memcg: fix prepare migration
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / sctp / socket.c
blob007e8baba0891c1c5b52e6e8bc2fadcc6edd96e5
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 #include <linux/types.h>
61 #include <linux/kernel.h>
62 #include <linux/wait.h>
63 #include <linux/time.h>
64 #include <linux/ip.h>
65 #include <linux/capability.h>
66 #include <linux/fcntl.h>
67 #include <linux/poll.h>
68 #include <linux/init.h>
69 #include <linux/crypto.h>
70 #include <linux/slab.h>
72 #include <net/ip.h>
73 #include <net/icmp.h>
74 #include <net/route.h>
75 #include <net/ipv6.h>
76 #include <net/inet_common.h>
78 #include <linux/socket.h> /* for sa_family_t */
79 #include <net/sock.h>
80 #include <net/sctp/sctp.h>
81 #include <net/sctp/sm.h>
83 /* WARNING: Please do not remove the SCTP_STATIC attribute to
84 * any of the functions below as they are used to export functions
85 * used by a project regression testsuite.
88 /* Forward declarations for internal helper functions. */
89 static int sctp_writeable(struct sock *sk);
90 static void sctp_wfree(struct sk_buff *skb);
91 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
92 size_t msg_len);
93 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
94 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
95 static int sctp_wait_for_accept(struct sock *sk, long timeo);
96 static void sctp_wait_for_close(struct sock *sk, long timeo);
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);
109 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
111 extern struct kmem_cache *sctp_bucket_cachep;
112 extern int sysctl_sctp_mem[3];
113 extern int sysctl_sctp_rmem[3];
114 extern int sysctl_sctp_wmem[3];
116 static int sctp_memory_pressure;
117 static atomic_t sctp_memory_allocated;
118 struct percpu_counter sctp_sockets_allocated;
120 static void sctp_enter_memory_pressure(struct sock *sk)
122 sctp_memory_pressure = 1;
126 /* Get the sndbuf space available at the time on the association. */
127 static inline int sctp_wspace(struct sctp_association *asoc)
129 int amt;
131 if (asoc->ep->sndbuf_policy)
132 amt = asoc->sndbuf_used;
133 else
134 amt = sk_wmem_alloc_get(asoc->base.sk);
136 if (amt >= asoc->base.sk->sk_sndbuf) {
137 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
138 amt = 0;
139 else {
140 amt = sk_stream_wspace(asoc->base.sk);
141 if (amt < 0)
142 amt = 0;
144 } else {
145 amt = asoc->base.sk->sk_sndbuf - amt;
147 return amt;
150 /* Increment the used sndbuf space count of the corresponding association by
151 * the size of the outgoing data chunk.
152 * Also, set the skb destructor for sndbuf accounting later.
154 * Since it is always 1-1 between chunk and skb, and also a new skb is always
155 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
156 * destructor in the data chunk skb for the purpose of the sndbuf space
157 * tracking.
159 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
161 struct sctp_association *asoc = chunk->asoc;
162 struct sock *sk = asoc->base.sk;
164 /* The sndbuf space is tracked per association. */
165 sctp_association_hold(asoc);
167 skb_set_owner_w(chunk->skb, sk);
169 chunk->skb->destructor = sctp_wfree;
170 /* Save the chunk pointer in skb for sctp_wfree to use later. */
171 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
173 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
174 sizeof(struct sk_buff) +
175 sizeof(struct sctp_chunk);
177 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
178 sk->sk_wmem_queued += chunk->skb->truesize;
179 sk_mem_charge(sk, chunk->skb->truesize);
182 /* Verify that this is a valid address. */
183 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
184 int len)
186 struct sctp_af *af;
188 /* Verify basic sockaddr. */
189 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
190 if (!af)
191 return -EINVAL;
193 /* Is this a valid SCTP address? */
194 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
195 return -EINVAL;
197 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
198 return -EINVAL;
200 return 0;
203 /* Look up the association by its id. If this is not a UDP-style
204 * socket, the ID field is always ignored.
206 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
208 struct sctp_association *asoc = NULL;
210 /* If this is not a UDP-style socket, assoc id should be ignored. */
211 if (!sctp_style(sk, UDP)) {
212 /* Return NULL if the socket state is not ESTABLISHED. It
213 * could be a TCP-style listening socket or a socket which
214 * hasn't yet called connect() to establish an association.
216 if (!sctp_sstate(sk, ESTABLISHED))
217 return NULL;
219 /* Get the first and the only association from the list. */
220 if (!list_empty(&sctp_sk(sk)->ep->asocs))
221 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
222 struct sctp_association, asocs);
223 return asoc;
226 /* Otherwise this is a UDP-style socket. */
227 if (!id || (id == (sctp_assoc_t)-1))
228 return NULL;
230 spin_lock_bh(&sctp_assocs_id_lock);
231 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
232 spin_unlock_bh(&sctp_assocs_id_lock);
234 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
235 return NULL;
237 return asoc;
240 /* Look up the transport from an address and an assoc id. If both address and
241 * id are specified, the associations matching the address and the id should be
242 * the same.
244 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
245 struct sockaddr_storage *addr,
246 sctp_assoc_t id)
248 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
249 struct sctp_transport *transport;
250 union sctp_addr *laddr = (union sctp_addr *)addr;
252 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
253 laddr,
254 &transport);
256 if (!addr_asoc)
257 return NULL;
259 id_asoc = sctp_id2assoc(sk, id);
260 if (id_asoc && (id_asoc != addr_asoc))
261 return NULL;
263 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
264 (union sctp_addr *)addr);
266 return transport;
269 /* API 3.1.2 bind() - UDP Style Syntax
270 * The syntax of bind() is,
272 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
274 * sd - the socket descriptor returned by socket().
275 * addr - the address structure (struct sockaddr_in or struct
276 * sockaddr_in6 [RFC 2553]),
277 * addr_len - the size of the address structure.
279 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
281 int retval = 0;
283 sctp_lock_sock(sk);
285 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
286 sk, addr, addr_len);
288 /* Disallow binding twice. */
289 if (!sctp_sk(sk)->ep->base.bind_addr.port)
290 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
291 addr_len);
292 else
293 retval = -EINVAL;
295 sctp_release_sock(sk);
297 return retval;
300 static long sctp_get_port_local(struct sock *, union sctp_addr *);
302 /* Verify this is a valid sockaddr. */
303 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
304 union sctp_addr *addr, int len)
306 struct sctp_af *af;
308 /* Check minimum size. */
309 if (len < sizeof (struct sockaddr))
310 return NULL;
312 /* V4 mapped address are really of AF_INET family */
313 if (addr->sa.sa_family == AF_INET6 &&
314 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
315 if (!opt->pf->af_supported(AF_INET, opt))
316 return NULL;
317 } else {
318 /* Does this PF support this AF? */
319 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
320 return NULL;
323 /* If we get this far, af is valid. */
324 af = sctp_get_af_specific(addr->sa.sa_family);
326 if (len < af->sockaddr_len)
327 return NULL;
329 return af;
332 /* Bind a local address either to an endpoint or to an association. */
333 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
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 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
346 sk, addr, len);
347 return -EINVAL;
350 snum = ntohs(addr->v4.sin_port);
352 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
353 ", port: %d, new port: %d, len: %d)\n",
355 addr,
356 bp->port, snum,
357 len);
359 /* PF specific bind() address verification. */
360 if (!sp->pf->bind_verify(sp, addr))
361 return -EADDRNOTAVAIL;
363 /* We must either be unbound, or bind to the same port.
364 * It's OK to allow 0 ports if we are already bound.
365 * We'll just inhert an already bound port in this case
367 if (bp->port) {
368 if (!snum)
369 snum = bp->port;
370 else if (snum != bp->port) {
371 SCTP_DEBUG_PRINTK("sctp_do_bind:"
372 " New port %d does not match existing port "
373 "%d.\n", snum, bp->port);
374 return -EINVAL;
378 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
379 return -EACCES;
381 /* See if the address matches any of the addresses we may have
382 * already bound before checking against other endpoints.
384 if (sctp_bind_addr_match(bp, addr, sp))
385 return -EINVAL;
387 /* Make sure we are allowed to bind here.
388 * The function sctp_get_port_local() does duplicate address
389 * detection.
391 addr->v4.sin_port = htons(snum);
392 if ((ret = sctp_get_port_local(sk, addr))) {
393 return -EADDRINUSE;
396 /* Refresh ephemeral port. */
397 if (!bp->port)
398 bp->port = inet_sk(sk)->inet_num;
400 /* Add the address to the bind address list.
401 * Use GFP_ATOMIC since BHs will be disabled.
403 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
405 /* Copy back into socket for getsockname() use. */
406 if (!ret) {
407 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
408 af->to_sk_saddr(addr, sk);
411 return ret;
414 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
416 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
417 * at any one time. If a sender, after sending an ASCONF chunk, decides
418 * it needs to transfer another ASCONF Chunk, it MUST wait until the
419 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
420 * subsequent ASCONF. Note this restriction binds each side, so at any
421 * time two ASCONF may be in-transit on any given association (one sent
422 * from each endpoint).
424 static int sctp_send_asconf(struct sctp_association *asoc,
425 struct sctp_chunk *chunk)
427 int retval = 0;
429 /* If there is an outstanding ASCONF chunk, queue it for later
430 * transmission.
432 if (asoc->addip_last_asconf) {
433 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
434 goto out;
437 /* Hold the chunk until an ASCONF_ACK is received. */
438 sctp_chunk_hold(chunk);
439 retval = sctp_primitive_ASCONF(asoc, chunk);
440 if (retval)
441 sctp_chunk_free(chunk);
442 else
443 asoc->addip_last_asconf = chunk;
445 out:
446 return retval;
449 /* Add a list of addresses as bind addresses to local endpoint or
450 * association.
452 * Basically run through each address specified in the addrs/addrcnt
453 * array/length pair, determine if it is IPv6 or IPv4 and call
454 * sctp_do_bind() on it.
456 * If any of them fails, then the operation will be reversed and the
457 * ones that were added will be removed.
459 * Only sctp_setsockopt_bindx() is supposed to call this function.
461 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
463 int cnt;
464 int retval = 0;
465 void *addr_buf;
466 struct sockaddr *sa_addr;
467 struct sctp_af *af;
469 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
470 sk, addrs, addrcnt);
472 addr_buf = addrs;
473 for (cnt = 0; cnt < addrcnt; cnt++) {
474 /* The list may contain either IPv4 or IPv6 address;
475 * determine the address length for walking thru the list.
477 sa_addr = (struct sockaddr *)addr_buf;
478 af = sctp_get_af_specific(sa_addr->sa_family);
479 if (!af) {
480 retval = -EINVAL;
481 goto err_bindx_add;
484 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
485 af->sockaddr_len);
487 addr_buf += af->sockaddr_len;
489 err_bindx_add:
490 if (retval < 0) {
491 /* Failed. Cleanup the ones that have been added */
492 if (cnt > 0)
493 sctp_bindx_rem(sk, addrs, cnt);
494 return retval;
498 return retval;
501 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
502 * associations that are part of the endpoint indicating that a list of local
503 * addresses are added to the endpoint.
505 * If any of the addresses is already in the bind address list of the
506 * association, we do not send the chunk for that association. But it will not
507 * affect other associations.
509 * Only sctp_setsockopt_bindx() is supposed to call this function.
511 static int sctp_send_asconf_add_ip(struct sock *sk,
512 struct sockaddr *addrs,
513 int addrcnt)
515 struct sctp_sock *sp;
516 struct sctp_endpoint *ep;
517 struct sctp_association *asoc;
518 struct sctp_bind_addr *bp;
519 struct sctp_chunk *chunk;
520 struct sctp_sockaddr_entry *laddr;
521 union sctp_addr *addr;
522 union sctp_addr saveaddr;
523 void *addr_buf;
524 struct sctp_af *af;
525 struct list_head *p;
526 int i;
527 int retval = 0;
529 if (!sctp_addip_enable)
530 return retval;
532 sp = sctp_sk(sk);
533 ep = sp->ep;
535 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
536 __func__, sk, addrs, addrcnt);
538 list_for_each_entry(asoc, &ep->asocs, asocs) {
540 if (!asoc->peer.asconf_capable)
541 continue;
543 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
544 continue;
546 if (!sctp_state(asoc, ESTABLISHED))
547 continue;
549 /* Check if any address in the packed array of addresses is
550 * in the bind address list of the association. If so,
551 * do not send the asconf chunk to its peer, but continue with
552 * other associations.
554 addr_buf = addrs;
555 for (i = 0; i < addrcnt; i++) {
556 addr = (union sctp_addr *)addr_buf;
557 af = sctp_get_af_specific(addr->v4.sin_family);
558 if (!af) {
559 retval = -EINVAL;
560 goto out;
563 if (sctp_assoc_lookup_laddr(asoc, addr))
564 break;
566 addr_buf += af->sockaddr_len;
568 if (i < addrcnt)
569 continue;
571 /* Use the first valid address in bind addr list of
572 * association as Address Parameter of ASCONF CHUNK.
574 bp = &asoc->base.bind_addr;
575 p = bp->address_list.next;
576 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
577 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
578 addrcnt, SCTP_PARAM_ADD_IP);
579 if (!chunk) {
580 retval = -ENOMEM;
581 goto out;
584 retval = sctp_send_asconf(asoc, chunk);
585 if (retval)
586 goto out;
588 /* Add the new addresses to the bind address list with
589 * use_as_src set to 0.
591 addr_buf = addrs;
592 for (i = 0; i < addrcnt; i++) {
593 addr = (union sctp_addr *)addr_buf;
594 af = sctp_get_af_specific(addr->v4.sin_family);
595 memcpy(&saveaddr, addr, af->sockaddr_len);
596 retval = sctp_add_bind_addr(bp, &saveaddr,
597 SCTP_ADDR_NEW, GFP_ATOMIC);
598 addr_buf += af->sockaddr_len;
602 out:
603 return retval;
606 /* Remove a list of addresses from bind addresses list. Do not remove the
607 * last address.
609 * Basically run through each address specified in the addrs/addrcnt
610 * array/length pair, determine if it is IPv6 or IPv4 and call
611 * sctp_del_bind() on it.
613 * If any of them fails, then the operation will be reversed and the
614 * ones that were removed will be added back.
616 * At least one address has to be left; if only one address is
617 * available, the operation will return -EBUSY.
619 * Only sctp_setsockopt_bindx() is supposed to call this function.
621 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
623 struct sctp_sock *sp = sctp_sk(sk);
624 struct sctp_endpoint *ep = sp->ep;
625 int cnt;
626 struct sctp_bind_addr *bp = &ep->base.bind_addr;
627 int retval = 0;
628 void *addr_buf;
629 union sctp_addr *sa_addr;
630 struct sctp_af *af;
632 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
633 sk, addrs, addrcnt);
635 addr_buf = addrs;
636 for (cnt = 0; cnt < addrcnt; cnt++) {
637 /* If the bind address list is empty or if there is only one
638 * bind address, there is nothing more to be removed (we need
639 * at least one address here).
641 if (list_empty(&bp->address_list) ||
642 (sctp_list_single_entry(&bp->address_list))) {
643 retval = -EBUSY;
644 goto err_bindx_rem;
647 sa_addr = (union sctp_addr *)addr_buf;
648 af = sctp_get_af_specific(sa_addr->sa.sa_family);
649 if (!af) {
650 retval = -EINVAL;
651 goto err_bindx_rem;
654 if (!af->addr_valid(sa_addr, sp, NULL)) {
655 retval = -EADDRNOTAVAIL;
656 goto err_bindx_rem;
659 if (sa_addr->v4.sin_port != htons(bp->port)) {
660 retval = -EINVAL;
661 goto err_bindx_rem;
664 /* FIXME - There is probably a need to check if sk->sk_saddr and
665 * sk->sk_rcv_addr are currently set to one of the addresses to
666 * be removed. This is something which needs to be looked into
667 * when we are fixing the outstanding issues with multi-homing
668 * socket routing and failover schemes. Refer to comments in
669 * sctp_do_bind(). -daisy
671 retval = sctp_del_bind_addr(bp, sa_addr);
673 addr_buf += af->sockaddr_len;
674 err_bindx_rem:
675 if (retval < 0) {
676 /* Failed. Add the ones that has been removed back */
677 if (cnt > 0)
678 sctp_bindx_add(sk, addrs, cnt);
679 return retval;
683 return retval;
686 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
687 * the associations that are part of the endpoint indicating that a list of
688 * local addresses are removed from the endpoint.
690 * If any of the addresses is already in the bind address list of the
691 * association, we do not send the chunk for that association. But it will not
692 * affect other associations.
694 * Only sctp_setsockopt_bindx() is supposed to call this function.
696 static int sctp_send_asconf_del_ip(struct sock *sk,
697 struct sockaddr *addrs,
698 int addrcnt)
700 struct sctp_sock *sp;
701 struct sctp_endpoint *ep;
702 struct sctp_association *asoc;
703 struct sctp_transport *transport;
704 struct sctp_bind_addr *bp;
705 struct sctp_chunk *chunk;
706 union sctp_addr *laddr;
707 void *addr_buf;
708 struct sctp_af *af;
709 struct sctp_sockaddr_entry *saddr;
710 int i;
711 int retval = 0;
713 if (!sctp_addip_enable)
714 return retval;
716 sp = sctp_sk(sk);
717 ep = sp->ep;
719 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
720 __func__, sk, addrs, addrcnt);
722 list_for_each_entry(asoc, &ep->asocs, asocs) {
724 if (!asoc->peer.asconf_capable)
725 continue;
727 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
728 continue;
730 if (!sctp_state(asoc, ESTABLISHED))
731 continue;
733 /* Check if any address in the packed array of addresses is
734 * not present in the bind address list of the association.
735 * If so, do not send the asconf chunk to its peer, but
736 * continue with other associations.
738 addr_buf = addrs;
739 for (i = 0; i < addrcnt; i++) {
740 laddr = (union sctp_addr *)addr_buf;
741 af = sctp_get_af_specific(laddr->v4.sin_family);
742 if (!af) {
743 retval = -EINVAL;
744 goto out;
747 if (!sctp_assoc_lookup_laddr(asoc, laddr))
748 break;
750 addr_buf += af->sockaddr_len;
752 if (i < addrcnt)
753 continue;
755 /* Find one address in the association's bind address list
756 * that is not in the packed array of addresses. This is to
757 * make sure that we do not delete all the addresses in the
758 * association.
760 bp = &asoc->base.bind_addr;
761 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
762 addrcnt, sp);
763 if (!laddr)
764 continue;
766 /* We do not need RCU protection throughout this loop
767 * because this is done under a socket lock from the
768 * setsockopt call.
770 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
771 SCTP_PARAM_DEL_IP);
772 if (!chunk) {
773 retval = -ENOMEM;
774 goto out;
777 /* Reset use_as_src flag for the addresses in the bind address
778 * list that are to be deleted.
780 addr_buf = addrs;
781 for (i = 0; i < addrcnt; i++) {
782 laddr = (union sctp_addr *)addr_buf;
783 af = sctp_get_af_specific(laddr->v4.sin_family);
784 list_for_each_entry(saddr, &bp->address_list, list) {
785 if (sctp_cmp_addr_exact(&saddr->a, laddr))
786 saddr->state = SCTP_ADDR_DEL;
788 addr_buf += af->sockaddr_len;
791 /* Update the route and saddr entries for all the transports
792 * as some of the addresses in the bind address list are
793 * about to be deleted and cannot be used as source addresses.
795 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
796 transports) {
797 dst_release(transport->dst);
798 sctp_transport_route(transport, NULL,
799 sctp_sk(asoc->base.sk));
802 retval = sctp_send_asconf(asoc, chunk);
804 out:
805 return retval;
808 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
810 * API 8.1
811 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
812 * int flags);
814 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
815 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
816 * or IPv6 addresses.
818 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
819 * Section 3.1.2 for this usage.
821 * addrs is a pointer to an array of one or more socket addresses. Each
822 * address is contained in its appropriate structure (i.e. struct
823 * sockaddr_in or struct sockaddr_in6) the family of the address type
824 * must be used to distinguish the address length (note that this
825 * representation is termed a "packed array" of addresses). The caller
826 * specifies the number of addresses in the array with addrcnt.
828 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
829 * -1, and sets errno to the appropriate error code.
831 * For SCTP, the port given in each socket address must be the same, or
832 * sctp_bindx() will fail, setting errno to EINVAL.
834 * The flags parameter is formed from the bitwise OR of zero or more of
835 * the following currently defined flags:
837 * SCTP_BINDX_ADD_ADDR
839 * SCTP_BINDX_REM_ADDR
841 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
842 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
843 * addresses from the association. The two flags are mutually exclusive;
844 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
845 * not remove all addresses from an association; sctp_bindx() will
846 * reject such an attempt with EINVAL.
848 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
849 * additional addresses with an endpoint after calling bind(). Or use
850 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
851 * socket is associated with so that no new association accepted will be
852 * associated with those addresses. If the endpoint supports dynamic
853 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
854 * endpoint to send the appropriate message to the peer to change the
855 * peers address lists.
857 * Adding and removing addresses from a connected association is
858 * optional functionality. Implementations that do not support this
859 * functionality should return EOPNOTSUPP.
861 * Basically do nothing but copying the addresses from user to kernel
862 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
863 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
864 * from userspace.
866 * We don't use copy_from_user() for optimization: we first do the
867 * sanity checks (buffer size -fast- and access check-healthy
868 * pointer); if all of those succeed, then we can alloc the memory
869 * (expensive operation) needed to copy the data to kernel. Then we do
870 * the copying without checking the user space area
871 * (__copy_from_user()).
873 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
874 * it.
876 * sk The sk of the socket
877 * addrs The pointer to the addresses in user land
878 * addrssize Size of the addrs buffer
879 * op Operation to perform (add or remove, see the flags of
880 * sctp_bindx)
882 * Returns 0 if ok, <0 errno code on error.
884 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
885 struct sockaddr __user *addrs,
886 int addrs_size, int op)
888 struct sockaddr *kaddrs;
889 int err;
890 int addrcnt = 0;
891 int walk_size = 0;
892 struct sockaddr *sa_addr;
893 void *addr_buf;
894 struct sctp_af *af;
896 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
897 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
899 if (unlikely(addrs_size <= 0))
900 return -EINVAL;
902 /* Check the user passed a healthy pointer. */
903 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
904 return -EFAULT;
906 /* Alloc space for the address array in kernel memory. */
907 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
908 if (unlikely(!kaddrs))
909 return -ENOMEM;
911 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
912 kfree(kaddrs);
913 return -EFAULT;
916 /* Walk through the addrs buffer and count the number of addresses. */
917 addr_buf = kaddrs;
918 while (walk_size < addrs_size) {
919 sa_addr = (struct sockaddr *)addr_buf;
920 af = sctp_get_af_specific(sa_addr->sa_family);
922 /* If the address family is not supported or if this address
923 * causes the address buffer to overflow return EINVAL.
925 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
926 kfree(kaddrs);
927 return -EINVAL;
929 addrcnt++;
930 addr_buf += af->sockaddr_len;
931 walk_size += af->sockaddr_len;
934 /* Do the work. */
935 switch (op) {
936 case SCTP_BINDX_ADD_ADDR:
937 err = sctp_bindx_add(sk, kaddrs, addrcnt);
938 if (err)
939 goto out;
940 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
941 break;
943 case SCTP_BINDX_REM_ADDR:
944 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
945 if (err)
946 goto out;
947 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
948 break;
950 default:
951 err = -EINVAL;
952 break;
955 out:
956 kfree(kaddrs);
958 return err;
961 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
963 * Common routine for handling connect() and sctp_connectx().
964 * Connect will come in with just a single address.
966 static int __sctp_connect(struct sock* sk,
967 struct sockaddr *kaddrs,
968 int addrs_size,
969 sctp_assoc_t *assoc_id)
971 struct sctp_sock *sp;
972 struct sctp_endpoint *ep;
973 struct sctp_association *asoc = NULL;
974 struct sctp_association *asoc2;
975 struct sctp_transport *transport;
976 union sctp_addr to;
977 struct sctp_af *af;
978 sctp_scope_t scope;
979 long timeo;
980 int err = 0;
981 int addrcnt = 0;
982 int walk_size = 0;
983 union sctp_addr *sa_addr = NULL;
984 void *addr_buf;
985 unsigned short port;
986 unsigned int f_flags = 0;
988 sp = sctp_sk(sk);
989 ep = sp->ep;
991 /* connect() cannot be done on a socket that is already in ESTABLISHED
992 * state - UDP-style peeled off socket or a TCP-style socket that
993 * is already connected.
994 * It cannot be done even on a TCP-style listening socket.
996 if (sctp_sstate(sk, ESTABLISHED) ||
997 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
998 err = -EISCONN;
999 goto out_free;
1002 /* Walk through the addrs buffer and count the number of addresses. */
1003 addr_buf = kaddrs;
1004 while (walk_size < addrs_size) {
1005 sa_addr = (union sctp_addr *)addr_buf;
1006 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1007 port = ntohs(sa_addr->v4.sin_port);
1009 /* If the address family is not supported or if this address
1010 * causes the address buffer to overflow return EINVAL.
1012 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1013 err = -EINVAL;
1014 goto out_free;
1017 /* Save current address so we can work with it */
1018 memcpy(&to, sa_addr, af->sockaddr_len);
1020 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1021 if (err)
1022 goto out_free;
1024 /* Make sure the destination port is correctly set
1025 * in all addresses.
1027 if (asoc && asoc->peer.port && asoc->peer.port != port)
1028 goto out_free;
1031 /* Check if there already is a matching association on the
1032 * endpoint (other than the one created here).
1034 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1035 if (asoc2 && asoc2 != asoc) {
1036 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1037 err = -EISCONN;
1038 else
1039 err = -EALREADY;
1040 goto out_free;
1043 /* If we could not find a matching association on the endpoint,
1044 * make sure that there is no peeled-off association matching
1045 * the peer address even on another socket.
1047 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1048 err = -EADDRNOTAVAIL;
1049 goto out_free;
1052 if (!asoc) {
1053 /* If a bind() or sctp_bindx() is not called prior to
1054 * an sctp_connectx() call, the system picks an
1055 * ephemeral port and will choose an address set
1056 * equivalent to binding with a wildcard address.
1058 if (!ep->base.bind_addr.port) {
1059 if (sctp_autobind(sk)) {
1060 err = -EAGAIN;
1061 goto out_free;
1063 } else {
1065 * If an unprivileged user inherits a 1-many
1066 * style socket with open associations on a
1067 * privileged port, it MAY be permitted to
1068 * accept new associations, but it SHOULD NOT
1069 * be permitted to open new associations.
1071 if (ep->base.bind_addr.port < PROT_SOCK &&
1072 !capable(CAP_NET_BIND_SERVICE)) {
1073 err = -EACCES;
1074 goto out_free;
1078 scope = sctp_scope(&to);
1079 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1080 if (!asoc) {
1081 err = -ENOMEM;
1082 goto out_free;
1085 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1086 GFP_KERNEL);
1087 if (err < 0) {
1088 goto out_free;
1093 /* Prime the peer's transport structures. */
1094 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1095 SCTP_UNKNOWN);
1096 if (!transport) {
1097 err = -ENOMEM;
1098 goto out_free;
1101 addrcnt++;
1102 addr_buf += af->sockaddr_len;
1103 walk_size += af->sockaddr_len;
1106 /* In case the user of sctp_connectx() wants an association
1107 * id back, assign one now.
1109 if (assoc_id) {
1110 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1111 if (err < 0)
1112 goto out_free;
1115 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1116 if (err < 0) {
1117 goto out_free;
1120 /* Initialize sk's dport and daddr for getpeername() */
1121 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1122 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1123 af->to_sk_daddr(sa_addr, sk);
1124 sk->sk_err = 0;
1126 /* in-kernel sockets don't generally have a file allocated to them
1127 * if all they do is call sock_create_kern().
1129 if (sk->sk_socket->file)
1130 f_flags = sk->sk_socket->file->f_flags;
1132 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1134 err = sctp_wait_for_connect(asoc, &timeo);
1135 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1136 *assoc_id = asoc->assoc_id;
1138 /* Don't free association on exit. */
1139 asoc = NULL;
1141 out_free:
1143 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1144 " kaddrs: %p err: %d\n",
1145 asoc, kaddrs, err);
1146 if (asoc)
1147 sctp_association_free(asoc);
1148 return err;
1151 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1153 * API 8.9
1154 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1155 * sctp_assoc_t *asoc);
1157 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1158 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1159 * or IPv6 addresses.
1161 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1162 * Section 3.1.2 for this usage.
1164 * addrs is a pointer to an array of one or more socket addresses. Each
1165 * address is contained in its appropriate structure (i.e. struct
1166 * sockaddr_in or struct sockaddr_in6) the family of the address type
1167 * must be used to distengish the address length (note that this
1168 * representation is termed a "packed array" of addresses). The caller
1169 * specifies the number of addresses in the array with addrcnt.
1171 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1172 * the association id of the new association. On failure, sctp_connectx()
1173 * returns -1, and sets errno to the appropriate error code. The assoc_id
1174 * is not touched by the kernel.
1176 * For SCTP, the port given in each socket address must be the same, or
1177 * sctp_connectx() will fail, setting errno to EINVAL.
1179 * An application can use sctp_connectx to initiate an association with
1180 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1181 * allows a caller to specify multiple addresses at which a peer can be
1182 * reached. The way the SCTP stack uses the list of addresses to set up
1183 * the association is implementation dependant. This function only
1184 * specifies that the stack will try to make use of all the addresses in
1185 * the list when needed.
1187 * Note that the list of addresses passed in is only used for setting up
1188 * the association. It does not necessarily equal the set of addresses
1189 * the peer uses for the resulting association. If the caller wants to
1190 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1191 * retrieve them after the association has been set up.
1193 * Basically do nothing but copying the addresses from user to kernel
1194 * land and invoking either sctp_connectx(). This is used for tunneling
1195 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1197 * We don't use copy_from_user() for optimization: we first do the
1198 * sanity checks (buffer size -fast- and access check-healthy
1199 * pointer); if all of those succeed, then we can alloc the memory
1200 * (expensive operation) needed to copy the data to kernel. Then we do
1201 * the copying without checking the user space area
1202 * (__copy_from_user()).
1204 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1205 * it.
1207 * sk The sk of the socket
1208 * addrs The pointer to the addresses in user land
1209 * addrssize Size of the addrs buffer
1211 * Returns >=0 if ok, <0 errno code on error.
1213 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1214 struct sockaddr __user *addrs,
1215 int addrs_size,
1216 sctp_assoc_t *assoc_id)
1218 int err = 0;
1219 struct sockaddr *kaddrs;
1221 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1222 __func__, sk, addrs, addrs_size);
1224 if (unlikely(addrs_size <= 0))
1225 return -EINVAL;
1227 /* Check the user passed a healthy pointer. */
1228 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1229 return -EFAULT;
1231 /* Alloc space for the address array in kernel memory. */
1232 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1233 if (unlikely(!kaddrs))
1234 return -ENOMEM;
1236 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1237 err = -EFAULT;
1238 } else {
1239 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1242 kfree(kaddrs);
1244 return err;
1248 * This is an older interface. It's kept for backward compatibility
1249 * to the option that doesn't provide association id.
1251 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1252 struct sockaddr __user *addrs,
1253 int addrs_size)
1255 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1259 * New interface for the API. The since the API is done with a socket
1260 * option, to make it simple we feed back the association id is as a return
1261 * indication to the call. Error is always negative and association id is
1262 * always positive.
1264 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1265 struct sockaddr __user *addrs,
1266 int addrs_size)
1268 sctp_assoc_t assoc_id = 0;
1269 int err = 0;
1271 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1273 if (err)
1274 return err;
1275 else
1276 return assoc_id;
1280 * New (hopefully final) interface for the API.
1281 * We use the sctp_getaddrs_old structure so that use-space library
1282 * can avoid any unnecessary allocations. The only defferent part
1283 * is that we store the actual length of the address buffer into the
1284 * addrs_num structure member. That way we can re-use the existing
1285 * code.
1287 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1288 char __user *optval,
1289 int __user *optlen)
1291 struct sctp_getaddrs_old param;
1292 sctp_assoc_t assoc_id = 0;
1293 int err = 0;
1295 if (len < sizeof(param))
1296 return -EINVAL;
1298 if (copy_from_user(&param, optval, sizeof(param)))
1299 return -EFAULT;
1301 err = __sctp_setsockopt_connectx(sk,
1302 (struct sockaddr __user *)param.addrs,
1303 param.addr_num, &assoc_id);
1305 if (err == 0 || err == -EINPROGRESS) {
1306 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1307 return -EFAULT;
1308 if (put_user(sizeof(assoc_id), optlen))
1309 return -EFAULT;
1312 return err;
1315 /* API 3.1.4 close() - UDP Style Syntax
1316 * Applications use close() to perform graceful shutdown (as described in
1317 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1318 * by a UDP-style socket.
1320 * The syntax is
1322 * ret = close(int sd);
1324 * sd - the socket descriptor of the associations to be closed.
1326 * To gracefully shutdown a specific association represented by the
1327 * UDP-style socket, an application should use the sendmsg() call,
1328 * passing no user data, but including the appropriate flag in the
1329 * ancillary data (see Section xxxx).
1331 * If sd in the close() call is a branched-off socket representing only
1332 * one association, the shutdown is performed on that association only.
1334 * 4.1.6 close() - TCP Style Syntax
1336 * Applications use close() to gracefully close down an association.
1338 * The syntax is:
1340 * int close(int sd);
1342 * sd - the socket descriptor of the association to be closed.
1344 * After an application calls close() on a socket descriptor, no further
1345 * socket operations will succeed on that descriptor.
1347 * API 7.1.4 SO_LINGER
1349 * An application using the TCP-style socket can use this option to
1350 * perform the SCTP ABORT primitive. The linger option structure is:
1352 * struct linger {
1353 * int l_onoff; // option on/off
1354 * int l_linger; // linger time
1355 * };
1357 * To enable the option, set l_onoff to 1. If the l_linger value is set
1358 * to 0, calling close() is the same as the ABORT primitive. If the
1359 * value is set to a negative value, the setsockopt() call will return
1360 * an error. If the value is set to a positive value linger_time, the
1361 * close() can be blocked for at most linger_time ms. If the graceful
1362 * shutdown phase does not finish during this period, close() will
1363 * return but the graceful shutdown phase continues in the system.
1365 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1367 struct sctp_endpoint *ep;
1368 struct sctp_association *asoc;
1369 struct list_head *pos, *temp;
1371 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1373 sctp_lock_sock(sk);
1374 sk->sk_shutdown = SHUTDOWN_MASK;
1375 sk->sk_state = SCTP_SS_CLOSING;
1377 ep = sctp_sk(sk)->ep;
1379 /* Walk all associations on an endpoint. */
1380 list_for_each_safe(pos, temp, &ep->asocs) {
1381 asoc = list_entry(pos, struct sctp_association, asocs);
1383 if (sctp_style(sk, TCP)) {
1384 /* A closed association can still be in the list if
1385 * it belongs to a TCP-style listening socket that is
1386 * not yet accepted. If so, free it. If not, send an
1387 * ABORT or SHUTDOWN based on the linger options.
1389 if (sctp_state(asoc, CLOSED)) {
1390 sctp_unhash_established(asoc);
1391 sctp_association_free(asoc);
1392 continue;
1396 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1397 struct sctp_chunk *chunk;
1399 chunk = sctp_make_abort_user(asoc, NULL, 0);
1400 if (chunk)
1401 sctp_primitive_ABORT(asoc, chunk);
1402 } else
1403 sctp_primitive_SHUTDOWN(asoc, NULL);
1406 /* Clean up any skbs sitting on the receive queue. */
1407 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1408 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1410 /* On a TCP-style socket, block for at most linger_time if set. */
1411 if (sctp_style(sk, TCP) && timeout)
1412 sctp_wait_for_close(sk, timeout);
1414 /* This will run the backlog queue. */
1415 sctp_release_sock(sk);
1417 /* Supposedly, no process has access to the socket, but
1418 * the net layers still may.
1420 sctp_local_bh_disable();
1421 sctp_bh_lock_sock(sk);
1423 /* Hold the sock, since sk_common_release() will put sock_put()
1424 * and we have just a little more cleanup.
1426 sock_hold(sk);
1427 sk_common_release(sk);
1429 sctp_bh_unlock_sock(sk);
1430 sctp_local_bh_enable();
1432 sock_put(sk);
1434 SCTP_DBG_OBJCNT_DEC(sock);
1437 /* Handle EPIPE error. */
1438 static int sctp_error(struct sock *sk, int flags, int err)
1440 if (err == -EPIPE)
1441 err = sock_error(sk) ? : -EPIPE;
1442 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1443 send_sig(SIGPIPE, current, 0);
1444 return err;
1447 /* API 3.1.3 sendmsg() - UDP Style Syntax
1449 * An application uses sendmsg() and recvmsg() calls to transmit data to
1450 * and receive data from its peer.
1452 * ssize_t sendmsg(int socket, const struct msghdr *message,
1453 * int flags);
1455 * socket - the socket descriptor of the endpoint.
1456 * message - pointer to the msghdr structure which contains a single
1457 * user message and possibly some ancillary data.
1459 * See Section 5 for complete description of the data
1460 * structures.
1462 * flags - flags sent or received with the user message, see Section
1463 * 5 for complete description of the flags.
1465 * Note: This function could use a rewrite especially when explicit
1466 * connect support comes in.
1468 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1470 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1472 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1473 struct msghdr *msg, size_t msg_len)
1475 struct sctp_sock *sp;
1476 struct sctp_endpoint *ep;
1477 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1478 struct sctp_transport *transport, *chunk_tp;
1479 struct sctp_chunk *chunk;
1480 union sctp_addr to;
1481 struct sockaddr *msg_name = NULL;
1482 struct sctp_sndrcvinfo default_sinfo = { 0 };
1483 struct sctp_sndrcvinfo *sinfo;
1484 struct sctp_initmsg *sinit;
1485 sctp_assoc_t associd = 0;
1486 sctp_cmsgs_t cmsgs = { NULL };
1487 int err;
1488 sctp_scope_t scope;
1489 long timeo;
1490 __u16 sinfo_flags = 0;
1491 struct sctp_datamsg *datamsg;
1492 int msg_flags = msg->msg_flags;
1494 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1495 sk, msg, msg_len);
1497 err = 0;
1498 sp = sctp_sk(sk);
1499 ep = sp->ep;
1501 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1503 /* We cannot send a message over a TCP-style listening socket. */
1504 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1505 err = -EPIPE;
1506 goto out_nounlock;
1509 /* Parse out the SCTP CMSGs. */
1510 err = sctp_msghdr_parse(msg, &cmsgs);
1512 if (err) {
1513 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1514 goto out_nounlock;
1517 /* Fetch the destination address for this packet. This
1518 * address only selects the association--it is not necessarily
1519 * the address we will send to.
1520 * For a peeled-off socket, msg_name is ignored.
1522 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1523 int msg_namelen = msg->msg_namelen;
1525 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1526 msg_namelen);
1527 if (err)
1528 return err;
1530 if (msg_namelen > sizeof(to))
1531 msg_namelen = sizeof(to);
1532 memcpy(&to, msg->msg_name, msg_namelen);
1533 msg_name = msg->msg_name;
1536 sinfo = cmsgs.info;
1537 sinit = cmsgs.init;
1539 /* Did the user specify SNDRCVINFO? */
1540 if (sinfo) {
1541 sinfo_flags = sinfo->sinfo_flags;
1542 associd = sinfo->sinfo_assoc_id;
1545 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1546 msg_len, sinfo_flags);
1548 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1549 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1550 err = -EINVAL;
1551 goto out_nounlock;
1554 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1555 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1556 * If SCTP_ABORT is set, the message length could be non zero with
1557 * the msg_iov set to the user abort reason.
1559 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1560 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1561 err = -EINVAL;
1562 goto out_nounlock;
1565 /* If SCTP_ADDR_OVER is set, there must be an address
1566 * specified in msg_name.
1568 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1569 err = -EINVAL;
1570 goto out_nounlock;
1573 transport = NULL;
1575 SCTP_DEBUG_PRINTK("About to look up association.\n");
1577 sctp_lock_sock(sk);
1579 /* If a msg_name has been specified, assume this is to be used. */
1580 if (msg_name) {
1581 /* Look for a matching association on the endpoint. */
1582 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1583 if (!asoc) {
1584 /* If we could not find a matching association on the
1585 * endpoint, make sure that it is not a TCP-style
1586 * socket that already has an association or there is
1587 * no peeled-off association on another socket.
1589 if ((sctp_style(sk, TCP) &&
1590 sctp_sstate(sk, ESTABLISHED)) ||
1591 sctp_endpoint_is_peeled_off(ep, &to)) {
1592 err = -EADDRNOTAVAIL;
1593 goto out_unlock;
1596 } else {
1597 asoc = sctp_id2assoc(sk, associd);
1598 if (!asoc) {
1599 err = -EPIPE;
1600 goto out_unlock;
1604 if (asoc) {
1605 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1607 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1608 * socket that has an association in CLOSED state. This can
1609 * happen when an accepted socket has an association that is
1610 * already CLOSED.
1612 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1613 err = -EPIPE;
1614 goto out_unlock;
1617 if (sinfo_flags & SCTP_EOF) {
1618 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1619 asoc);
1620 sctp_primitive_SHUTDOWN(asoc, NULL);
1621 err = 0;
1622 goto out_unlock;
1624 if (sinfo_flags & SCTP_ABORT) {
1626 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1627 if (!chunk) {
1628 err = -ENOMEM;
1629 goto out_unlock;
1632 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1633 sctp_primitive_ABORT(asoc, chunk);
1634 err = 0;
1635 goto out_unlock;
1639 /* Do we need to create the association? */
1640 if (!asoc) {
1641 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1643 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1644 err = -EINVAL;
1645 goto out_unlock;
1648 /* Check for invalid stream against the stream counts,
1649 * either the default or the user specified stream counts.
1651 if (sinfo) {
1652 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1653 /* Check against the defaults. */
1654 if (sinfo->sinfo_stream >=
1655 sp->initmsg.sinit_num_ostreams) {
1656 err = -EINVAL;
1657 goto out_unlock;
1659 } else {
1660 /* Check against the requested. */
1661 if (sinfo->sinfo_stream >=
1662 sinit->sinit_num_ostreams) {
1663 err = -EINVAL;
1664 goto out_unlock;
1670 * API 3.1.2 bind() - UDP Style Syntax
1671 * If a bind() or sctp_bindx() is not called prior to a
1672 * sendmsg() call that initiates a new association, the
1673 * system picks an ephemeral port and will choose an address
1674 * set equivalent to binding with a wildcard address.
1676 if (!ep->base.bind_addr.port) {
1677 if (sctp_autobind(sk)) {
1678 err = -EAGAIN;
1679 goto out_unlock;
1681 } else {
1683 * If an unprivileged user inherits a one-to-many
1684 * style socket with open associations on a privileged
1685 * port, it MAY be permitted to accept new associations,
1686 * but it SHOULD NOT be permitted to open new
1687 * associations.
1689 if (ep->base.bind_addr.port < PROT_SOCK &&
1690 !capable(CAP_NET_BIND_SERVICE)) {
1691 err = -EACCES;
1692 goto out_unlock;
1696 scope = sctp_scope(&to);
1697 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1698 if (!new_asoc) {
1699 err = -ENOMEM;
1700 goto out_unlock;
1702 asoc = new_asoc;
1703 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1704 if (err < 0) {
1705 err = -ENOMEM;
1706 goto out_free;
1709 /* If the SCTP_INIT ancillary data is specified, set all
1710 * the association init values accordingly.
1712 if (sinit) {
1713 if (sinit->sinit_num_ostreams) {
1714 asoc->c.sinit_num_ostreams =
1715 sinit->sinit_num_ostreams;
1717 if (sinit->sinit_max_instreams) {
1718 asoc->c.sinit_max_instreams =
1719 sinit->sinit_max_instreams;
1721 if (sinit->sinit_max_attempts) {
1722 asoc->max_init_attempts
1723 = sinit->sinit_max_attempts;
1725 if (sinit->sinit_max_init_timeo) {
1726 asoc->max_init_timeo =
1727 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1731 /* Prime the peer's transport structures. */
1732 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1733 if (!transport) {
1734 err = -ENOMEM;
1735 goto out_free;
1739 /* ASSERT: we have a valid association at this point. */
1740 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1742 if (!sinfo) {
1743 /* If the user didn't specify SNDRCVINFO, make up one with
1744 * some defaults.
1746 default_sinfo.sinfo_stream = asoc->default_stream;
1747 default_sinfo.sinfo_flags = asoc->default_flags;
1748 default_sinfo.sinfo_ppid = asoc->default_ppid;
1749 default_sinfo.sinfo_context = asoc->default_context;
1750 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1751 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1752 sinfo = &default_sinfo;
1755 /* API 7.1.7, the sndbuf size per association bounds the
1756 * maximum size of data that can be sent in a single send call.
1758 if (msg_len > sk->sk_sndbuf) {
1759 err = -EMSGSIZE;
1760 goto out_free;
1763 if (asoc->pmtu_pending)
1764 sctp_assoc_pending_pmtu(asoc);
1766 /* If fragmentation is disabled and the message length exceeds the
1767 * association fragmentation point, return EMSGSIZE. The I-D
1768 * does not specify what this error is, but this looks like
1769 * a great fit.
1771 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1772 err = -EMSGSIZE;
1773 goto out_free;
1776 if (sinfo) {
1777 /* Check for invalid stream. */
1778 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1779 err = -EINVAL;
1780 goto out_free;
1784 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1785 if (!sctp_wspace(asoc)) {
1786 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1787 if (err)
1788 goto out_free;
1791 /* If an address is passed with the sendto/sendmsg call, it is used
1792 * to override the primary destination address in the TCP model, or
1793 * when SCTP_ADDR_OVER flag is set in the UDP model.
1795 if ((sctp_style(sk, TCP) && msg_name) ||
1796 (sinfo_flags & SCTP_ADDR_OVER)) {
1797 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1798 if (!chunk_tp) {
1799 err = -EINVAL;
1800 goto out_free;
1802 } else
1803 chunk_tp = NULL;
1805 /* Auto-connect, if we aren't connected already. */
1806 if (sctp_state(asoc, CLOSED)) {
1807 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1808 if (err < 0)
1809 goto out_free;
1810 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1813 /* Break the message into multiple chunks of maximum size. */
1814 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1815 if (!datamsg) {
1816 err = -ENOMEM;
1817 goto out_free;
1820 /* Now send the (possibly) fragmented message. */
1821 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1822 sctp_chunk_hold(chunk);
1824 /* Do accounting for the write space. */
1825 sctp_set_owner_w(chunk);
1827 chunk->transport = chunk_tp;
1830 /* Send it to the lower layers. Note: all chunks
1831 * must either fail or succeed. The lower layer
1832 * works that way today. Keep it that way or this
1833 * breaks.
1835 err = sctp_primitive_SEND(asoc, datamsg);
1836 /* Did the lower layer accept the chunk? */
1837 if (err)
1838 sctp_datamsg_free(datamsg);
1839 else
1840 sctp_datamsg_put(datamsg);
1842 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1844 if (err)
1845 goto out_free;
1846 else
1847 err = msg_len;
1849 /* If we are already past ASSOCIATE, the lower
1850 * layers are responsible for association cleanup.
1852 goto out_unlock;
1854 out_free:
1855 if (new_asoc)
1856 sctp_association_free(asoc);
1857 out_unlock:
1858 sctp_release_sock(sk);
1860 out_nounlock:
1861 return sctp_error(sk, msg_flags, err);
1863 #if 0
1864 do_sock_err:
1865 if (msg_len)
1866 err = msg_len;
1867 else
1868 err = sock_error(sk);
1869 goto out;
1871 do_interrupted:
1872 if (msg_len)
1873 err = msg_len;
1874 goto out;
1875 #endif /* 0 */
1878 /* This is an extended version of skb_pull() that removes the data from the
1879 * start of a skb even when data is spread across the list of skb's in the
1880 * frag_list. len specifies the total amount of data that needs to be removed.
1881 * when 'len' bytes could be removed from the skb, it returns 0.
1882 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1883 * could not be removed.
1885 static int sctp_skb_pull(struct sk_buff *skb, int len)
1887 struct sk_buff *list;
1888 int skb_len = skb_headlen(skb);
1889 int rlen;
1891 if (len <= skb_len) {
1892 __skb_pull(skb, len);
1893 return 0;
1895 len -= skb_len;
1896 __skb_pull(skb, skb_len);
1898 skb_walk_frags(skb, list) {
1899 rlen = sctp_skb_pull(list, len);
1900 skb->len -= (len-rlen);
1901 skb->data_len -= (len-rlen);
1903 if (!rlen)
1904 return 0;
1906 len = rlen;
1909 return len;
1912 /* API 3.1.3 recvmsg() - UDP Style Syntax
1914 * ssize_t recvmsg(int socket, struct msghdr *message,
1915 * int flags);
1917 * socket - the socket descriptor of the endpoint.
1918 * message - pointer to the msghdr structure which contains a single
1919 * user message and possibly some ancillary data.
1921 * See Section 5 for complete description of the data
1922 * structures.
1924 * flags - flags sent or received with the user message, see Section
1925 * 5 for complete description of the flags.
1927 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1929 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1930 struct msghdr *msg, size_t len, int noblock,
1931 int flags, int *addr_len)
1933 struct sctp_ulpevent *event = NULL;
1934 struct sctp_sock *sp = sctp_sk(sk);
1935 struct sk_buff *skb;
1936 int copied;
1937 int err = 0;
1938 int skb_len;
1940 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1941 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1942 "len", len, "knoblauch", noblock,
1943 "flags", flags, "addr_len", addr_len);
1945 sctp_lock_sock(sk);
1947 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1948 err = -ENOTCONN;
1949 goto out;
1952 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1953 if (!skb)
1954 goto out;
1956 /* Get the total length of the skb including any skb's in the
1957 * frag_list.
1959 skb_len = skb->len;
1961 copied = skb_len;
1962 if (copied > len)
1963 copied = len;
1965 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1967 event = sctp_skb2event(skb);
1969 if (err)
1970 goto out_free;
1972 sock_recv_ts_and_drops(msg, sk, skb);
1973 if (sctp_ulpevent_is_notification(event)) {
1974 msg->msg_flags |= MSG_NOTIFICATION;
1975 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1976 } else {
1977 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1980 /* Check if we allow SCTP_SNDRCVINFO. */
1981 if (sp->subscribe.sctp_data_io_event)
1982 sctp_ulpevent_read_sndrcvinfo(event, msg);
1983 #if 0
1984 /* FIXME: we should be calling IP/IPv6 layers. */
1985 if (sk->sk_protinfo.af_inet.cmsg_flags)
1986 ip_cmsg_recv(msg, skb);
1987 #endif
1989 err = copied;
1991 /* If skb's length exceeds the user's buffer, update the skb and
1992 * push it back to the receive_queue so that the next call to
1993 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1995 if (skb_len > copied) {
1996 msg->msg_flags &= ~MSG_EOR;
1997 if (flags & MSG_PEEK)
1998 goto out_free;
1999 sctp_skb_pull(skb, copied);
2000 skb_queue_head(&sk->sk_receive_queue, skb);
2002 /* When only partial message is copied to the user, increase
2003 * rwnd by that amount. If all the data in the skb is read,
2004 * rwnd is updated when the event is freed.
2006 if (!sctp_ulpevent_is_notification(event))
2007 sctp_assoc_rwnd_increase(event->asoc, copied);
2008 goto out;
2009 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2010 (event->msg_flags & MSG_EOR))
2011 msg->msg_flags |= MSG_EOR;
2012 else
2013 msg->msg_flags &= ~MSG_EOR;
2015 out_free:
2016 if (flags & MSG_PEEK) {
2017 /* Release the skb reference acquired after peeking the skb in
2018 * sctp_skb_recv_datagram().
2020 kfree_skb(skb);
2021 } else {
2022 /* Free the event which includes releasing the reference to
2023 * the owner of the skb, freeing the skb and updating the
2024 * rwnd.
2026 sctp_ulpevent_free(event);
2028 out:
2029 sctp_release_sock(sk);
2030 return err;
2033 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2035 * This option is a on/off flag. If enabled no SCTP message
2036 * fragmentation will be performed. Instead if a message being sent
2037 * exceeds the current PMTU size, the message will NOT be sent and
2038 * instead a error will be indicated to the user.
2040 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2041 char __user *optval,
2042 unsigned int optlen)
2044 int val;
2046 if (optlen < sizeof(int))
2047 return -EINVAL;
2049 if (get_user(val, (int __user *)optval))
2050 return -EFAULT;
2052 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2054 return 0;
2057 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2058 unsigned int optlen)
2060 if (optlen > sizeof(struct sctp_event_subscribe))
2061 return -EINVAL;
2062 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2063 return -EFAULT;
2064 return 0;
2067 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2069 * This socket option is applicable to the UDP-style socket only. When
2070 * set it will cause associations that are idle for more than the
2071 * specified number of seconds to automatically close. An association
2072 * being idle is defined an association that has NOT sent or received
2073 * user data. The special value of '0' indicates that no automatic
2074 * close of any associations should be performed. The option expects an
2075 * integer defining the number of seconds of idle time before an
2076 * association is closed.
2078 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2079 unsigned int optlen)
2081 struct sctp_sock *sp = sctp_sk(sk);
2083 /* Applicable to UDP-style socket only */
2084 if (sctp_style(sk, TCP))
2085 return -EOPNOTSUPP;
2086 if (optlen != sizeof(int))
2087 return -EINVAL;
2088 if (copy_from_user(&sp->autoclose, optval, optlen))
2089 return -EFAULT;
2090 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2091 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2093 return 0;
2096 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2098 * Applications can enable or disable heartbeats for any peer address of
2099 * an association, modify an address's heartbeat interval, force a
2100 * heartbeat to be sent immediately, and adjust the address's maximum
2101 * number of retransmissions sent before an address is considered
2102 * unreachable. The following structure is used to access and modify an
2103 * address's parameters:
2105 * struct sctp_paddrparams {
2106 * sctp_assoc_t spp_assoc_id;
2107 * struct sockaddr_storage spp_address;
2108 * uint32_t spp_hbinterval;
2109 * uint16_t spp_pathmaxrxt;
2110 * uint32_t spp_pathmtu;
2111 * uint32_t spp_sackdelay;
2112 * uint32_t spp_flags;
2113 * };
2115 * spp_assoc_id - (one-to-many style socket) This is filled in the
2116 * application, and identifies the association for
2117 * this query.
2118 * spp_address - This specifies which address is of interest.
2119 * spp_hbinterval - This contains the value of the heartbeat interval,
2120 * in milliseconds. If a value of zero
2121 * is present in this field then no changes are to
2122 * be made to this parameter.
2123 * spp_pathmaxrxt - This contains the maximum number of
2124 * retransmissions before this address shall be
2125 * considered unreachable. If a value of zero
2126 * is present in this field then no changes are to
2127 * be made to this parameter.
2128 * spp_pathmtu - When Path MTU discovery is disabled the value
2129 * specified here will be the "fixed" path mtu.
2130 * Note that if the spp_address field is empty
2131 * then all associations on this address will
2132 * have this fixed path mtu set upon them.
2134 * spp_sackdelay - When delayed sack is enabled, this value specifies
2135 * the number of milliseconds that sacks will be delayed
2136 * for. This value will apply to all addresses of an
2137 * association if the spp_address field is empty. Note
2138 * also, that if delayed sack is enabled and this
2139 * value is set to 0, no change is made to the last
2140 * recorded delayed sack timer value.
2142 * spp_flags - These flags are used to control various features
2143 * on an association. The flag field may contain
2144 * zero or more of the following options.
2146 * SPP_HB_ENABLE - Enable heartbeats on the
2147 * specified address. Note that if the address
2148 * field is empty all addresses for the association
2149 * have heartbeats enabled upon them.
2151 * SPP_HB_DISABLE - Disable heartbeats on the
2152 * speicifed address. Note that if the address
2153 * field is empty all addresses for the association
2154 * will have their heartbeats disabled. Note also
2155 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2156 * mutually exclusive, only one of these two should
2157 * be specified. Enabling both fields will have
2158 * undetermined results.
2160 * SPP_HB_DEMAND - Request a user initiated heartbeat
2161 * to be made immediately.
2163 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2164 * heartbeat delayis to be set to the value of 0
2165 * milliseconds.
2167 * SPP_PMTUD_ENABLE - This field will enable PMTU
2168 * discovery upon the specified address. Note that
2169 * if the address feild is empty then all addresses
2170 * on the association are effected.
2172 * SPP_PMTUD_DISABLE - This field will disable PMTU
2173 * discovery upon the specified address. Note that
2174 * if the address feild is empty then all addresses
2175 * on the association are effected. Not also that
2176 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2177 * exclusive. Enabling both will have undetermined
2178 * results.
2180 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2181 * on delayed sack. The time specified in spp_sackdelay
2182 * is used to specify the sack delay for this address. Note
2183 * that if spp_address is empty then all addresses will
2184 * enable delayed sack and take on the sack delay
2185 * value specified in spp_sackdelay.
2186 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2187 * off delayed sack. If the spp_address field is blank then
2188 * delayed sack is disabled for the entire association. Note
2189 * also that this field is mutually exclusive to
2190 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2191 * results.
2193 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2194 struct sctp_transport *trans,
2195 struct sctp_association *asoc,
2196 struct sctp_sock *sp,
2197 int hb_change,
2198 int pmtud_change,
2199 int sackdelay_change)
2201 int error;
2203 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2204 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2205 if (error)
2206 return error;
2209 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2210 * this field is ignored. Note also that a value of zero indicates
2211 * the current setting should be left unchanged.
2213 if (params->spp_flags & SPP_HB_ENABLE) {
2215 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2216 * set. This lets us use 0 value when this flag
2217 * is set.
2219 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2220 params->spp_hbinterval = 0;
2222 if (params->spp_hbinterval ||
2223 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2224 if (trans) {
2225 trans->hbinterval =
2226 msecs_to_jiffies(params->spp_hbinterval);
2227 } else if (asoc) {
2228 asoc->hbinterval =
2229 msecs_to_jiffies(params->spp_hbinterval);
2230 } else {
2231 sp->hbinterval = params->spp_hbinterval;
2236 if (hb_change) {
2237 if (trans) {
2238 trans->param_flags =
2239 (trans->param_flags & ~SPP_HB) | hb_change;
2240 } else if (asoc) {
2241 asoc->param_flags =
2242 (asoc->param_flags & ~SPP_HB) | hb_change;
2243 } else {
2244 sp->param_flags =
2245 (sp->param_flags & ~SPP_HB) | hb_change;
2249 /* When Path MTU discovery is disabled the value specified here will
2250 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2251 * include the flag SPP_PMTUD_DISABLE for this field to have any
2252 * effect).
2254 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2255 if (trans) {
2256 trans->pathmtu = params->spp_pathmtu;
2257 sctp_assoc_sync_pmtu(asoc);
2258 } else if (asoc) {
2259 asoc->pathmtu = params->spp_pathmtu;
2260 sctp_frag_point(asoc, params->spp_pathmtu);
2261 } else {
2262 sp->pathmtu = params->spp_pathmtu;
2266 if (pmtud_change) {
2267 if (trans) {
2268 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2269 (params->spp_flags & SPP_PMTUD_ENABLE);
2270 trans->param_flags =
2271 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2272 if (update) {
2273 sctp_transport_pmtu(trans);
2274 sctp_assoc_sync_pmtu(asoc);
2276 } else if (asoc) {
2277 asoc->param_flags =
2278 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2279 } else {
2280 sp->param_flags =
2281 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2285 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2286 * value of this field is ignored. Note also that a value of zero
2287 * indicates the current setting should be left unchanged.
2289 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2290 if (trans) {
2291 trans->sackdelay =
2292 msecs_to_jiffies(params->spp_sackdelay);
2293 } else if (asoc) {
2294 asoc->sackdelay =
2295 msecs_to_jiffies(params->spp_sackdelay);
2296 } else {
2297 sp->sackdelay = params->spp_sackdelay;
2301 if (sackdelay_change) {
2302 if (trans) {
2303 trans->param_flags =
2304 (trans->param_flags & ~SPP_SACKDELAY) |
2305 sackdelay_change;
2306 } else if (asoc) {
2307 asoc->param_flags =
2308 (asoc->param_flags & ~SPP_SACKDELAY) |
2309 sackdelay_change;
2310 } else {
2311 sp->param_flags =
2312 (sp->param_flags & ~SPP_SACKDELAY) |
2313 sackdelay_change;
2317 /* Note that a value of zero indicates the current setting should be
2318 left unchanged.
2320 if (params->spp_pathmaxrxt) {
2321 if (trans) {
2322 trans->pathmaxrxt = params->spp_pathmaxrxt;
2323 } else if (asoc) {
2324 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2325 } else {
2326 sp->pathmaxrxt = params->spp_pathmaxrxt;
2330 return 0;
2333 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2334 char __user *optval,
2335 unsigned int optlen)
2337 struct sctp_paddrparams params;
2338 struct sctp_transport *trans = NULL;
2339 struct sctp_association *asoc = NULL;
2340 struct sctp_sock *sp = sctp_sk(sk);
2341 int error;
2342 int hb_change, pmtud_change, sackdelay_change;
2344 if (optlen != sizeof(struct sctp_paddrparams))
2345 return - EINVAL;
2347 if (copy_from_user(&params, optval, optlen))
2348 return -EFAULT;
2350 /* Validate flags and value parameters. */
2351 hb_change = params.spp_flags & SPP_HB;
2352 pmtud_change = params.spp_flags & SPP_PMTUD;
2353 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2355 if (hb_change == SPP_HB ||
2356 pmtud_change == SPP_PMTUD ||
2357 sackdelay_change == SPP_SACKDELAY ||
2358 params.spp_sackdelay > 500 ||
2359 (params.spp_pathmtu &&
2360 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2361 return -EINVAL;
2363 /* If an address other than INADDR_ANY is specified, and
2364 * no transport is found, then the request is invalid.
2366 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2367 trans = sctp_addr_id2transport(sk, &params.spp_address,
2368 params.spp_assoc_id);
2369 if (!trans)
2370 return -EINVAL;
2373 /* Get association, if assoc_id != 0 and the socket is a one
2374 * to many style socket, and an association was not found, then
2375 * the id was invalid.
2377 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2378 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2379 return -EINVAL;
2381 /* Heartbeat demand can only be sent on a transport or
2382 * association, but not a socket.
2384 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2385 return -EINVAL;
2387 /* Process parameters. */
2388 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2389 hb_change, pmtud_change,
2390 sackdelay_change);
2392 if (error)
2393 return error;
2395 /* If changes are for association, also apply parameters to each
2396 * transport.
2398 if (!trans && asoc) {
2399 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2400 transports) {
2401 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2402 hb_change, pmtud_change,
2403 sackdelay_change);
2407 return 0;
2411 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2413 * This option will effect the way delayed acks are performed. This
2414 * option allows you to get or set the delayed ack time, in
2415 * milliseconds. It also allows changing the delayed ack frequency.
2416 * Changing the frequency to 1 disables the delayed sack algorithm. If
2417 * the assoc_id is 0, then this sets or gets the endpoints default
2418 * values. If the assoc_id field is non-zero, then the set or get
2419 * effects the specified association for the one to many model (the
2420 * assoc_id field is ignored by the one to one model). Note that if
2421 * sack_delay or sack_freq are 0 when setting this option, then the
2422 * current values will remain unchanged.
2424 * struct sctp_sack_info {
2425 * sctp_assoc_t sack_assoc_id;
2426 * uint32_t sack_delay;
2427 * uint32_t sack_freq;
2428 * };
2430 * sack_assoc_id - This parameter, indicates which association the user
2431 * is performing an action upon. Note that if this field's value is
2432 * zero then the endpoints default value is changed (effecting future
2433 * associations only).
2435 * sack_delay - This parameter contains the number of milliseconds that
2436 * the user is requesting the delayed ACK timer be set to. Note that
2437 * this value is defined in the standard to be between 200 and 500
2438 * milliseconds.
2440 * sack_freq - This parameter contains the number of packets that must
2441 * be received before a sack is sent without waiting for the delay
2442 * timer to expire. The default value for this is 2, setting this
2443 * value to 1 will disable the delayed sack algorithm.
2446 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2447 char __user *optval, unsigned int optlen)
2449 struct sctp_sack_info params;
2450 struct sctp_transport *trans = NULL;
2451 struct sctp_association *asoc = NULL;
2452 struct sctp_sock *sp = sctp_sk(sk);
2454 if (optlen == sizeof(struct sctp_sack_info)) {
2455 if (copy_from_user(&params, optval, optlen))
2456 return -EFAULT;
2458 if (params.sack_delay == 0 && params.sack_freq == 0)
2459 return 0;
2460 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2461 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2462 "in delayed_ack socket option deprecated\n");
2463 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2464 if (copy_from_user(&params, optval, optlen))
2465 return -EFAULT;
2467 if (params.sack_delay == 0)
2468 params.sack_freq = 1;
2469 else
2470 params.sack_freq = 0;
2471 } else
2472 return - EINVAL;
2474 /* Validate value parameter. */
2475 if (params.sack_delay > 500)
2476 return -EINVAL;
2478 /* Get association, if sack_assoc_id != 0 and the socket is a one
2479 * to many style socket, and an association was not found, then
2480 * the id was invalid.
2482 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2483 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2484 return -EINVAL;
2486 if (params.sack_delay) {
2487 if (asoc) {
2488 asoc->sackdelay =
2489 msecs_to_jiffies(params.sack_delay);
2490 asoc->param_flags =
2491 (asoc->param_flags & ~SPP_SACKDELAY) |
2492 SPP_SACKDELAY_ENABLE;
2493 } else {
2494 sp->sackdelay = params.sack_delay;
2495 sp->param_flags =
2496 (sp->param_flags & ~SPP_SACKDELAY) |
2497 SPP_SACKDELAY_ENABLE;
2501 if (params.sack_freq == 1) {
2502 if (asoc) {
2503 asoc->param_flags =
2504 (asoc->param_flags & ~SPP_SACKDELAY) |
2505 SPP_SACKDELAY_DISABLE;
2506 } else {
2507 sp->param_flags =
2508 (sp->param_flags & ~SPP_SACKDELAY) |
2509 SPP_SACKDELAY_DISABLE;
2511 } else if (params.sack_freq > 1) {
2512 if (asoc) {
2513 asoc->sackfreq = params.sack_freq;
2514 asoc->param_flags =
2515 (asoc->param_flags & ~SPP_SACKDELAY) |
2516 SPP_SACKDELAY_ENABLE;
2517 } else {
2518 sp->sackfreq = params.sack_freq;
2519 sp->param_flags =
2520 (sp->param_flags & ~SPP_SACKDELAY) |
2521 SPP_SACKDELAY_ENABLE;
2525 /* If change is for association, also apply to each transport. */
2526 if (asoc) {
2527 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2528 transports) {
2529 if (params.sack_delay) {
2530 trans->sackdelay =
2531 msecs_to_jiffies(params.sack_delay);
2532 trans->param_flags =
2533 (trans->param_flags & ~SPP_SACKDELAY) |
2534 SPP_SACKDELAY_ENABLE;
2536 if (params.sack_freq == 1) {
2537 trans->param_flags =
2538 (trans->param_flags & ~SPP_SACKDELAY) |
2539 SPP_SACKDELAY_DISABLE;
2540 } else if (params.sack_freq > 1) {
2541 trans->sackfreq = params.sack_freq;
2542 trans->param_flags =
2543 (trans->param_flags & ~SPP_SACKDELAY) |
2544 SPP_SACKDELAY_ENABLE;
2549 return 0;
2552 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2554 * Applications can specify protocol parameters for the default association
2555 * initialization. The option name argument to setsockopt() and getsockopt()
2556 * is SCTP_INITMSG.
2558 * Setting initialization parameters is effective only on an unconnected
2559 * socket (for UDP-style sockets only future associations are effected
2560 * by the change). With TCP-style sockets, this option is inherited by
2561 * sockets derived from a listener socket.
2563 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2565 struct sctp_initmsg sinit;
2566 struct sctp_sock *sp = sctp_sk(sk);
2568 if (optlen != sizeof(struct sctp_initmsg))
2569 return -EINVAL;
2570 if (copy_from_user(&sinit, optval, optlen))
2571 return -EFAULT;
2573 if (sinit.sinit_num_ostreams)
2574 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2575 if (sinit.sinit_max_instreams)
2576 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2577 if (sinit.sinit_max_attempts)
2578 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2579 if (sinit.sinit_max_init_timeo)
2580 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2582 return 0;
2586 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2588 * Applications that wish to use the sendto() system call may wish to
2589 * specify a default set of parameters that would normally be supplied
2590 * through the inclusion of ancillary data. This socket option allows
2591 * such an application to set the default sctp_sndrcvinfo structure.
2592 * The application that wishes to use this socket option simply passes
2593 * in to this call the sctp_sndrcvinfo structure defined in Section
2594 * 5.2.2) The input parameters accepted by this call include
2595 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2596 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2597 * to this call if the caller is using the UDP model.
2599 static int sctp_setsockopt_default_send_param(struct sock *sk,
2600 char __user *optval,
2601 unsigned int optlen)
2603 struct sctp_sndrcvinfo info;
2604 struct sctp_association *asoc;
2605 struct sctp_sock *sp = sctp_sk(sk);
2607 if (optlen != sizeof(struct sctp_sndrcvinfo))
2608 return -EINVAL;
2609 if (copy_from_user(&info, optval, optlen))
2610 return -EFAULT;
2612 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2613 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2614 return -EINVAL;
2616 if (asoc) {
2617 asoc->default_stream = info.sinfo_stream;
2618 asoc->default_flags = info.sinfo_flags;
2619 asoc->default_ppid = info.sinfo_ppid;
2620 asoc->default_context = info.sinfo_context;
2621 asoc->default_timetolive = info.sinfo_timetolive;
2622 } else {
2623 sp->default_stream = info.sinfo_stream;
2624 sp->default_flags = info.sinfo_flags;
2625 sp->default_ppid = info.sinfo_ppid;
2626 sp->default_context = info.sinfo_context;
2627 sp->default_timetolive = info.sinfo_timetolive;
2630 return 0;
2633 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2635 * Requests that the local SCTP stack use the enclosed peer address as
2636 * the association primary. The enclosed address must be one of the
2637 * association peer's addresses.
2639 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2640 unsigned int optlen)
2642 struct sctp_prim prim;
2643 struct sctp_transport *trans;
2645 if (optlen != sizeof(struct sctp_prim))
2646 return -EINVAL;
2648 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2649 return -EFAULT;
2651 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2652 if (!trans)
2653 return -EINVAL;
2655 sctp_assoc_set_primary(trans->asoc, trans);
2657 return 0;
2661 * 7.1.5 SCTP_NODELAY
2663 * Turn on/off any Nagle-like algorithm. This means that packets are
2664 * generally sent as soon as possible and no unnecessary delays are
2665 * introduced, at the cost of more packets in the network. Expects an
2666 * integer boolean flag.
2668 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2669 unsigned int optlen)
2671 int val;
2673 if (optlen < sizeof(int))
2674 return -EINVAL;
2675 if (get_user(val, (int __user *)optval))
2676 return -EFAULT;
2678 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2679 return 0;
2684 * 7.1.1 SCTP_RTOINFO
2686 * The protocol parameters used to initialize and bound retransmission
2687 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2688 * and modify these parameters.
2689 * All parameters are time values, in milliseconds. A value of 0, when
2690 * modifying the parameters, indicates that the current value should not
2691 * be changed.
2694 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2696 struct sctp_rtoinfo rtoinfo;
2697 struct sctp_association *asoc;
2699 if (optlen != sizeof (struct sctp_rtoinfo))
2700 return -EINVAL;
2702 if (copy_from_user(&rtoinfo, optval, optlen))
2703 return -EFAULT;
2705 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2707 /* Set the values to the specific association */
2708 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2709 return -EINVAL;
2711 if (asoc) {
2712 if (rtoinfo.srto_initial != 0)
2713 asoc->rto_initial =
2714 msecs_to_jiffies(rtoinfo.srto_initial);
2715 if (rtoinfo.srto_max != 0)
2716 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2717 if (rtoinfo.srto_min != 0)
2718 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2719 } else {
2720 /* If there is no association or the association-id = 0
2721 * set the values to the endpoint.
2723 struct sctp_sock *sp = sctp_sk(sk);
2725 if (rtoinfo.srto_initial != 0)
2726 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2727 if (rtoinfo.srto_max != 0)
2728 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2729 if (rtoinfo.srto_min != 0)
2730 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2733 return 0;
2738 * 7.1.2 SCTP_ASSOCINFO
2740 * This option is used to tune the maximum retransmission attempts
2741 * of the association.
2742 * Returns an error if the new association retransmission value is
2743 * greater than the sum of the retransmission value of the peer.
2744 * See [SCTP] for more information.
2747 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2750 struct sctp_assocparams assocparams;
2751 struct sctp_association *asoc;
2753 if (optlen != sizeof(struct sctp_assocparams))
2754 return -EINVAL;
2755 if (copy_from_user(&assocparams, optval, optlen))
2756 return -EFAULT;
2758 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2760 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2761 return -EINVAL;
2763 /* Set the values to the specific association */
2764 if (asoc) {
2765 if (assocparams.sasoc_asocmaxrxt != 0) {
2766 __u32 path_sum = 0;
2767 int paths = 0;
2768 struct sctp_transport *peer_addr;
2770 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2771 transports) {
2772 path_sum += peer_addr->pathmaxrxt;
2773 paths++;
2776 /* Only validate asocmaxrxt if we have more than
2777 * one path/transport. We do this because path
2778 * retransmissions are only counted when we have more
2779 * then one path.
2781 if (paths > 1 &&
2782 assocparams.sasoc_asocmaxrxt > path_sum)
2783 return -EINVAL;
2785 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2788 if (assocparams.sasoc_cookie_life != 0) {
2789 asoc->cookie_life.tv_sec =
2790 assocparams.sasoc_cookie_life / 1000;
2791 asoc->cookie_life.tv_usec =
2792 (assocparams.sasoc_cookie_life % 1000)
2793 * 1000;
2795 } else {
2796 /* Set the values to the endpoint */
2797 struct sctp_sock *sp = sctp_sk(sk);
2799 if (assocparams.sasoc_asocmaxrxt != 0)
2800 sp->assocparams.sasoc_asocmaxrxt =
2801 assocparams.sasoc_asocmaxrxt;
2802 if (assocparams.sasoc_cookie_life != 0)
2803 sp->assocparams.sasoc_cookie_life =
2804 assocparams.sasoc_cookie_life;
2806 return 0;
2810 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2812 * This socket option is a boolean flag which turns on or off mapped V4
2813 * addresses. If this option is turned on and the socket is type
2814 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2815 * If this option is turned off, then no mapping will be done of V4
2816 * addresses and a user will receive both PF_INET6 and PF_INET type
2817 * addresses on the socket.
2819 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2821 int val;
2822 struct sctp_sock *sp = sctp_sk(sk);
2824 if (optlen < sizeof(int))
2825 return -EINVAL;
2826 if (get_user(val, (int __user *)optval))
2827 return -EFAULT;
2828 if (val)
2829 sp->v4mapped = 1;
2830 else
2831 sp->v4mapped = 0;
2833 return 0;
2837 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2838 * This option will get or set the maximum size to put in any outgoing
2839 * SCTP DATA chunk. If a message is larger than this size it will be
2840 * fragmented by SCTP into the specified size. Note that the underlying
2841 * SCTP implementation may fragment into smaller sized chunks when the
2842 * PMTU of the underlying association is smaller than the value set by
2843 * the user. The default value for this option is '0' which indicates
2844 * the user is NOT limiting fragmentation and only the PMTU will effect
2845 * SCTP's choice of DATA chunk size. Note also that values set larger
2846 * than the maximum size of an IP datagram will effectively let SCTP
2847 * control fragmentation (i.e. the same as setting this option to 0).
2849 * The following structure is used to access and modify this parameter:
2851 * struct sctp_assoc_value {
2852 * sctp_assoc_t assoc_id;
2853 * uint32_t assoc_value;
2854 * };
2856 * assoc_id: This parameter is ignored for one-to-one style sockets.
2857 * For one-to-many style sockets this parameter indicates which
2858 * association the user is performing an action upon. Note that if
2859 * this field's value is zero then the endpoints default value is
2860 * changed (effecting future associations only).
2861 * assoc_value: This parameter specifies the maximum size in bytes.
2863 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2865 struct sctp_assoc_value params;
2866 struct sctp_association *asoc;
2867 struct sctp_sock *sp = sctp_sk(sk);
2868 int val;
2870 if (optlen == sizeof(int)) {
2871 printk(KERN_WARNING
2872 "SCTP: Use of int in maxseg socket option deprecated\n");
2873 printk(KERN_WARNING
2874 "SCTP: Use struct sctp_assoc_value instead\n");
2875 if (copy_from_user(&val, optval, optlen))
2876 return -EFAULT;
2877 params.assoc_id = 0;
2878 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2879 if (copy_from_user(&params, optval, optlen))
2880 return -EFAULT;
2881 val = params.assoc_value;
2882 } else
2883 return -EINVAL;
2885 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2886 return -EINVAL;
2888 asoc = sctp_id2assoc(sk, params.assoc_id);
2889 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2890 return -EINVAL;
2892 if (asoc) {
2893 if (val == 0) {
2894 val = asoc->pathmtu;
2895 val -= sp->pf->af->net_header_len;
2896 val -= sizeof(struct sctphdr) +
2897 sizeof(struct sctp_data_chunk);
2899 asoc->user_frag = val;
2900 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2901 } else {
2902 sp->user_frag = val;
2905 return 0;
2910 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2912 * Requests that the peer mark the enclosed address as the association
2913 * primary. The enclosed address must be one of the association's
2914 * locally bound addresses. The following structure is used to make a
2915 * set primary request:
2917 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2918 unsigned int optlen)
2920 struct sctp_sock *sp;
2921 struct sctp_endpoint *ep;
2922 struct sctp_association *asoc = NULL;
2923 struct sctp_setpeerprim prim;
2924 struct sctp_chunk *chunk;
2925 int err;
2927 sp = sctp_sk(sk);
2928 ep = sp->ep;
2930 if (!sctp_addip_enable)
2931 return -EPERM;
2933 if (optlen != sizeof(struct sctp_setpeerprim))
2934 return -EINVAL;
2936 if (copy_from_user(&prim, optval, optlen))
2937 return -EFAULT;
2939 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2940 if (!asoc)
2941 return -EINVAL;
2943 if (!asoc->peer.asconf_capable)
2944 return -EPERM;
2946 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2947 return -EPERM;
2949 if (!sctp_state(asoc, ESTABLISHED))
2950 return -ENOTCONN;
2952 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2953 return -EADDRNOTAVAIL;
2955 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2956 chunk = sctp_make_asconf_set_prim(asoc,
2957 (union sctp_addr *)&prim.sspp_addr);
2958 if (!chunk)
2959 return -ENOMEM;
2961 err = sctp_send_asconf(asoc, chunk);
2963 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2965 return err;
2968 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2969 unsigned int optlen)
2971 struct sctp_setadaptation adaptation;
2973 if (optlen != sizeof(struct sctp_setadaptation))
2974 return -EINVAL;
2975 if (copy_from_user(&adaptation, optval, optlen))
2976 return -EFAULT;
2978 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2980 return 0;
2984 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2986 * The context field in the sctp_sndrcvinfo structure is normally only
2987 * used when a failed message is retrieved holding the value that was
2988 * sent down on the actual send call. This option allows the setting of
2989 * a default context on an association basis that will be received on
2990 * reading messages from the peer. This is especially helpful in the
2991 * one-2-many model for an application to keep some reference to an
2992 * internal state machine that is processing messages on the
2993 * association. Note that the setting of this value only effects
2994 * received messages from the peer and does not effect the value that is
2995 * saved with outbound messages.
2997 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2998 unsigned int optlen)
3000 struct sctp_assoc_value params;
3001 struct sctp_sock *sp;
3002 struct sctp_association *asoc;
3004 if (optlen != sizeof(struct sctp_assoc_value))
3005 return -EINVAL;
3006 if (copy_from_user(&params, optval, optlen))
3007 return -EFAULT;
3009 sp = sctp_sk(sk);
3011 if (params.assoc_id != 0) {
3012 asoc = sctp_id2assoc(sk, params.assoc_id);
3013 if (!asoc)
3014 return -EINVAL;
3015 asoc->default_rcv_context = params.assoc_value;
3016 } else {
3017 sp->default_rcv_context = params.assoc_value;
3020 return 0;
3024 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3026 * This options will at a minimum specify if the implementation is doing
3027 * fragmented interleave. Fragmented interleave, for a one to many
3028 * socket, is when subsequent calls to receive a message may return
3029 * parts of messages from different associations. Some implementations
3030 * may allow you to turn this value on or off. If so, when turned off,
3031 * no fragment interleave will occur (which will cause a head of line
3032 * blocking amongst multiple associations sharing the same one to many
3033 * socket). When this option is turned on, then each receive call may
3034 * come from a different association (thus the user must receive data
3035 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3036 * association each receive belongs to.
3038 * This option takes a boolean value. A non-zero value indicates that
3039 * fragmented interleave is on. A value of zero indicates that
3040 * fragmented interleave is off.
3042 * Note that it is important that an implementation that allows this
3043 * option to be turned on, have it off by default. Otherwise an unaware
3044 * application using the one to many model may become confused and act
3045 * incorrectly.
3047 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3048 char __user *optval,
3049 unsigned int optlen)
3051 int val;
3053 if (optlen != sizeof(int))
3054 return -EINVAL;
3055 if (get_user(val, (int __user *)optval))
3056 return -EFAULT;
3058 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3060 return 0;
3064 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3065 * (SCTP_PARTIAL_DELIVERY_POINT)
3067 * This option will set or get the SCTP partial delivery point. This
3068 * point is the size of a message where the partial delivery API will be
3069 * invoked to help free up rwnd space for the peer. Setting this to a
3070 * lower value will cause partial deliveries to happen more often. The
3071 * calls argument is an integer that sets or gets the partial delivery
3072 * point. Note also that the call will fail if the user attempts to set
3073 * this value larger than the socket receive buffer size.
3075 * Note that any single message having a length smaller than or equal to
3076 * the SCTP partial delivery point will be delivered in one single read
3077 * call as long as the user provided buffer is large enough to hold the
3078 * message.
3080 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3081 char __user *optval,
3082 unsigned int optlen)
3084 u32 val;
3086 if (optlen != sizeof(u32))
3087 return -EINVAL;
3088 if (get_user(val, (int __user *)optval))
3089 return -EFAULT;
3091 /* Note: We double the receive buffer from what the user sets
3092 * it to be, also initial rwnd is based on rcvbuf/2.
3094 if (val > (sk->sk_rcvbuf >> 1))
3095 return -EINVAL;
3097 sctp_sk(sk)->pd_point = val;
3099 return 0; /* is this the right error code? */
3103 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3105 * This option will allow a user to change the maximum burst of packets
3106 * that can be emitted by this association. Note that the default value
3107 * is 4, and some implementations may restrict this setting so that it
3108 * can only be lowered.
3110 * NOTE: This text doesn't seem right. Do this on a socket basis with
3111 * future associations inheriting the socket value.
3113 static int sctp_setsockopt_maxburst(struct sock *sk,
3114 char __user *optval,
3115 unsigned int optlen)
3117 struct sctp_assoc_value params;
3118 struct sctp_sock *sp;
3119 struct sctp_association *asoc;
3120 int val;
3121 int assoc_id = 0;
3123 if (optlen == sizeof(int)) {
3124 printk(KERN_WARNING
3125 "SCTP: Use of int in max_burst socket option deprecated\n");
3126 printk(KERN_WARNING
3127 "SCTP: Use struct sctp_assoc_value instead\n");
3128 if (copy_from_user(&val, optval, optlen))
3129 return -EFAULT;
3130 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3131 if (copy_from_user(&params, optval, optlen))
3132 return -EFAULT;
3133 val = params.assoc_value;
3134 assoc_id = params.assoc_id;
3135 } else
3136 return -EINVAL;
3138 sp = sctp_sk(sk);
3140 if (assoc_id != 0) {
3141 asoc = sctp_id2assoc(sk, assoc_id);
3142 if (!asoc)
3143 return -EINVAL;
3144 asoc->max_burst = val;
3145 } else
3146 sp->max_burst = val;
3148 return 0;
3152 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3154 * This set option adds a chunk type that the user is requesting to be
3155 * received only in an authenticated way. Changes to the list of chunks
3156 * will only effect future associations on the socket.
3158 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3159 char __user *optval,
3160 unsigned int optlen)
3162 struct sctp_authchunk val;
3164 if (!sctp_auth_enable)
3165 return -EACCES;
3167 if (optlen != sizeof(struct sctp_authchunk))
3168 return -EINVAL;
3169 if (copy_from_user(&val, optval, optlen))
3170 return -EFAULT;
3172 switch (val.sauth_chunk) {
3173 case SCTP_CID_INIT:
3174 case SCTP_CID_INIT_ACK:
3175 case SCTP_CID_SHUTDOWN_COMPLETE:
3176 case SCTP_CID_AUTH:
3177 return -EINVAL;
3180 /* add this chunk id to the endpoint */
3181 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3185 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3187 * This option gets or sets the list of HMAC algorithms that the local
3188 * endpoint requires the peer to use.
3190 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3191 char __user *optval,
3192 unsigned int optlen)
3194 struct sctp_hmacalgo *hmacs;
3195 u32 idents;
3196 int err;
3198 if (!sctp_auth_enable)
3199 return -EACCES;
3201 if (optlen < sizeof(struct sctp_hmacalgo))
3202 return -EINVAL;
3204 hmacs = kmalloc(optlen, GFP_KERNEL);
3205 if (!hmacs)
3206 return -ENOMEM;
3208 if (copy_from_user(hmacs, optval, optlen)) {
3209 err = -EFAULT;
3210 goto out;
3213 idents = hmacs->shmac_num_idents;
3214 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3215 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3216 err = -EINVAL;
3217 goto out;
3220 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3221 out:
3222 kfree(hmacs);
3223 return err;
3227 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3229 * This option will set a shared secret key which is used to build an
3230 * association shared key.
3232 static int sctp_setsockopt_auth_key(struct sock *sk,
3233 char __user *optval,
3234 unsigned int optlen)
3236 struct sctp_authkey *authkey;
3237 struct sctp_association *asoc;
3238 int ret;
3240 if (!sctp_auth_enable)
3241 return -EACCES;
3243 if (optlen <= sizeof(struct sctp_authkey))
3244 return -EINVAL;
3246 authkey = kmalloc(optlen, GFP_KERNEL);
3247 if (!authkey)
3248 return -ENOMEM;
3250 if (copy_from_user(authkey, optval, optlen)) {
3251 ret = -EFAULT;
3252 goto out;
3255 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3256 ret = -EINVAL;
3257 goto out;
3260 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3261 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3262 ret = -EINVAL;
3263 goto out;
3266 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3267 out:
3268 kfree(authkey);
3269 return ret;
3273 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3275 * This option will get or set the active shared key to be used to build
3276 * the association shared key.
3278 static int sctp_setsockopt_active_key(struct sock *sk,
3279 char __user *optval,
3280 unsigned int optlen)
3282 struct sctp_authkeyid val;
3283 struct sctp_association *asoc;
3285 if (!sctp_auth_enable)
3286 return -EACCES;
3288 if (optlen != sizeof(struct sctp_authkeyid))
3289 return -EINVAL;
3290 if (copy_from_user(&val, optval, optlen))
3291 return -EFAULT;
3293 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3294 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3295 return -EINVAL;
3297 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3298 val.scact_keynumber);
3302 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3304 * This set option will delete a shared secret key from use.
3306 static int sctp_setsockopt_del_key(struct sock *sk,
3307 char __user *optval,
3308 unsigned int optlen)
3310 struct sctp_authkeyid val;
3311 struct sctp_association *asoc;
3313 if (!sctp_auth_enable)
3314 return -EACCES;
3316 if (optlen != sizeof(struct sctp_authkeyid))
3317 return -EINVAL;
3318 if (copy_from_user(&val, optval, optlen))
3319 return -EFAULT;
3321 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3322 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3323 return -EINVAL;
3325 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3326 val.scact_keynumber);
3331 /* API 6.2 setsockopt(), getsockopt()
3333 * Applications use setsockopt() and getsockopt() to set or retrieve
3334 * socket options. Socket options are used to change the default
3335 * behavior of sockets calls. They are described in Section 7.
3337 * The syntax is:
3339 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3340 * int __user *optlen);
3341 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3342 * int optlen);
3344 * sd - the socket descript.
3345 * level - set to IPPROTO_SCTP for all SCTP options.
3346 * optname - the option name.
3347 * optval - the buffer to store the value of the option.
3348 * optlen - the size of the buffer.
3350 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3351 char __user *optval, unsigned int optlen)
3353 int retval = 0;
3355 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3356 sk, optname);
3358 /* I can hardly begin to describe how wrong this is. This is
3359 * so broken as to be worse than useless. The API draft
3360 * REALLY is NOT helpful here... I am not convinced that the
3361 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3362 * are at all well-founded.
3364 if (level != SOL_SCTP) {
3365 struct sctp_af *af = sctp_sk(sk)->pf->af;
3366 retval = af->setsockopt(sk, level, optname, optval, optlen);
3367 goto out_nounlock;
3370 sctp_lock_sock(sk);
3372 switch (optname) {
3373 case SCTP_SOCKOPT_BINDX_ADD:
3374 /* 'optlen' is the size of the addresses buffer. */
3375 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3376 optlen, SCTP_BINDX_ADD_ADDR);
3377 break;
3379 case SCTP_SOCKOPT_BINDX_REM:
3380 /* 'optlen' is the size of the addresses buffer. */
3381 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3382 optlen, SCTP_BINDX_REM_ADDR);
3383 break;
3385 case SCTP_SOCKOPT_CONNECTX_OLD:
3386 /* 'optlen' is the size of the addresses buffer. */
3387 retval = sctp_setsockopt_connectx_old(sk,
3388 (struct sockaddr __user *)optval,
3389 optlen);
3390 break;
3392 case SCTP_SOCKOPT_CONNECTX:
3393 /* 'optlen' is the size of the addresses buffer. */
3394 retval = sctp_setsockopt_connectx(sk,
3395 (struct sockaddr __user *)optval,
3396 optlen);
3397 break;
3399 case SCTP_DISABLE_FRAGMENTS:
3400 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3401 break;
3403 case SCTP_EVENTS:
3404 retval = sctp_setsockopt_events(sk, optval, optlen);
3405 break;
3407 case SCTP_AUTOCLOSE:
3408 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3409 break;
3411 case SCTP_PEER_ADDR_PARAMS:
3412 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3413 break;
3415 case SCTP_DELAYED_ACK:
3416 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3417 break;
3418 case SCTP_PARTIAL_DELIVERY_POINT:
3419 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3420 break;
3422 case SCTP_INITMSG:
3423 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3424 break;
3425 case SCTP_DEFAULT_SEND_PARAM:
3426 retval = sctp_setsockopt_default_send_param(sk, optval,
3427 optlen);
3428 break;
3429 case SCTP_PRIMARY_ADDR:
3430 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3431 break;
3432 case SCTP_SET_PEER_PRIMARY_ADDR:
3433 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3434 break;
3435 case SCTP_NODELAY:
3436 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3437 break;
3438 case SCTP_RTOINFO:
3439 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3440 break;
3441 case SCTP_ASSOCINFO:
3442 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3443 break;
3444 case SCTP_I_WANT_MAPPED_V4_ADDR:
3445 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3446 break;
3447 case SCTP_MAXSEG:
3448 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3449 break;
3450 case SCTP_ADAPTATION_LAYER:
3451 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3452 break;
3453 case SCTP_CONTEXT:
3454 retval = sctp_setsockopt_context(sk, optval, optlen);
3455 break;
3456 case SCTP_FRAGMENT_INTERLEAVE:
3457 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3458 break;
3459 case SCTP_MAX_BURST:
3460 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3461 break;
3462 case SCTP_AUTH_CHUNK:
3463 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3464 break;
3465 case SCTP_HMAC_IDENT:
3466 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3467 break;
3468 case SCTP_AUTH_KEY:
3469 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3470 break;
3471 case SCTP_AUTH_ACTIVE_KEY:
3472 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3473 break;
3474 case SCTP_AUTH_DELETE_KEY:
3475 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3476 break;
3477 default:
3478 retval = -ENOPROTOOPT;
3479 break;
3482 sctp_release_sock(sk);
3484 out_nounlock:
3485 return retval;
3488 /* API 3.1.6 connect() - UDP Style Syntax
3490 * An application may use the connect() call in the UDP model to initiate an
3491 * association without sending data.
3493 * The syntax is:
3495 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3497 * sd: the socket descriptor to have a new association added to.
3499 * nam: the address structure (either struct sockaddr_in or struct
3500 * sockaddr_in6 defined in RFC2553 [7]).
3502 * len: the size of the address.
3504 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3505 int addr_len)
3507 int err = 0;
3508 struct sctp_af *af;
3510 sctp_lock_sock(sk);
3512 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3513 __func__, sk, addr, addr_len);
3515 /* Validate addr_len before calling common connect/connectx routine. */
3516 af = sctp_get_af_specific(addr->sa_family);
3517 if (!af || addr_len < af->sockaddr_len) {
3518 err = -EINVAL;
3519 } else {
3520 /* Pass correct addr len to common routine (so it knows there
3521 * is only one address being passed.
3523 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3526 sctp_release_sock(sk);
3527 return err;
3530 /* FIXME: Write comments. */
3531 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3533 return -EOPNOTSUPP; /* STUB */
3536 /* 4.1.4 accept() - TCP Style Syntax
3538 * Applications use accept() call to remove an established SCTP
3539 * association from the accept queue of the endpoint. A new socket
3540 * descriptor will be returned from accept() to represent the newly
3541 * formed association.
3543 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3545 struct sctp_sock *sp;
3546 struct sctp_endpoint *ep;
3547 struct sock *newsk = NULL;
3548 struct sctp_association *asoc;
3549 long timeo;
3550 int error = 0;
3552 sctp_lock_sock(sk);
3554 sp = sctp_sk(sk);
3555 ep = sp->ep;
3557 if (!sctp_style(sk, TCP)) {
3558 error = -EOPNOTSUPP;
3559 goto out;
3562 if (!sctp_sstate(sk, LISTENING)) {
3563 error = -EINVAL;
3564 goto out;
3567 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3569 error = sctp_wait_for_accept(sk, timeo);
3570 if (error)
3571 goto out;
3573 /* We treat the list of associations on the endpoint as the accept
3574 * queue and pick the first association on the list.
3576 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3578 newsk = sp->pf->create_accept_sk(sk, asoc);
3579 if (!newsk) {
3580 error = -ENOMEM;
3581 goto out;
3584 /* Populate the fields of the newsk from the oldsk and migrate the
3585 * asoc to the newsk.
3587 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3589 out:
3590 sctp_release_sock(sk);
3591 *err = error;
3592 return newsk;
3595 /* The SCTP ioctl handler. */
3596 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3598 return -ENOIOCTLCMD;
3601 /* This is the function which gets called during socket creation to
3602 * initialized the SCTP-specific portion of the sock.
3603 * The sock structure should already be zero-filled memory.
3605 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3607 struct sctp_endpoint *ep;
3608 struct sctp_sock *sp;
3610 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3612 sp = sctp_sk(sk);
3614 /* Initialize the SCTP per socket area. */
3615 switch (sk->sk_type) {
3616 case SOCK_SEQPACKET:
3617 sp->type = SCTP_SOCKET_UDP;
3618 break;
3619 case SOCK_STREAM:
3620 sp->type = SCTP_SOCKET_TCP;
3621 break;
3622 default:
3623 return -ESOCKTNOSUPPORT;
3626 /* Initialize default send parameters. These parameters can be
3627 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3629 sp->default_stream = 0;
3630 sp->default_ppid = 0;
3631 sp->default_flags = 0;
3632 sp->default_context = 0;
3633 sp->default_timetolive = 0;
3635 sp->default_rcv_context = 0;
3636 sp->max_burst = sctp_max_burst;
3638 /* Initialize default setup parameters. These parameters
3639 * can be modified with the SCTP_INITMSG socket option or
3640 * overridden by the SCTP_INIT CMSG.
3642 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3643 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3644 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3645 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3647 /* Initialize default RTO related parameters. These parameters can
3648 * be modified for with the SCTP_RTOINFO socket option.
3650 sp->rtoinfo.srto_initial = sctp_rto_initial;
3651 sp->rtoinfo.srto_max = sctp_rto_max;
3652 sp->rtoinfo.srto_min = sctp_rto_min;
3654 /* Initialize default association related parameters. These parameters
3655 * can be modified with the SCTP_ASSOCINFO socket option.
3657 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3658 sp->assocparams.sasoc_number_peer_destinations = 0;
3659 sp->assocparams.sasoc_peer_rwnd = 0;
3660 sp->assocparams.sasoc_local_rwnd = 0;
3661 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3663 /* Initialize default event subscriptions. By default, all the
3664 * options are off.
3666 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3668 /* Default Peer Address Parameters. These defaults can
3669 * be modified via SCTP_PEER_ADDR_PARAMS
3671 sp->hbinterval = sctp_hb_interval;
3672 sp->pathmaxrxt = sctp_max_retrans_path;
3673 sp->pathmtu = 0; // allow default discovery
3674 sp->sackdelay = sctp_sack_timeout;
3675 sp->sackfreq = 2;
3676 sp->param_flags = SPP_HB_ENABLE |
3677 SPP_PMTUD_ENABLE |
3678 SPP_SACKDELAY_ENABLE;
3680 /* If enabled no SCTP message fragmentation will be performed.
3681 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3683 sp->disable_fragments = 0;
3685 /* Enable Nagle algorithm by default. */
3686 sp->nodelay = 0;
3688 /* Enable by default. */
3689 sp->v4mapped = 1;
3691 /* Auto-close idle associations after the configured
3692 * number of seconds. A value of 0 disables this
3693 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3694 * for UDP-style sockets only.
3696 sp->autoclose = 0;
3698 /* User specified fragmentation limit. */
3699 sp->user_frag = 0;
3701 sp->adaptation_ind = 0;
3703 sp->pf = sctp_get_pf_specific(sk->sk_family);
3705 /* Control variables for partial data delivery. */
3706 atomic_set(&sp->pd_mode, 0);
3707 skb_queue_head_init(&sp->pd_lobby);
3708 sp->frag_interleave = 0;
3710 /* Create a per socket endpoint structure. Even if we
3711 * change the data structure relationships, this may still
3712 * be useful for storing pre-connect address information.
3714 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3715 if (!ep)
3716 return -ENOMEM;
3718 sp->ep = ep;
3719 sp->hmac = NULL;
3721 SCTP_DBG_OBJCNT_INC(sock);
3722 percpu_counter_inc(&sctp_sockets_allocated);
3724 /* Set socket backlog limit. */
3725 sk->sk_backlog.limit = sysctl_sctp_rmem[1];
3727 local_bh_disable();
3728 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3729 local_bh_enable();
3731 return 0;
3734 /* Cleanup any SCTP per socket resources. */
3735 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3737 struct sctp_endpoint *ep;
3739 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3741 /* Release our hold on the endpoint. */
3742 ep = sctp_sk(sk)->ep;
3743 sctp_endpoint_free(ep);
3744 percpu_counter_dec(&sctp_sockets_allocated);
3745 local_bh_disable();
3746 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3747 local_bh_enable();
3750 /* API 4.1.7 shutdown() - TCP Style Syntax
3751 * int shutdown(int socket, int how);
3753 * sd - the socket descriptor of the association to be closed.
3754 * how - Specifies the type of shutdown. The values are
3755 * as follows:
3756 * SHUT_RD
3757 * Disables further receive operations. No SCTP
3758 * protocol action is taken.
3759 * SHUT_WR
3760 * Disables further send operations, and initiates
3761 * the SCTP shutdown sequence.
3762 * SHUT_RDWR
3763 * Disables further send and receive operations
3764 * and initiates the SCTP shutdown sequence.
3766 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3768 struct sctp_endpoint *ep;
3769 struct sctp_association *asoc;
3771 if (!sctp_style(sk, TCP))
3772 return;
3774 if (how & SEND_SHUTDOWN) {
3775 ep = sctp_sk(sk)->ep;
3776 if (!list_empty(&ep->asocs)) {
3777 asoc = list_entry(ep->asocs.next,
3778 struct sctp_association, asocs);
3779 sctp_primitive_SHUTDOWN(asoc, NULL);
3784 /* 7.2.1 Association Status (SCTP_STATUS)
3786 * Applications can retrieve current status information about an
3787 * association, including association state, peer receiver window size,
3788 * number of unacked data chunks, and number of data chunks pending
3789 * receipt. This information is read-only.
3791 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3792 char __user *optval,
3793 int __user *optlen)
3795 struct sctp_status status;
3796 struct sctp_association *asoc = NULL;
3797 struct sctp_transport *transport;
3798 sctp_assoc_t associd;
3799 int retval = 0;
3801 if (len < sizeof(status)) {
3802 retval = -EINVAL;
3803 goto out;
3806 len = sizeof(status);
3807 if (copy_from_user(&status, optval, len)) {
3808 retval = -EFAULT;
3809 goto out;
3812 associd = status.sstat_assoc_id;
3813 asoc = sctp_id2assoc(sk, associd);
3814 if (!asoc) {
3815 retval = -EINVAL;
3816 goto out;
3819 transport = asoc->peer.primary_path;
3821 status.sstat_assoc_id = sctp_assoc2id(asoc);
3822 status.sstat_state = asoc->state;
3823 status.sstat_rwnd = asoc->peer.rwnd;
3824 status.sstat_unackdata = asoc->unack_data;
3826 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3827 status.sstat_instrms = asoc->c.sinit_max_instreams;
3828 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3829 status.sstat_fragmentation_point = asoc->frag_point;
3830 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3831 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3832 transport->af_specific->sockaddr_len);
3833 /* Map ipv4 address into v4-mapped-on-v6 address. */
3834 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3835 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3836 status.sstat_primary.spinfo_state = transport->state;
3837 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3838 status.sstat_primary.spinfo_srtt = transport->srtt;
3839 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3840 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3842 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3843 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3845 if (put_user(len, optlen)) {
3846 retval = -EFAULT;
3847 goto out;
3850 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3851 len, status.sstat_state, status.sstat_rwnd,
3852 status.sstat_assoc_id);
3854 if (copy_to_user(optval, &status, len)) {
3855 retval = -EFAULT;
3856 goto out;
3859 out:
3860 return (retval);
3864 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3866 * Applications can retrieve information about a specific peer address
3867 * of an association, including its reachability state, congestion
3868 * window, and retransmission timer values. This information is
3869 * read-only.
3871 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3872 char __user *optval,
3873 int __user *optlen)
3875 struct sctp_paddrinfo pinfo;
3876 struct sctp_transport *transport;
3877 int retval = 0;
3879 if (len < sizeof(pinfo)) {
3880 retval = -EINVAL;
3881 goto out;
3884 len = sizeof(pinfo);
3885 if (copy_from_user(&pinfo, optval, len)) {
3886 retval = -EFAULT;
3887 goto out;
3890 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3891 pinfo.spinfo_assoc_id);
3892 if (!transport)
3893 return -EINVAL;
3895 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3896 pinfo.spinfo_state = transport->state;
3897 pinfo.spinfo_cwnd = transport->cwnd;
3898 pinfo.spinfo_srtt = transport->srtt;
3899 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3900 pinfo.spinfo_mtu = transport->pathmtu;
3902 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3903 pinfo.spinfo_state = SCTP_ACTIVE;
3905 if (put_user(len, optlen)) {
3906 retval = -EFAULT;
3907 goto out;
3910 if (copy_to_user(optval, &pinfo, len)) {
3911 retval = -EFAULT;
3912 goto out;
3915 out:
3916 return (retval);
3919 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3921 * This option is a on/off flag. If enabled no SCTP message
3922 * fragmentation will be performed. Instead if a message being sent
3923 * exceeds the current PMTU size, the message will NOT be sent and
3924 * instead a error will be indicated to the user.
3926 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3927 char __user *optval, int __user *optlen)
3929 int val;
3931 if (len < sizeof(int))
3932 return -EINVAL;
3934 len = sizeof(int);
3935 val = (sctp_sk(sk)->disable_fragments == 1);
3936 if (put_user(len, optlen))
3937 return -EFAULT;
3938 if (copy_to_user(optval, &val, len))
3939 return -EFAULT;
3940 return 0;
3943 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3945 * This socket option is used to specify various notifications and
3946 * ancillary data the user wishes to receive.
3948 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3949 int __user *optlen)
3951 if (len < sizeof(struct sctp_event_subscribe))
3952 return -EINVAL;
3953 len = sizeof(struct sctp_event_subscribe);
3954 if (put_user(len, optlen))
3955 return -EFAULT;
3956 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3957 return -EFAULT;
3958 return 0;
3961 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3963 * This socket option is applicable to the UDP-style socket only. When
3964 * set it will cause associations that are idle for more than the
3965 * specified number of seconds to automatically close. An association
3966 * being idle is defined an association that has NOT sent or received
3967 * user data. The special value of '0' indicates that no automatic
3968 * close of any associations should be performed. The option expects an
3969 * integer defining the number of seconds of idle time before an
3970 * association is closed.
3972 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3974 /* Applicable to UDP-style socket only */
3975 if (sctp_style(sk, TCP))
3976 return -EOPNOTSUPP;
3977 if (len < sizeof(int))
3978 return -EINVAL;
3979 len = sizeof(int);
3980 if (put_user(len, optlen))
3981 return -EFAULT;
3982 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3983 return -EFAULT;
3984 return 0;
3987 /* Helper routine to branch off an association to a new socket. */
3988 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3989 struct socket **sockp)
3991 struct sock *sk = asoc->base.sk;
3992 struct socket *sock;
3993 struct sctp_af *af;
3994 int err = 0;
3996 /* An association cannot be branched off from an already peeled-off
3997 * socket, nor is this supported for tcp style sockets.
3999 if (!sctp_style(sk, UDP))
4000 return -EINVAL;
4002 /* Create a new socket. */
4003 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4004 if (err < 0)
4005 return err;
4007 sctp_copy_sock(sock->sk, sk, asoc);
4009 /* Make peeled-off sockets more like 1-1 accepted sockets.
4010 * Set the daddr and initialize id to something more random
4012 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4013 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4015 /* Populate the fields of the newsk from the oldsk and migrate the
4016 * asoc to the newsk.
4018 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4020 *sockp = sock;
4022 return err;
4025 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4027 sctp_peeloff_arg_t peeloff;
4028 struct socket *newsock;
4029 int retval = 0;
4030 struct sctp_association *asoc;
4032 if (len < sizeof(sctp_peeloff_arg_t))
4033 return -EINVAL;
4034 len = sizeof(sctp_peeloff_arg_t);
4035 if (copy_from_user(&peeloff, optval, len))
4036 return -EFAULT;
4038 asoc = sctp_id2assoc(sk, peeloff.associd);
4039 if (!asoc) {
4040 retval = -EINVAL;
4041 goto out;
4044 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4046 retval = sctp_do_peeloff(asoc, &newsock);
4047 if (retval < 0)
4048 goto out;
4050 /* Map the socket to an unused fd that can be returned to the user. */
4051 retval = sock_map_fd(newsock, 0);
4052 if (retval < 0) {
4053 sock_release(newsock);
4054 goto out;
4057 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4058 __func__, sk, asoc, newsock->sk, retval);
4060 /* Return the fd mapped to the new socket. */
4061 peeloff.sd = retval;
4062 if (put_user(len, optlen))
4063 return -EFAULT;
4064 if (copy_to_user(optval, &peeloff, len))
4065 retval = -EFAULT;
4067 out:
4068 return retval;
4071 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4073 * Applications can enable or disable heartbeats for any peer address of
4074 * an association, modify an address's heartbeat interval, force a
4075 * heartbeat to be sent immediately, and adjust the address's maximum
4076 * number of retransmissions sent before an address is considered
4077 * unreachable. The following structure is used to access and modify an
4078 * address's parameters:
4080 * struct sctp_paddrparams {
4081 * sctp_assoc_t spp_assoc_id;
4082 * struct sockaddr_storage spp_address;
4083 * uint32_t spp_hbinterval;
4084 * uint16_t spp_pathmaxrxt;
4085 * uint32_t spp_pathmtu;
4086 * uint32_t spp_sackdelay;
4087 * uint32_t spp_flags;
4088 * };
4090 * spp_assoc_id - (one-to-many style socket) This is filled in the
4091 * application, and identifies the association for
4092 * this query.
4093 * spp_address - This specifies which address is of interest.
4094 * spp_hbinterval - This contains the value of the heartbeat interval,
4095 * in milliseconds. If a value of zero
4096 * is present in this field then no changes are to
4097 * be made to this parameter.
4098 * spp_pathmaxrxt - This contains the maximum number of
4099 * retransmissions before this address shall be
4100 * considered unreachable. If a value of zero
4101 * is present in this field then no changes are to
4102 * be made to this parameter.
4103 * spp_pathmtu - When Path MTU discovery is disabled the value
4104 * specified here will be the "fixed" path mtu.
4105 * Note that if the spp_address field is empty
4106 * then all associations on this address will
4107 * have this fixed path mtu set upon them.
4109 * spp_sackdelay - When delayed sack is enabled, this value specifies
4110 * the number of milliseconds that sacks will be delayed
4111 * for. This value will apply to all addresses of an
4112 * association if the spp_address field is empty. Note
4113 * also, that if delayed sack is enabled and this
4114 * value is set to 0, no change is made to the last
4115 * recorded delayed sack timer value.
4117 * spp_flags - These flags are used to control various features
4118 * on an association. The flag field may contain
4119 * zero or more of the following options.
4121 * SPP_HB_ENABLE - Enable heartbeats on the
4122 * specified address. Note that if the address
4123 * field is empty all addresses for the association
4124 * have heartbeats enabled upon them.
4126 * SPP_HB_DISABLE - Disable heartbeats on the
4127 * speicifed address. Note that if the address
4128 * field is empty all addresses for the association
4129 * will have their heartbeats disabled. Note also
4130 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4131 * mutually exclusive, only one of these two should
4132 * be specified. Enabling both fields will have
4133 * undetermined results.
4135 * SPP_HB_DEMAND - Request a user initiated heartbeat
4136 * to be made immediately.
4138 * SPP_PMTUD_ENABLE - This field will enable PMTU
4139 * discovery upon the specified address. Note that
4140 * if the address feild is empty then all addresses
4141 * on the association are effected.
4143 * SPP_PMTUD_DISABLE - This field will disable PMTU
4144 * discovery upon the specified address. Note that
4145 * if the address feild is empty then all addresses
4146 * on the association are effected. Not also that
4147 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4148 * exclusive. Enabling both will have undetermined
4149 * results.
4151 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4152 * on delayed sack. The time specified in spp_sackdelay
4153 * is used to specify the sack delay for this address. Note
4154 * that if spp_address is empty then all addresses will
4155 * enable delayed sack and take on the sack delay
4156 * value specified in spp_sackdelay.
4157 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4158 * off delayed sack. If the spp_address field is blank then
4159 * delayed sack is disabled for the entire association. Note
4160 * also that this field is mutually exclusive to
4161 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4162 * results.
4164 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4165 char __user *optval, int __user *optlen)
4167 struct sctp_paddrparams params;
4168 struct sctp_transport *trans = NULL;
4169 struct sctp_association *asoc = NULL;
4170 struct sctp_sock *sp = sctp_sk(sk);
4172 if (len < sizeof(struct sctp_paddrparams))
4173 return -EINVAL;
4174 len = sizeof(struct sctp_paddrparams);
4175 if (copy_from_user(&params, optval, len))
4176 return -EFAULT;
4178 /* If an address other than INADDR_ANY is specified, and
4179 * no transport is found, then the request is invalid.
4181 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4182 trans = sctp_addr_id2transport(sk, &params.spp_address,
4183 params.spp_assoc_id);
4184 if (!trans) {
4185 SCTP_DEBUG_PRINTK("Failed no transport\n");
4186 return -EINVAL;
4190 /* Get association, if assoc_id != 0 and the socket is a one
4191 * to many style socket, and an association was not found, then
4192 * the id was invalid.
4194 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4195 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4196 SCTP_DEBUG_PRINTK("Failed no association\n");
4197 return -EINVAL;
4200 if (trans) {
4201 /* Fetch transport values. */
4202 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4203 params.spp_pathmtu = trans->pathmtu;
4204 params.spp_pathmaxrxt = trans->pathmaxrxt;
4205 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4207 /*draft-11 doesn't say what to return in spp_flags*/
4208 params.spp_flags = trans->param_flags;
4209 } else if (asoc) {
4210 /* Fetch association values. */
4211 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4212 params.spp_pathmtu = asoc->pathmtu;
4213 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4214 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4216 /*draft-11 doesn't say what to return in spp_flags*/
4217 params.spp_flags = asoc->param_flags;
4218 } else {
4219 /* Fetch socket values. */
4220 params.spp_hbinterval = sp->hbinterval;
4221 params.spp_pathmtu = sp->pathmtu;
4222 params.spp_sackdelay = sp->sackdelay;
4223 params.spp_pathmaxrxt = sp->pathmaxrxt;
4225 /*draft-11 doesn't say what to return in spp_flags*/
4226 params.spp_flags = sp->param_flags;
4229 if (copy_to_user(optval, &params, len))
4230 return -EFAULT;
4232 if (put_user(len, optlen))
4233 return -EFAULT;
4235 return 0;
4239 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4241 * This option will effect the way delayed acks are performed. This
4242 * option allows you to get or set the delayed ack time, in
4243 * milliseconds. It also allows changing the delayed ack frequency.
4244 * Changing the frequency to 1 disables the delayed sack algorithm. If
4245 * the assoc_id is 0, then this sets or gets the endpoints default
4246 * values. If the assoc_id field is non-zero, then the set or get
4247 * effects the specified association for the one to many model (the
4248 * assoc_id field is ignored by the one to one model). Note that if
4249 * sack_delay or sack_freq are 0 when setting this option, then the
4250 * current values will remain unchanged.
4252 * struct sctp_sack_info {
4253 * sctp_assoc_t sack_assoc_id;
4254 * uint32_t sack_delay;
4255 * uint32_t sack_freq;
4256 * };
4258 * sack_assoc_id - This parameter, indicates which association the user
4259 * is performing an action upon. Note that if this field's value is
4260 * zero then the endpoints default value is changed (effecting future
4261 * associations only).
4263 * sack_delay - This parameter contains the number of milliseconds that
4264 * the user is requesting the delayed ACK timer be set to. Note that
4265 * this value is defined in the standard to be between 200 and 500
4266 * milliseconds.
4268 * sack_freq - This parameter contains the number of packets that must
4269 * be received before a sack is sent without waiting for the delay
4270 * timer to expire. The default value for this is 2, setting this
4271 * value to 1 will disable the delayed sack algorithm.
4273 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4274 char __user *optval,
4275 int __user *optlen)
4277 struct sctp_sack_info params;
4278 struct sctp_association *asoc = NULL;
4279 struct sctp_sock *sp = sctp_sk(sk);
4281 if (len >= sizeof(struct sctp_sack_info)) {
4282 len = sizeof(struct sctp_sack_info);
4284 if (copy_from_user(&params, optval, len))
4285 return -EFAULT;
4286 } else if (len == sizeof(struct sctp_assoc_value)) {
4287 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4288 "in delayed_ack socket option deprecated\n");
4289 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4290 if (copy_from_user(&params, optval, len))
4291 return -EFAULT;
4292 } else
4293 return - EINVAL;
4295 /* Get association, if sack_assoc_id != 0 and the socket is a one
4296 * to many style socket, and an association was not found, then
4297 * the id was invalid.
4299 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4300 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4301 return -EINVAL;
4303 if (asoc) {
4304 /* Fetch association values. */
4305 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4306 params.sack_delay = jiffies_to_msecs(
4307 asoc->sackdelay);
4308 params.sack_freq = asoc->sackfreq;
4310 } else {
4311 params.sack_delay = 0;
4312 params.sack_freq = 1;
4314 } else {
4315 /* Fetch socket values. */
4316 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4317 params.sack_delay = sp->sackdelay;
4318 params.sack_freq = sp->sackfreq;
4319 } else {
4320 params.sack_delay = 0;
4321 params.sack_freq = 1;
4325 if (copy_to_user(optval, &params, len))
4326 return -EFAULT;
4328 if (put_user(len, optlen))
4329 return -EFAULT;
4331 return 0;
4334 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4336 * Applications can specify protocol parameters for the default association
4337 * initialization. The option name argument to setsockopt() and getsockopt()
4338 * is SCTP_INITMSG.
4340 * Setting initialization parameters is effective only on an unconnected
4341 * socket (for UDP-style sockets only future associations are effected
4342 * by the change). With TCP-style sockets, this option is inherited by
4343 * sockets derived from a listener socket.
4345 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4347 if (len < sizeof(struct sctp_initmsg))
4348 return -EINVAL;
4349 len = sizeof(struct sctp_initmsg);
4350 if (put_user(len, optlen))
4351 return -EFAULT;
4352 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4353 return -EFAULT;
4354 return 0;
4358 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4359 char __user *optval, int __user *optlen)
4361 struct sctp_association *asoc;
4362 int cnt = 0;
4363 struct sctp_getaddrs getaddrs;
4364 struct sctp_transport *from;
4365 void __user *to;
4366 union sctp_addr temp;
4367 struct sctp_sock *sp = sctp_sk(sk);
4368 int addrlen;
4369 size_t space_left;
4370 int bytes_copied;
4372 if (len < sizeof(struct sctp_getaddrs))
4373 return -EINVAL;
4375 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4376 return -EFAULT;
4378 /* For UDP-style sockets, id specifies the association to query. */
4379 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4380 if (!asoc)
4381 return -EINVAL;
4383 to = optval + offsetof(struct sctp_getaddrs,addrs);
4384 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4386 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4387 transports) {
4388 memcpy(&temp, &from->ipaddr, sizeof(temp));
4389 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4390 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4391 if (space_left < addrlen)
4392 return -ENOMEM;
4393 if (copy_to_user(to, &temp, addrlen))
4394 return -EFAULT;
4395 to += addrlen;
4396 cnt++;
4397 space_left -= addrlen;
4400 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4401 return -EFAULT;
4402 bytes_copied = ((char __user *)to) - optval;
4403 if (put_user(bytes_copied, optlen))
4404 return -EFAULT;
4406 return 0;
4409 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4410 size_t space_left, int *bytes_copied)
4412 struct sctp_sockaddr_entry *addr;
4413 union sctp_addr temp;
4414 int cnt = 0;
4415 int addrlen;
4417 rcu_read_lock();
4418 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4419 if (!addr->valid)
4420 continue;
4422 if ((PF_INET == sk->sk_family) &&
4423 (AF_INET6 == addr->a.sa.sa_family))
4424 continue;
4425 if ((PF_INET6 == sk->sk_family) &&
4426 inet_v6_ipv6only(sk) &&
4427 (AF_INET == addr->a.sa.sa_family))
4428 continue;
4429 memcpy(&temp, &addr->a, sizeof(temp));
4430 if (!temp.v4.sin_port)
4431 temp.v4.sin_port = htons(port);
4433 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4434 &temp);
4435 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4436 if (space_left < addrlen) {
4437 cnt = -ENOMEM;
4438 break;
4440 memcpy(to, &temp, addrlen);
4442 to += addrlen;
4443 cnt ++;
4444 space_left -= addrlen;
4445 *bytes_copied += addrlen;
4447 rcu_read_unlock();
4449 return cnt;
4453 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4454 char __user *optval, int __user *optlen)
4456 struct sctp_bind_addr *bp;
4457 struct sctp_association *asoc;
4458 int cnt = 0;
4459 struct sctp_getaddrs getaddrs;
4460 struct sctp_sockaddr_entry *addr;
4461 void __user *to;
4462 union sctp_addr temp;
4463 struct sctp_sock *sp = sctp_sk(sk);
4464 int addrlen;
4465 int err = 0;
4466 size_t space_left;
4467 int bytes_copied = 0;
4468 void *addrs;
4469 void *buf;
4471 if (len < sizeof(struct sctp_getaddrs))
4472 return -EINVAL;
4474 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4475 return -EFAULT;
4478 * For UDP-style sockets, id specifies the association to query.
4479 * If the id field is set to the value '0' then the locally bound
4480 * addresses are returned without regard to any particular
4481 * association.
4483 if (0 == getaddrs.assoc_id) {
4484 bp = &sctp_sk(sk)->ep->base.bind_addr;
4485 } else {
4486 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4487 if (!asoc)
4488 return -EINVAL;
4489 bp = &asoc->base.bind_addr;
4492 to = optval + offsetof(struct sctp_getaddrs,addrs);
4493 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4495 addrs = kmalloc(space_left, GFP_KERNEL);
4496 if (!addrs)
4497 return -ENOMEM;
4499 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4500 * addresses from the global local address list.
4502 if (sctp_list_single_entry(&bp->address_list)) {
4503 addr = list_entry(bp->address_list.next,
4504 struct sctp_sockaddr_entry, list);
4505 if (sctp_is_any(sk, &addr->a)) {
4506 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4507 space_left, &bytes_copied);
4508 if (cnt < 0) {
4509 err = cnt;
4510 goto out;
4512 goto copy_getaddrs;
4516 buf = addrs;
4517 /* Protection on the bound address list is not needed since
4518 * in the socket option context we hold a socket lock and
4519 * thus the bound address list can't change.
4521 list_for_each_entry(addr, &bp->address_list, list) {
4522 memcpy(&temp, &addr->a, sizeof(temp));
4523 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4524 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4525 if (space_left < addrlen) {
4526 err = -ENOMEM; /*fixme: right error?*/
4527 goto out;
4529 memcpy(buf, &temp, addrlen);
4530 buf += addrlen;
4531 bytes_copied += addrlen;
4532 cnt ++;
4533 space_left -= addrlen;
4536 copy_getaddrs:
4537 if (copy_to_user(to, addrs, bytes_copied)) {
4538 err = -EFAULT;
4539 goto out;
4541 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4542 err = -EFAULT;
4543 goto out;
4545 if (put_user(bytes_copied, optlen))
4546 err = -EFAULT;
4547 out:
4548 kfree(addrs);
4549 return err;
4552 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4554 * Requests that the local SCTP stack use the enclosed peer address as
4555 * the association primary. The enclosed address must be one of the
4556 * association peer's addresses.
4558 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4559 char __user *optval, int __user *optlen)
4561 struct sctp_prim prim;
4562 struct sctp_association *asoc;
4563 struct sctp_sock *sp = sctp_sk(sk);
4565 if (len < sizeof(struct sctp_prim))
4566 return -EINVAL;
4568 len = sizeof(struct sctp_prim);
4570 if (copy_from_user(&prim, optval, len))
4571 return -EFAULT;
4573 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4574 if (!asoc)
4575 return -EINVAL;
4577 if (!asoc->peer.primary_path)
4578 return -ENOTCONN;
4580 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4581 asoc->peer.primary_path->af_specific->sockaddr_len);
4583 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4584 (union sctp_addr *)&prim.ssp_addr);
4586 if (put_user(len, optlen))
4587 return -EFAULT;
4588 if (copy_to_user(optval, &prim, len))
4589 return -EFAULT;
4591 return 0;
4595 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4597 * Requests that the local endpoint set the specified Adaptation Layer
4598 * Indication parameter for all future INIT and INIT-ACK exchanges.
4600 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4601 char __user *optval, int __user *optlen)
4603 struct sctp_setadaptation adaptation;
4605 if (len < sizeof(struct sctp_setadaptation))
4606 return -EINVAL;
4608 len = sizeof(struct sctp_setadaptation);
4610 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4612 if (put_user(len, optlen))
4613 return -EFAULT;
4614 if (copy_to_user(optval, &adaptation, len))
4615 return -EFAULT;
4617 return 0;
4622 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4624 * Applications that wish to use the sendto() system call may wish to
4625 * specify a default set of parameters that would normally be supplied
4626 * through the inclusion of ancillary data. This socket option allows
4627 * such an application to set the default sctp_sndrcvinfo structure.
4630 * The application that wishes to use this socket option simply passes
4631 * in to this call the sctp_sndrcvinfo structure defined in Section
4632 * 5.2.2) The input parameters accepted by this call include
4633 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4634 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4635 * to this call if the caller is using the UDP model.
4637 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4639 static int sctp_getsockopt_default_send_param(struct sock *sk,
4640 int len, char __user *optval,
4641 int __user *optlen)
4643 struct sctp_sndrcvinfo info;
4644 struct sctp_association *asoc;
4645 struct sctp_sock *sp = sctp_sk(sk);
4647 if (len < sizeof(struct sctp_sndrcvinfo))
4648 return -EINVAL;
4650 len = sizeof(struct sctp_sndrcvinfo);
4652 if (copy_from_user(&info, optval, len))
4653 return -EFAULT;
4655 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4656 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4657 return -EINVAL;
4659 if (asoc) {
4660 info.sinfo_stream = asoc->default_stream;
4661 info.sinfo_flags = asoc->default_flags;
4662 info.sinfo_ppid = asoc->default_ppid;
4663 info.sinfo_context = asoc->default_context;
4664 info.sinfo_timetolive = asoc->default_timetolive;
4665 } else {
4666 info.sinfo_stream = sp->default_stream;
4667 info.sinfo_flags = sp->default_flags;
4668 info.sinfo_ppid = sp->default_ppid;
4669 info.sinfo_context = sp->default_context;
4670 info.sinfo_timetolive = sp->default_timetolive;
4673 if (put_user(len, optlen))
4674 return -EFAULT;
4675 if (copy_to_user(optval, &info, len))
4676 return -EFAULT;
4678 return 0;
4683 * 7.1.5 SCTP_NODELAY
4685 * Turn on/off any Nagle-like algorithm. This means that packets are
4686 * generally sent as soon as possible and no unnecessary delays are
4687 * introduced, at the cost of more packets in the network. Expects an
4688 * integer boolean flag.
4691 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4692 char __user *optval, int __user *optlen)
4694 int val;
4696 if (len < sizeof(int))
4697 return -EINVAL;
4699 len = sizeof(int);
4700 val = (sctp_sk(sk)->nodelay == 1);
4701 if (put_user(len, optlen))
4702 return -EFAULT;
4703 if (copy_to_user(optval, &val, len))
4704 return -EFAULT;
4705 return 0;
4710 * 7.1.1 SCTP_RTOINFO
4712 * The protocol parameters used to initialize and bound retransmission
4713 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4714 * and modify these parameters.
4715 * All parameters are time values, in milliseconds. A value of 0, when
4716 * modifying the parameters, indicates that the current value should not
4717 * be changed.
4720 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4721 char __user *optval,
4722 int __user *optlen) {
4723 struct sctp_rtoinfo rtoinfo;
4724 struct sctp_association *asoc;
4726 if (len < sizeof (struct sctp_rtoinfo))
4727 return -EINVAL;
4729 len = sizeof(struct sctp_rtoinfo);
4731 if (copy_from_user(&rtoinfo, optval, len))
4732 return -EFAULT;
4734 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4736 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4737 return -EINVAL;
4739 /* Values corresponding to the specific association. */
4740 if (asoc) {
4741 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4742 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4743 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4744 } else {
4745 /* Values corresponding to the endpoint. */
4746 struct sctp_sock *sp = sctp_sk(sk);
4748 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4749 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4750 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4753 if (put_user(len, optlen))
4754 return -EFAULT;
4756 if (copy_to_user(optval, &rtoinfo, len))
4757 return -EFAULT;
4759 return 0;
4764 * 7.1.2 SCTP_ASSOCINFO
4766 * This option is used to tune the maximum retransmission attempts
4767 * of the association.
4768 * Returns an error if the new association retransmission value is
4769 * greater than the sum of the retransmission value of the peer.
4770 * See [SCTP] for more information.
4773 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4774 char __user *optval,
4775 int __user *optlen)
4778 struct sctp_assocparams assocparams;
4779 struct sctp_association *asoc;
4780 struct list_head *pos;
4781 int cnt = 0;
4783 if (len < sizeof (struct sctp_assocparams))
4784 return -EINVAL;
4786 len = sizeof(struct sctp_assocparams);
4788 if (copy_from_user(&assocparams, optval, len))
4789 return -EFAULT;
4791 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4793 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4794 return -EINVAL;
4796 /* Values correspoinding to the specific association */
4797 if (asoc) {
4798 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4799 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4800 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4801 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4802 * 1000) +
4803 (asoc->cookie_life.tv_usec
4804 / 1000);
4806 list_for_each(pos, &asoc->peer.transport_addr_list) {
4807 cnt ++;
4810 assocparams.sasoc_number_peer_destinations = cnt;
4811 } else {
4812 /* Values corresponding to the endpoint */
4813 struct sctp_sock *sp = sctp_sk(sk);
4815 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4816 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4817 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4818 assocparams.sasoc_cookie_life =
4819 sp->assocparams.sasoc_cookie_life;
4820 assocparams.sasoc_number_peer_destinations =
4821 sp->assocparams.
4822 sasoc_number_peer_destinations;
4825 if (put_user(len, optlen))
4826 return -EFAULT;
4828 if (copy_to_user(optval, &assocparams, len))
4829 return -EFAULT;
4831 return 0;
4835 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4837 * This socket option is a boolean flag which turns on or off mapped V4
4838 * addresses. If this option is turned on and the socket is type
4839 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4840 * If this option is turned off, then no mapping will be done of V4
4841 * addresses and a user will receive both PF_INET6 and PF_INET type
4842 * addresses on the socket.
4844 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4845 char __user *optval, int __user *optlen)
4847 int val;
4848 struct sctp_sock *sp = sctp_sk(sk);
4850 if (len < sizeof(int))
4851 return -EINVAL;
4853 len = sizeof(int);
4854 val = sp->v4mapped;
4855 if (put_user(len, optlen))
4856 return -EFAULT;
4857 if (copy_to_user(optval, &val, len))
4858 return -EFAULT;
4860 return 0;
4864 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4865 * (chapter and verse is quoted at sctp_setsockopt_context())
4867 static int sctp_getsockopt_context(struct sock *sk, int len,
4868 char __user *optval, int __user *optlen)
4870 struct sctp_assoc_value params;
4871 struct sctp_sock *sp;
4872 struct sctp_association *asoc;
4874 if (len < sizeof(struct sctp_assoc_value))
4875 return -EINVAL;
4877 len = sizeof(struct sctp_assoc_value);
4879 if (copy_from_user(&params, optval, len))
4880 return -EFAULT;
4882 sp = sctp_sk(sk);
4884 if (params.assoc_id != 0) {
4885 asoc = sctp_id2assoc(sk, params.assoc_id);
4886 if (!asoc)
4887 return -EINVAL;
4888 params.assoc_value = asoc->default_rcv_context;
4889 } else {
4890 params.assoc_value = sp->default_rcv_context;
4893 if (put_user(len, optlen))
4894 return -EFAULT;
4895 if (copy_to_user(optval, &params, len))
4896 return -EFAULT;
4898 return 0;
4902 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4903 * This option will get or set the maximum size to put in any outgoing
4904 * SCTP DATA chunk. If a message is larger than this size it will be
4905 * fragmented by SCTP into the specified size. Note that the underlying
4906 * SCTP implementation may fragment into smaller sized chunks when the
4907 * PMTU of the underlying association is smaller than the value set by
4908 * the user. The default value for this option is '0' which indicates
4909 * the user is NOT limiting fragmentation and only the PMTU will effect
4910 * SCTP's choice of DATA chunk size. Note also that values set larger
4911 * than the maximum size of an IP datagram will effectively let SCTP
4912 * control fragmentation (i.e. the same as setting this option to 0).
4914 * The following structure is used to access and modify this parameter:
4916 * struct sctp_assoc_value {
4917 * sctp_assoc_t assoc_id;
4918 * uint32_t assoc_value;
4919 * };
4921 * assoc_id: This parameter is ignored for one-to-one style sockets.
4922 * For one-to-many style sockets this parameter indicates which
4923 * association the user is performing an action upon. Note that if
4924 * this field's value is zero then the endpoints default value is
4925 * changed (effecting future associations only).
4926 * assoc_value: This parameter specifies the maximum size in bytes.
4928 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4929 char __user *optval, int __user *optlen)
4931 struct sctp_assoc_value params;
4932 struct sctp_association *asoc;
4934 if (len == sizeof(int)) {
4935 printk(KERN_WARNING
4936 "SCTP: Use of int in maxseg socket option deprecated\n");
4937 printk(KERN_WARNING
4938 "SCTP: Use struct sctp_assoc_value instead\n");
4939 params.assoc_id = 0;
4940 } else if (len >= sizeof(struct sctp_assoc_value)) {
4941 len = sizeof(struct sctp_assoc_value);
4942 if (copy_from_user(&params, optval, sizeof(params)))
4943 return -EFAULT;
4944 } else
4945 return -EINVAL;
4947 asoc = sctp_id2assoc(sk, params.assoc_id);
4948 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4949 return -EINVAL;
4951 if (asoc)
4952 params.assoc_value = asoc->frag_point;
4953 else
4954 params.assoc_value = sctp_sk(sk)->user_frag;
4956 if (put_user(len, optlen))
4957 return -EFAULT;
4958 if (len == sizeof(int)) {
4959 if (copy_to_user(optval, &params.assoc_value, len))
4960 return -EFAULT;
4961 } else {
4962 if (copy_to_user(optval, &params, len))
4963 return -EFAULT;
4966 return 0;
4970 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4971 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4973 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4974 char __user *optval, int __user *optlen)
4976 int val;
4978 if (len < sizeof(int))
4979 return -EINVAL;
4981 len = sizeof(int);
4983 val = sctp_sk(sk)->frag_interleave;
4984 if (put_user(len, optlen))
4985 return -EFAULT;
4986 if (copy_to_user(optval, &val, len))
4987 return -EFAULT;
4989 return 0;
4993 * 7.1.25. Set or Get the sctp partial delivery point
4994 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4996 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4997 char __user *optval,
4998 int __user *optlen)
5000 u32 val;
5002 if (len < sizeof(u32))
5003 return -EINVAL;
5005 len = sizeof(u32);
5007 val = sctp_sk(sk)->pd_point;
5008 if (put_user(len, optlen))
5009 return -EFAULT;
5010 if (copy_to_user(optval, &val, len))
5011 return -EFAULT;
5013 return -ENOTSUPP;
5017 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5018 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5020 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5021 char __user *optval,
5022 int __user *optlen)
5024 struct sctp_assoc_value params;
5025 struct sctp_sock *sp;
5026 struct sctp_association *asoc;
5028 if (len == sizeof(int)) {
5029 printk(KERN_WARNING
5030 "SCTP: Use of int in max_burst socket option deprecated\n");
5031 printk(KERN_WARNING
5032 "SCTP: Use struct sctp_assoc_value instead\n");
5033 params.assoc_id = 0;
5034 } else if (len >= sizeof(struct sctp_assoc_value)) {
5035 len = sizeof(struct sctp_assoc_value);
5036 if (copy_from_user(&params, optval, len))
5037 return -EFAULT;
5038 } else
5039 return -EINVAL;
5041 sp = sctp_sk(sk);
5043 if (params.assoc_id != 0) {
5044 asoc = sctp_id2assoc(sk, params.assoc_id);
5045 if (!asoc)
5046 return -EINVAL;
5047 params.assoc_value = asoc->max_burst;
5048 } else
5049 params.assoc_value = sp->max_burst;
5051 if (len == sizeof(int)) {
5052 if (copy_to_user(optval, &params.assoc_value, len))
5053 return -EFAULT;
5054 } else {
5055 if (copy_to_user(optval, &params, len))
5056 return -EFAULT;
5059 return 0;
5063 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5064 char __user *optval, int __user *optlen)
5066 struct sctp_hmacalgo __user *p = (void __user *)optval;
5067 struct sctp_hmac_algo_param *hmacs;
5068 __u16 data_len = 0;
5069 u32 num_idents;
5071 if (!sctp_auth_enable)
5072 return -EACCES;
5074 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5075 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5077 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5078 return -EINVAL;
5080 len = sizeof(struct sctp_hmacalgo) + data_len;
5081 num_idents = data_len / sizeof(u16);
5083 if (put_user(len, optlen))
5084 return -EFAULT;
5085 if (put_user(num_idents, &p->shmac_num_idents))
5086 return -EFAULT;
5087 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5088 return -EFAULT;
5089 return 0;
5092 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5093 char __user *optval, int __user *optlen)
5095 struct sctp_authkeyid val;
5096 struct sctp_association *asoc;
5098 if (!sctp_auth_enable)
5099 return -EACCES;
5101 if (len < sizeof(struct sctp_authkeyid))
5102 return -EINVAL;
5103 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5104 return -EFAULT;
5106 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5107 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5108 return -EINVAL;
5110 if (asoc)
5111 val.scact_keynumber = asoc->active_key_id;
5112 else
5113 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5115 len = sizeof(struct sctp_authkeyid);
5116 if (put_user(len, optlen))
5117 return -EFAULT;
5118 if (copy_to_user(optval, &val, len))
5119 return -EFAULT;
5121 return 0;
5124 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5125 char __user *optval, int __user *optlen)
5127 struct sctp_authchunks __user *p = (void __user *)optval;
5128 struct sctp_authchunks val;
5129 struct sctp_association *asoc;
5130 struct sctp_chunks_param *ch;
5131 u32 num_chunks = 0;
5132 char __user *to;
5134 if (!sctp_auth_enable)
5135 return -EACCES;
5137 if (len < sizeof(struct sctp_authchunks))
5138 return -EINVAL;
5140 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5141 return -EFAULT;
5143 to = p->gauth_chunks;
5144 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5145 if (!asoc)
5146 return -EINVAL;
5148 ch = asoc->peer.peer_chunks;
5149 if (!ch)
5150 goto num;
5152 /* See if the user provided enough room for all the data */
5153 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5154 if (len < num_chunks)
5155 return -EINVAL;
5157 if (copy_to_user(to, ch->chunks, num_chunks))
5158 return -EFAULT;
5159 num:
5160 len = sizeof(struct sctp_authchunks) + num_chunks;
5161 if (put_user(len, optlen)) return -EFAULT;
5162 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5163 return -EFAULT;
5164 return 0;
5167 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5168 char __user *optval, int __user *optlen)
5170 struct sctp_authchunks __user *p = (void __user *)optval;
5171 struct sctp_authchunks val;
5172 struct sctp_association *asoc;
5173 struct sctp_chunks_param *ch;
5174 u32 num_chunks = 0;
5175 char __user *to;
5177 if (!sctp_auth_enable)
5178 return -EACCES;
5180 if (len < sizeof(struct sctp_authchunks))
5181 return -EINVAL;
5183 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5184 return -EFAULT;
5186 to = p->gauth_chunks;
5187 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5188 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5189 return -EINVAL;
5191 if (asoc)
5192 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5193 else
5194 ch = sctp_sk(sk)->ep->auth_chunk_list;
5196 if (!ch)
5197 goto num;
5199 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5200 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5201 return -EINVAL;
5203 if (copy_to_user(to, ch->chunks, num_chunks))
5204 return -EFAULT;
5205 num:
5206 len = sizeof(struct sctp_authchunks) + num_chunks;
5207 if (put_user(len, optlen))
5208 return -EFAULT;
5209 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5210 return -EFAULT;
5212 return 0;
5216 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5217 * This option gets the current number of associations that are attached
5218 * to a one-to-many style socket. The option value is an uint32_t.
5220 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5221 char __user *optval, int __user *optlen)
5223 struct sctp_sock *sp = sctp_sk(sk);
5224 struct sctp_association *asoc;
5225 u32 val = 0;
5227 if (sctp_style(sk, TCP))
5228 return -EOPNOTSUPP;
5230 if (len < sizeof(u32))
5231 return -EINVAL;
5233 len = sizeof(u32);
5235 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5236 val++;
5239 if (put_user(len, optlen))
5240 return -EFAULT;
5241 if (copy_to_user(optval, &val, len))
5242 return -EFAULT;
5244 return 0;
5247 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5248 char __user *optval, int __user *optlen)
5250 int retval = 0;
5251 int len;
5253 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5254 sk, optname);
5256 /* I can hardly begin to describe how wrong this is. This is
5257 * so broken as to be worse than useless. The API draft
5258 * REALLY is NOT helpful here... I am not convinced that the
5259 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5260 * are at all well-founded.
5262 if (level != SOL_SCTP) {
5263 struct sctp_af *af = sctp_sk(sk)->pf->af;
5265 retval = af->getsockopt(sk, level, optname, optval, optlen);
5266 return retval;
5269 if (get_user(len, optlen))
5270 return -EFAULT;
5272 sctp_lock_sock(sk);
5274 switch (optname) {
5275 case SCTP_STATUS:
5276 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5277 break;
5278 case SCTP_DISABLE_FRAGMENTS:
5279 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5280 optlen);
5281 break;
5282 case SCTP_EVENTS:
5283 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5284 break;
5285 case SCTP_AUTOCLOSE:
5286 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5287 break;
5288 case SCTP_SOCKOPT_PEELOFF:
5289 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5290 break;
5291 case SCTP_PEER_ADDR_PARAMS:
5292 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5293 optlen);
5294 break;
5295 case SCTP_DELAYED_ACK:
5296 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5297 optlen);
5298 break;
5299 case SCTP_INITMSG:
5300 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5301 break;
5302 case SCTP_GET_PEER_ADDRS:
5303 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5304 optlen);
5305 break;
5306 case SCTP_GET_LOCAL_ADDRS:
5307 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5308 optlen);
5309 break;
5310 case SCTP_SOCKOPT_CONNECTX3:
5311 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5312 break;
5313 case SCTP_DEFAULT_SEND_PARAM:
5314 retval = sctp_getsockopt_default_send_param(sk, len,
5315 optval, optlen);
5316 break;
5317 case SCTP_PRIMARY_ADDR:
5318 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5319 break;
5320 case SCTP_NODELAY:
5321 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5322 break;
5323 case SCTP_RTOINFO:
5324 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5325 break;
5326 case SCTP_ASSOCINFO:
5327 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5328 break;
5329 case SCTP_I_WANT_MAPPED_V4_ADDR:
5330 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5331 break;
5332 case SCTP_MAXSEG:
5333 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5334 break;
5335 case SCTP_GET_PEER_ADDR_INFO:
5336 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5337 optlen);
5338 break;
5339 case SCTP_ADAPTATION_LAYER:
5340 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5341 optlen);
5342 break;
5343 case SCTP_CONTEXT:
5344 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5345 break;
5346 case SCTP_FRAGMENT_INTERLEAVE:
5347 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5348 optlen);
5349 break;
5350 case SCTP_PARTIAL_DELIVERY_POINT:
5351 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5352 optlen);
5353 break;
5354 case SCTP_MAX_BURST:
5355 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5356 break;
5357 case SCTP_AUTH_KEY:
5358 case SCTP_AUTH_CHUNK:
5359 case SCTP_AUTH_DELETE_KEY:
5360 retval = -EOPNOTSUPP;
5361 break;
5362 case SCTP_HMAC_IDENT:
5363 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5364 break;
5365 case SCTP_AUTH_ACTIVE_KEY:
5366 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5367 break;
5368 case SCTP_PEER_AUTH_CHUNKS:
5369 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5370 optlen);
5371 break;
5372 case SCTP_LOCAL_AUTH_CHUNKS:
5373 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5374 optlen);
5375 break;
5376 case SCTP_GET_ASSOC_NUMBER:
5377 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5378 break;
5379 default:
5380 retval = -ENOPROTOOPT;
5381 break;
5384 sctp_release_sock(sk);
5385 return retval;
5388 static void sctp_hash(struct sock *sk)
5390 /* STUB */
5393 static void sctp_unhash(struct sock *sk)
5395 /* STUB */
5398 /* Check if port is acceptable. Possibly find first available port.
5400 * The port hash table (contained in the 'global' SCTP protocol storage
5401 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5402 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5403 * list (the list number is the port number hashed out, so as you
5404 * would expect from a hash function, all the ports in a given list have
5405 * such a number that hashes out to the same list number; you were
5406 * expecting that, right?); so each list has a set of ports, with a
5407 * link to the socket (struct sock) that uses it, the port number and
5408 * a fastreuse flag (FIXME: NPI ipg).
5410 static struct sctp_bind_bucket *sctp_bucket_create(
5411 struct sctp_bind_hashbucket *head, unsigned short snum);
5413 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5415 struct sctp_bind_hashbucket *head; /* hash list */
5416 struct sctp_bind_bucket *pp; /* hash list port iterator */
5417 struct hlist_node *node;
5418 unsigned short snum;
5419 int ret;
5421 snum = ntohs(addr->v4.sin_port);
5423 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5424 sctp_local_bh_disable();
5426 if (snum == 0) {
5427 /* Search for an available port. */
5428 int low, high, remaining, index;
5429 unsigned int rover;
5431 inet_get_local_port_range(&low, &high);
5432 remaining = (high - low) + 1;
5433 rover = net_random() % remaining + low;
5435 do {
5436 rover++;
5437 if ((rover < low) || (rover > high))
5438 rover = low;
5439 index = sctp_phashfn(rover);
5440 head = &sctp_port_hashtable[index];
5441 sctp_spin_lock(&head->lock);
5442 sctp_for_each_hentry(pp, node, &head->chain)
5443 if (pp->port == rover)
5444 goto next;
5445 break;
5446 next:
5447 sctp_spin_unlock(&head->lock);
5448 } while (--remaining > 0);
5450 /* Exhausted local port range during search? */
5451 ret = 1;
5452 if (remaining <= 0)
5453 goto fail;
5455 /* OK, here is the one we will use. HEAD (the port
5456 * hash table list entry) is non-NULL and we hold it's
5457 * mutex.
5459 snum = rover;
5460 } else {
5461 /* We are given an specific port number; we verify
5462 * that it is not being used. If it is used, we will
5463 * exahust the search in the hash list corresponding
5464 * to the port number (snum) - we detect that with the
5465 * port iterator, pp being NULL.
5467 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5468 sctp_spin_lock(&head->lock);
5469 sctp_for_each_hentry(pp, node, &head->chain) {
5470 if (pp->port == snum)
5471 goto pp_found;
5474 pp = NULL;
5475 goto pp_not_found;
5476 pp_found:
5477 if (!hlist_empty(&pp->owner)) {
5478 /* We had a port hash table hit - there is an
5479 * available port (pp != NULL) and it is being
5480 * used by other socket (pp->owner not empty); that other
5481 * socket is going to be sk2.
5483 int reuse = sk->sk_reuse;
5484 struct sock *sk2;
5485 struct hlist_node *node;
5487 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5488 if (pp->fastreuse && sk->sk_reuse &&
5489 sk->sk_state != SCTP_SS_LISTENING)
5490 goto success;
5492 /* Run through the list of sockets bound to the port
5493 * (pp->port) [via the pointers bind_next and
5494 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5495 * we get the endpoint they describe and run through
5496 * the endpoint's list of IP (v4 or v6) addresses,
5497 * comparing each of the addresses with the address of
5498 * the socket sk. If we find a match, then that means
5499 * that this port/socket (sk) combination are already
5500 * in an endpoint.
5502 sk_for_each_bound(sk2, node, &pp->owner) {
5503 struct sctp_endpoint *ep2;
5504 ep2 = sctp_sk(sk2)->ep;
5506 if (sk == sk2 ||
5507 (reuse && sk2->sk_reuse &&
5508 sk2->sk_state != SCTP_SS_LISTENING))
5509 continue;
5511 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5512 sctp_sk(sk2), sctp_sk(sk))) {
5513 ret = (long)sk2;
5514 goto fail_unlock;
5517 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5519 pp_not_found:
5520 /* If there was a hash table miss, create a new port. */
5521 ret = 1;
5522 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5523 goto fail_unlock;
5525 /* In either case (hit or miss), make sure fastreuse is 1 only
5526 * if sk->sk_reuse is too (that is, if the caller requested
5527 * SO_REUSEADDR on this socket -sk-).
5529 if (hlist_empty(&pp->owner)) {
5530 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5531 pp->fastreuse = 1;
5532 else
5533 pp->fastreuse = 0;
5534 } else if (pp->fastreuse &&
5535 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5536 pp->fastreuse = 0;
5538 /* We are set, so fill up all the data in the hash table
5539 * entry, tie the socket list information with the rest of the
5540 * sockets FIXME: Blurry, NPI (ipg).
5542 success:
5543 if (!sctp_sk(sk)->bind_hash) {
5544 inet_sk(sk)->inet_num = snum;
5545 sk_add_bind_node(sk, &pp->owner);
5546 sctp_sk(sk)->bind_hash = pp;
5548 ret = 0;
5550 fail_unlock:
5551 sctp_spin_unlock(&head->lock);
5553 fail:
5554 sctp_local_bh_enable();
5555 return ret;
5558 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5559 * port is requested.
5561 static int sctp_get_port(struct sock *sk, unsigned short snum)
5563 long ret;
5564 union sctp_addr addr;
5565 struct sctp_af *af = sctp_sk(sk)->pf->af;
5567 /* Set up a dummy address struct from the sk. */
5568 af->from_sk(&addr, sk);
5569 addr.v4.sin_port = htons(snum);
5571 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5572 ret = sctp_get_port_local(sk, &addr);
5574 return (ret ? 1 : 0);
5578 * Move a socket to LISTENING state.
5580 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5582 struct sctp_sock *sp = sctp_sk(sk);
5583 struct sctp_endpoint *ep = sp->ep;
5584 struct crypto_hash *tfm = NULL;
5586 /* Allocate HMAC for generating cookie. */
5587 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5588 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5589 if (IS_ERR(tfm)) {
5590 if (net_ratelimit()) {
5591 printk(KERN_INFO
5592 "SCTP: failed to load transform for %s: %ld\n",
5593 sctp_hmac_alg, PTR_ERR(tfm));
5595 return -ENOSYS;
5597 sctp_sk(sk)->hmac = tfm;
5601 * If a bind() or sctp_bindx() is not called prior to a listen()
5602 * call that allows new associations to be accepted, the system
5603 * picks an ephemeral port and will choose an address set equivalent
5604 * to binding with a wildcard address.
5606 * This is not currently spelled out in the SCTP sockets
5607 * extensions draft, but follows the practice as seen in TCP
5608 * sockets.
5611 sk->sk_state = SCTP_SS_LISTENING;
5612 if (!ep->base.bind_addr.port) {
5613 if (sctp_autobind(sk))
5614 return -EAGAIN;
5615 } else {
5616 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5617 sk->sk_state = SCTP_SS_CLOSED;
5618 return -EADDRINUSE;
5622 sk->sk_max_ack_backlog = backlog;
5623 sctp_hash_endpoint(ep);
5624 return 0;
5628 * 4.1.3 / 5.1.3 listen()
5630 * By default, new associations are not accepted for UDP style sockets.
5631 * An application uses listen() to mark a socket as being able to
5632 * accept new associations.
5634 * On TCP style sockets, applications use listen() to ready the SCTP
5635 * endpoint for accepting inbound associations.
5637 * On both types of endpoints a backlog of '0' disables listening.
5639 * Move a socket to LISTENING state.
5641 int sctp_inet_listen(struct socket *sock, int backlog)
5643 struct sock *sk = sock->sk;
5644 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5645 int err = -EINVAL;
5647 if (unlikely(backlog < 0))
5648 return err;
5650 sctp_lock_sock(sk);
5652 /* Peeled-off sockets are not allowed to listen(). */
5653 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5654 goto out;
5656 if (sock->state != SS_UNCONNECTED)
5657 goto out;
5659 /* If backlog is zero, disable listening. */
5660 if (!backlog) {
5661 if (sctp_sstate(sk, CLOSED))
5662 goto out;
5664 err = 0;
5665 sctp_unhash_endpoint(ep);
5666 sk->sk_state = SCTP_SS_CLOSED;
5667 if (sk->sk_reuse)
5668 sctp_sk(sk)->bind_hash->fastreuse = 1;
5669 goto out;
5672 /* If we are already listening, just update the backlog */
5673 if (sctp_sstate(sk, LISTENING))
5674 sk->sk_max_ack_backlog = backlog;
5675 else {
5676 err = sctp_listen_start(sk, backlog);
5677 if (err)
5678 goto out;
5681 err = 0;
5682 out:
5683 sctp_release_sock(sk);
5684 return err;
5688 * This function is done by modeling the current datagram_poll() and the
5689 * tcp_poll(). Note that, based on these implementations, we don't
5690 * lock the socket in this function, even though it seems that,
5691 * ideally, locking or some other mechanisms can be used to ensure
5692 * the integrity of the counters (sndbuf and wmem_alloc) used
5693 * in this place. We assume that we don't need locks either until proven
5694 * otherwise.
5696 * Another thing to note is that we include the Async I/O support
5697 * here, again, by modeling the current TCP/UDP code. We don't have
5698 * a good way to test with it yet.
5700 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5702 struct sock *sk = sock->sk;
5703 struct sctp_sock *sp = sctp_sk(sk);
5704 unsigned int mask;
5706 poll_wait(file, sk->sk_sleep, wait);
5708 /* A TCP-style listening socket becomes readable when the accept queue
5709 * is not empty.
5711 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5712 return (!list_empty(&sp->ep->asocs)) ?
5713 (POLLIN | POLLRDNORM) : 0;
5715 mask = 0;
5717 /* Is there any exceptional events? */
5718 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5719 mask |= POLLERR;
5720 if (sk->sk_shutdown & RCV_SHUTDOWN)
5721 mask |= POLLRDHUP;
5722 if (sk->sk_shutdown == SHUTDOWN_MASK)
5723 mask |= POLLHUP;
5725 /* Is it readable? Reconsider this code with TCP-style support. */
5726 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5727 (sk->sk_shutdown & RCV_SHUTDOWN))
5728 mask |= POLLIN | POLLRDNORM;
5730 /* The association is either gone or not ready. */
5731 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5732 return mask;
5734 /* Is it writable? */
5735 if (sctp_writeable(sk)) {
5736 mask |= POLLOUT | POLLWRNORM;
5737 } else {
5738 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5740 * Since the socket is not locked, the buffer
5741 * might be made available after the writeable check and
5742 * before the bit is set. This could cause a lost I/O
5743 * signal. tcp_poll() has a race breaker for this race
5744 * condition. Based on their implementation, we put
5745 * in the following code to cover it as well.
5747 if (sctp_writeable(sk))
5748 mask |= POLLOUT | POLLWRNORM;
5750 return mask;
5753 /********************************************************************
5754 * 2nd Level Abstractions
5755 ********************************************************************/
5757 static struct sctp_bind_bucket *sctp_bucket_create(
5758 struct sctp_bind_hashbucket *head, unsigned short snum)
5760 struct sctp_bind_bucket *pp;
5762 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5763 if (pp) {
5764 SCTP_DBG_OBJCNT_INC(bind_bucket);
5765 pp->port = snum;
5766 pp->fastreuse = 0;
5767 INIT_HLIST_HEAD(&pp->owner);
5768 hlist_add_head(&pp->node, &head->chain);
5770 return pp;
5773 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5774 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5776 if (pp && hlist_empty(&pp->owner)) {
5777 __hlist_del(&pp->node);
5778 kmem_cache_free(sctp_bucket_cachep, pp);
5779 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5783 /* Release this socket's reference to a local port. */
5784 static inline void __sctp_put_port(struct sock *sk)
5786 struct sctp_bind_hashbucket *head =
5787 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5788 struct sctp_bind_bucket *pp;
5790 sctp_spin_lock(&head->lock);
5791 pp = sctp_sk(sk)->bind_hash;
5792 __sk_del_bind_node(sk);
5793 sctp_sk(sk)->bind_hash = NULL;
5794 inet_sk(sk)->inet_num = 0;
5795 sctp_bucket_destroy(pp);
5796 sctp_spin_unlock(&head->lock);
5799 void sctp_put_port(struct sock *sk)
5801 sctp_local_bh_disable();
5802 __sctp_put_port(sk);
5803 sctp_local_bh_enable();
5807 * The system picks an ephemeral port and choose an address set equivalent
5808 * to binding with a wildcard address.
5809 * One of those addresses will be the primary address for the association.
5810 * This automatically enables the multihoming capability of SCTP.
5812 static int sctp_autobind(struct sock *sk)
5814 union sctp_addr autoaddr;
5815 struct sctp_af *af;
5816 __be16 port;
5818 /* Initialize a local sockaddr structure to INADDR_ANY. */
5819 af = sctp_sk(sk)->pf->af;
5821 port = htons(inet_sk(sk)->inet_num);
5822 af->inaddr_any(&autoaddr, port);
5824 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5827 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5829 * From RFC 2292
5830 * 4.2 The cmsghdr Structure *
5832 * When ancillary data is sent or received, any number of ancillary data
5833 * objects can be specified by the msg_control and msg_controllen members of
5834 * the msghdr structure, because each object is preceded by
5835 * a cmsghdr structure defining the object's length (the cmsg_len member).
5836 * Historically Berkeley-derived implementations have passed only one object
5837 * at a time, but this API allows multiple objects to be
5838 * passed in a single call to sendmsg() or recvmsg(). The following example
5839 * shows two ancillary data objects in a control buffer.
5841 * |<--------------------------- msg_controllen -------------------------->|
5842 * | |
5844 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5846 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5847 * | | |
5849 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5851 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5852 * | | | | |
5854 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5855 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5857 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5859 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5863 * msg_control
5864 * points here
5866 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5867 sctp_cmsgs_t *cmsgs)
5869 struct cmsghdr *cmsg;
5870 struct msghdr *my_msg = (struct msghdr *)msg;
5872 for (cmsg = CMSG_FIRSTHDR(msg);
5873 cmsg != NULL;
5874 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5875 if (!CMSG_OK(my_msg, cmsg))
5876 return -EINVAL;
5878 /* Should we parse this header or ignore? */
5879 if (cmsg->cmsg_level != IPPROTO_SCTP)
5880 continue;
5882 /* Strictly check lengths following example in SCM code. */
5883 switch (cmsg->cmsg_type) {
5884 case SCTP_INIT:
5885 /* SCTP Socket API Extension
5886 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5888 * This cmsghdr structure provides information for
5889 * initializing new SCTP associations with sendmsg().
5890 * The SCTP_INITMSG socket option uses this same data
5891 * structure. This structure is not used for
5892 * recvmsg().
5894 * cmsg_level cmsg_type cmsg_data[]
5895 * ------------ ------------ ----------------------
5896 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5898 if (cmsg->cmsg_len !=
5899 CMSG_LEN(sizeof(struct sctp_initmsg)))
5900 return -EINVAL;
5901 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5902 break;
5904 case SCTP_SNDRCV:
5905 /* SCTP Socket API Extension
5906 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5908 * This cmsghdr structure specifies SCTP options for
5909 * sendmsg() and describes SCTP header information
5910 * about a received message through recvmsg().
5912 * cmsg_level cmsg_type cmsg_data[]
5913 * ------------ ------------ ----------------------
5914 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5916 if (cmsg->cmsg_len !=
5917 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5918 return -EINVAL;
5920 cmsgs->info =
5921 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5923 /* Minimally, validate the sinfo_flags. */
5924 if (cmsgs->info->sinfo_flags &
5925 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5926 SCTP_ABORT | SCTP_EOF))
5927 return -EINVAL;
5928 break;
5930 default:
5931 return -EINVAL;
5934 return 0;
5938 * Wait for a packet..
5939 * Note: This function is the same function as in core/datagram.c
5940 * with a few modifications to make lksctp work.
5942 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5944 int error;
5945 DEFINE_WAIT(wait);
5947 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5949 /* Socket errors? */
5950 error = sock_error(sk);
5951 if (error)
5952 goto out;
5954 if (!skb_queue_empty(&sk->sk_receive_queue))
5955 goto ready;
5957 /* Socket shut down? */
5958 if (sk->sk_shutdown & RCV_SHUTDOWN)
5959 goto out;
5961 /* Sequenced packets can come disconnected. If so we report the
5962 * problem.
5964 error = -ENOTCONN;
5966 /* Is there a good reason to think that we may receive some data? */
5967 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5968 goto out;
5970 /* Handle signals. */
5971 if (signal_pending(current))
5972 goto interrupted;
5974 /* Let another process have a go. Since we are going to sleep
5975 * anyway. Note: This may cause odd behaviors if the message
5976 * does not fit in the user's buffer, but this seems to be the
5977 * only way to honor MSG_DONTWAIT realistically.
5979 sctp_release_sock(sk);
5980 *timeo_p = schedule_timeout(*timeo_p);
5981 sctp_lock_sock(sk);
5983 ready:
5984 finish_wait(sk->sk_sleep, &wait);
5985 return 0;
5987 interrupted:
5988 error = sock_intr_errno(*timeo_p);
5990 out:
5991 finish_wait(sk->sk_sleep, &wait);
5992 *err = error;
5993 return error;
5996 /* Receive a datagram.
5997 * Note: This is pretty much the same routine as in core/datagram.c
5998 * with a few changes to make lksctp work.
6000 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6001 int noblock, int *err)
6003 int error;
6004 struct sk_buff *skb;
6005 long timeo;
6007 timeo = sock_rcvtimeo(sk, noblock);
6009 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6010 timeo, MAX_SCHEDULE_TIMEOUT);
6012 do {
6013 /* Again only user level code calls this function,
6014 * so nothing interrupt level
6015 * will suddenly eat the receive_queue.
6017 * Look at current nfs client by the way...
6018 * However, this function was corrent in any case. 8)
6020 if (flags & MSG_PEEK) {
6021 spin_lock_bh(&sk->sk_receive_queue.lock);
6022 skb = skb_peek(&sk->sk_receive_queue);
6023 if (skb)
6024 atomic_inc(&skb->users);
6025 spin_unlock_bh(&sk->sk_receive_queue.lock);
6026 } else {
6027 skb = skb_dequeue(&sk->sk_receive_queue);
6030 if (skb)
6031 return skb;
6033 /* Caller is allowed not to check sk->sk_err before calling. */
6034 error = sock_error(sk);
6035 if (error)
6036 goto no_packet;
6038 if (sk->sk_shutdown & RCV_SHUTDOWN)
6039 break;
6041 /* User doesn't want to wait. */
6042 error = -EAGAIN;
6043 if (!timeo)
6044 goto no_packet;
6045 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6047 return NULL;
6049 no_packet:
6050 *err = error;
6051 return NULL;
6054 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6055 static void __sctp_write_space(struct sctp_association *asoc)
6057 struct sock *sk = asoc->base.sk;
6058 struct socket *sock = sk->sk_socket;
6060 if ((sctp_wspace(asoc) > 0) && sock) {
6061 if (waitqueue_active(&asoc->wait))
6062 wake_up_interruptible(&asoc->wait);
6064 if (sctp_writeable(sk)) {
6065 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6066 wake_up_interruptible(sk->sk_sleep);
6068 /* Note that we try to include the Async I/O support
6069 * here by modeling from the current TCP/UDP code.
6070 * We have not tested with it yet.
6072 if (sock->fasync_list &&
6073 !(sk->sk_shutdown & SEND_SHUTDOWN))
6074 sock_wake_async(sock,
6075 SOCK_WAKE_SPACE, POLL_OUT);
6080 /* Do accounting for the sndbuf space.
6081 * Decrement the used sndbuf space of the corresponding association by the
6082 * data size which was just transmitted(freed).
6084 static void sctp_wfree(struct sk_buff *skb)
6086 struct sctp_association *asoc;
6087 struct sctp_chunk *chunk;
6088 struct sock *sk;
6090 /* Get the saved chunk pointer. */
6091 chunk = *((struct sctp_chunk **)(skb->cb));
6092 asoc = chunk->asoc;
6093 sk = asoc->base.sk;
6094 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6095 sizeof(struct sk_buff) +
6096 sizeof(struct sctp_chunk);
6098 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6101 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6103 sk->sk_wmem_queued -= skb->truesize;
6104 sk_mem_uncharge(sk, skb->truesize);
6106 sock_wfree(skb);
6107 __sctp_write_space(asoc);
6109 sctp_association_put(asoc);
6112 /* Do accounting for the receive space on the socket.
6113 * Accounting for the association is done in ulpevent.c
6114 * We set this as a destructor for the cloned data skbs so that
6115 * accounting is done at the correct time.
6117 void sctp_sock_rfree(struct sk_buff *skb)
6119 struct sock *sk = skb->sk;
6120 struct sctp_ulpevent *event = sctp_skb2event(skb);
6122 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6125 * Mimic the behavior of sock_rfree
6127 sk_mem_uncharge(sk, event->rmem_len);
6131 /* Helper function to wait for space in the sndbuf. */
6132 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6133 size_t msg_len)
6135 struct sock *sk = asoc->base.sk;
6136 int err = 0;
6137 long current_timeo = *timeo_p;
6138 DEFINE_WAIT(wait);
6140 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6141 asoc, (long)(*timeo_p), msg_len);
6143 /* Increment the association's refcnt. */
6144 sctp_association_hold(asoc);
6146 /* Wait on the association specific sndbuf space. */
6147 for (;;) {
6148 prepare_to_wait_exclusive(&asoc->wait, &wait,
6149 TASK_INTERRUPTIBLE);
6150 if (!*timeo_p)
6151 goto do_nonblock;
6152 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6153 asoc->base.dead)
6154 goto do_error;
6155 if (signal_pending(current))
6156 goto do_interrupted;
6157 if (msg_len <= sctp_wspace(asoc))
6158 break;
6160 /* Let another process have a go. Since we are going
6161 * to sleep anyway.
6163 sctp_release_sock(sk);
6164 current_timeo = schedule_timeout(current_timeo);
6165 BUG_ON(sk != asoc->base.sk);
6166 sctp_lock_sock(sk);
6168 *timeo_p = current_timeo;
6171 out:
6172 finish_wait(&asoc->wait, &wait);
6174 /* Release the association's refcnt. */
6175 sctp_association_put(asoc);
6177 return err;
6179 do_error:
6180 err = -EPIPE;
6181 goto out;
6183 do_interrupted:
6184 err = sock_intr_errno(*timeo_p);
6185 goto out;
6187 do_nonblock:
6188 err = -EAGAIN;
6189 goto out;
6192 /* If socket sndbuf has changed, wake up all per association waiters. */
6193 void sctp_write_space(struct sock *sk)
6195 struct sctp_association *asoc;
6197 /* Wake up the tasks in each wait queue. */
6198 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6199 __sctp_write_space(asoc);
6203 /* Is there any sndbuf space available on the socket?
6205 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6206 * associations on the same socket. For a UDP-style socket with
6207 * multiple associations, it is possible for it to be "unwriteable"
6208 * prematurely. I assume that this is acceptable because
6209 * a premature "unwriteable" is better than an accidental "writeable" which
6210 * would cause an unwanted block under certain circumstances. For the 1-1
6211 * UDP-style sockets or TCP-style sockets, this code should work.
6212 * - Daisy
6214 static int sctp_writeable(struct sock *sk)
6216 int amt = 0;
6218 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6219 if (amt < 0)
6220 amt = 0;
6221 return amt;
6224 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6225 * returns immediately with EINPROGRESS.
6227 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6229 struct sock *sk = asoc->base.sk;
6230 int err = 0;
6231 long current_timeo = *timeo_p;
6232 DEFINE_WAIT(wait);
6234 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6235 (long)(*timeo_p));
6237 /* Increment the association's refcnt. */
6238 sctp_association_hold(asoc);
6240 for (;;) {
6241 prepare_to_wait_exclusive(&asoc->wait, &wait,
6242 TASK_INTERRUPTIBLE);
6243 if (!*timeo_p)
6244 goto do_nonblock;
6245 if (sk->sk_shutdown & RCV_SHUTDOWN)
6246 break;
6247 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6248 asoc->base.dead)
6249 goto do_error;
6250 if (signal_pending(current))
6251 goto do_interrupted;
6253 if (sctp_state(asoc, ESTABLISHED))
6254 break;
6256 /* Let another process have a go. Since we are going
6257 * to sleep anyway.
6259 sctp_release_sock(sk);
6260 current_timeo = schedule_timeout(current_timeo);
6261 sctp_lock_sock(sk);
6263 *timeo_p = current_timeo;
6266 out:
6267 finish_wait(&asoc->wait, &wait);
6269 /* Release the association's refcnt. */
6270 sctp_association_put(asoc);
6272 return err;
6274 do_error:
6275 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6276 err = -ETIMEDOUT;
6277 else
6278 err = -ECONNREFUSED;
6279 goto out;
6281 do_interrupted:
6282 err = sock_intr_errno(*timeo_p);
6283 goto out;
6285 do_nonblock:
6286 err = -EINPROGRESS;
6287 goto out;
6290 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6292 struct sctp_endpoint *ep;
6293 int err = 0;
6294 DEFINE_WAIT(wait);
6296 ep = sctp_sk(sk)->ep;
6299 for (;;) {
6300 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6301 TASK_INTERRUPTIBLE);
6303 if (list_empty(&ep->asocs)) {
6304 sctp_release_sock(sk);
6305 timeo = schedule_timeout(timeo);
6306 sctp_lock_sock(sk);
6309 err = -EINVAL;
6310 if (!sctp_sstate(sk, LISTENING))
6311 break;
6313 err = 0;
6314 if (!list_empty(&ep->asocs))
6315 break;
6317 err = sock_intr_errno(timeo);
6318 if (signal_pending(current))
6319 break;
6321 err = -EAGAIN;
6322 if (!timeo)
6323 break;
6326 finish_wait(sk->sk_sleep, &wait);
6328 return err;
6331 static void sctp_wait_for_close(struct sock *sk, long timeout)
6333 DEFINE_WAIT(wait);
6335 do {
6336 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6337 if (list_empty(&sctp_sk(sk)->ep->asocs))
6338 break;
6339 sctp_release_sock(sk);
6340 timeout = schedule_timeout(timeout);
6341 sctp_lock_sock(sk);
6342 } while (!signal_pending(current) && timeout);
6344 finish_wait(sk->sk_sleep, &wait);
6347 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6349 struct sk_buff *frag;
6351 if (!skb->data_len)
6352 goto done;
6354 /* Don't forget the fragments. */
6355 skb_walk_frags(skb, frag)
6356 sctp_skb_set_owner_r_frag(frag, sk);
6358 done:
6359 sctp_skb_set_owner_r(skb, sk);
6362 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6363 struct sctp_association *asoc)
6365 struct inet_sock *inet = inet_sk(sk);
6366 struct inet_sock *newinet;
6368 newsk->sk_type = sk->sk_type;
6369 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6370 newsk->sk_flags = sk->sk_flags;
6371 newsk->sk_no_check = sk->sk_no_check;
6372 newsk->sk_reuse = sk->sk_reuse;
6374 newsk->sk_shutdown = sk->sk_shutdown;
6375 newsk->sk_destruct = inet_sock_destruct;
6376 newsk->sk_family = sk->sk_family;
6377 newsk->sk_protocol = IPPROTO_SCTP;
6378 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6379 newsk->sk_sndbuf = sk->sk_sndbuf;
6380 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6381 newsk->sk_lingertime = sk->sk_lingertime;
6382 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6383 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6385 newinet = inet_sk(newsk);
6387 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6388 * getsockname() and getpeername()
6390 newinet->inet_sport = inet->inet_sport;
6391 newinet->inet_saddr = inet->inet_saddr;
6392 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6393 newinet->inet_dport = htons(asoc->peer.port);
6394 newinet->pmtudisc = inet->pmtudisc;
6395 newinet->inet_id = asoc->next_tsn ^ jiffies;
6397 newinet->uc_ttl = inet->uc_ttl;
6398 newinet->mc_loop = 1;
6399 newinet->mc_ttl = 1;
6400 newinet->mc_index = 0;
6401 newinet->mc_list = NULL;
6404 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6405 * and its messages to the newsk.
6407 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6408 struct sctp_association *assoc,
6409 sctp_socket_type_t type)
6411 struct sctp_sock *oldsp = sctp_sk(oldsk);
6412 struct sctp_sock *newsp = sctp_sk(newsk);
6413 struct sctp_bind_bucket *pp; /* hash list port iterator */
6414 struct sctp_endpoint *newep = newsp->ep;
6415 struct sk_buff *skb, *tmp;
6416 struct sctp_ulpevent *event;
6417 struct sctp_bind_hashbucket *head;
6419 /* Migrate socket buffer sizes and all the socket level options to the
6420 * new socket.
6422 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6423 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6424 /* Brute force copy old sctp opt. */
6425 inet_sk_copy_descendant(newsk, oldsk);
6427 /* Restore the ep value that was overwritten with the above structure
6428 * copy.
6430 newsp->ep = newep;
6431 newsp->hmac = NULL;
6433 /* Hook this new socket in to the bind_hash list. */
6434 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6435 sctp_local_bh_disable();
6436 sctp_spin_lock(&head->lock);
6437 pp = sctp_sk(oldsk)->bind_hash;
6438 sk_add_bind_node(newsk, &pp->owner);
6439 sctp_sk(newsk)->bind_hash = pp;
6440 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6441 sctp_spin_unlock(&head->lock);
6442 sctp_local_bh_enable();
6444 /* Copy the bind_addr list from the original endpoint to the new
6445 * endpoint so that we can handle restarts properly
6447 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6448 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6450 /* Move any messages in the old socket's receive queue that are for the
6451 * peeled off association to the new socket's receive queue.
6453 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6454 event = sctp_skb2event(skb);
6455 if (event->asoc == assoc) {
6456 __skb_unlink(skb, &oldsk->sk_receive_queue);
6457 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6458 sctp_skb_set_owner_r_frag(skb, newsk);
6462 /* Clean up any messages pending delivery due to partial
6463 * delivery. Three cases:
6464 * 1) No partial deliver; no work.
6465 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6466 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6468 skb_queue_head_init(&newsp->pd_lobby);
6469 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6471 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6472 struct sk_buff_head *queue;
6474 /* Decide which queue to move pd_lobby skbs to. */
6475 if (assoc->ulpq.pd_mode) {
6476 queue = &newsp->pd_lobby;
6477 } else
6478 queue = &newsk->sk_receive_queue;
6480 /* Walk through the pd_lobby, looking for skbs that
6481 * need moved to the new socket.
6483 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6484 event = sctp_skb2event(skb);
6485 if (event->asoc == assoc) {
6486 __skb_unlink(skb, &oldsp->pd_lobby);
6487 __skb_queue_tail(queue, skb);
6488 sctp_skb_set_owner_r_frag(skb, newsk);
6492 /* Clear up any skbs waiting for the partial
6493 * delivery to finish.
6495 if (assoc->ulpq.pd_mode)
6496 sctp_clear_pd(oldsk, NULL);
6500 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6501 sctp_skb_set_owner_r_frag(skb, newsk);
6503 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6504 sctp_skb_set_owner_r_frag(skb, newsk);
6506 /* Set the type of socket to indicate that it is peeled off from the
6507 * original UDP-style socket or created with the accept() call on a
6508 * TCP-style socket..
6510 newsp->type = type;
6512 /* Mark the new socket "in-use" by the user so that any packets
6513 * that may arrive on the association after we've moved it are
6514 * queued to the backlog. This prevents a potential race between
6515 * backlog processing on the old socket and new-packet processing
6516 * on the new socket.
6518 * The caller has just allocated newsk so we can guarantee that other
6519 * paths won't try to lock it and then oldsk.
6521 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6522 sctp_assoc_migrate(assoc, newsk);
6524 /* If the association on the newsk is already closed before accept()
6525 * is called, set RCV_SHUTDOWN flag.
6527 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6528 newsk->sk_shutdown |= RCV_SHUTDOWN;
6530 newsk->sk_state = SCTP_SS_ESTABLISHED;
6531 sctp_release_sock(newsk);
6535 /* This proto struct describes the ULP interface for SCTP. */
6536 struct proto sctp_prot = {
6537 .name = "SCTP",
6538 .owner = THIS_MODULE,
6539 .close = sctp_close,
6540 .connect = sctp_connect,
6541 .disconnect = sctp_disconnect,
6542 .accept = sctp_accept,
6543 .ioctl = sctp_ioctl,
6544 .init = sctp_init_sock,
6545 .destroy = sctp_destroy_sock,
6546 .shutdown = sctp_shutdown,
6547 .setsockopt = sctp_setsockopt,
6548 .getsockopt = sctp_getsockopt,
6549 .sendmsg = sctp_sendmsg,
6550 .recvmsg = sctp_recvmsg,
6551 .bind = sctp_bind,
6552 .backlog_rcv = sctp_backlog_rcv,
6553 .hash = sctp_hash,
6554 .unhash = sctp_unhash,
6555 .get_port = sctp_get_port,
6556 .obj_size = sizeof(struct sctp_sock),
6557 .sysctl_mem = sysctl_sctp_mem,
6558 .sysctl_rmem = sysctl_sctp_rmem,
6559 .sysctl_wmem = sysctl_sctp_wmem,
6560 .memory_pressure = &sctp_memory_pressure,
6561 .enter_memory_pressure = sctp_enter_memory_pressure,
6562 .memory_allocated = &sctp_memory_allocated,
6563 .sockets_allocated = &sctp_sockets_allocated,
6566 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6568 struct proto sctpv6_prot = {
6569 .name = "SCTPv6",
6570 .owner = THIS_MODULE,
6571 .close = sctp_close,
6572 .connect = sctp_connect,
6573 .disconnect = sctp_disconnect,
6574 .accept = sctp_accept,
6575 .ioctl = sctp_ioctl,
6576 .init = sctp_init_sock,
6577 .destroy = sctp_destroy_sock,
6578 .shutdown = sctp_shutdown,
6579 .setsockopt = sctp_setsockopt,
6580 .getsockopt = sctp_getsockopt,
6581 .sendmsg = sctp_sendmsg,
6582 .recvmsg = sctp_recvmsg,
6583 .bind = sctp_bind,
6584 .backlog_rcv = sctp_backlog_rcv,
6585 .hash = sctp_hash,
6586 .unhash = sctp_unhash,
6587 .get_port = sctp_get_port,
6588 .obj_size = sizeof(struct sctp6_sock),
6589 .sysctl_mem = sysctl_sctp_mem,
6590 .sysctl_rmem = sysctl_sctp_rmem,
6591 .sysctl_wmem = sysctl_sctp_wmem,
6592 .memory_pressure = &sctp_memory_pressure,
6593 .enter_memory_pressure = sctp_enter_memory_pressure,
6594 .memory_allocated = &sctp_memory_allocated,
6595 .sockets_allocated = &sctp_sockets_allocated,
6597 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */