon them.
4386 * spp_sackdelay - When delayed sack is enabled, this value specifies
4387 * the number of milliseconds that sacks will be delayed
4388 * for. This value will apply to all addresses of an
4389 * association if the spp_address field is empty. Note
4390 * also, that if delayed sack is enabled and this
4391 * value is set to 0, no change is made to the last
4392 * recorded delayed sack timer value.
4394 * spp_flags - These flags are used to control various features
4395 * on an association. The flag field may contain
4396 * zero or more of the following options.
4398 * SPP_HB_ENABLE - Enable heartbeats on the
4399 * specified address. Note that if the address
4400 * field is empty all addresses for the association
4401 * have heartbeats enabled upon them.
4403 * SPP_HB_DISABLE - Disable heartbeats on the
4404 * speicifed address. Note that if the address
4405 * field is empty all addresses for the association
4406 * will have their heartbeats disabled. Note also
4407 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4408 * mutually exclusive, only one of these two should
4409 * be specified. Enabling both fields will have
4410 * undetermined results.
4412 * SPP_HB_DEMAND - Request a user initiated heartbeat
4413 * to be made immediately.
4415 * SPP_PMTUD_ENABLE - This field will enable PMTU
4416 * discovery upon the specified address. Note that
4417 * if the address feild is empty then all addresses
4418 * on the association are effected.
4420 * SPP_PMTUD_DISABLE - This field will disable PMTU
4421 * discovery upon the specified address. Note that
4422 * if the address feild is empty then all addresses
4423 * on the association are effected. Not also that
4424 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4425 * exclusive. Enabling both will have undetermined
4426 * results.
4428 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4429 * on delayed sack. The time specified in spp_sackdelay
4430 * is used to specify the sack delay for this address. Note
4431 * that if spp_address is empty then all addresses will
4432 * enable delayed sack and take on the sack delay
4433 * value specified in spp_sackdelay.
4434 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4435 * off delayed sack. If the spp_address field is blank then
4436 * delayed sack is disabled for the entire association. Note
4437 * also that this field is mutually exclusive to
4438 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4439 * results.
4441 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4442 char __user *optval, int __user *optlen)
4444 struct sctp_paddrparams params;
4445 struct sctp_transport *trans = NULL;
4446 struct sctp_association *asoc = NULL;
4447 struct sctp_sock *sp = sctp_sk(sk);
4449 if (len < sizeof(struct sctp_paddrparams))
4450 return -EINVAL;
4451 len = sizeof(struct sctp_paddrparams);
4452 if (copy_from_user(&params, optval, len))
4453 return -EFAULT;
4455 /* If an address other than INADDR_ANY is specified, and
4456 * no transport is found, then the request is invalid.
4458 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4459 trans = sctp_addr_id2transport(sk, &params.spp_address,
4460 params.spp_assoc_id);
4461 if (!trans) {
4462 pr_debug("%s: failed no transport\n", __func__);
4463 return -EINVAL;
4467 /* Get association, if assoc_id != 0 and the socket is a one
4468 * to many style socket, and an association was not found, then
4469 * the id was invalid.
4471 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4472 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4473 pr_debug("%s: failed no association\n", __func__);
4474 return -EINVAL;
4477 if (trans) {
4478 /* Fetch transport values. */
4479 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4480 params.spp_pathmtu = trans->pathmtu;
4481 params.spp_pathmaxrxt = trans->pathmaxrxt;
4482 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4484 /*draft-11 doesn't say what to return in spp_flags*/
4485 params.spp_flags = trans->param_flags;
4486 } else if (asoc) {
4487 /* Fetch association values. */
4488 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4489 params.spp_pathmtu = asoc->pathmtu;
4490 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4491 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4493 /*draft-11 doesn't say what to return in spp_flags*/
4494 params.spp_flags = asoc->param_flags;
4495 } else {
4496 /* Fetch socket values. */
4497 params.spp_hbinterval = sp->hbinterval;
4498 params.spp_pathmtu = sp->pathmtu;
4499 params.spp_sackdelay = sp->sackdelay;
4500 params.spp_pathmaxrxt = sp->pathmaxrxt;
4502 /*draft-11 doesn't say what to return in spp_flags*/
4503 params.spp_flags = sp->param_flags;
4506 if (copy_to_user(optval, &params, len))
4507 return -EFAULT;
4509 if (put_user(len, optlen))
4510 return -EFAULT;
4512 return 0;
4516 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4518 * This option will effect the way delayed acks are performed. This
4519 * option allows you to get or set the delayed ack time, in
4520 * milliseconds. It also allows changing the delayed ack frequency.
4521 * Changing the frequency to 1 disables the delayed sack algorithm. If
4522 * the assoc_id is 0, then this sets or gets the endpoints default
4523 * values. If the assoc_id field is non-zero, then the set or get
4524 * effects the specified association for the one to many model (the
4525 * assoc_id field is ignored by the one to one model). Note that if
4526 * sack_delay or sack_freq are 0 when setting this option, then the
4527 * current values will remain unchanged.
4529 * struct sctp_sack_info {
4530 * sctp_assoc_t sack_assoc_id;
4531 * uint32_t sack_delay;
4532 * uint32_t sack_freq;
4533 * };
4535 * sack_assoc_id - This parameter, indicates which association the user
4536 * is performing an action upon. Note that if this field's value is
4537 * zero then the endpoints default value is changed (effecting future
4538 * associations only).
4540 * sack_delay - This parameter contains the number of milliseconds that
4541 * the user is requesting the delayed ACK timer be set to. Note that
4542 * this value is defined in the standard to be between 200 and 500
4543 * milliseconds.
4545 * sack_freq - This parameter contains the number of packets that must
4546 * be received before a sack is sent without waiting for the delay
4547 * timer to expire. The default value for this is 2, setting this
4548 * value to 1 will disable the delayed sack algorithm.
4550 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4551 char __user *optval,
4552 int __user *optlen)
4554 struct sctp_sack_info params;
4555 struct sctp_association *asoc = NULL;
4556 struct sctp_sock *sp = sctp_sk(sk);
4558 if (len >= sizeof(struct sctp_sack_info)) {
4559 len = sizeof(struct sctp_sack_info);
4561 if (copy_from_user(&params, optval, len))
4562 return -EFAULT;
4563 } else if (len == sizeof(struct sctp_assoc_value)) {
4564 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4565 pr_warn("Use struct sctp_sack_info instead\n");
4566 if (copy_from_user(&params, optval, len))
4567 return -EFAULT;
4568 } else
4569 return - EINVAL;
4571 /* Get association, if sack_assoc_id != 0 and the socket is a one
4572 * to many style socket, and an association was not found, then
4573 * the id was invalid.
4575 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4576 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4577 return -EINVAL;
4579 if (asoc) {
4580 /* Fetch association values. */
4581 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4582 params.sack_delay = jiffies_to_msecs(
4583 asoc->sackdelay);
4584 params.sack_freq = asoc->sackfreq;
4586 } else {
4587 params.sack_delay = 0;
4588 params.sack_freq = 1;
4590 } else {
4591 /* Fetch socket values. */
4592 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4593 params.sack_delay = sp->sackdelay;
4594 params.sack_freq = sp->sackfreq;
4595 } else {
4596 params.sack_delay = 0;
4597 params.sack_freq = 1;
4601 if (copy_to_user(optval, &params, len))
4602 return -EFAULT;
4604 if (put_user(len, optlen))
4605 return -EFAULT;
4607 return 0;
4610 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4612 * Applications can specify protocol parameters for the default association
4613 * initialization. The option name argument to setsockopt() and getsockopt()
4614 * is SCTP_INITMSG.
4616 * Setting initialization parameters is effective only on an unconnected
4617 * socket (for UDP-style sockets only future associations are effected
4618 * by the change). With TCP-style sockets, this option is inherited by
4619 * sockets derived from a listener socket.
4621 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4623 if (len < sizeof(struct sctp_initmsg))
4624 return -EINVAL;
4625 len = sizeof(struct sctp_initmsg);
4626 if (put_user(len, optlen))
4627 return -EFAULT;
4628 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4629 return -EFAULT;
4630 return 0;
4634 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4635 char __user *optval, int __user *optlen)
4637 struct sctp_association *asoc;
4638 int cnt = 0;
4639 struct sctp_getaddrs getaddrs;
4640 struct sctp_transport *from;
4641 void __user *to;
4642 union sctp_addr temp;
4643 struct sctp_sock *sp = sctp_sk(sk);
4644 int addrlen;
4645 size_t space_left;
4646 int bytes_copied;
4648 if (len < sizeof(struct sctp_getaddrs))
4649 return -EINVAL;
4651 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4652 return -EFAULT;
4654 /* For UDP-style sockets, id specifies the association to query. */
4655 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4656 if (!asoc)
4657 return -EINVAL;
4659 to = optval + offsetof(struct sctp_getaddrs,addrs);
4660 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4662 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4663 transports) {
4664 memcpy(&temp, &from->ipaddr, sizeof(temp));
4665 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4666 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4667 if (space_left < addrlen)
4668 return -ENOMEM;
4669 if (copy_to_user(to, &temp, addrlen))
4670 return -EFAULT;
4671 to += addrlen;
4672 cnt++;
4673 space_left -= addrlen;
4676 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4677 return -EFAULT;
4678 bytes_copied = ((char __user *)to) - optval;
4679 if (put_user(bytes_copied, optlen))
4680 return -EFAULT;
4682 return 0;
4685 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4686 size_t space_left, int *bytes_copied)
4688 struct sctp_sockaddr_entry *addr;
4689 union sctp_addr temp;
4690 int cnt = 0;
4691 int addrlen;
4692 struct net *net = sock_net(sk);
4694 rcu_read_lock();
4695 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4696 if (!addr->valid)
4697 continue;
4699 if ((PF_INET == sk->sk_family) &&
4700 (AF_INET6 == addr->a.sa.sa_family))
4701 continue;
4702 if ((PF_INET6 == sk->sk_family) &&
4703 inet_v6_ipv6only(sk) &&
4704 (AF_INET == addr->a.sa.sa_family))
4705 continue;
4706 memcpy(&temp, &addr->a, sizeof(temp));
4707 if (!temp.v4.sin_port)
4708 temp.v4.sin_port = htons(port);
4710 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4711 &temp);
4712 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4713 if (space_left < addrlen) {
4714 cnt = -ENOMEM;
4715 break;
4717 memcpy(to, &temp, addrlen);
4719 to += addrlen;
4720 cnt ++;
4721 space_left -= addrlen;
4722 *bytes_copied += addrlen;
4724 rcu_read_unlock();
4726 return cnt;
4730 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4731 char __user *optval, int __user *optlen)
4733 struct sctp_bind_addr *bp;
4734 struct sctp_association *asoc;
4735 int cnt = 0;
4736 struct sctp_getaddrs getaddrs;
4737 struct sctp_sockaddr_entry *addr;
4738 void __user *to;
4739 union sctp_addr temp;
4740 struct sctp_sock *sp = sctp_sk(sk);
4741 int addrlen;
4742 int err = 0;
4743 size_t space_left;
4744 int bytes_copied = 0;
4745 void *addrs;
4746 void *buf;
4748 if (len < sizeof(struct sctp_getaddrs))
4749 return -EINVAL;
4751 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4752 return -EFAULT;
4755 * For UDP-style sockets, id specifies the association to query.
4756 * If the id field is set to the value '0' then the locally bound
4757 * addresses are returned without regard to any particular
4758 * association.
4760 if (0 == getaddrs.assoc_id) {
4761 bp = &sctp_sk(sk)->ep->base.bind_addr;
4762 } else {
4763 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4764 if (!asoc)
4765 return -EINVAL;
4766 bp = &asoc->base.bind_addr;
4769 to = optval + offsetof(struct sctp_getaddrs,addrs);
4770 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4772 addrs = kmalloc(space_left, GFP_KERNEL);
4773 if (!addrs)
4774 return -ENOMEM;
4776 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4777 * addresses from the global local address list.
4779 if (sctp_list_single_entry(&bp->address_list)) {
4780 addr = list_entry(bp->address_list.next,
4781 struct sctp_sockaddr_entry, list);
4782 if (sctp_is_any(sk, &addr->a)) {
4783 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4784 space_left, &bytes_copied);
4785 if (cnt < 0) {
4786 err = cnt;
4787 goto out;
4789 goto copy_getaddrs;
4793 buf = addrs;
4794 /* Protection on the bound address list is not needed since
4795 * in the socket option context we hold a socket lock and
4796 * thus the bound address list can't change.
4798 list_for_each_entry(addr, &bp->address_list, list) {
4799 memcpy(&temp, &addr->a, sizeof(temp));
4800 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4801 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4802 if (space_left < addrlen) {
4803 err = -ENOMEM; /*fixme: right error?*/
4804 goto out;
4806 memcpy(buf, &temp, addrlen);
4807 buf += addrlen;
4808 bytes_copied += addrlen;
4809 cnt ++;
4810 space_left -= addrlen;
4813 copy_getaddrs:
4814 if (copy_to_user(to, addrs, bytes_copied)) {
4815 err = -EFAULT;
4816 goto out;
4818 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4819 err = -EFAULT;
4820 goto out;
4822 if (put_user(bytes_copied, optlen))
4823 err = -EFAULT;
4824 out:
4825 kfree(addrs);
4826 return err;
4829 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4831 * Requests that the local SCTP stack use the enclosed peer address as
4832 * the association primary. The enclosed address must be one of the
4833 * association peer's addresses.
4835 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4836 char __user *optval, int __user *optlen)
4838 struct sctp_prim prim;
4839 struct sctp_association *asoc;
4840 struct sctp_sock *sp = sctp_sk(sk);
4842 if (len < sizeof(struct sctp_prim))
4843 return -EINVAL;
4845 len = sizeof(struct sctp_prim);
4847 if (copy_from_user(&prim, optval, len))
4848 return -EFAULT;
4850 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4851 if (!asoc)
4852 return -EINVAL;
4854 if (!asoc->peer.primary_path)
4855 return -ENOTCONN;
4857 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4858 asoc->peer.primary_path->af_specific->sockaddr_len);
4860 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4861 (union sctp_addr *)&prim.ssp_addr);
4863 if (put_user(len, optlen))
4864 return -EFAULT;
4865 if (copy_to_user(optval, &prim, len))
4866 return -EFAULT;
4868 return 0;
4872 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4874 * Requests that the local endpoint set the specified Adaptation Layer
4875 * Indication parameter for all future INIT and INIT-ACK exchanges.
4877 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4878 char __user *optval, int __user *optlen)
4880 struct sctp_setadaptation adaptation;
4882 if (len < sizeof(struct sctp_setadaptation))
4883 return -EINVAL;
4885 len = sizeof(struct sctp_setadaptation);
4887 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4889 if (put_user(len, optlen))
4890 return -EFAULT;
4891 if (copy_to_user(optval, &adaptation, len))
4892 return -EFAULT;
4894 return 0;
4899 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4901 * Applications that wish to use the sendto() system call may wish to
4902 * specify a default set of parameters that would normally be supplied
4903 * through the inclusion of ancillary data. This socket option allows
4904 * such an application to set the default sctp_sndrcvinfo structure.
4907 * The application that wishes to use this socket option simply passes
4908 * in to this call the sctp_sndrcvinfo structure defined in Section
4909 * 5.2.2) The input parameters accepted by this call include
4910 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4911 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4912 * to this call if the caller is using the UDP model.
4914 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4916 static int sctp_getsockopt_default_send_param(struct sock *sk,
4917 int len, char __user *optval,
4918 int __user *optlen)
4920 struct sctp_sndrcvinfo info;
4921 struct sctp_association *asoc;
4922 struct sctp_sock *sp = sctp_sk(sk);
4924 if (len < sizeof(struct sctp_sndrcvinfo))
4925 return -EINVAL;
4927 len = sizeof(struct sctp_sndrcvinfo);
4929 if (copy_from_user(&info, optval, len))
4930 return -EFAULT;
4932 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4933 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4934 return -EINVAL;
4936 if (asoc) {
4937 info.sinfo_stream = asoc->default_stream;
4938 info.sinfo_flags = asoc->default_flags;
4939 info.sinfo_ppid = asoc->default_ppid;
4940 info.sinfo_context = asoc->default_context;
4941 info.sinfo_timetolive = asoc->default_timetolive;
4942 } else {
4943 info.sinfo_stream = sp->default_stream;
4944 info.sinfo_flags = sp->default_flags;
4945 info.sinfo_ppid = sp->default_ppid;
4946 info.sinfo_context = sp->default_context;
4947 info.sinfo_timetolive = sp->default_timetolive;
4950 if (put_user(len, optlen))
4951 return -EFAULT;
4952 if (copy_to_user(optval, &info, len))
4953 return -EFAULT;
4955 return 0;
4960 * 7.1.5 SCTP_NODELAY
4962 * Turn on/off any Nagle-like algorithm. This means that packets are
4963 * generally sent as soon as possible and no unnecessary delays are
4964 * introduced, at the cost of more packets in the network. Expects an
4965 * integer boolean flag.
4968 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4969 char __user *optval, int __user *optlen)
4971 int val;
4973 if (len < sizeof(int))
4974 return -EINVAL;
4976 len = sizeof(int);
4977 val = (sctp_sk(sk)->nodelay == 1);
4978 if (put_user(len, optlen))
4979 return -EFAULT;
4980 if (copy_to_user(optval, &val, len))
4981 return -EFAULT;
4982 return 0;
4987 * 7.1.1 SCTP_RTOINFO
4989 * The protocol parameters used to initialize and bound retransmission
4990 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4991 * and modify these parameters.
4992 * All parameters are time values, in milliseconds. A value of 0, when
4993 * modifying the parameters, indicates that the current value should not
4994 * be changed.
4997 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4998 char __user *optval,
4999 int __user *optlen) {
5000 struct sctp_rtoinfo rtoinfo;
5001 struct sctp_association *asoc;
5003 if (len < sizeof (struct sctp_rtoinfo))
5004 return -EINVAL;
5006 len = sizeof(struct sctp_rtoinfo);
5008 if (copy_from_user(&rtoinfo, optval, len))
5009 return -EFAULT;
5011 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5013 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5014 return -EINVAL;
5016 /* Values corresponding to the specific association. */
5017 if (asoc) {
5018 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5019 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5020 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5021 } else {
5022 /* Values corresponding to the endpoint. */
5023 struct sctp_sock *sp = sctp_sk(sk);
5025 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5026 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5027 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5030 if (put_user(len, optlen))
5031 return -EFAULT;
5033 if (copy_to_user(optval, &rtoinfo, len))
5034 return -EFAULT;
5036 return 0;
5041 * 7.1.2 SCTP_ASSOCINFO
5043 * This option is used to tune the maximum retransmission attempts
5044 * of the association.
5045 * Returns an error if the new association retransmission value is
5046 * greater than the sum of the retransmission value of the peer.
5047 * See [SCTP] for more information.
5050 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5051 char __user *optval,
5052 int __user *optlen)
5055 struct sctp_assocparams assocparams;
5056 struct sctp_association *asoc;
5057 struct list_head *pos;
5058 int cnt = 0;
5060 if (len < sizeof (struct sctp_assocparams))
5061 return -EINVAL;
5063 len = sizeof(struct sctp_assocparams);
5065 if (copy_from_user(&assocparams, optval, len))
5066 return -EFAULT;
5068 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5070 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5071 return -EINVAL;
5073 /* Values correspoinding to the specific association */
5074 if (asoc) {
5075 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5076 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5077 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5078 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5080 list_for_each(pos, &asoc->peer.transport_addr_list) {
5081 cnt ++;
5084 assocparams.sasoc_number_peer_destinations = cnt;
5085 } else {
5086 /* Values corresponding to the endpoint */
5087 struct sctp_sock *sp = sctp_sk(sk);
5089 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5090 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5091 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5092 assocparams.sasoc_cookie_life =
5093 sp->assocparams.sasoc_cookie_life;
5094 assocparams.sasoc_number_peer_destinations =
5095 sp->assocparams.
5096 sasoc_number_peer_destinations;
5099 if (put_user(len, optlen))
5100 return -EFAULT;
5102 if (copy_to_user(optval, &assocparams, len))
5103 return -EFAULT;
5105 return 0;
5109 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5111 * This socket option is a boolean flag which turns on or off mapped V4
5112 * addresses. If this option is turned on and the socket is type
5113 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5114 * If this option is turned off, then no mapping will be done of V4
5115 * addresses and a user will receive both PF_INET6 and PF_INET type
5116 * addresses on the socket.
5118 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5119 char __user *optval, int __user *optlen)
5121 int val;
5122 struct sctp_sock *sp = sctp_sk(sk);
5124 if (len < sizeof(int))
5125 return -EINVAL;
5127 len = sizeof(int);
5128 val = sp->v4mapped;
5129 if (put_user(len, optlen))
5130 return -EFAULT;
5131 if (copy_to_user(optval, &val, len))
5132 return -EFAULT;
5134 return 0;
5138 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5139 * (chapter and verse is quoted at sctp_setsockopt_context())
5141 static int sctp_getsockopt_context(struct sock *sk, int len,
5142 char __user *optval, int __user *optlen)
5144 struct sctp_assoc_value params;
5145 struct sctp_sock *sp;
5146 struct sctp_association *asoc;
5148 if (len < sizeof(struct sctp_assoc_value))
5149 return -EINVAL;
5151 len = sizeof(struct sctp_assoc_value);
5153 if (copy_from_user(&params, optval, len))
5154 return -EFAULT;
5156 sp = sctp_sk(sk);
5158 if (params.assoc_id != 0) {
5159 asoc = sctp_id2assoc(sk, params.assoc_id);
5160 if (!asoc)
5161 return -EINVAL;
5162 params.assoc_value = asoc->default_rcv_context;
5163 } else {
5164 params.assoc_value = sp->default_rcv_context;
5167 if (put_user(len, optlen))
5168 return -EFAULT;
5169 if (copy_to_user(optval, &params, len))
5170 return -EFAULT;
5172 return 0;
5176 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5177 * This option will get or set the maximum size to put in any outgoing
5178 * SCTP DATA chunk. If a message is larger than this size it will be
5179 * fragmented by SCTP into the specified size. Note that the underlying
5180 * SCTP implementation may fragment into smaller sized chunks when the
5181 * PMTU of the underlying association is smaller than the value set by
5182 * the user. The default value for this option is '0' which indicates
5183 * the user is NOT limiting fragmentation and only the PMTU will effect
5184 * SCTP's choice of DATA chunk size. Note also that values set larger
5185 * than the maximum size of an IP datagram will effectively let SCTP
5186 * control fragmentation (i.e. the same as setting this option to 0).
5188 * The following structure is used to access and modify this parameter:
5190 * struct sctp_assoc_value {
5191 * sctp_assoc_t assoc_id;
5192 * uint32_t assoc_value;
5193 * };
5195 * assoc_id: This parameter is ignored for one-to-one style sockets.
5196 * For one-to-many style sockets this parameter indicates which
5197 * association the user is performing an action upon. Note that if
5198 * this field's value is zero then the endpoints default value is
5199 * changed (effecting future associations only).
5200 * assoc_value: This parameter specifies the maximum size in bytes.
5202 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5203 char __user *optval, int __user *optlen)
5205 struct sctp_assoc_value params;
5206 struct sctp_association *asoc;
5208 if (len == sizeof(int)) {
5209 pr_warn("Use of int in maxseg socket option deprecated\n");
5210 pr_warn("Use struct sctp_assoc_value instead\n");
5211 params.assoc_id = 0;
5212 } else if (len >= sizeof(struct sctp_assoc_value)) {
5213 len = sizeof(struct sctp_assoc_value);
5214 if (copy_from_user(&params, optval, sizeof(params)))
5215 return -EFAULT;
5216 } else
5217 return -EINVAL;
5219 asoc = sctp_id2assoc(sk, params.assoc_id);
5220 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5221 return -EINVAL;
5223 if (asoc)
5224 params.assoc_value = asoc->frag_point;
5225 else
5226 params.assoc_value = sctp_sk(sk)->user_frag;
5228 if (put_user(len, optlen))
5229 return -EFAULT;
5230 if (len == sizeof(int)) {
5231 if (copy_to_user(optval, &params.assoc_value, len))
5232 return -EFAULT;
5233 } else {
5234 if (copy_to_user(optval, &params, len))
5235 return -EFAULT;
5238 return 0;
5242 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5243 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5245 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5246 char __user *optval, int __user *optlen)
5248 int val;
5250 if (len < sizeof(int))
5251 return -EINVAL;
5253 len = sizeof(int);
5255 val = sctp_sk(sk)->frag_interleave;
5256 if (put_user(len, optlen))
5257 return -EFAULT;
5258 if (copy_to_user(optval, &val, len))
5259 return -EFAULT;
5261 return 0;
5265 * 7.1.25. Set or Get the sctp partial delivery point
5266 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5268 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5269 char __user *optval,
5270 int __user *optlen)
5272 u32 val;
5274 if (len < sizeof(u32))
5275 return -EINVAL;
5277 len = sizeof(u32);
5279 val = sctp_sk(sk)->pd_point;
5280 if (put_user(len, optlen))
5281 return -EFAULT;
5282 if (copy_to_user(optval, &val, len))
5283 return -EFAULT;
5285 return 0;
5289 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5290 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5292 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5293 char __user *optval,
5294 int __user *optlen)
5296 struct sctp_assoc_value params;
5297 struct sctp_sock *sp;
5298 struct sctp_association *asoc;
5300 if (len == sizeof(int)) {
5301 pr_warn("Use of int in max_burst socket option deprecated\n");
5302 pr_warn("Use struct sctp_assoc_value instead\n");
5303 params.assoc_id = 0;
5304 } else if (len >= sizeof(struct sctp_assoc_value)) {
5305 len = sizeof(struct sctp_assoc_value);
5306 if (copy_from_user(&params, optval, len))
5307 return -EFAULT;
5308 } else
5309 return -EINVAL;
5311 sp = sctp_sk(sk);
5313 if (params.assoc_id != 0) {
5314 asoc = sctp_id2assoc(sk, params.assoc_id);
5315 if (!asoc)
5316 return -EINVAL;
5317 params.assoc_value = asoc->max_burst;
5318 } else
5319 params.assoc_value = sp->max_burst;
5321 if (len == sizeof(int)) {
5322 if (copy_to_user(optval, &params.assoc_value, len))
5323 return -EFAULT;
5324 } else {
5325 if (copy_to_user(optval, &params, len))
5326 return -EFAULT;
5329 return 0;
5333 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5334 char __user *optval, int __user *optlen)
5336 struct net *net = sock_net(sk);
5337 struct sctp_hmacalgo __user *p = (void __user *)optval;
5338 struct sctp_hmac_algo_param *hmacs;
5339 __u16 data_len = 0;
5340 u32 num_idents;
5342 if (!net->sctp.auth_enable)
5343 return -EACCES;
5345 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5346 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5348 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5349 return -EINVAL;
5351 len = sizeof(struct sctp_hmacalgo) + data_len;
5352 num_idents = data_len / sizeof(u16);
5354 if (put_user(len, optlen))
5355 return -EFAULT;
5356 if (put_user(num_idents, &p->shmac_num_idents))
5357 return -EFAULT;
5358 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5359 return -EFAULT;
5360 return 0;
5363 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5364 char __user *optval, int __user *optlen)
5366 struct net *net = sock_net(sk);
5367 struct sctp_authkeyid val;
5368 struct sctp_association *asoc;
5370 if (!net->sctp.auth_enable)
5371 return -EACCES;
5373 if (len < sizeof(struct sctp_authkeyid))
5374 return -EINVAL;
5375 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5376 return -EFAULT;
5378 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5379 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5380 return -EINVAL;
5382 if (asoc)
5383 val.scact_keynumber = asoc->active_key_id;
5384 else
5385 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5387 len = sizeof(struct sctp_authkeyid);
5388 if (put_user(len, optlen))
5389 return -EFAULT;
5390 if (copy_to_user(optval, &val, len))
5391 return -EFAULT;
5393 return 0;
5396 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5397 char __user *optval, int __user *optlen)
5399 struct net *net = sock_net(sk);
5400 struct sctp_authchunks __user *p = (void __user *)optval;
5401 struct sctp_authchunks val;
5402 struct sctp_association *asoc;
5403 struct sctp_chunks_param *ch;
5404 u32 num_chunks = 0;
5405 char __user *to;
5407 if (!net->sctp.auth_enable)
5408 return -EACCES;
5410 if (len < sizeof(struct sctp_authchunks))
5411 return -EINVAL;
5413 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5414 return -EFAULT;
5416 to = p->gauth_chunks;
5417 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5418 if (!asoc)
5419 return -EINVAL;
5421 ch = asoc->peer.peer_chunks;
5422 if (!ch)
5423 goto num;
5425 /* See if the user provided enough room for all the data */
5426 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5427 if (len < num_chunks)
5428 return -EINVAL;
5430 if (copy_to_user(to, ch->chunks, num_chunks))
5431 return -EFAULT;
5432 num:
5433 len = sizeof(struct sctp_authchunks) + num_chunks;
5434 if (put_user(len, optlen)) return -EFAULT;
5435 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5436 return -EFAULT;
5437 return 0;
5440 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5441 char __user *optval, int __user *optlen)
5443 struct net *net = sock_net(sk);
5444 struct sctp_authchunks __user *p = (void __user *)optval;
5445 struct sctp_authchunks val;
5446 struct sctp_association *asoc;
5447 struct sctp_chunks_param *ch;
5448 u32 num_chunks = 0;
5449 char __user *to;
5451 if (!net->sctp.auth_enable)
5452 return -EACCES;
5454 if (len < sizeof(struct sctp_authchunks))
5455 return -EINVAL;
5457 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5458 return -EFAULT;
5460 to = p->gauth_chunks;
5461 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5462 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5463 return -EINVAL;
5465 if (asoc)
5466 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5467 else
5468 ch = sctp_sk(sk)->ep->auth_chunk_list;
5470 if (!ch)
5471 goto num;
5473 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5474 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5475 return -EINVAL;
5477 if (copy_to_user(to, ch->chunks, num_chunks))
5478 return -EFAULT;
5479 num:
5480 len = sizeof(struct sctp_authchunks) + num_chunks;
5481 if (put_user(len, optlen))
5482 return -EFAULT;
5483 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5484 return -EFAULT;
5486 return 0;
5490 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5491 * This option gets the current number of associations that are attached
5492 * to a one-to-many style socket. The option value is an uint32_t.
5494 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5495 char __user *optval, int __user *optlen)
5497 struct sctp_sock *sp = sctp_sk(sk);
5498 struct sctp_association *asoc;
5499 u32 val = 0;
5501 if (sctp_style(sk, TCP))
5502 return -EOPNOTSUPP;
5504 if (len < sizeof(u32))
5505 return -EINVAL;
5507 len = sizeof(u32);
5509 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5510 val++;
5513 if (put_user(len, optlen))
5514 return -EFAULT;
5515 if (copy_to_user(optval, &val, len))
5516 return -EFAULT;
5518 return 0;
5522 * 8.1.23 SCTP_AUTO_ASCONF
5523 * See the corresponding setsockopt entry as description
5525 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5526 char __user *optval, int __user *optlen)
5528 int val = 0;
5530 if (len < sizeof(int))
5531 return -EINVAL;
5533 len = sizeof(int);
5534 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5535 val = 1;
5536 if (put_user(len, optlen))
5537 return -EFAULT;
5538 if (copy_to_user(optval, &val, len))
5539 return -EFAULT;
5540 return 0;
5544 * 8.2.6. Get the Current Identifiers of Associations
5545 * (SCTP_GET_ASSOC_ID_LIST)
5547 * This option gets the current list of SCTP association identifiers of
5548 * the SCTP associations handled by a one-to-many style socket.
5550 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5551 char __user *optval, int __user *optlen)
5553 struct sctp_sock *sp = sctp_sk(sk);
5554 struct sctp_association *asoc;
5555 struct sctp_assoc_ids *ids;
5556 u32 num = 0;
5558 if (sctp_style(sk, TCP))
5559 return -EOPNOTSUPP;
5561 if (len < sizeof(struct sctp_assoc_ids))
5562 return -EINVAL;
5564 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5565 num++;
5568 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5569 return -EINVAL;
5571 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5573 ids = kmalloc(len, GFP_KERNEL);
5574 if (unlikely(!ids))
5575 return -ENOMEM;
5577 ids->gaids_number_of_ids = num;
5578 num = 0;
5579 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5580 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5583 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5584 kfree(ids);
5585 return -EFAULT;
5588 kfree(ids);
5589 return 0;
5593 * SCTP_PEER_ADDR_THLDS
5595 * This option allows us to fetch the partially failed threshold for one or all
5596 * transports in an association. See Section 6.1 of:
5597 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5599 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5600 char __user *optval,
5601 int len,
5602 int __user *optlen)
5604 struct sctp_paddrthlds val;
5605 struct sctp_transport *trans;
5606 struct sctp_association *asoc;
5608 if (len < sizeof(struct sctp_paddrthlds))
5609 return -EINVAL;
5610 len = sizeof(struct sctp_paddrthlds);
5611 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5612 return -EFAULT;
5614 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5615 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5616 if (!asoc)
5617 return -ENOENT;
5619 val.spt_pathpfthld = asoc->pf_retrans;
5620 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5621 } else {
5622 trans = sctp_addr_id2transport(sk, &val.spt_address,
5623 val.spt_assoc_id);
5624 if (!trans)
5625 return -ENOENT;
5627 val.spt_pathmaxrxt = trans->pathmaxrxt;
5628 val.spt_pathpfthld = trans->pf_retrans;
5631 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5632 return -EFAULT;
5634 return 0;
5638 * SCTP_GET_ASSOC_STATS
5640 * This option retrieves local per endpoint statistics. It is modeled
5641 * after OpenSolaris' implementation
5643 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5644 char __user *optval,
5645 int __user *optlen)
5647 struct sctp_assoc_stats sas;
5648 struct sctp_association *asoc = NULL;
5650 /* User must provide at least the assoc id */
5651 if (len < sizeof(sctp_assoc_t))
5652 return -EINVAL;
5654 /* Allow the struct to grow and fill in as much as possible */
5655 len = min_t(size_t, len, sizeof(sas));
5657 if (copy_from_user(&sas, optval, len))
5658 return -EFAULT;
5660 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5661 if (!asoc)
5662 return -EINVAL;
5664 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5665 sas.sas_gapcnt = asoc->stats.gapcnt;
5666 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5667 sas.sas_osacks = asoc->stats.osacks;
5668 sas.sas_isacks = asoc->stats.isacks;
5669 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5670 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5671 sas.sas_oodchunks = asoc->stats.oodchunks;
5672 sas.sas_iodchunks = asoc->stats.iodchunks;
5673 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5674 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5675 sas.sas_idupchunks = asoc->stats.idupchunks;
5676 sas.sas_opackets = asoc->stats.opackets;
5677 sas.sas_ipackets = asoc->stats.ipackets;
5679 /* New high max rto observed, will return 0 if not a single
5680 * RTO update took place. obs_rto_ipaddr will be bogus
5681 * in such a case
5683 sas.sas_maxrto = asoc->stats.max_obs_rto;
5684 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5685 sizeof(struct sockaddr_storage));
5687 /* Mark beginning of a new observation period */
5688 asoc->stats.max_obs_rto = asoc->rto_min;
5690 if (put_user(len, optlen))
5691 return -EFAULT;
5693 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5695 if (copy_to_user(optval, &sas, len))
5696 return -EFAULT;
5698 return 0;
5701 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5702 char __user *optval, int __user *optlen)
5704 int retval = 0;
5705 int len;
5707 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5709 /* I can hardly begin to describe how wrong this is. This is
5710 * so broken as to be worse than useless. The API draft
5711 * REALLY is NOT helpful here... I am not convinced that the
5712 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5713 * are at all well-founded.
5715 if (level != SOL_SCTP) {
5716 struct sctp_af *af = sctp_sk(sk)->pf->af;
5718 retval = af->getsockopt(sk, level, optname, optval, optlen);
5719 return retval;
5722 if (get_user(len, optlen))
5723 return -EFAULT;
5725 sctp_lock_sock(sk);
5727 switch (optname) {
5728 case SCTP_STATUS:
5729 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5730 break;
5731 case SCTP_DISABLE_FRAGMENTS:
5732 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5733 optlen);
5734 break;
5735 case SCTP_EVENTS:
5736 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5737 break;
5738 case SCTP_AUTOCLOSE:
5739 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5740 break;
5741 case SCTP_SOCKOPT_PEELOFF:
5742 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5743 break;
5744 case SCTP_PEER_ADDR_PARAMS:
5745 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5746 optlen);
5747 break;
5748 case SCTP_DELAYED_SACK:
5749 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5750 optlen);
5751 break;
5752 case SCTP_INITMSG:
5753 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5754 break;
5755 case SCTP_GET_PEER_ADDRS:
5756 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5757 optlen);
5758 break;
5759 case SCTP_GET_LOCAL_ADDRS:
5760 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5761 optlen);
5762 break;
5763 case SCTP_SOCKOPT_CONNECTX3:
5764 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5765 break;
5766 case SCTP_DEFAULT_SEND_PARAM:
5767 retval = sctp_getsockopt_default_send_param(sk, len,
5768 optval, optlen);
5769 break;
5770 case SCTP_PRIMARY_ADDR:
5771 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5772 break;
5773 case SCTP_NODELAY:
5774 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5775 break;
5776 case SCTP_RTOINFO:
5777 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5778 break;
5779 case SCTP_ASSOCINFO:
5780 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5781 break;
5782 case SCTP_I_WANT_MAPPED_V4_ADDR:
5783 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5784 break;
5785 case SCTP_MAXSEG:
5786 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5787 break;
5788 case SCTP_GET_PEER_ADDR_INFO:
5789 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5790 optlen);
5791 break;
5792 case SCTP_ADAPTATION_LAYER:
5793 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5794 optlen);
5795 break;
5796 case SCTP_CONTEXT:
5797 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5798 break;
5799 case SCTP_FRAGMENT_INTERLEAVE:
5800 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5801 optlen);
5802 break;
5803 case SCTP_PARTIAL_DELIVERY_POINT:
5804 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5805 optlen);
5806 break;
5807 case SCTP_MAX_BURST:
5808 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5809 break;
5810 case SCTP_AUTH_KEY:
5811 case SCTP_AUTH_CHUNK:
5812 case SCTP_AUTH_DELETE_KEY:
5813 retval = -EOPNOTSUPP;
5814 break;
5815 case SCTP_HMAC_IDENT:
5816 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5817 break;
5818 case SCTP_AUTH_ACTIVE_KEY:
5819 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5820 break;
5821 case SCTP_PEER_AUTH_CHUNKS:
5822 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5823 optlen);
5824 break;
5825 case SCTP_LOCAL_AUTH_CHUNKS:
5826 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5827 optlen);
5828 break;
5829 case SCTP_GET_ASSOC_NUMBER:
5830 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5831 break;
5832 case SCTP_GET_ASSOC_ID_LIST:
5833 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5834 break;
5835 case SCTP_AUTO_ASCONF:
5836 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5837 break;
5838 case SCTP_PEER_ADDR_THLDS:
5839 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5840 break;
5841 case SCTP_GET_ASSOC_STATS:
5842 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5843 break;
5844 default:
5845 retval = -ENOPROTOOPT;
5846 break;
5849 sctp_release_sock(sk);
5850 return retval;
5853 static void sctp_hash(struct sock *sk)
5855 /* STUB */
5858 static void sctp_unhash(struct sock *sk)
5860 /* STUB */
5863 /* Check if port is acceptable. Possibly find first available port.
5865 * The port hash table (contained in the 'global' SCTP protocol storage
5866 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5867 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5868 * list (the list number is the port number hashed out, so as you
5869 * would expect from a hash function, all the ports in a given list have
5870 * such a number that hashes out to the same list number; you were
5871 * expecting that, right?); so each list has a set of ports, with a
5872 * link to the socket (struct sock) that uses it, the port number and
5873 * a fastreuse flag (FIXME: NPI ipg).
5875 static struct sctp_bind_bucket *sctp_bucket_create(
5876 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5878 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5880 struct sctp_bind_hashbucket *head; /* hash list */
5881 struct sctp_bind_bucket *pp;
5882 unsigned short snum;
5883 int ret;
5885 snum = ntohs(addr->v4.sin_port);
5887 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5889 sctp_local_bh_disable();
5891 if (snum == 0) {
5892 /* Search for an available port. */
5893 int low, high, remaining, index;
5894 unsigned int rover;
5896 inet_get_local_port_range(&low, &high);
5897 remaining = (high - low) + 1;
5898 rover = net_random() % remaining + low;
5900 do {
5901 rover++;
5902 if ((rover < low) || (rover > high))
5903 rover = low;
5904 if (inet_is_reserved_local_port(rover))
5905 continue;
5906 index = sctp_phashfn(sock_net(sk), rover);
5907 head = &sctp_port_hashtable[index];
5908 sctp_spin_lock(&head->lock);
5909 sctp_for_each_hentry(pp, &head->chain)
5910 if ((pp->port == rover) &&
5911 net_eq(sock_net(sk), pp->net))
5912 goto next;
5913 break;
5914 next:
5915 sctp_spin_unlock(&head->lock);
5916 } while (--remaining > 0);
5918 /* Exhausted local port range during search? */
5919 ret = 1;
5920 if (remaining <= 0)
5921 goto fail;
5923 /* OK, here is the one we will use. HEAD (the port
5924 * hash table list entry) is non-NULL and we hold it's
5925 * mutex.
5927 snum = rover;
5928 } else {
5929 /* We are given an specific port number; we verify
5930 * that it is not being used. If it is used, we will
5931 * exahust the search in the hash list corresponding
5932 * to the port number (snum) - we detect that with the
5933 * port iterator, pp being NULL.
5935 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5936 sctp_spin_lock(&head->lock);
5937 sctp_for_each_hentry(pp, &head->chain) {
5938 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5939 goto pp_found;
5942 pp = NULL;
5943 goto pp_not_found;
5944 pp_found:
5945 if (!hlist_empty(&pp->owner)) {
5946 /* We had a port hash table hit - there is an
5947 * available port (pp != NULL) and it is being
5948 * used by other socket (pp->owner not empty); that other
5949 * socket is going to be sk2.
5951 int reuse = sk->sk_reuse;
5952 struct sock *sk2;
5954 pr_debug("%s: found a possible match\n", __func__);
5956 if (pp->fastreuse && sk->sk_reuse &&
5957 sk->sk_state != SCTP_SS_LISTENING)
5958 goto success;
5960 /* Run through the list of sockets bound to the port
5961 * (pp->port) [via the pointers bind_next and
5962 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5963 * we get the endpoint they describe and run through
5964 * the endpoint's list of IP (v4 or v6) addresses,
5965 * comparing each of the addresses with the address of
5966 * the socket sk. If we find a match, then that means
5967 * that this port/socket (sk) combination are already
5968 * in an endpoint.
5970 sk_for_each_bound(sk2, &pp->owner) {
5971 struct sctp_endpoint *ep2;
5972 ep2 = sctp_sk(sk2)->ep;
5974 if (sk == sk2 ||
5975 (reuse && sk2->sk_reuse &&
5976 sk2->sk_state != SCTP_SS_LISTENING))
5977 continue;
5979 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5980 sctp_sk(sk2), sctp_sk(sk))) {
5981 ret = (long)sk2;
5982 goto fail_unlock;
5986 pr_debug("%s: found a match\n", __func__);
5988 pp_not_found:
5989 /* If there was a hash table miss, create a new port. */
5990 ret = 1;
5991 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
5992 goto fail_unlock;
5994 /* In either case (hit or miss), make sure fastreuse is 1 only
5995 * if sk->sk_reuse is too (that is, if the caller requested
5996 * SO_REUSEADDR on this socket -sk-).
5998 if (hlist_empty(&pp->owner)) {
5999 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6000 pp->fastreuse = 1;
6001 else
6002 pp->fastreuse = 0;
6003 } else if (pp->fastreuse &&
6004 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6005 pp->fastreuse = 0;
6007 /* We are set, so fill up all the data in the hash table
6008 * entry, tie the socket list information with the rest of the
6009 * sockets FIXME: Blurry, NPI (ipg).
6011 success:
6012 if (!sctp_sk(sk)->bind_hash) {
6013 inet_sk(sk)->inet_num = snum;
6014 sk_add_bind_node(sk, &pp->owner);
6015 sctp_sk(sk)->bind_hash = pp;
6017 ret = 0;
6019 fail_unlock:
6020 sctp_spin_unlock(&head->lock);
6022 fail:
6023 sctp_local_bh_enable();
6024 return ret;
6027 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6028 * port is requested.
6030 static int sctp_get_port(struct sock *sk, unsigned short snum)
6032 union sctp_addr addr;
6033 struct sctp_af *af = sctp_sk(sk)->pf->af;
6035 /* Set up a dummy address struct from the sk. */
6036 af->from_sk(&addr, sk);
6037 addr.v4.sin_port = htons(snum);
6039 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6040 return !!sctp_get_port_local(sk, &addr);
6044 * Move a socket to LISTENING state.
6046 static int sctp_listen_start(struct sock *sk, int backlog)
6048 struct sctp_sock *sp = sctp_sk(sk);
6049 struct sctp_endpoint *ep = sp->ep;
6050 struct crypto_hash *tfm = NULL;
6051 char alg[32];
6053 /* Allocate HMAC for generating cookie. */
6054 if (!sp->hmac && sp->sctp_hmac_alg) {
6055 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6056 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6057 if (IS_ERR(tfm)) {
6058 net_info_ratelimited("failed to load transform for %s: %ld\n",
6059 sp->sctp_hmac_alg, PTR_ERR(tfm));
6060 return -ENOSYS;
6062 sctp_sk(sk)->hmac = tfm;
6066 * If a bind() or sctp_bindx() is not called prior to a listen()
6067 * call that allows new associations to be accepted, the system
6068 * picks an ephemeral port and will choose an address set equivalent
6069 * to binding with a wildcard address.
6071 * This is not currently spelled out in the SCTP sockets
6072 * extensions draft, but follows the practice as seen in TCP
6073 * sockets.
6076 sk->sk_state = SCTP_SS_LISTENING;
6077 if (!ep->base.bind_addr.port) {
6078 if (sctp_autobind(sk))
6079 return -EAGAIN;
6080 } else {
6081 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6082 sk->sk_state = SCTP_SS_CLOSED;
6083 return -EADDRINUSE;
6087 sk->sk_max_ack_backlog = backlog;
6088 sctp_hash_endpoint(ep);
6089 return 0;
6093 * 4.1.3 / 5.1.3 listen()
6095 * By default, new associations are not accepted for UDP style sockets.
6096 * An application uses listen() to mark a socket as being able to
6097 * accept new associations.
6099 * On TCP style sockets, applications use listen() to ready the SCTP
6100 * endpoint for accepting inbound associations.
6102 * On both types of endpoints a backlog of '0' disables listening.
6104 * Move a socket to LISTENING state.
6106 int sctp_inet_listen(struct socket *sock, int backlog)
6108 struct sock *sk = sock->sk;
6109 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6110 int err = -EINVAL;
6112 if (unlikely(backlog < 0))
6113 return err;
6115 sctp_lock_sock(sk);
6117 /* Peeled-off sockets are not allowed to listen(). */
6118 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6119 goto out;
6121 if (sock->state != SS_UNCONNECTED)
6122 goto out;
6124 /* If backlog is zero, disable listening. */
6125 if (!backlog) {
6126 if (sctp_sstate(sk, CLOSED))
6127 goto out;
6129 err = 0;
6130 sctp_unhash_endpoint(ep);
6131 sk->sk_state = SCTP_SS_CLOSED;
6132 if (sk->sk_reuse)
6133 sctp_sk(sk)->bind_hash->fastreuse = 1;
6134 goto out;
6137 /* If we are already listening, just update the backlog */
6138 if (sctp_sstate(sk, LISTENING))
6139 sk->sk_max_ack_backlog = backlog;
6140 else {
6141 err = sctp_listen_start(sk, backlog);
6142 if (err)
6143 goto out;
6146 err = 0;
6147 out:
6148 sctp_release_sock(sk);
6149 return err;
6153 * This function is done by modeling the current datagram_poll() and the
6154 * tcp_poll(). Note that, based on these implementations, we don't
6155 * lock the socket in this function, even though it seems that,
6156 * ideally, locking or some other mechanisms can be used to ensure
6157 * the integrity of the counters (sndbuf and wmem_alloc) used
6158 * in this place. We assume that we don't need locks either until proven
6159 * otherwise.
6161 * Another thing to note is that we include the Async I/O support
6162 * here, again, by modeling the current TCP/UDP code. We don't have
6163 * a good way to test with it yet.
6165 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6167 struct sock *sk = sock->sk;
6168 struct sctp_sock *sp = sctp_sk(sk);
6169 unsigned int mask;
6171 poll_wait(file, sk_sleep(sk), wait);
6173 /* A TCP-style listening socket becomes readable when the accept queue
6174 * is not empty.
6176 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6177 return (!list_empty(&sp->ep->asocs)) ?
6178 (POLLIN | POLLRDNORM) : 0;
6180 mask = 0;
6182 /* Is there any exceptional events? */
6183 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6184 mask |= POLLERR |
6185 sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0;
6186 if (sk->sk_shutdown & RCV_SHUTDOWN)
6187 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6188 if (sk->sk_shutdown == SHUTDOWN_MASK)
6189 mask |= POLLHUP;
6191 /* Is it readable? Reconsider this code with TCP-style support. */
6192 if (!skb_queue_empty(&sk->sk_receive_queue))
6193 mask |= POLLIN | POLLRDNORM;
6195 /* The association is either gone or not ready. */
6196 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6197 return mask;
6199 /* Is it writable? */
6200 if (sctp_writeable(sk)) {
6201 mask |= POLLOUT | POLLWRNORM;
6202 } else {
6203 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6205 * Since the socket is not locked, the buffer
6206 * might be made available after the writeable check and
6207 * before the bit is set. This could cause a lost I/O
6208 * signal. tcp_poll() has a race breaker for this race
6209 * condition. Based on their implementation, we put
6210 * in the following code to cover it as well.
6212 if (sctp_writeable(sk))
6213 mask |= POLLOUT | POLLWRNORM;
6215 return mask;
6218 /********************************************************************
6219 * 2nd Level Abstractions
6220 ********************************************************************/
6222 static struct sctp_bind_bucket *sctp_bucket_create(
6223 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6225 struct sctp_bind_bucket *pp;
6227 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6228 if (pp) {
6229 SCTP_DBG_OBJCNT_INC(bind_bucket);
6230 pp->port = snum;
6231 pp->fastreuse = 0;
6232 INIT_HLIST_HEAD(&pp->owner);
6233 pp->net = net;
6234 hlist_add_head(&pp->node, &head->chain);
6236 return pp;
6239 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6240 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6242 if (pp && hlist_empty(&pp->owner)) {
6243 __hlist_del(&pp->node);
6244 kmem_cache_free(sctp_bucket_cachep, pp);
6245 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6249 /* Release this socket's reference to a local port. */
6250 static inline void __sctp_put_port(struct sock *sk)
6252 struct sctp_bind_hashbucket *head =
6253 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6254 inet_sk(sk)->inet_num)];
6255 struct sctp_bind_bucket *pp;
6257 sctp_spin_lock(&head->lock);
6258 pp = sctp_sk(sk)->bind_hash;
6259 __sk_del_bind_node(sk);
6260 sctp_sk(sk)->bind_hash = NULL;
6261 inet_sk(sk)->inet_num = 0;
6262 sctp_bucket_destroy(pp);
6263 sctp_spin_unlock(&head->lock);
6266 void sctp_put_port(struct sock *sk)
6268 sctp_local_bh_disable();
6269 __sctp_put_port(sk);
6270 sctp_local_bh_enable();
6274 * The system picks an ephemeral port and choose an address set equivalent
6275 * to binding with a wildcard address.
6276 * One of those addresses will be the primary address for the association.
6277 * This automatically enables the multihoming capability of SCTP.
6279 static int sctp_autobind(struct sock *sk)
6281 union sctp_addr autoaddr;
6282 struct sctp_af *af;
6283 __be16 port;
6285 /* Initialize a local sockaddr structure to INADDR_ANY. */
6286 af = sctp_sk(sk)->pf->af;
6288 port = htons(inet_sk(sk)->inet_num);
6289 af->inaddr_any(&autoaddr, port);
6291 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6294 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6296 * From RFC 2292
6297 * 4.2 The cmsghdr Structure *
6299 * When ancillary data is sent or received, any number of ancillary data
6300 * objects can be specified by the msg_control and msg_controllen members of
6301 * the msghdr structure, because each object is preceded by
6302 * a cmsghdr structure defining the object's length (the cmsg_len member).
6303 * Historically Berkeley-derived implementations have passed only one object
6304 * at a time, but this API allows multiple objects to be
6305 * passed in a single call to sendmsg() or recvmsg(). The following example
6306 * shows two ancillary data objects in a control buffer.
6308 * |<--------------------------- msg_controllen -------------------------->|
6309 * | |
6311 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6313 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6314 * | | |
6316 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6318 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6319 * | | | | |
6321 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6322 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6324 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6326 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6330 * msg_control
6331 * points here
6333 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6335 struct cmsghdr *cmsg;
6336 struct msghdr *my_msg = (struct msghdr *)msg;
6338 for (cmsg = CMSG_FIRSTHDR(msg);
6339 cmsg != NULL;
6340 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6341 if (!CMSG_OK(my_msg, cmsg))
6342 return -EINVAL;
6344 /* Should we parse this header or ignore? */
6345 if (cmsg->cmsg_level != IPPROTO_SCTP)
6346 continue;
6348 /* Strictly check lengths following example in SCM code. */
6349 switch (cmsg->cmsg_type) {
6350 case SCTP_INIT:
6351 /* SCTP Socket API Extension
6352 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6354 * This cmsghdr structure provides information for
6355 * initializing new SCTP associations with sendmsg().
6356 * The SCTP_INITMSG socket option uses this same data
6357 * structure. This structure is not used for
6358 * recvmsg().
6360 * cmsg_level cmsg_type cmsg_data[]
6361 * ------------ ------------ ----------------------
6362 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6364 if (cmsg->cmsg_len !=
6365 CMSG_LEN(sizeof(struct sctp_initmsg)))
6366 return -EINVAL;
6367 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6368 break;
6370 case SCTP_SNDRCV:
6371 /* SCTP Socket API Extension
6372 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6374 * This cmsghdr structure specifies SCTP options for
6375 * sendmsg() and describes SCTP header information
6376 * about a received message through recvmsg().
6378 * cmsg_level cmsg_type cmsg_data[]
6379 * ------------ ------------ ----------------------
6380 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6382 if (cmsg->cmsg_len !=
6383 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6384 return -EINVAL;
6386 cmsgs->info =
6387 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6389 /* Minimally, validate the sinfo_flags. */
6390 if (cmsgs->info->sinfo_flags &
6391 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6392 SCTP_ABORT | SCTP_EOF))
6393 return -EINVAL;
6394 break;
6396 default:
6397 return -EINVAL;
6400 return 0;
6404 * Wait for a packet..
6405 * Note: This function is the same function as in core/datagram.c
6406 * with a few modifications to make lksctp work.
6408 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6410 int error;
6411 DEFINE_WAIT(wait);
6413 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6415 /* Socket errors? */
6416 error = sock_error(sk);
6417 if (error)
6418 goto out;
6420 if (!skb_queue_empty(&sk->sk_receive_queue))
6421 goto ready;
6423 /* Socket shut down? */
6424 if (sk->sk_shutdown & RCV_SHUTDOWN)
6425 goto out;
6427 /* Sequenced packets can come disconnected. If so we report the
6428 * problem.
6430 error = -ENOTCONN;
6432 /* Is there a good reason to think that we may receive some data? */
6433 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6434 goto out;
6436 /* Handle signals. */
6437 if (signal_pending(current))
6438 goto interrupted;
6440 /* Let another process have a go. Since we are going to sleep
6441 * anyway. Note: This may cause odd behaviors if the message
6442 * does not fit in the user's buffer, but this seems to be the
6443 * only way to honor MSG_DONTWAIT realistically.
6445 sctp_release_sock(sk);
6446 *timeo_p = schedule_timeout(*timeo_p);
6447 sctp_lock_sock(sk);
6449 ready:
6450 finish_wait(sk_sleep(sk), &wait);
6451 return 0;
6453 interrupted:
6454 error = sock_intr_errno(*timeo_p);
6456 out:
6457 finish_wait(sk_sleep(sk), &wait);
6458 *err = error;
6459 return error;
6462 /* Receive a datagram.
6463 * Note: This is pretty much the same routine as in core/datagram.c
6464 * with a few changes to make lksctp work.
6466 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6467 int noblock, int *err)
6469 int error;
6470 struct sk_buff *skb;
6471 long timeo;
6473 timeo = sock_rcvtimeo(sk, noblock);
6475 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6476 MAX_SCHEDULE_TIMEOUT);
6478 do {
6479 /* Again only user level code calls this function,
6480 * so nothing interrupt level
6481 * will suddenly eat the receive_queue.
6483 * Look at current nfs client by the way...
6484 * However, this function was correct in any case. 8)
6486 if (flags & MSG_PEEK) {
6487 spin_lock_bh(&sk->sk_receive_queue.lock);
6488 skb = skb_peek(&sk->sk_receive_queue);
6489 if (skb)
6490 atomic_inc(&skb->users);
6491 spin_unlock_bh(&sk->sk_receive_queue.lock);
6492 } else {
6493 skb = skb_dequeue(&sk->sk_receive_queue);
6496 if (skb)
6497 return skb;
6499 /* Caller is allowed not to check sk->sk_err before calling. */
6500 error = sock_error(sk);
6501 if (error)
6502 goto no_packet;
6504 if (sk->sk_shutdown & RCV_SHUTDOWN)
6505 break;
6507 /* User doesn't want to wait. */
6508 error = -EAGAIN;
6509 if (!timeo)
6510 goto no_packet;
6511 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6513 return NULL;
6515 no_packet:
6516 *err = error;
6517 return NULL;
6520 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6521 static void __sctp_write_space(struct sctp_association *asoc)
6523 struct sock *sk = asoc->base.sk;
6524 struct socket *sock = sk->sk_socket;
6526 if ((sctp_wspace(asoc) > 0) && sock) {
6527 if (waitqueue_active(&asoc->wait))
6528 wake_up_interruptible(&asoc->wait);
6530 if (sctp_writeable(sk)) {
6531 wait_queue_head_t *wq = sk_sleep(sk);
6533 if (wq && waitqueue_active(wq))
6534 wake_up_interruptible(wq);
6536 /* Note that we try to include the Async I/O support
6537 * here by modeling from the current TCP/UDP code.
6538 * We have not tested with it yet.
6540 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6541 sock_wake_async(sock,
6542 SOCK_WAKE_SPACE, POLL_OUT);
6547 /* Do accounting for the sndbuf space.
6548 * Decrement the used sndbuf space of the corresponding association by the
6549 * data size which was just transmitted(freed).
6551 static void sctp_wfree(struct sk_buff *skb)
6553 struct sctp_association *asoc;
6554 struct sctp_chunk *chunk;
6555 struct sock *sk;
6557 /* Get the saved chunk pointer. */
6558 chunk = *((struct sctp_chunk **)(skb->cb));
6559 asoc = chunk->asoc;
6560 sk = asoc->base.sk;
6561 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6562 sizeof(struct sk_buff) +
6563 sizeof(struct sctp_chunk);
6565 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6568 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6570 sk->sk_wmem_queued -= skb->truesize;
6571 sk_mem_uncharge(sk, skb->truesize);
6573 sock_wfree(skb);
6574 __sctp_write_space(asoc);
6576 sctp_association_put(asoc);
6579 /* Do accounting for the receive space on the socket.
6580 * Accounting for the association is done in ulpevent.c
6581 * We set this as a destructor for the cloned data skbs so that
6582 * accounting is done at the correct time.
6584 void sctp_sock_rfree(struct sk_buff *skb)
6586 struct sock *sk = skb->sk;
6587 struct sctp_ulpevent *event = sctp_skb2event(skb);
6589 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6592 * Mimic the behavior of sock_rfree
6594 sk_mem_uncharge(sk, event->rmem_len);
6598 /* Helper function to wait for space in the sndbuf. */
6599 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6600 size_t msg_len)
6602 struct sock *sk = asoc->base.sk;
6603 int err = 0;
6604 long current_timeo = *timeo_p;
6605 DEFINE_WAIT(wait);
6607 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6608 *timeo_p, msg_len);
6610 /* Increment the association's refcnt. */
6611 sctp_association_hold(asoc);
6613 /* Wait on the association specific sndbuf space. */
6614 for (;;) {
6615 prepare_to_wait_exclusive(&asoc->wait, &wait,
6616 TASK_INTERRUPTIBLE);
6617 if (!*timeo_p)
6618 goto do_nonblock;
6619 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6620 asoc->base.dead)
6621 goto do_error;
6622 if (signal_pending(current))
6623 goto do_interrupted;
6624 if (msg_len <= sctp_wspace(asoc))
6625 break;
6627 /* Let another process have a go. Since we are going
6628 * to sleep anyway.
6630 sctp_release_sock(sk);
6631 current_timeo = schedule_timeout(current_timeo);
6632 BUG_ON(sk != asoc->base.sk);
6633 sctp_lock_sock(sk);
6635 *timeo_p = current_timeo;
6638 out:
6639 finish_wait(&asoc->wait, &wait);
6641 /* Release the association's refcnt. */
6642 sctp_association_put(asoc);
6644 return err;
6646 do_error:
6647 err = -EPIPE;
6648 goto out;
6650 do_interrupted:
6651 err = sock_intr_errno(*timeo_p);
6652 goto out;
6654 do_nonblock:
6655 err = -EAGAIN;
6656 goto out;
6659 void sctp_data_ready(struct sock *sk, int len)
6661 struct socket_wq *wq;
6663 rcu_read_lock();
6664 wq = rcu_dereference(sk->sk_wq);
6665 if (wq_has_sleeper(wq))
6666 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6667 POLLRDNORM | POLLRDBAND);
6668 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6669 rcu_read_unlock();
6672 /* If socket sndbuf has changed, wake up all per association waiters. */
6673 void sctp_write_space(struct sock *sk)
6675 struct sctp_association *asoc;
6677 /* Wake up the tasks in each wait queue. */
6678 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6679 __sctp_write_space(asoc);
6683 /* Is there any sndbuf space available on the socket?
6685 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6686 * associations on the same socket. For a UDP-style socket with
6687 * multiple associations, it is possible for it to be "unwriteable"
6688 * prematurely. I assume that this is acceptable because
6689 * a premature "unwriteable" is better than an accidental "writeable" which
6690 * would cause an unwanted block under certain circumstances. For the 1-1
6691 * UDP-style sockets or TCP-style sockets, this code should work.
6692 * - Daisy
6694 static int sctp_writeable(struct sock *sk)
6696 int amt = 0;
6698 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6699 if (amt < 0)
6700 amt = 0;
6701 return amt;
6704 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6705 * returns immediately with EINPROGRESS.
6707 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6709 struct sock *sk = asoc->base.sk;
6710 int err = 0;
6711 long current_timeo = *timeo_p;
6712 DEFINE_WAIT(wait);
6714 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6716 /* Increment the association's refcnt. */
6717 sctp_association_hold(asoc);
6719 for (;;) {
6720 prepare_to_wait_exclusive(&asoc->wait, &wait,
6721 TASK_INTERRUPTIBLE);
6722 if (!*timeo_p)
6723 goto do_nonblock;
6724 if (sk->sk_shutdown & RCV_SHUTDOWN)
6725 break;
6726 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6727 asoc->base.dead)
6728 goto do_error;
6729 if (signal_pending(current))
6730 goto do_interrupted;
6732 if (sctp_state(asoc, ESTABLISHED))
6733 break;
6735 /* Let another process have a go. Since we are going
6736 * to sleep anyway.
6738 sctp_release_sock(sk);
6739 current_timeo = schedule_timeout(current_timeo);
6740 sctp_lock_sock(sk);
6742 *timeo_p = current_timeo;
6745 out:
6746 finish_wait(&asoc->wait, &wait);
6748 /* Release the association's refcnt. */
6749 sctp_association_put(asoc);
6751 return err;
6753 do_error:
6754 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6755 err = -ETIMEDOUT;
6756 else
6757 err = -ECONNREFUSED;
6758 goto out;
6760 do_interrupted:
6761 err = sock_intr_errno(*timeo_p);
6762 goto out;
6764 do_nonblock:
6765 err = -EINPROGRESS;
6766 goto out;
6769 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6771 struct sctp_endpoint *ep;
6772 int err = 0;
6773 DEFINE_WAIT(wait);
6775 ep = sctp_sk(sk)->ep;
6778 for (;;) {
6779 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6780 TASK_INTERRUPTIBLE);
6782 if (list_empty(&ep->asocs)) {
6783 sctp_release_sock(sk);
6784 timeo = schedule_timeout(timeo);
6785 sctp_lock_sock(sk);
6788 err = -EINVAL;
6789 if (!sctp_sstate(sk, LISTENING))
6790 break;
6792 err = 0;
6793 if (!list_empty(&ep->asocs))
6794 break;
6796 err = sock_intr_errno(timeo);
6797 if (signal_pending(current))
6798 break;
6800 err = -EAGAIN;
6801 if (!timeo)
6802 break;
6805 finish_wait(sk_sleep(sk), &wait);
6807 return err;
6810 static void sctp_wait_for_close(struct sock *sk, long timeout)
6812 DEFINE_WAIT(wait);
6814 do {
6815 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6816 if (list_empty(&sctp_sk(sk)->ep->asocs))
6817 break;
6818 sctp_release_sock(sk);
6819 timeout = schedule_timeout(timeout);
6820 sctp_lock_sock(sk);
6821 } while (!signal_pending(current) && timeout);
6823 finish_wait(sk_sleep(sk), &wait);
6826 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6828 struct sk_buff *frag;
6830 if (!skb->data_len)
6831 goto done;
6833 /* Don't forget the fragments. */
6834 skb_walk_frags(skb, frag)
6835 sctp_skb_set_owner_r_frag(frag, sk);
6837 done:
6838 sctp_skb_set_owner_r(skb, sk);
6841 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6842 struct sctp_association *asoc)
6844 struct inet_sock *inet = inet_sk(sk);
6845 struct inet_sock *newinet;
6847 newsk->sk_type = sk->sk_type;
6848 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6849 newsk->sk_flags = sk->sk_flags;
6850 newsk->sk_no_check = sk->sk_no_check;
6851 newsk->sk_reuse = sk->sk_reuse;
6853 newsk->sk_shutdown = sk->sk_shutdown;
6854 newsk->sk_destruct = sctp_destruct_sock;
6855 newsk->sk_family = sk->sk_family;
6856 newsk->sk_protocol = IPPROTO_SCTP;
6857 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6858 newsk->sk_sndbuf = sk->sk_sndbuf;
6859 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6860 newsk->sk_lingertime = sk->sk_lingertime;
6861 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6862 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6864 newinet = inet_sk(newsk);
6866 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6867 * getsockname() and getpeername()
6869 newinet->inet_sport = inet->inet_sport;
6870 newinet->inet_saddr = inet->inet_saddr;
6871 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6872 newinet->inet_dport = htons(asoc->peer.port);
6873 newinet->pmtudisc = inet->pmtudisc;
6874 newinet->inet_id = asoc->next_tsn ^ jiffies;
6876 newinet->uc_ttl = inet->uc_ttl;
6877 newinet->mc_loop = 1;
6878 newinet->mc_ttl = 1;
6879 newinet->mc_index = 0;
6880 newinet->mc_list = NULL;
6883 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6884 * and its messages to the newsk.
6886 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6887 struct sctp_association *assoc,
6888 sctp_socket_type_t type)
6890 struct sctp_sock *oldsp = sctp_sk(oldsk);
6891 struct sctp_sock *newsp = sctp_sk(newsk);
6892 struct sctp_bind_bucket *pp; /* hash list port iterator */
6893 struct sctp_endpoint *newep = newsp->ep;
6894 struct sk_buff *skb, *tmp;
6895 struct sctp_ulpevent *event;
6896 struct sctp_bind_hashbucket *head;
6897 struct list_head tmplist;
6899 /* Migrate socket buffer sizes and all the socket level options to the
6900 * new socket.
6902 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6903 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6904 /* Brute force copy old sctp opt. */
6905 if (oldsp->do_auto_asconf) {
6906 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6907 inet_sk_copy_descendant(newsk, oldsk);
6908 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6909 } else
6910 inet_sk_copy_descendant(newsk, oldsk);
6912 /* Restore the ep value that was overwritten with the above structure
6913 * copy.
6915 newsp->ep = newep;
6916 newsp->hmac = NULL;
6918 /* Hook this new socket in to the bind_hash list. */
6919 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
6920 inet_sk(oldsk)->inet_num)];
6921 sctp_local_bh_disable();
6922 sctp_spin_lock(&head->lock);
6923 pp = sctp_sk(oldsk)->bind_hash;
6924 sk_add_bind_node(newsk, &pp->owner);
6925 sctp_sk(newsk)->bind_hash = pp;
6926 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6927 sctp_spin_unlock(&head->lock);
6928 sctp_local_bh_enable();
6930 /* Copy the bind_addr list from the original endpoint to the new
6931 * endpoint so that we can handle restarts properly
6933 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6934 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6936 /* Move any messages in the old socket's receive queue that are for the
6937 * peeled off association to the new socket's receive queue.
6939 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6940 event = sctp_skb2event(skb);
6941 if (event->asoc == assoc) {
6942 __skb_unlink(skb, &oldsk->sk_receive_queue);
6943 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6944 sctp_skb_set_owner_r_frag(skb, newsk);
6948 /* Clean up any messages pending delivery due to partial
6949 * delivery. Three cases:
6950 * 1) No partial deliver; no work.
6951 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6952 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6954 skb_queue_head_init(&newsp->pd_lobby);
6955 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6957 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6958 struct sk_buff_head *queue;
6960 /* Decide which queue to move pd_lobby skbs to. */
6961 if (assoc->ulpq.pd_mode) {
6962 queue = &newsp->pd_lobby;
6963 } else
6964 queue = &newsk->sk_receive_queue;
6966 /* Walk through the pd_lobby, looking for skbs that
6967 * need moved to the new socket.
6969 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6970 event = sctp_skb2event(skb);
6971 if (event->asoc == assoc) {
6972 __skb_unlink(skb, &oldsp->pd_lobby);
6973 __skb_queue_tail(queue, skb);
6974 sctp_skb_set_owner_r_frag(skb, newsk);
6978 /* Clear up any skbs waiting for the partial
6979 * delivery to finish.
6981 if (assoc->ulpq.pd_mode)
6982 sctp_clear_pd(oldsk, NULL);
6986 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6987 sctp_skb_set_owner_r_frag(skb, newsk);
6989 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6990 sctp_skb_set_owner_r_frag(skb, newsk);
6992 /* Set the type of socket to indicate that it is peeled off from the
6993 * original UDP-style socket or created with the accept() call on a
6994 * TCP-style socket..
6996 newsp->type = type;
6998 /* Mark the new socket "in-use" by the user so that any packets
6999 * that may arrive on the association after we've moved it are
7000 * queued to the backlog. This prevents a potential race between
7001 * backlog processing on the old socket and new-packet processing
7002 * on the new socket.
7004 * The caller has just allocated newsk so we can guarantee that other
7005 * paths won't try to lock it and then oldsk.
7007 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7008 sctp_assoc_migrate(assoc, newsk);
7010 /* If the association on the newsk is already closed before accept()
7011 * is called, set RCV_SHUTDOWN flag.
7013 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7014 newsk->sk_shutdown |= RCV_SHUTDOWN;
7016 newsk->sk_state = SCTP_SS_ESTABLISHED;
7017 sctp_release_sock(newsk);
7021 /* This proto struct describes the ULP interface for SCTP. */
7022 struct proto sctp_prot = {
7023 .name = "SCTP",
7024 .owner = THIS_MODULE,
7025 .close = sctp_close,
7026 .connect = sctp_connect,
7027 .disconnect = sctp_disconnect,
7028 .accept = sctp_accept,
7029 .ioctl = sctp_ioctl,
7030 .init = sctp_init_sock,
7031 .destroy = sctp_destroy_sock,
7032 .shutdown = sctp_shutdown,
7033 .setsockopt = sctp_setsockopt,
7034 .getsockopt = sctp_getsockopt,
7035 .sendmsg = sctp_sendmsg,
7036 .recvmsg = sctp_recvmsg,
7037 .bind = sctp_bind,
7038 .backlog_rcv = sctp_backlog_rcv,
7039 .hash = sctp_hash,
7040 .unhash = sctp_unhash,
7041 .get_port = sctp_get_port,
7042 .obj_size = sizeof(struct sctp_sock),
7043 .sysctl_mem = sysctl_sctp_mem,
7044 .sysctl_rmem = sysctl_sctp_rmem,
7045 .sysctl_wmem = sysctl_sctp_wmem,
7046 .memory_pressure = &sctp_memory_pressure,
7047 .enter_memory_pressure = sctp_enter_memory_pressure,
7048 .memory_allocated = &sctp_memory_allocated,
7049 .sockets_allocated = &sctp_sockets_allocated,
7052 #if IS_ENABLED(CONFIG_IPV6)
7054 struct proto sctpv6_prot = {
7055 .name = "SCTPv6",
7056 .owner = THIS_MODULE,
7057 .close = sctp_close,
7058 .connect = sctp_connect,
7059 .disconnect = sctp_disconnect,
7060 .accept = sctp_accept,
7061 .ioctl = sctp_ioctl,
7062 .init = sctp_init_sock,
7063 .destroy = sctp_destroy_sock,
7064 .shutdown = sctp_shutdown,
7065 .setsockopt = sctp_setsockopt,
7066 .getsockopt = sctp_getsockopt,
7067 .sendmsg = sctp_sendmsg,
7068 .recvmsg = sctp_recvmsg,
7069 .bind = sctp_bind,
7070 .backlog_rcv = sctp_backlog_rcv,
7071 .hash = sctp_hash,
7072 .unhash = sctp_unhash,
7073 .get_port = sctp_get_port,
7074 .obj_size = sizeof(struct sctp6_sock),
7075 .sysctl_mem = sysctl_sctp_mem,
7076 .sysctl_rmem = sysctl_sctp_rmem,
7077 .sysctl_wmem = sysctl_sctp_wmem,
7078 .memory_pressure = &sctp_memory_pressure,
7079 .enter_memory_pressure = sctp_enter_memory_pressure,
7080 .memory_allocated = &sctp_memory_allocated,
7081 .sockets_allocated = &sctp_sockets_allocated,
7083 #endif /* IS_ENABLED(CONFIG_IPV6) */