fs/btrfs/inode.c: Add missing IS_ERR test
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / sctp / socket.c
blobfbb70770ad05d05807d25b5527e5616a621f58e8
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 if (walk_size + sizeof(sa_family_t) > addrs_size) {
920 kfree(kaddrs);
921 return -EINVAL;
924 sa_addr = (struct sockaddr *)addr_buf;
925 af = sctp_get_af_specific(sa_addr->sa_family);
927 /* If the address family is not supported or if this address
928 * causes the address buffer to overflow return EINVAL.
930 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
931 kfree(kaddrs);
932 return -EINVAL;
934 addrcnt++;
935 addr_buf += af->sockaddr_len;
936 walk_size += af->sockaddr_len;
939 /* Do the work. */
940 switch (op) {
941 case SCTP_BINDX_ADD_ADDR:
942 err = sctp_bindx_add(sk, kaddrs, addrcnt);
943 if (err)
944 goto out;
945 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
946 break;
948 case SCTP_BINDX_REM_ADDR:
949 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
950 if (err)
951 goto out;
952 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
953 break;
955 default:
956 err = -EINVAL;
957 break;
960 out:
961 kfree(kaddrs);
963 return err;
966 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
968 * Common routine for handling connect() and sctp_connectx().
969 * Connect will come in with just a single address.
971 static int __sctp_connect(struct sock* sk,
972 struct sockaddr *kaddrs,
973 int addrs_size,
974 sctp_assoc_t *assoc_id)
976 struct sctp_sock *sp;
977 struct sctp_endpoint *ep;
978 struct sctp_association *asoc = NULL;
979 struct sctp_association *asoc2;
980 struct sctp_transport *transport;
981 union sctp_addr to;
982 struct sctp_af *af;
983 sctp_scope_t scope;
984 long timeo;
985 int err = 0;
986 int addrcnt = 0;
987 int walk_size = 0;
988 union sctp_addr *sa_addr = NULL;
989 void *addr_buf;
990 unsigned short port;
991 unsigned int f_flags = 0;
993 sp = sctp_sk(sk);
994 ep = sp->ep;
996 /* connect() cannot be done on a socket that is already in ESTABLISHED
997 * state - UDP-style peeled off socket or a TCP-style socket that
998 * is already connected.
999 * It cannot be done even on a TCP-style listening socket.
1001 if (sctp_sstate(sk, ESTABLISHED) ||
1002 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1003 err = -EISCONN;
1004 goto out_free;
1007 /* Walk through the addrs buffer and count the number of addresses. */
1008 addr_buf = kaddrs;
1009 while (walk_size < addrs_size) {
1010 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1011 err = -EINVAL;
1012 goto out_free;
1015 sa_addr = (union sctp_addr *)addr_buf;
1016 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1018 /* If the address family is not supported or if this address
1019 * causes the address buffer to overflow return EINVAL.
1021 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1022 err = -EINVAL;
1023 goto out_free;
1026 port = ntohs(sa_addr->v4.sin_port);
1028 /* Save current address so we can work with it */
1029 memcpy(&to, sa_addr, af->sockaddr_len);
1031 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1032 if (err)
1033 goto out_free;
1035 /* Make sure the destination port is correctly set
1036 * in all addresses.
1038 if (asoc && asoc->peer.port && asoc->peer.port != port)
1039 goto out_free;
1042 /* Check if there already is a matching association on the
1043 * endpoint (other than the one created here).
1045 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1046 if (asoc2 && asoc2 != asoc) {
1047 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1048 err = -EISCONN;
1049 else
1050 err = -EALREADY;
1051 goto out_free;
1054 /* If we could not find a matching association on the endpoint,
1055 * make sure that there is no peeled-off association matching
1056 * the peer address even on another socket.
1058 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1059 err = -EADDRNOTAVAIL;
1060 goto out_free;
1063 if (!asoc) {
1064 /* If a bind() or sctp_bindx() is not called prior to
1065 * an sctp_connectx() call, the system picks an
1066 * ephemeral port and will choose an address set
1067 * equivalent to binding with a wildcard address.
1069 if (!ep->base.bind_addr.port) {
1070 if (sctp_autobind(sk)) {
1071 err = -EAGAIN;
1072 goto out_free;
1074 } else {
1076 * If an unprivileged user inherits a 1-many
1077 * style socket with open associations on a
1078 * privileged port, it MAY be permitted to
1079 * accept new associations, but it SHOULD NOT
1080 * be permitted to open new associations.
1082 if (ep->base.bind_addr.port < PROT_SOCK &&
1083 !capable(CAP_NET_BIND_SERVICE)) {
1084 err = -EACCES;
1085 goto out_free;
1089 scope = sctp_scope(&to);
1090 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1091 if (!asoc) {
1092 err = -ENOMEM;
1093 goto out_free;
1096 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1097 GFP_KERNEL);
1098 if (err < 0) {
1099 goto out_free;
1104 /* Prime the peer's transport structures. */
1105 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1106 SCTP_UNKNOWN);
1107 if (!transport) {
1108 err = -ENOMEM;
1109 goto out_free;
1112 addrcnt++;
1113 addr_buf += af->sockaddr_len;
1114 walk_size += af->sockaddr_len;
1117 /* In case the user of sctp_connectx() wants an association
1118 * id back, assign one now.
1120 if (assoc_id) {
1121 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1122 if (err < 0)
1123 goto out_free;
1126 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1127 if (err < 0) {
1128 goto out_free;
1131 /* Initialize sk's dport and daddr for getpeername() */
1132 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1133 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1134 af->to_sk_daddr(sa_addr, sk);
1135 sk->sk_err = 0;
1137 /* in-kernel sockets don't generally have a file allocated to them
1138 * if all they do is call sock_create_kern().
1140 if (sk->sk_socket->file)
1141 f_flags = sk->sk_socket->file->f_flags;
1143 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1145 err = sctp_wait_for_connect(asoc, &timeo);
1146 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1147 *assoc_id = asoc->assoc_id;
1149 /* Don't free association on exit. */
1150 asoc = NULL;
1152 out_free:
1154 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1155 " kaddrs: %p err: %d\n",
1156 asoc, kaddrs, err);
1157 if (asoc)
1158 sctp_association_free(asoc);
1159 return err;
1162 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1164 * API 8.9
1165 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1166 * sctp_assoc_t *asoc);
1168 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1169 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1170 * or IPv6 addresses.
1172 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1173 * Section 3.1.2 for this usage.
1175 * addrs is a pointer to an array of one or more socket addresses. Each
1176 * address is contained in its appropriate structure (i.e. struct
1177 * sockaddr_in or struct sockaddr_in6) the family of the address type
1178 * must be used to distengish the address length (note that this
1179 * representation is termed a "packed array" of addresses). The caller
1180 * specifies the number of addresses in the array with addrcnt.
1182 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1183 * the association id of the new association. On failure, sctp_connectx()
1184 * returns -1, and sets errno to the appropriate error code. The assoc_id
1185 * is not touched by the kernel.
1187 * For SCTP, the port given in each socket address must be the same, or
1188 * sctp_connectx() will fail, setting errno to EINVAL.
1190 * An application can use sctp_connectx to initiate an association with
1191 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1192 * allows a caller to specify multiple addresses at which a peer can be
1193 * reached. The way the SCTP stack uses the list of addresses to set up
1194 * the association is implementation dependant. This function only
1195 * specifies that the stack will try to make use of all the addresses in
1196 * the list when needed.
1198 * Note that the list of addresses passed in is only used for setting up
1199 * the association. It does not necessarily equal the set of addresses
1200 * the peer uses for the resulting association. If the caller wants to
1201 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1202 * retrieve them after the association has been set up.
1204 * Basically do nothing but copying the addresses from user to kernel
1205 * land and invoking either sctp_connectx(). This is used for tunneling
1206 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1208 * We don't use copy_from_user() for optimization: we first do the
1209 * sanity checks (buffer size -fast- and access check-healthy
1210 * pointer); if all of those succeed, then we can alloc the memory
1211 * (expensive operation) needed to copy the data to kernel. Then we do
1212 * the copying without checking the user space area
1213 * (__copy_from_user()).
1215 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1216 * it.
1218 * sk The sk of the socket
1219 * addrs The pointer to the addresses in user land
1220 * addrssize Size of the addrs buffer
1222 * Returns >=0 if ok, <0 errno code on error.
1224 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1225 struct sockaddr __user *addrs,
1226 int addrs_size,
1227 sctp_assoc_t *assoc_id)
1229 int err = 0;
1230 struct sockaddr *kaddrs;
1232 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1233 __func__, sk, addrs, addrs_size);
1235 if (unlikely(addrs_size <= 0))
1236 return -EINVAL;
1238 /* Check the user passed a healthy pointer. */
1239 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1240 return -EFAULT;
1242 /* Alloc space for the address array in kernel memory. */
1243 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1244 if (unlikely(!kaddrs))
1245 return -ENOMEM;
1247 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1248 err = -EFAULT;
1249 } else {
1250 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1253 kfree(kaddrs);
1255 return err;
1259 * This is an older interface. It's kept for backward compatibility
1260 * to the option that doesn't provide association id.
1262 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1263 struct sockaddr __user *addrs,
1264 int addrs_size)
1266 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1270 * New interface for the API. The since the API is done with a socket
1271 * option, to make it simple we feed back the association id is as a return
1272 * indication to the call. Error is always negative and association id is
1273 * always positive.
1275 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1276 struct sockaddr __user *addrs,
1277 int addrs_size)
1279 sctp_assoc_t assoc_id = 0;
1280 int err = 0;
1282 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1284 if (err)
1285 return err;
1286 else
1287 return assoc_id;
1291 * New (hopefully final) interface for the API.
1292 * We use the sctp_getaddrs_old structure so that use-space library
1293 * can avoid any unnecessary allocations. The only defferent part
1294 * is that we store the actual length of the address buffer into the
1295 * addrs_num structure member. That way we can re-use the existing
1296 * code.
1298 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1299 char __user *optval,
1300 int __user *optlen)
1302 struct sctp_getaddrs_old param;
1303 sctp_assoc_t assoc_id = 0;
1304 int err = 0;
1306 if (len < sizeof(param))
1307 return -EINVAL;
1309 if (copy_from_user(&param, optval, sizeof(param)))
1310 return -EFAULT;
1312 err = __sctp_setsockopt_connectx(sk,
1313 (struct sockaddr __user *)param.addrs,
1314 param.addr_num, &assoc_id);
1316 if (err == 0 || err == -EINPROGRESS) {
1317 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1318 return -EFAULT;
1319 if (put_user(sizeof(assoc_id), optlen))
1320 return -EFAULT;
1323 return err;
1326 /* API 3.1.4 close() - UDP Style Syntax
1327 * Applications use close() to perform graceful shutdown (as described in
1328 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1329 * by a UDP-style socket.
1331 * The syntax is
1333 * ret = close(int sd);
1335 * sd - the socket descriptor of the associations to be closed.
1337 * To gracefully shutdown a specific association represented by the
1338 * UDP-style socket, an application should use the sendmsg() call,
1339 * passing no user data, but including the appropriate flag in the
1340 * ancillary data (see Section xxxx).
1342 * If sd in the close() call is a branched-off socket representing only
1343 * one association, the shutdown is performed on that association only.
1345 * 4.1.6 close() - TCP Style Syntax
1347 * Applications use close() to gracefully close down an association.
1349 * The syntax is:
1351 * int close(int sd);
1353 * sd - the socket descriptor of the association to be closed.
1355 * After an application calls close() on a socket descriptor, no further
1356 * socket operations will succeed on that descriptor.
1358 * API 7.1.4 SO_LINGER
1360 * An application using the TCP-style socket can use this option to
1361 * perform the SCTP ABORT primitive. The linger option structure is:
1363 * struct linger {
1364 * int l_onoff; // option on/off
1365 * int l_linger; // linger time
1366 * };
1368 * To enable the option, set l_onoff to 1. If the l_linger value is set
1369 * to 0, calling close() is the same as the ABORT primitive. If the
1370 * value is set to a negative value, the setsockopt() call will return
1371 * an error. If the value is set to a positive value linger_time, the
1372 * close() can be blocked for at most linger_time ms. If the graceful
1373 * shutdown phase does not finish during this period, close() will
1374 * return but the graceful shutdown phase continues in the system.
1376 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1378 struct sctp_endpoint *ep;
1379 struct sctp_association *asoc;
1380 struct list_head *pos, *temp;
1382 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1384 sctp_lock_sock(sk);
1385 sk->sk_shutdown = SHUTDOWN_MASK;
1386 sk->sk_state = SCTP_SS_CLOSING;
1388 ep = sctp_sk(sk)->ep;
1390 /* Walk all associations on an endpoint. */
1391 list_for_each_safe(pos, temp, &ep->asocs) {
1392 asoc = list_entry(pos, struct sctp_association, asocs);
1394 if (sctp_style(sk, TCP)) {
1395 /* A closed association can still be in the list if
1396 * it belongs to a TCP-style listening socket that is
1397 * not yet accepted. If so, free it. If not, send an
1398 * ABORT or SHUTDOWN based on the linger options.
1400 if (sctp_state(asoc, CLOSED)) {
1401 sctp_unhash_established(asoc);
1402 sctp_association_free(asoc);
1403 continue;
1407 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1408 struct sctp_chunk *chunk;
1410 chunk = sctp_make_abort_user(asoc, NULL, 0);
1411 if (chunk)
1412 sctp_primitive_ABORT(asoc, chunk);
1413 } else
1414 sctp_primitive_SHUTDOWN(asoc, NULL);
1417 /* Clean up any skbs sitting on the receive queue. */
1418 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1419 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1421 /* On a TCP-style socket, block for at most linger_time if set. */
1422 if (sctp_style(sk, TCP) && timeout)
1423 sctp_wait_for_close(sk, timeout);
1425 /* This will run the backlog queue. */
1426 sctp_release_sock(sk);
1428 /* Supposedly, no process has access to the socket, but
1429 * the net layers still may.
1431 sctp_local_bh_disable();
1432 sctp_bh_lock_sock(sk);
1434 /* Hold the sock, since sk_common_release() will put sock_put()
1435 * and we have just a little more cleanup.
1437 sock_hold(sk);
1438 sk_common_release(sk);
1440 sctp_bh_unlock_sock(sk);
1441 sctp_local_bh_enable();
1443 sock_put(sk);
1445 SCTP_DBG_OBJCNT_DEC(sock);
1448 /* Handle EPIPE error. */
1449 static int sctp_error(struct sock *sk, int flags, int err)
1451 if (err == -EPIPE)
1452 err = sock_error(sk) ? : -EPIPE;
1453 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1454 send_sig(SIGPIPE, current, 0);
1455 return err;
1458 /* API 3.1.3 sendmsg() - UDP Style Syntax
1460 * An application uses sendmsg() and recvmsg() calls to transmit data to
1461 * and receive data from its peer.
1463 * ssize_t sendmsg(int socket, const struct msghdr *message,
1464 * int flags);
1466 * socket - the socket descriptor of the endpoint.
1467 * message - pointer to the msghdr structure which contains a single
1468 * user message and possibly some ancillary data.
1470 * See Section 5 for complete description of the data
1471 * structures.
1473 * flags - flags sent or received with the user message, see Section
1474 * 5 for complete description of the flags.
1476 * Note: This function could use a rewrite especially when explicit
1477 * connect support comes in.
1479 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1481 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1483 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1484 struct msghdr *msg, size_t msg_len)
1486 struct sctp_sock *sp;
1487 struct sctp_endpoint *ep;
1488 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1489 struct sctp_transport *transport, *chunk_tp;
1490 struct sctp_chunk *chunk;
1491 union sctp_addr to;
1492 struct sockaddr *msg_name = NULL;
1493 struct sctp_sndrcvinfo default_sinfo = { 0 };
1494 struct sctp_sndrcvinfo *sinfo;
1495 struct sctp_initmsg *sinit;
1496 sctp_assoc_t associd = 0;
1497 sctp_cmsgs_t cmsgs = { NULL };
1498 int err;
1499 sctp_scope_t scope;
1500 long timeo;
1501 __u16 sinfo_flags = 0;
1502 struct sctp_datamsg *datamsg;
1503 int msg_flags = msg->msg_flags;
1505 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1506 sk, msg, msg_len);
1508 err = 0;
1509 sp = sctp_sk(sk);
1510 ep = sp->ep;
1512 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1514 /* We cannot send a message over a TCP-style listening socket. */
1515 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1516 err = -EPIPE;
1517 goto out_nounlock;
1520 /* Parse out the SCTP CMSGs. */
1521 err = sctp_msghdr_parse(msg, &cmsgs);
1523 if (err) {
1524 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1525 goto out_nounlock;
1528 /* Fetch the destination address for this packet. This
1529 * address only selects the association--it is not necessarily
1530 * the address we will send to.
1531 * For a peeled-off socket, msg_name is ignored.
1533 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1534 int msg_namelen = msg->msg_namelen;
1536 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1537 msg_namelen);
1538 if (err)
1539 return err;
1541 if (msg_namelen > sizeof(to))
1542 msg_namelen = sizeof(to);
1543 memcpy(&to, msg->msg_name, msg_namelen);
1544 msg_name = msg->msg_name;
1547 sinfo = cmsgs.info;
1548 sinit = cmsgs.init;
1550 /* Did the user specify SNDRCVINFO? */
1551 if (sinfo) {
1552 sinfo_flags = sinfo->sinfo_flags;
1553 associd = sinfo->sinfo_assoc_id;
1556 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1557 msg_len, sinfo_flags);
1559 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1560 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1561 err = -EINVAL;
1562 goto out_nounlock;
1565 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1566 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1567 * If SCTP_ABORT is set, the message length could be non zero with
1568 * the msg_iov set to the user abort reason.
1570 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1571 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1572 err = -EINVAL;
1573 goto out_nounlock;
1576 /* If SCTP_ADDR_OVER is set, there must be an address
1577 * specified in msg_name.
1579 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1580 err = -EINVAL;
1581 goto out_nounlock;
1584 transport = NULL;
1586 SCTP_DEBUG_PRINTK("About to look up association.\n");
1588 sctp_lock_sock(sk);
1590 /* If a msg_name has been specified, assume this is to be used. */
1591 if (msg_name) {
1592 /* Look for a matching association on the endpoint. */
1593 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1594 if (!asoc) {
1595 /* If we could not find a matching association on the
1596 * endpoint, make sure that it is not a TCP-style
1597 * socket that already has an association or there is
1598 * no peeled-off association on another socket.
1600 if ((sctp_style(sk, TCP) &&
1601 sctp_sstate(sk, ESTABLISHED)) ||
1602 sctp_endpoint_is_peeled_off(ep, &to)) {
1603 err = -EADDRNOTAVAIL;
1604 goto out_unlock;
1607 } else {
1608 asoc = sctp_id2assoc(sk, associd);
1609 if (!asoc) {
1610 err = -EPIPE;
1611 goto out_unlock;
1615 if (asoc) {
1616 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1618 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1619 * socket that has an association in CLOSED state. This can
1620 * happen when an accepted socket has an association that is
1621 * already CLOSED.
1623 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1624 err = -EPIPE;
1625 goto out_unlock;
1628 if (sinfo_flags & SCTP_EOF) {
1629 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1630 asoc);
1631 sctp_primitive_SHUTDOWN(asoc, NULL);
1632 err = 0;
1633 goto out_unlock;
1635 if (sinfo_flags & SCTP_ABORT) {
1637 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1638 if (!chunk) {
1639 err = -ENOMEM;
1640 goto out_unlock;
1643 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1644 sctp_primitive_ABORT(asoc, chunk);
1645 err = 0;
1646 goto out_unlock;
1650 /* Do we need to create the association? */
1651 if (!asoc) {
1652 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1654 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1655 err = -EINVAL;
1656 goto out_unlock;
1659 /* Check for invalid stream against the stream counts,
1660 * either the default or the user specified stream counts.
1662 if (sinfo) {
1663 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1664 /* Check against the defaults. */
1665 if (sinfo->sinfo_stream >=
1666 sp->initmsg.sinit_num_ostreams) {
1667 err = -EINVAL;
1668 goto out_unlock;
1670 } else {
1671 /* Check against the requested. */
1672 if (sinfo->sinfo_stream >=
1673 sinit->sinit_num_ostreams) {
1674 err = -EINVAL;
1675 goto out_unlock;
1681 * API 3.1.2 bind() - UDP Style Syntax
1682 * If a bind() or sctp_bindx() is not called prior to a
1683 * sendmsg() call that initiates a new association, the
1684 * system picks an ephemeral port and will choose an address
1685 * set equivalent to binding with a wildcard address.
1687 if (!ep->base.bind_addr.port) {
1688 if (sctp_autobind(sk)) {
1689 err = -EAGAIN;
1690 goto out_unlock;
1692 } else {
1694 * If an unprivileged user inherits a one-to-many
1695 * style socket with open associations on a privileged
1696 * port, it MAY be permitted to accept new associations,
1697 * but it SHOULD NOT be permitted to open new
1698 * associations.
1700 if (ep->base.bind_addr.port < PROT_SOCK &&
1701 !capable(CAP_NET_BIND_SERVICE)) {
1702 err = -EACCES;
1703 goto out_unlock;
1707 scope = sctp_scope(&to);
1708 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1709 if (!new_asoc) {
1710 err = -ENOMEM;
1711 goto out_unlock;
1713 asoc = new_asoc;
1714 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1715 if (err < 0) {
1716 err = -ENOMEM;
1717 goto out_free;
1720 /* If the SCTP_INIT ancillary data is specified, set all
1721 * the association init values accordingly.
1723 if (sinit) {
1724 if (sinit->sinit_num_ostreams) {
1725 asoc->c.sinit_num_ostreams =
1726 sinit->sinit_num_ostreams;
1728 if (sinit->sinit_max_instreams) {
1729 asoc->c.sinit_max_instreams =
1730 sinit->sinit_max_instreams;
1732 if (sinit->sinit_max_attempts) {
1733 asoc->max_init_attempts
1734 = sinit->sinit_max_attempts;
1736 if (sinit->sinit_max_init_timeo) {
1737 asoc->max_init_timeo =
1738 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1742 /* Prime the peer's transport structures. */
1743 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1744 if (!transport) {
1745 err = -ENOMEM;
1746 goto out_free;
1750 /* ASSERT: we have a valid association at this point. */
1751 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1753 if (!sinfo) {
1754 /* If the user didn't specify SNDRCVINFO, make up one with
1755 * some defaults.
1757 default_sinfo.sinfo_stream = asoc->default_stream;
1758 default_sinfo.sinfo_flags = asoc->default_flags;
1759 default_sinfo.sinfo_ppid = asoc->default_ppid;
1760 default_sinfo.sinfo_context = asoc->default_context;
1761 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1762 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1763 sinfo = &default_sinfo;
1766 /* API 7.1.7, the sndbuf size per association bounds the
1767 * maximum size of data that can be sent in a single send call.
1769 if (msg_len > sk->sk_sndbuf) {
1770 err = -EMSGSIZE;
1771 goto out_free;
1774 if (asoc->pmtu_pending)
1775 sctp_assoc_pending_pmtu(asoc);
1777 /* If fragmentation is disabled and the message length exceeds the
1778 * association fragmentation point, return EMSGSIZE. The I-D
1779 * does not specify what this error is, but this looks like
1780 * a great fit.
1782 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1783 err = -EMSGSIZE;
1784 goto out_free;
1787 if (sinfo) {
1788 /* Check for invalid stream. */
1789 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1790 err = -EINVAL;
1791 goto out_free;
1795 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1796 if (!sctp_wspace(asoc)) {
1797 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1798 if (err)
1799 goto out_free;
1802 /* If an address is passed with the sendto/sendmsg call, it is used
1803 * to override the primary destination address in the TCP model, or
1804 * when SCTP_ADDR_OVER flag is set in the UDP model.
1806 if ((sctp_style(sk, TCP) && msg_name) ||
1807 (sinfo_flags & SCTP_ADDR_OVER)) {
1808 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1809 if (!chunk_tp) {
1810 err = -EINVAL;
1811 goto out_free;
1813 } else
1814 chunk_tp = NULL;
1816 /* Auto-connect, if we aren't connected already. */
1817 if (sctp_state(asoc, CLOSED)) {
1818 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1819 if (err < 0)
1820 goto out_free;
1821 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1824 /* Break the message into multiple chunks of maximum size. */
1825 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1826 if (!datamsg) {
1827 err = -ENOMEM;
1828 goto out_free;
1831 /* Now send the (possibly) fragmented message. */
1832 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1833 sctp_chunk_hold(chunk);
1835 /* Do accounting for the write space. */
1836 sctp_set_owner_w(chunk);
1838 chunk->transport = chunk_tp;
1841 /* Send it to the lower layers. Note: all chunks
1842 * must either fail or succeed. The lower layer
1843 * works that way today. Keep it that way or this
1844 * breaks.
1846 err = sctp_primitive_SEND(asoc, datamsg);
1847 /* Did the lower layer accept the chunk? */
1848 if (err)
1849 sctp_datamsg_free(datamsg);
1850 else
1851 sctp_datamsg_put(datamsg);
1853 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1855 if (err)
1856 goto out_free;
1857 else
1858 err = msg_len;
1860 /* If we are already past ASSOCIATE, the lower
1861 * layers are responsible for association cleanup.
1863 goto out_unlock;
1865 out_free:
1866 if (new_asoc)
1867 sctp_association_free(asoc);
1868 out_unlock:
1869 sctp_release_sock(sk);
1871 out_nounlock:
1872 return sctp_error(sk, msg_flags, err);
1874 #if 0
1875 do_sock_err:
1876 if (msg_len)
1877 err = msg_len;
1878 else
1879 err = sock_error(sk);
1880 goto out;
1882 do_interrupted:
1883 if (msg_len)
1884 err = msg_len;
1885 goto out;
1886 #endif /* 0 */
1889 /* This is an extended version of skb_pull() that removes the data from the
1890 * start of a skb even when data is spread across the list of skb's in the
1891 * frag_list. len specifies the total amount of data that needs to be removed.
1892 * when 'len' bytes could be removed from the skb, it returns 0.
1893 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1894 * could not be removed.
1896 static int sctp_skb_pull(struct sk_buff *skb, int len)
1898 struct sk_buff *list;
1899 int skb_len = skb_headlen(skb);
1900 int rlen;
1902 if (len <= skb_len) {
1903 __skb_pull(skb, len);
1904 return 0;
1906 len -= skb_len;
1907 __skb_pull(skb, skb_len);
1909 skb_walk_frags(skb, list) {
1910 rlen = sctp_skb_pull(list, len);
1911 skb->len -= (len-rlen);
1912 skb->data_len -= (len-rlen);
1914 if (!rlen)
1915 return 0;
1917 len = rlen;
1920 return len;
1923 /* API 3.1.3 recvmsg() - UDP Style Syntax
1925 * ssize_t recvmsg(int socket, struct msghdr *message,
1926 * int flags);
1928 * socket - the socket descriptor of the endpoint.
1929 * message - pointer to the msghdr structure which contains a single
1930 * user message and possibly some ancillary data.
1932 * See Section 5 for complete description of the data
1933 * structures.
1935 * flags - flags sent or received with the user message, see Section
1936 * 5 for complete description of the flags.
1938 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1940 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1941 struct msghdr *msg, size_t len, int noblock,
1942 int flags, int *addr_len)
1944 struct sctp_ulpevent *event = NULL;
1945 struct sctp_sock *sp = sctp_sk(sk);
1946 struct sk_buff *skb;
1947 int copied;
1948 int err = 0;
1949 int skb_len;
1951 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1952 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1953 "len", len, "knoblauch", noblock,
1954 "flags", flags, "addr_len", addr_len);
1956 sctp_lock_sock(sk);
1958 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1959 err = -ENOTCONN;
1960 goto out;
1963 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1964 if (!skb)
1965 goto out;
1967 /* Get the total length of the skb including any skb's in the
1968 * frag_list.
1970 skb_len = skb->len;
1972 copied = skb_len;
1973 if (copied > len)
1974 copied = len;
1976 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1978 event = sctp_skb2event(skb);
1980 if (err)
1981 goto out_free;
1983 sock_recv_ts_and_drops(msg, sk, skb);
1984 if (sctp_ulpevent_is_notification(event)) {
1985 msg->msg_flags |= MSG_NOTIFICATION;
1986 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1987 } else {
1988 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1991 /* Check if we allow SCTP_SNDRCVINFO. */
1992 if (sp->subscribe.sctp_data_io_event)
1993 sctp_ulpevent_read_sndrcvinfo(event, msg);
1994 #if 0
1995 /* FIXME: we should be calling IP/IPv6 layers. */
1996 if (sk->sk_protinfo.af_inet.cmsg_flags)
1997 ip_cmsg_recv(msg, skb);
1998 #endif
2000 err = copied;
2002 /* If skb's length exceeds the user's buffer, update the skb and
2003 * push it back to the receive_queue so that the next call to
2004 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2006 if (skb_len > copied) {
2007 msg->msg_flags &= ~MSG_EOR;
2008 if (flags & MSG_PEEK)
2009 goto out_free;
2010 sctp_skb_pull(skb, copied);
2011 skb_queue_head(&sk->sk_receive_queue, skb);
2013 /* When only partial message is copied to the user, increase
2014 * rwnd by that amount. If all the data in the skb is read,
2015 * rwnd is updated when the event is freed.
2017 if (!sctp_ulpevent_is_notification(event))
2018 sctp_assoc_rwnd_increase(event->asoc, copied);
2019 goto out;
2020 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2021 (event->msg_flags & MSG_EOR))
2022 msg->msg_flags |= MSG_EOR;
2023 else
2024 msg->msg_flags &= ~MSG_EOR;
2026 out_free:
2027 if (flags & MSG_PEEK) {
2028 /* Release the skb reference acquired after peeking the skb in
2029 * sctp_skb_recv_datagram().
2031 kfree_skb(skb);
2032 } else {
2033 /* Free the event which includes releasing the reference to
2034 * the owner of the skb, freeing the skb and updating the
2035 * rwnd.
2037 sctp_ulpevent_free(event);
2039 out:
2040 sctp_release_sock(sk);
2041 return err;
2044 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2046 * This option is a on/off flag. If enabled no SCTP message
2047 * fragmentation will be performed. Instead if a message being sent
2048 * exceeds the current PMTU size, the message will NOT be sent and
2049 * instead a error will be indicated to the user.
2051 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2052 char __user *optval,
2053 unsigned int optlen)
2055 int val;
2057 if (optlen < sizeof(int))
2058 return -EINVAL;
2060 if (get_user(val, (int __user *)optval))
2061 return -EFAULT;
2063 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2065 return 0;
2068 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2069 unsigned int optlen)
2071 if (optlen > sizeof(struct sctp_event_subscribe))
2072 return -EINVAL;
2073 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2074 return -EFAULT;
2075 return 0;
2078 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2080 * This socket option is applicable to the UDP-style socket only. When
2081 * set it will cause associations that are idle for more than the
2082 * specified number of seconds to automatically close. An association
2083 * being idle is defined an association that has NOT sent or received
2084 * user data. The special value of '0' indicates that no automatic
2085 * close of any associations should be performed. The option expects an
2086 * integer defining the number of seconds of idle time before an
2087 * association is closed.
2089 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2090 unsigned int optlen)
2092 struct sctp_sock *sp = sctp_sk(sk);
2094 /* Applicable to UDP-style socket only */
2095 if (sctp_style(sk, TCP))
2096 return -EOPNOTSUPP;
2097 if (optlen != sizeof(int))
2098 return -EINVAL;
2099 if (copy_from_user(&sp->autoclose, optval, optlen))
2100 return -EFAULT;
2101 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2102 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2104 return 0;
2107 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2109 * Applications can enable or disable heartbeats for any peer address of
2110 * an association, modify an address's heartbeat interval, force a
2111 * heartbeat to be sent immediately, and adjust the address's maximum
2112 * number of retransmissions sent before an address is considered
2113 * unreachable. The following structure is used to access and modify an
2114 * address's parameters:
2116 * struct sctp_paddrparams {
2117 * sctp_assoc_t spp_assoc_id;
2118 * struct sockaddr_storage spp_address;
2119 * uint32_t spp_hbinterval;
2120 * uint16_t spp_pathmaxrxt;
2121 * uint32_t spp_pathmtu;
2122 * uint32_t spp_sackdelay;
2123 * uint32_t spp_flags;
2124 * };
2126 * spp_assoc_id - (one-to-many style socket) This is filled in the
2127 * application, and identifies the association for
2128 * this query.
2129 * spp_address - This specifies which address is of interest.
2130 * spp_hbinterval - This contains the value of the heartbeat interval,
2131 * in milliseconds. If a value of zero
2132 * is present in this field then no changes are to
2133 * be made to this parameter.
2134 * spp_pathmaxrxt - This contains the maximum number of
2135 * retransmissions before this address shall be
2136 * considered unreachable. If a value of zero
2137 * is present in this field then no changes are to
2138 * be made to this parameter.
2139 * spp_pathmtu - When Path MTU discovery is disabled the value
2140 * specified here will be the "fixed" path mtu.
2141 * Note that if the spp_address field is empty
2142 * then all associations on this address will
2143 * have this fixed path mtu set upon them.
2145 * spp_sackdelay - When delayed sack is enabled, this value specifies
2146 * the number of milliseconds that sacks will be delayed
2147 * for. This value will apply to all addresses of an
2148 * association if the spp_address field is empty. Note
2149 * also, that if delayed sack is enabled and this
2150 * value is set to 0, no change is made to the last
2151 * recorded delayed sack timer value.
2153 * spp_flags - These flags are used to control various features
2154 * on an association. The flag field may contain
2155 * zero or more of the following options.
2157 * SPP_HB_ENABLE - Enable heartbeats on the
2158 * specified address. Note that if the address
2159 * field is empty all addresses for the association
2160 * have heartbeats enabled upon them.
2162 * SPP_HB_DISABLE - Disable heartbeats on the
2163 * speicifed address. Note that if the address
2164 * field is empty all addresses for the association
2165 * will have their heartbeats disabled. Note also
2166 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2167 * mutually exclusive, only one of these two should
2168 * be specified. Enabling both fields will have
2169 * undetermined results.
2171 * SPP_HB_DEMAND - Request a user initiated heartbeat
2172 * to be made immediately.
2174 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2175 * heartbeat delayis to be set to the value of 0
2176 * milliseconds.
2178 * SPP_PMTUD_ENABLE - This field will enable PMTU
2179 * discovery upon the specified address. Note that
2180 * if the address feild is empty then all addresses
2181 * on the association are effected.
2183 * SPP_PMTUD_DISABLE - This field will disable PMTU
2184 * discovery upon the specified address. Note that
2185 * if the address feild is empty then all addresses
2186 * on the association are effected. Not also that
2187 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2188 * exclusive. Enabling both will have undetermined
2189 * results.
2191 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2192 * on delayed sack. The time specified in spp_sackdelay
2193 * is used to specify the sack delay for this address. Note
2194 * that if spp_address is empty then all addresses will
2195 * enable delayed sack and take on the sack delay
2196 * value specified in spp_sackdelay.
2197 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2198 * off delayed sack. If the spp_address field is blank then
2199 * delayed sack is disabled for the entire association. Note
2200 * also that this field is mutually exclusive to
2201 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2202 * results.
2204 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2205 struct sctp_transport *trans,
2206 struct sctp_association *asoc,
2207 struct sctp_sock *sp,
2208 int hb_change,
2209 int pmtud_change,
2210 int sackdelay_change)
2212 int error;
2214 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2215 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2216 if (error)
2217 return error;
2220 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2221 * this field is ignored. Note also that a value of zero indicates
2222 * the current setting should be left unchanged.
2224 if (params->spp_flags & SPP_HB_ENABLE) {
2226 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2227 * set. This lets us use 0 value when this flag
2228 * is set.
2230 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2231 params->spp_hbinterval = 0;
2233 if (params->spp_hbinterval ||
2234 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2235 if (trans) {
2236 trans->hbinterval =
2237 msecs_to_jiffies(params->spp_hbinterval);
2238 } else if (asoc) {
2239 asoc->hbinterval =
2240 msecs_to_jiffies(params->spp_hbinterval);
2241 } else {
2242 sp->hbinterval = params->spp_hbinterval;
2247 if (hb_change) {
2248 if (trans) {
2249 trans->param_flags =
2250 (trans->param_flags & ~SPP_HB) | hb_change;
2251 } else if (asoc) {
2252 asoc->param_flags =
2253 (asoc->param_flags & ~SPP_HB) | hb_change;
2254 } else {
2255 sp->param_flags =
2256 (sp->param_flags & ~SPP_HB) | hb_change;
2260 /* When Path MTU discovery is disabled the value specified here will
2261 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2262 * include the flag SPP_PMTUD_DISABLE for this field to have any
2263 * effect).
2265 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2266 if (trans) {
2267 trans->pathmtu = params->spp_pathmtu;
2268 sctp_assoc_sync_pmtu(asoc);
2269 } else if (asoc) {
2270 asoc->pathmtu = params->spp_pathmtu;
2271 sctp_frag_point(asoc, params->spp_pathmtu);
2272 } else {
2273 sp->pathmtu = params->spp_pathmtu;
2277 if (pmtud_change) {
2278 if (trans) {
2279 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2280 (params->spp_flags & SPP_PMTUD_ENABLE);
2281 trans->param_flags =
2282 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2283 if (update) {
2284 sctp_transport_pmtu(trans);
2285 sctp_assoc_sync_pmtu(asoc);
2287 } else if (asoc) {
2288 asoc->param_flags =
2289 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2290 } else {
2291 sp->param_flags =
2292 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2296 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2297 * value of this field is ignored. Note also that a value of zero
2298 * indicates the current setting should be left unchanged.
2300 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2301 if (trans) {
2302 trans->sackdelay =
2303 msecs_to_jiffies(params->spp_sackdelay);
2304 } else if (asoc) {
2305 asoc->sackdelay =
2306 msecs_to_jiffies(params->spp_sackdelay);
2307 } else {
2308 sp->sackdelay = params->spp_sackdelay;
2312 if (sackdelay_change) {
2313 if (trans) {
2314 trans->param_flags =
2315 (trans->param_flags & ~SPP_SACKDELAY) |
2316 sackdelay_change;
2317 } else if (asoc) {
2318 asoc->param_flags =
2319 (asoc->param_flags & ~SPP_SACKDELAY) |
2320 sackdelay_change;
2321 } else {
2322 sp->param_flags =
2323 (sp->param_flags & ~SPP_SACKDELAY) |
2324 sackdelay_change;
2328 /* Note that a value of zero indicates the current setting should be
2329 left unchanged.
2331 if (params->spp_pathmaxrxt) {
2332 if (trans) {
2333 trans->pathmaxrxt = params->spp_pathmaxrxt;
2334 } else if (asoc) {
2335 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2336 } else {
2337 sp->pathmaxrxt = params->spp_pathmaxrxt;
2341 return 0;
2344 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2345 char __user *optval,
2346 unsigned int optlen)
2348 struct sctp_paddrparams params;
2349 struct sctp_transport *trans = NULL;
2350 struct sctp_association *asoc = NULL;
2351 struct sctp_sock *sp = sctp_sk(sk);
2352 int error;
2353 int hb_change, pmtud_change, sackdelay_change;
2355 if (optlen != sizeof(struct sctp_paddrparams))
2356 return - EINVAL;
2358 if (copy_from_user(&params, optval, optlen))
2359 return -EFAULT;
2361 /* Validate flags and value parameters. */
2362 hb_change = params.spp_flags & SPP_HB;
2363 pmtud_change = params.spp_flags & SPP_PMTUD;
2364 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2366 if (hb_change == SPP_HB ||
2367 pmtud_change == SPP_PMTUD ||
2368 sackdelay_change == SPP_SACKDELAY ||
2369 params.spp_sackdelay > 500 ||
2370 (params.spp_pathmtu &&
2371 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2372 return -EINVAL;
2374 /* If an address other than INADDR_ANY is specified, and
2375 * no transport is found, then the request is invalid.
2377 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2378 trans = sctp_addr_id2transport(sk, &params.spp_address,
2379 params.spp_assoc_id);
2380 if (!trans)
2381 return -EINVAL;
2384 /* Get association, if assoc_id != 0 and the socket is a one
2385 * to many style socket, and an association was not found, then
2386 * the id was invalid.
2388 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2389 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2390 return -EINVAL;
2392 /* Heartbeat demand can only be sent on a transport or
2393 * association, but not a socket.
2395 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2396 return -EINVAL;
2398 /* Process parameters. */
2399 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2400 hb_change, pmtud_change,
2401 sackdelay_change);
2403 if (error)
2404 return error;
2406 /* If changes are for association, also apply parameters to each
2407 * transport.
2409 if (!trans && asoc) {
2410 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2411 transports) {
2412 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2413 hb_change, pmtud_change,
2414 sackdelay_change);
2418 return 0;
2422 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2424 * This option will effect the way delayed acks are performed. This
2425 * option allows you to get or set the delayed ack time, in
2426 * milliseconds. It also allows changing the delayed ack frequency.
2427 * Changing the frequency to 1 disables the delayed sack algorithm. If
2428 * the assoc_id is 0, then this sets or gets the endpoints default
2429 * values. If the assoc_id field is non-zero, then the set or get
2430 * effects the specified association for the one to many model (the
2431 * assoc_id field is ignored by the one to one model). Note that if
2432 * sack_delay or sack_freq are 0 when setting this option, then the
2433 * current values will remain unchanged.
2435 * struct sctp_sack_info {
2436 * sctp_assoc_t sack_assoc_id;
2437 * uint32_t sack_delay;
2438 * uint32_t sack_freq;
2439 * };
2441 * sack_assoc_id - This parameter, indicates which association the user
2442 * is performing an action upon. Note that if this field's value is
2443 * zero then the endpoints default value is changed (effecting future
2444 * associations only).
2446 * sack_delay - This parameter contains the number of milliseconds that
2447 * the user is requesting the delayed ACK timer be set to. Note that
2448 * this value is defined in the standard to be between 200 and 500
2449 * milliseconds.
2451 * sack_freq - This parameter contains the number of packets that must
2452 * be received before a sack is sent without waiting for the delay
2453 * timer to expire. The default value for this is 2, setting this
2454 * value to 1 will disable the delayed sack algorithm.
2457 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2458 char __user *optval, unsigned int optlen)
2460 struct sctp_sack_info params;
2461 struct sctp_transport *trans = NULL;
2462 struct sctp_association *asoc = NULL;
2463 struct sctp_sock *sp = sctp_sk(sk);
2465 if (optlen == sizeof(struct sctp_sack_info)) {
2466 if (copy_from_user(&params, optval, optlen))
2467 return -EFAULT;
2469 if (params.sack_delay == 0 && params.sack_freq == 0)
2470 return 0;
2471 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2472 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2473 "in delayed_ack socket option deprecated\n");
2474 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2475 if (copy_from_user(&params, optval, optlen))
2476 return -EFAULT;
2478 if (params.sack_delay == 0)
2479 params.sack_freq = 1;
2480 else
2481 params.sack_freq = 0;
2482 } else
2483 return - EINVAL;
2485 /* Validate value parameter. */
2486 if (params.sack_delay > 500)
2487 return -EINVAL;
2489 /* Get association, if sack_assoc_id != 0 and the socket is a one
2490 * to many style socket, and an association was not found, then
2491 * the id was invalid.
2493 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2494 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2495 return -EINVAL;
2497 if (params.sack_delay) {
2498 if (asoc) {
2499 asoc->sackdelay =
2500 msecs_to_jiffies(params.sack_delay);
2501 asoc->param_flags =
2502 (asoc->param_flags & ~SPP_SACKDELAY) |
2503 SPP_SACKDELAY_ENABLE;
2504 } else {
2505 sp->sackdelay = params.sack_delay;
2506 sp->param_flags =
2507 (sp->param_flags & ~SPP_SACKDELAY) |
2508 SPP_SACKDELAY_ENABLE;
2512 if (params.sack_freq == 1) {
2513 if (asoc) {
2514 asoc->param_flags =
2515 (asoc->param_flags & ~SPP_SACKDELAY) |
2516 SPP_SACKDELAY_DISABLE;
2517 } else {
2518 sp->param_flags =
2519 (sp->param_flags & ~SPP_SACKDELAY) |
2520 SPP_SACKDELAY_DISABLE;
2522 } else if (params.sack_freq > 1) {
2523 if (asoc) {
2524 asoc->sackfreq = params.sack_freq;
2525 asoc->param_flags =
2526 (asoc->param_flags & ~SPP_SACKDELAY) |
2527 SPP_SACKDELAY_ENABLE;
2528 } else {
2529 sp->sackfreq = params.sack_freq;
2530 sp->param_flags =
2531 (sp->param_flags & ~SPP_SACKDELAY) |
2532 SPP_SACKDELAY_ENABLE;
2536 /* If change is for association, also apply to each transport. */
2537 if (asoc) {
2538 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2539 transports) {
2540 if (params.sack_delay) {
2541 trans->sackdelay =
2542 msecs_to_jiffies(params.sack_delay);
2543 trans->param_flags =
2544 (trans->param_flags & ~SPP_SACKDELAY) |
2545 SPP_SACKDELAY_ENABLE;
2547 if (params.sack_freq == 1) {
2548 trans->param_flags =
2549 (trans->param_flags & ~SPP_SACKDELAY) |
2550 SPP_SACKDELAY_DISABLE;
2551 } else if (params.sack_freq > 1) {
2552 trans->sackfreq = params.sack_freq;
2553 trans->param_flags =
2554 (trans->param_flags & ~SPP_SACKDELAY) |
2555 SPP_SACKDELAY_ENABLE;
2560 return 0;
2563 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2565 * Applications can specify protocol parameters for the default association
2566 * initialization. The option name argument to setsockopt() and getsockopt()
2567 * is SCTP_INITMSG.
2569 * Setting initialization parameters is effective only on an unconnected
2570 * socket (for UDP-style sockets only future associations are effected
2571 * by the change). With TCP-style sockets, this option is inherited by
2572 * sockets derived from a listener socket.
2574 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2576 struct sctp_initmsg sinit;
2577 struct sctp_sock *sp = sctp_sk(sk);
2579 if (optlen != sizeof(struct sctp_initmsg))
2580 return -EINVAL;
2581 if (copy_from_user(&sinit, optval, optlen))
2582 return -EFAULT;
2584 if (sinit.sinit_num_ostreams)
2585 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2586 if (sinit.sinit_max_instreams)
2587 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2588 if (sinit.sinit_max_attempts)
2589 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2590 if (sinit.sinit_max_init_timeo)
2591 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2593 return 0;
2597 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2599 * Applications that wish to use the sendto() system call may wish to
2600 * specify a default set of parameters that would normally be supplied
2601 * through the inclusion of ancillary data. This socket option allows
2602 * such an application to set the default sctp_sndrcvinfo structure.
2603 * The application that wishes to use this socket option simply passes
2604 * in to this call the sctp_sndrcvinfo structure defined in Section
2605 * 5.2.2) The input parameters accepted by this call include
2606 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2607 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2608 * to this call if the caller is using the UDP model.
2610 static int sctp_setsockopt_default_send_param(struct sock *sk,
2611 char __user *optval,
2612 unsigned int optlen)
2614 struct sctp_sndrcvinfo info;
2615 struct sctp_association *asoc;
2616 struct sctp_sock *sp = sctp_sk(sk);
2618 if (optlen != sizeof(struct sctp_sndrcvinfo))
2619 return -EINVAL;
2620 if (copy_from_user(&info, optval, optlen))
2621 return -EFAULT;
2623 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2624 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2625 return -EINVAL;
2627 if (asoc) {
2628 asoc->default_stream = info.sinfo_stream;
2629 asoc->default_flags = info.sinfo_flags;
2630 asoc->default_ppid = info.sinfo_ppid;
2631 asoc->default_context = info.sinfo_context;
2632 asoc->default_timetolive = info.sinfo_timetolive;
2633 } else {
2634 sp->default_stream = info.sinfo_stream;
2635 sp->default_flags = info.sinfo_flags;
2636 sp->default_ppid = info.sinfo_ppid;
2637 sp->default_context = info.sinfo_context;
2638 sp->default_timetolive = info.sinfo_timetolive;
2641 return 0;
2644 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2646 * Requests that the local SCTP stack use the enclosed peer address as
2647 * the association primary. The enclosed address must be one of the
2648 * association peer's addresses.
2650 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2651 unsigned int optlen)
2653 struct sctp_prim prim;
2654 struct sctp_transport *trans;
2656 if (optlen != sizeof(struct sctp_prim))
2657 return -EINVAL;
2659 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2660 return -EFAULT;
2662 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2663 if (!trans)
2664 return -EINVAL;
2666 sctp_assoc_set_primary(trans->asoc, trans);
2668 return 0;
2672 * 7.1.5 SCTP_NODELAY
2674 * Turn on/off any Nagle-like algorithm. This means that packets are
2675 * generally sent as soon as possible and no unnecessary delays are
2676 * introduced, at the cost of more packets in the network. Expects an
2677 * integer boolean flag.
2679 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2680 unsigned int optlen)
2682 int val;
2684 if (optlen < sizeof(int))
2685 return -EINVAL;
2686 if (get_user(val, (int __user *)optval))
2687 return -EFAULT;
2689 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2690 return 0;
2695 * 7.1.1 SCTP_RTOINFO
2697 * The protocol parameters used to initialize and bound retransmission
2698 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2699 * and modify these parameters.
2700 * All parameters are time values, in milliseconds. A value of 0, when
2701 * modifying the parameters, indicates that the current value should not
2702 * be changed.
2705 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2707 struct sctp_rtoinfo rtoinfo;
2708 struct sctp_association *asoc;
2710 if (optlen != sizeof (struct sctp_rtoinfo))
2711 return -EINVAL;
2713 if (copy_from_user(&rtoinfo, optval, optlen))
2714 return -EFAULT;
2716 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2718 /* Set the values to the specific association */
2719 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2720 return -EINVAL;
2722 if (asoc) {
2723 if (rtoinfo.srto_initial != 0)
2724 asoc->rto_initial =
2725 msecs_to_jiffies(rtoinfo.srto_initial);
2726 if (rtoinfo.srto_max != 0)
2727 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2728 if (rtoinfo.srto_min != 0)
2729 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2730 } else {
2731 /* If there is no association or the association-id = 0
2732 * set the values to the endpoint.
2734 struct sctp_sock *sp = sctp_sk(sk);
2736 if (rtoinfo.srto_initial != 0)
2737 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2738 if (rtoinfo.srto_max != 0)
2739 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2740 if (rtoinfo.srto_min != 0)
2741 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2744 return 0;
2749 * 7.1.2 SCTP_ASSOCINFO
2751 * This option is used to tune the maximum retransmission attempts
2752 * of the association.
2753 * Returns an error if the new association retransmission value is
2754 * greater than the sum of the retransmission value of the peer.
2755 * See [SCTP] for more information.
2758 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2761 struct sctp_assocparams assocparams;
2762 struct sctp_association *asoc;
2764 if (optlen != sizeof(struct sctp_assocparams))
2765 return -EINVAL;
2766 if (copy_from_user(&assocparams, optval, optlen))
2767 return -EFAULT;
2769 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2771 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2772 return -EINVAL;
2774 /* Set the values to the specific association */
2775 if (asoc) {
2776 if (assocparams.sasoc_asocmaxrxt != 0) {
2777 __u32 path_sum = 0;
2778 int paths = 0;
2779 struct sctp_transport *peer_addr;
2781 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2782 transports) {
2783 path_sum += peer_addr->pathmaxrxt;
2784 paths++;
2787 /* Only validate asocmaxrxt if we have more than
2788 * one path/transport. We do this because path
2789 * retransmissions are only counted when we have more
2790 * then one path.
2792 if (paths > 1 &&
2793 assocparams.sasoc_asocmaxrxt > path_sum)
2794 return -EINVAL;
2796 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2799 if (assocparams.sasoc_cookie_life != 0) {
2800 asoc->cookie_life.tv_sec =
2801 assocparams.sasoc_cookie_life / 1000;
2802 asoc->cookie_life.tv_usec =
2803 (assocparams.sasoc_cookie_life % 1000)
2804 * 1000;
2806 } else {
2807 /* Set the values to the endpoint */
2808 struct sctp_sock *sp = sctp_sk(sk);
2810 if (assocparams.sasoc_asocmaxrxt != 0)
2811 sp->assocparams.sasoc_asocmaxrxt =
2812 assocparams.sasoc_asocmaxrxt;
2813 if (assocparams.sasoc_cookie_life != 0)
2814 sp->assocparams.sasoc_cookie_life =
2815 assocparams.sasoc_cookie_life;
2817 return 0;
2821 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2823 * This socket option is a boolean flag which turns on or off mapped V4
2824 * addresses. If this option is turned on and the socket is type
2825 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2826 * If this option is turned off, then no mapping will be done of V4
2827 * addresses and a user will receive both PF_INET6 and PF_INET type
2828 * addresses on the socket.
2830 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2832 int val;
2833 struct sctp_sock *sp = sctp_sk(sk);
2835 if (optlen < sizeof(int))
2836 return -EINVAL;
2837 if (get_user(val, (int __user *)optval))
2838 return -EFAULT;
2839 if (val)
2840 sp->v4mapped = 1;
2841 else
2842 sp->v4mapped = 0;
2844 return 0;
2848 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2849 * This option will get or set the maximum size to put in any outgoing
2850 * SCTP DATA chunk. If a message is larger than this size it will be
2851 * fragmented by SCTP into the specified size. Note that the underlying
2852 * SCTP implementation may fragment into smaller sized chunks when the
2853 * PMTU of the underlying association is smaller than the value set by
2854 * the user. The default value for this option is '0' which indicates
2855 * the user is NOT limiting fragmentation and only the PMTU will effect
2856 * SCTP's choice of DATA chunk size. Note also that values set larger
2857 * than the maximum size of an IP datagram will effectively let SCTP
2858 * control fragmentation (i.e. the same as setting this option to 0).
2860 * The following structure is used to access and modify this parameter:
2862 * struct sctp_assoc_value {
2863 * sctp_assoc_t assoc_id;
2864 * uint32_t assoc_value;
2865 * };
2867 * assoc_id: This parameter is ignored for one-to-one style sockets.
2868 * For one-to-many style sockets this parameter indicates which
2869 * association the user is performing an action upon. Note that if
2870 * this field's value is zero then the endpoints default value is
2871 * changed (effecting future associations only).
2872 * assoc_value: This parameter specifies the maximum size in bytes.
2874 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2876 struct sctp_assoc_value params;
2877 struct sctp_association *asoc;
2878 struct sctp_sock *sp = sctp_sk(sk);
2879 int val;
2881 if (optlen == sizeof(int)) {
2882 printk(KERN_WARNING
2883 "SCTP: Use of int in maxseg socket option deprecated\n");
2884 printk(KERN_WARNING
2885 "SCTP: Use struct sctp_assoc_value instead\n");
2886 if (copy_from_user(&val, optval, optlen))
2887 return -EFAULT;
2888 params.assoc_id = 0;
2889 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2890 if (copy_from_user(&params, optval, optlen))
2891 return -EFAULT;
2892 val = params.assoc_value;
2893 } else
2894 return -EINVAL;
2896 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2897 return -EINVAL;
2899 asoc = sctp_id2assoc(sk, params.assoc_id);
2900 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2901 return -EINVAL;
2903 if (asoc) {
2904 if (val == 0) {
2905 val = asoc->pathmtu;
2906 val -= sp->pf->af->net_header_len;
2907 val -= sizeof(struct sctphdr) +
2908 sizeof(struct sctp_data_chunk);
2910 asoc->user_frag = val;
2911 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2912 } else {
2913 sp->user_frag = val;
2916 return 0;
2921 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2923 * Requests that the peer mark the enclosed address as the association
2924 * primary. The enclosed address must be one of the association's
2925 * locally bound addresses. The following structure is used to make a
2926 * set primary request:
2928 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2929 unsigned int optlen)
2931 struct sctp_sock *sp;
2932 struct sctp_endpoint *ep;
2933 struct sctp_association *asoc = NULL;
2934 struct sctp_setpeerprim prim;
2935 struct sctp_chunk *chunk;
2936 int err;
2938 sp = sctp_sk(sk);
2939 ep = sp->ep;
2941 if (!sctp_addip_enable)
2942 return -EPERM;
2944 if (optlen != sizeof(struct sctp_setpeerprim))
2945 return -EINVAL;
2947 if (copy_from_user(&prim, optval, optlen))
2948 return -EFAULT;
2950 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2951 if (!asoc)
2952 return -EINVAL;
2954 if (!asoc->peer.asconf_capable)
2955 return -EPERM;
2957 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2958 return -EPERM;
2960 if (!sctp_state(asoc, ESTABLISHED))
2961 return -ENOTCONN;
2963 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2964 return -EADDRNOTAVAIL;
2966 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2967 chunk = sctp_make_asconf_set_prim(asoc,
2968 (union sctp_addr *)&prim.sspp_addr);
2969 if (!chunk)
2970 return -ENOMEM;
2972 err = sctp_send_asconf(asoc, chunk);
2974 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2976 return err;
2979 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2980 unsigned int optlen)
2982 struct sctp_setadaptation adaptation;
2984 if (optlen != sizeof(struct sctp_setadaptation))
2985 return -EINVAL;
2986 if (copy_from_user(&adaptation, optval, optlen))
2987 return -EFAULT;
2989 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2991 return 0;
2995 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2997 * The context field in the sctp_sndrcvinfo structure is normally only
2998 * used when a failed message is retrieved holding the value that was
2999 * sent down on the actual send call. This option allows the setting of
3000 * a default context on an association basis that will be received on
3001 * reading messages from the peer. This is especially helpful in the
3002 * one-2-many model for an application to keep some reference to an
3003 * internal state machine that is processing messages on the
3004 * association. Note that the setting of this value only effects
3005 * received messages from the peer and does not effect the value that is
3006 * saved with outbound messages.
3008 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3009 unsigned int optlen)
3011 struct sctp_assoc_value params;
3012 struct sctp_sock *sp;
3013 struct sctp_association *asoc;
3015 if (optlen != sizeof(struct sctp_assoc_value))
3016 return -EINVAL;
3017 if (copy_from_user(&params, optval, optlen))
3018 return -EFAULT;
3020 sp = sctp_sk(sk);
3022 if (params.assoc_id != 0) {
3023 asoc = sctp_id2assoc(sk, params.assoc_id);
3024 if (!asoc)
3025 return -EINVAL;
3026 asoc->default_rcv_context = params.assoc_value;
3027 } else {
3028 sp->default_rcv_context = params.assoc_value;
3031 return 0;
3035 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3037 * This options will at a minimum specify if the implementation is doing
3038 * fragmented interleave. Fragmented interleave, for a one to many
3039 * socket, is when subsequent calls to receive a message may return
3040 * parts of messages from different associations. Some implementations
3041 * may allow you to turn this value on or off. If so, when turned off,
3042 * no fragment interleave will occur (which will cause a head of line
3043 * blocking amongst multiple associations sharing the same one to many
3044 * socket). When this option is turned on, then each receive call may
3045 * come from a different association (thus the user must receive data
3046 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3047 * association each receive belongs to.
3049 * This option takes a boolean value. A non-zero value indicates that
3050 * fragmented interleave is on. A value of zero indicates that
3051 * fragmented interleave is off.
3053 * Note that it is important that an implementation that allows this
3054 * option to be turned on, have it off by default. Otherwise an unaware
3055 * application using the one to many model may become confused and act
3056 * incorrectly.
3058 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3059 char __user *optval,
3060 unsigned int optlen)
3062 int val;
3064 if (optlen != sizeof(int))
3065 return -EINVAL;
3066 if (get_user(val, (int __user *)optval))
3067 return -EFAULT;
3069 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3071 return 0;
3075 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3076 * (SCTP_PARTIAL_DELIVERY_POINT)
3078 * This option will set or get the SCTP partial delivery point. This
3079 * point is the size of a message where the partial delivery API will be
3080 * invoked to help free up rwnd space for the peer. Setting this to a
3081 * lower value will cause partial deliveries to happen more often. The
3082 * calls argument is an integer that sets or gets the partial delivery
3083 * point. Note also that the call will fail if the user attempts to set
3084 * this value larger than the socket receive buffer size.
3086 * Note that any single message having a length smaller than or equal to
3087 * the SCTP partial delivery point will be delivered in one single read
3088 * call as long as the user provided buffer is large enough to hold the
3089 * message.
3091 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3092 char __user *optval,
3093 unsigned int optlen)
3095 u32 val;
3097 if (optlen != sizeof(u32))
3098 return -EINVAL;
3099 if (get_user(val, (int __user *)optval))
3100 return -EFAULT;
3102 /* Note: We double the receive buffer from what the user sets
3103 * it to be, also initial rwnd is based on rcvbuf/2.
3105 if (val > (sk->sk_rcvbuf >> 1))
3106 return -EINVAL;
3108 sctp_sk(sk)->pd_point = val;
3110 return 0; /* is this the right error code? */
3114 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3116 * This option will allow a user to change the maximum burst of packets
3117 * that can be emitted by this association. Note that the default value
3118 * is 4, and some implementations may restrict this setting so that it
3119 * can only be lowered.
3121 * NOTE: This text doesn't seem right. Do this on a socket basis with
3122 * future associations inheriting the socket value.
3124 static int sctp_setsockopt_maxburst(struct sock *sk,
3125 char __user *optval,
3126 unsigned int optlen)
3128 struct sctp_assoc_value params;
3129 struct sctp_sock *sp;
3130 struct sctp_association *asoc;
3131 int val;
3132 int assoc_id = 0;
3134 if (optlen == sizeof(int)) {
3135 printk(KERN_WARNING
3136 "SCTP: Use of int in max_burst socket option deprecated\n");
3137 printk(KERN_WARNING
3138 "SCTP: Use struct sctp_assoc_value instead\n");
3139 if (copy_from_user(&val, optval, optlen))
3140 return -EFAULT;
3141 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3142 if (copy_from_user(&params, optval, optlen))
3143 return -EFAULT;
3144 val = params.assoc_value;
3145 assoc_id = params.assoc_id;
3146 } else
3147 return -EINVAL;
3149 sp = sctp_sk(sk);
3151 if (assoc_id != 0) {
3152 asoc = sctp_id2assoc(sk, assoc_id);
3153 if (!asoc)
3154 return -EINVAL;
3155 asoc->max_burst = val;
3156 } else
3157 sp->max_burst = val;
3159 return 0;
3163 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3165 * This set option adds a chunk type that the user is requesting to be
3166 * received only in an authenticated way. Changes to the list of chunks
3167 * will only effect future associations on the socket.
3169 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3170 char __user *optval,
3171 unsigned int optlen)
3173 struct sctp_authchunk val;
3175 if (!sctp_auth_enable)
3176 return -EACCES;
3178 if (optlen != sizeof(struct sctp_authchunk))
3179 return -EINVAL;
3180 if (copy_from_user(&val, optval, optlen))
3181 return -EFAULT;
3183 switch (val.sauth_chunk) {
3184 case SCTP_CID_INIT:
3185 case SCTP_CID_INIT_ACK:
3186 case SCTP_CID_SHUTDOWN_COMPLETE:
3187 case SCTP_CID_AUTH:
3188 return -EINVAL;
3191 /* add this chunk id to the endpoint */
3192 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3196 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3198 * This option gets or sets the list of HMAC algorithms that the local
3199 * endpoint requires the peer to use.
3201 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3202 char __user *optval,
3203 unsigned int optlen)
3205 struct sctp_hmacalgo *hmacs;
3206 u32 idents;
3207 int err;
3209 if (!sctp_auth_enable)
3210 return -EACCES;
3212 if (optlen < sizeof(struct sctp_hmacalgo))
3213 return -EINVAL;
3215 hmacs = kmalloc(optlen, GFP_KERNEL);
3216 if (!hmacs)
3217 return -ENOMEM;
3219 if (copy_from_user(hmacs, optval, optlen)) {
3220 err = -EFAULT;
3221 goto out;
3224 idents = hmacs->shmac_num_idents;
3225 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3226 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3227 err = -EINVAL;
3228 goto out;
3231 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3232 out:
3233 kfree(hmacs);
3234 return err;
3238 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3240 * This option will set a shared secret key which is used to build an
3241 * association shared key.
3243 static int sctp_setsockopt_auth_key(struct sock *sk,
3244 char __user *optval,
3245 unsigned int optlen)
3247 struct sctp_authkey *authkey;
3248 struct sctp_association *asoc;
3249 int ret;
3251 if (!sctp_auth_enable)
3252 return -EACCES;
3254 if (optlen <= sizeof(struct sctp_authkey))
3255 return -EINVAL;
3257 authkey = kmalloc(optlen, GFP_KERNEL);
3258 if (!authkey)
3259 return -ENOMEM;
3261 if (copy_from_user(authkey, optval, optlen)) {
3262 ret = -EFAULT;
3263 goto out;
3266 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3267 ret = -EINVAL;
3268 goto out;
3271 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3272 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3273 ret = -EINVAL;
3274 goto out;
3277 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3278 out:
3279 kfree(authkey);
3280 return ret;
3284 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3286 * This option will get or set the active shared key to be used to build
3287 * the association shared key.
3289 static int sctp_setsockopt_active_key(struct sock *sk,
3290 char __user *optval,
3291 unsigned int optlen)
3293 struct sctp_authkeyid val;
3294 struct sctp_association *asoc;
3296 if (!sctp_auth_enable)
3297 return -EACCES;
3299 if (optlen != sizeof(struct sctp_authkeyid))
3300 return -EINVAL;
3301 if (copy_from_user(&val, optval, optlen))
3302 return -EFAULT;
3304 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3305 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3306 return -EINVAL;
3308 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3309 val.scact_keynumber);
3313 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3315 * This set option will delete a shared secret key from use.
3317 static int sctp_setsockopt_del_key(struct sock *sk,
3318 char __user *optval,
3319 unsigned int optlen)
3321 struct sctp_authkeyid val;
3322 struct sctp_association *asoc;
3324 if (!sctp_auth_enable)
3325 return -EACCES;
3327 if (optlen != sizeof(struct sctp_authkeyid))
3328 return -EINVAL;
3329 if (copy_from_user(&val, optval, optlen))
3330 return -EFAULT;
3332 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3333 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3334 return -EINVAL;
3336 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3337 val.scact_keynumber);
3342 /* API 6.2 setsockopt(), getsockopt()
3344 * Applications use setsockopt() and getsockopt() to set or retrieve
3345 * socket options. Socket options are used to change the default
3346 * behavior of sockets calls. They are described in Section 7.
3348 * The syntax is:
3350 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3351 * int __user *optlen);
3352 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3353 * int optlen);
3355 * sd - the socket descript.
3356 * level - set to IPPROTO_SCTP for all SCTP options.
3357 * optname - the option name.
3358 * optval - the buffer to store the value of the option.
3359 * optlen - the size of the buffer.
3361 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3362 char __user *optval, unsigned int optlen)
3364 int retval = 0;
3366 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3367 sk, optname);
3369 /* I can hardly begin to describe how wrong this is. This is
3370 * so broken as to be worse than useless. The API draft
3371 * REALLY is NOT helpful here... I am not convinced that the
3372 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3373 * are at all well-founded.
3375 if (level != SOL_SCTP) {
3376 struct sctp_af *af = sctp_sk(sk)->pf->af;
3377 retval = af->setsockopt(sk, level, optname, optval, optlen);
3378 goto out_nounlock;
3381 sctp_lock_sock(sk);
3383 switch (optname) {
3384 case SCTP_SOCKOPT_BINDX_ADD:
3385 /* 'optlen' is the size of the addresses buffer. */
3386 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3387 optlen, SCTP_BINDX_ADD_ADDR);
3388 break;
3390 case SCTP_SOCKOPT_BINDX_REM:
3391 /* 'optlen' is the size of the addresses buffer. */
3392 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3393 optlen, SCTP_BINDX_REM_ADDR);
3394 break;
3396 case SCTP_SOCKOPT_CONNECTX_OLD:
3397 /* 'optlen' is the size of the addresses buffer. */
3398 retval = sctp_setsockopt_connectx_old(sk,
3399 (struct sockaddr __user *)optval,
3400 optlen);
3401 break;
3403 case SCTP_SOCKOPT_CONNECTX:
3404 /* 'optlen' is the size of the addresses buffer. */
3405 retval = sctp_setsockopt_connectx(sk,
3406 (struct sockaddr __user *)optval,
3407 optlen);
3408 break;
3410 case SCTP_DISABLE_FRAGMENTS:
3411 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3412 break;
3414 case SCTP_EVENTS:
3415 retval = sctp_setsockopt_events(sk, optval, optlen);
3416 break;
3418 case SCTP_AUTOCLOSE:
3419 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3420 break;
3422 case SCTP_PEER_ADDR_PARAMS:
3423 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3424 break;
3426 case SCTP_DELAYED_ACK:
3427 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3428 break;
3429 case SCTP_PARTIAL_DELIVERY_POINT:
3430 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3431 break;
3433 case SCTP_INITMSG:
3434 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3435 break;
3436 case SCTP_DEFAULT_SEND_PARAM:
3437 retval = sctp_setsockopt_default_send_param(sk, optval,
3438 optlen);
3439 break;
3440 case SCTP_PRIMARY_ADDR:
3441 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3442 break;
3443 case SCTP_SET_PEER_PRIMARY_ADDR:
3444 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3445 break;
3446 case SCTP_NODELAY:
3447 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3448 break;
3449 case SCTP_RTOINFO:
3450 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3451 break;
3452 case SCTP_ASSOCINFO:
3453 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3454 break;
3455 case SCTP_I_WANT_MAPPED_V4_ADDR:
3456 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3457 break;
3458 case SCTP_MAXSEG:
3459 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3460 break;
3461 case SCTP_ADAPTATION_LAYER:
3462 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3463 break;
3464 case SCTP_CONTEXT:
3465 retval = sctp_setsockopt_context(sk, optval, optlen);
3466 break;
3467 case SCTP_FRAGMENT_INTERLEAVE:
3468 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3469 break;
3470 case SCTP_MAX_BURST:
3471 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3472 break;
3473 case SCTP_AUTH_CHUNK:
3474 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3475 break;
3476 case SCTP_HMAC_IDENT:
3477 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3478 break;
3479 case SCTP_AUTH_KEY:
3480 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3481 break;
3482 case SCTP_AUTH_ACTIVE_KEY:
3483 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3484 break;
3485 case SCTP_AUTH_DELETE_KEY:
3486 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3487 break;
3488 default:
3489 retval = -ENOPROTOOPT;
3490 break;
3493 sctp_release_sock(sk);
3495 out_nounlock:
3496 return retval;
3499 /* API 3.1.6 connect() - UDP Style Syntax
3501 * An application may use the connect() call in the UDP model to initiate an
3502 * association without sending data.
3504 * The syntax is:
3506 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3508 * sd: the socket descriptor to have a new association added to.
3510 * nam: the address structure (either struct sockaddr_in or struct
3511 * sockaddr_in6 defined in RFC2553 [7]).
3513 * len: the size of the address.
3515 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3516 int addr_len)
3518 int err = 0;
3519 struct sctp_af *af;
3521 sctp_lock_sock(sk);
3523 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3524 __func__, sk, addr, addr_len);
3526 /* Validate addr_len before calling common connect/connectx routine. */
3527 af = sctp_get_af_specific(addr->sa_family);
3528 if (!af || addr_len < af->sockaddr_len) {
3529 err = -EINVAL;
3530 } else {
3531 /* Pass correct addr len to common routine (so it knows there
3532 * is only one address being passed.
3534 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3537 sctp_release_sock(sk);
3538 return err;
3541 /* FIXME: Write comments. */
3542 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3544 return -EOPNOTSUPP; /* STUB */
3547 /* 4.1.4 accept() - TCP Style Syntax
3549 * Applications use accept() call to remove an established SCTP
3550 * association from the accept queue of the endpoint. A new socket
3551 * descriptor will be returned from accept() to represent the newly
3552 * formed association.
3554 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3556 struct sctp_sock *sp;
3557 struct sctp_endpoint *ep;
3558 struct sock *newsk = NULL;
3559 struct sctp_association *asoc;
3560 long timeo;
3561 int error = 0;
3563 sctp_lock_sock(sk);
3565 sp = sctp_sk(sk);
3566 ep = sp->ep;
3568 if (!sctp_style(sk, TCP)) {
3569 error = -EOPNOTSUPP;
3570 goto out;
3573 if (!sctp_sstate(sk, LISTENING)) {
3574 error = -EINVAL;
3575 goto out;
3578 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3580 error = sctp_wait_for_accept(sk, timeo);
3581 if (error)
3582 goto out;
3584 /* We treat the list of associations on the endpoint as the accept
3585 * queue and pick the first association on the list.
3587 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3589 newsk = sp->pf->create_accept_sk(sk, asoc);
3590 if (!newsk) {
3591 error = -ENOMEM;
3592 goto out;
3595 /* Populate the fields of the newsk from the oldsk and migrate the
3596 * asoc to the newsk.
3598 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3600 out:
3601 sctp_release_sock(sk);
3602 *err = error;
3603 return newsk;
3606 /* The SCTP ioctl handler. */
3607 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3609 return -ENOIOCTLCMD;
3612 /* This is the function which gets called during socket creation to
3613 * initialized the SCTP-specific portion of the sock.
3614 * The sock structure should already be zero-filled memory.
3616 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3618 struct sctp_endpoint *ep;
3619 struct sctp_sock *sp;
3621 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3623 sp = sctp_sk(sk);
3625 /* Initialize the SCTP per socket area. */
3626 switch (sk->sk_type) {
3627 case SOCK_SEQPACKET:
3628 sp->type = SCTP_SOCKET_UDP;
3629 break;
3630 case SOCK_STREAM:
3631 sp->type = SCTP_SOCKET_TCP;
3632 break;
3633 default:
3634 return -ESOCKTNOSUPPORT;
3637 /* Initialize default send parameters. These parameters can be
3638 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3640 sp->default_stream = 0;
3641 sp->default_ppid = 0;
3642 sp->default_flags = 0;
3643 sp->default_context = 0;
3644 sp->default_timetolive = 0;
3646 sp->default_rcv_context = 0;
3647 sp->max_burst = sctp_max_burst;
3649 /* Initialize default setup parameters. These parameters
3650 * can be modified with the SCTP_INITMSG socket option or
3651 * overridden by the SCTP_INIT CMSG.
3653 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3654 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3655 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3656 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3658 /* Initialize default RTO related parameters. These parameters can
3659 * be modified for with the SCTP_RTOINFO socket option.
3661 sp->rtoinfo.srto_initial = sctp_rto_initial;
3662 sp->rtoinfo.srto_max = sctp_rto_max;
3663 sp->rtoinfo.srto_min = sctp_rto_min;
3665 /* Initialize default association related parameters. These parameters
3666 * can be modified with the SCTP_ASSOCINFO socket option.
3668 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3669 sp->assocparams.sasoc_number_peer_destinations = 0;
3670 sp->assocparams.sasoc_peer_rwnd = 0;
3671 sp->assocparams.sasoc_local_rwnd = 0;
3672 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3674 /* Initialize default event subscriptions. By default, all the
3675 * options are off.
3677 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3679 /* Default Peer Address Parameters. These defaults can
3680 * be modified via SCTP_PEER_ADDR_PARAMS
3682 sp->hbinterval = sctp_hb_interval;
3683 sp->pathmaxrxt = sctp_max_retrans_path;
3684 sp->pathmtu = 0; // allow default discovery
3685 sp->sackdelay = sctp_sack_timeout;
3686 sp->sackfreq = 2;
3687 sp->param_flags = SPP_HB_ENABLE |
3688 SPP_PMTUD_ENABLE |
3689 SPP_SACKDELAY_ENABLE;
3691 /* If enabled no SCTP message fragmentation will be performed.
3692 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3694 sp->disable_fragments = 0;
3696 /* Enable Nagle algorithm by default. */
3697 sp->nodelay = 0;
3699 /* Enable by default. */
3700 sp->v4mapped = 1;
3702 /* Auto-close idle associations after the configured
3703 * number of seconds. A value of 0 disables this
3704 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3705 * for UDP-style sockets only.
3707 sp->autoclose = 0;
3709 /* User specified fragmentation limit. */
3710 sp->user_frag = 0;
3712 sp->adaptation_ind = 0;
3714 sp->pf = sctp_get_pf_specific(sk->sk_family);
3716 /* Control variables for partial data delivery. */
3717 atomic_set(&sp->pd_mode, 0);
3718 skb_queue_head_init(&sp->pd_lobby);
3719 sp->frag_interleave = 0;
3721 /* Create a per socket endpoint structure. Even if we
3722 * change the data structure relationships, this may still
3723 * be useful for storing pre-connect address information.
3725 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3726 if (!ep)
3727 return -ENOMEM;
3729 sp->ep = ep;
3730 sp->hmac = NULL;
3732 SCTP_DBG_OBJCNT_INC(sock);
3734 local_bh_disable();
3735 percpu_counter_inc(&sctp_sockets_allocated);
3736 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3737 local_bh_enable();
3739 return 0;
3742 /* Cleanup any SCTP per socket resources. */
3743 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3745 struct sctp_endpoint *ep;
3747 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3749 /* Release our hold on the endpoint. */
3750 ep = sctp_sk(sk)->ep;
3751 sctp_endpoint_free(ep);
3752 local_bh_disable();
3753 percpu_counter_dec(&sctp_sockets_allocated);
3754 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3755 local_bh_enable();
3758 /* API 4.1.7 shutdown() - TCP Style Syntax
3759 * int shutdown(int socket, int how);
3761 * sd - the socket descriptor of the association to be closed.
3762 * how - Specifies the type of shutdown. The values are
3763 * as follows:
3764 * SHUT_RD
3765 * Disables further receive operations. No SCTP
3766 * protocol action is taken.
3767 * SHUT_WR
3768 * Disables further send operations, and initiates
3769 * the SCTP shutdown sequence.
3770 * SHUT_RDWR
3771 * Disables further send and receive operations
3772 * and initiates the SCTP shutdown sequence.
3774 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3776 struct sctp_endpoint *ep;
3777 struct sctp_association *asoc;
3779 if (!sctp_style(sk, TCP))
3780 return;
3782 if (how & SEND_SHUTDOWN) {
3783 ep = sctp_sk(sk)->ep;
3784 if (!list_empty(&ep->asocs)) {
3785 asoc = list_entry(ep->asocs.next,
3786 struct sctp_association, asocs);
3787 sctp_primitive_SHUTDOWN(asoc, NULL);
3792 /* 7.2.1 Association Status (SCTP_STATUS)
3794 * Applications can retrieve current status information about an
3795 * association, including association state, peer receiver window size,
3796 * number of unacked data chunks, and number of data chunks pending
3797 * receipt. This information is read-only.
3799 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3800 char __user *optval,
3801 int __user *optlen)
3803 struct sctp_status status;
3804 struct sctp_association *asoc = NULL;
3805 struct sctp_transport *transport;
3806 sctp_assoc_t associd;
3807 int retval = 0;
3809 if (len < sizeof(status)) {
3810 retval = -EINVAL;
3811 goto out;
3814 len = sizeof(status);
3815 if (copy_from_user(&status, optval, len)) {
3816 retval = -EFAULT;
3817 goto out;
3820 associd = status.sstat_assoc_id;
3821 asoc = sctp_id2assoc(sk, associd);
3822 if (!asoc) {
3823 retval = -EINVAL;
3824 goto out;
3827 transport = asoc->peer.primary_path;
3829 status.sstat_assoc_id = sctp_assoc2id(asoc);
3830 status.sstat_state = asoc->state;
3831 status.sstat_rwnd = asoc->peer.rwnd;
3832 status.sstat_unackdata = asoc->unack_data;
3834 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3835 status.sstat_instrms = asoc->c.sinit_max_instreams;
3836 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3837 status.sstat_fragmentation_point = asoc->frag_point;
3838 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3839 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3840 transport->af_specific->sockaddr_len);
3841 /* Map ipv4 address into v4-mapped-on-v6 address. */
3842 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3843 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3844 status.sstat_primary.spinfo_state = transport->state;
3845 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3846 status.sstat_primary.spinfo_srtt = transport->srtt;
3847 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3848 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3850 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3851 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3853 if (put_user(len, optlen)) {
3854 retval = -EFAULT;
3855 goto out;
3858 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3859 len, status.sstat_state, status.sstat_rwnd,
3860 status.sstat_assoc_id);
3862 if (copy_to_user(optval, &status, len)) {
3863 retval = -EFAULT;
3864 goto out;
3867 out:
3868 return (retval);
3872 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3874 * Applications can retrieve information about a specific peer address
3875 * of an association, including its reachability state, congestion
3876 * window, and retransmission timer values. This information is
3877 * read-only.
3879 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3880 char __user *optval,
3881 int __user *optlen)
3883 struct sctp_paddrinfo pinfo;
3884 struct sctp_transport *transport;
3885 int retval = 0;
3887 if (len < sizeof(pinfo)) {
3888 retval = -EINVAL;
3889 goto out;
3892 len = sizeof(pinfo);
3893 if (copy_from_user(&pinfo, optval, len)) {
3894 retval = -EFAULT;
3895 goto out;
3898 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3899 pinfo.spinfo_assoc_id);
3900 if (!transport)
3901 return -EINVAL;
3903 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3904 pinfo.spinfo_state = transport->state;
3905 pinfo.spinfo_cwnd = transport->cwnd;
3906 pinfo.spinfo_srtt = transport->srtt;
3907 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3908 pinfo.spinfo_mtu = transport->pathmtu;
3910 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3911 pinfo.spinfo_state = SCTP_ACTIVE;
3913 if (put_user(len, optlen)) {
3914 retval = -EFAULT;
3915 goto out;
3918 if (copy_to_user(optval, &pinfo, len)) {
3919 retval = -EFAULT;
3920 goto out;
3923 out:
3924 return (retval);
3927 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3929 * This option is a on/off flag. If enabled no SCTP message
3930 * fragmentation will be performed. Instead if a message being sent
3931 * exceeds the current PMTU size, the message will NOT be sent and
3932 * instead a error will be indicated to the user.
3934 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3935 char __user *optval, int __user *optlen)
3937 int val;
3939 if (len < sizeof(int))
3940 return -EINVAL;
3942 len = sizeof(int);
3943 val = (sctp_sk(sk)->disable_fragments == 1);
3944 if (put_user(len, optlen))
3945 return -EFAULT;
3946 if (copy_to_user(optval, &val, len))
3947 return -EFAULT;
3948 return 0;
3951 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3953 * This socket option is used to specify various notifications and
3954 * ancillary data the user wishes to receive.
3956 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3957 int __user *optlen)
3959 if (len < sizeof(struct sctp_event_subscribe))
3960 return -EINVAL;
3961 len = sizeof(struct sctp_event_subscribe);
3962 if (put_user(len, optlen))
3963 return -EFAULT;
3964 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3965 return -EFAULT;
3966 return 0;
3969 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3971 * This socket option is applicable to the UDP-style socket only. When
3972 * set it will cause associations that are idle for more than the
3973 * specified number of seconds to automatically close. An association
3974 * being idle is defined an association that has NOT sent or received
3975 * user data. The special value of '0' indicates that no automatic
3976 * close of any associations should be performed. The option expects an
3977 * integer defining the number of seconds of idle time before an
3978 * association is closed.
3980 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3982 /* Applicable to UDP-style socket only */
3983 if (sctp_style(sk, TCP))
3984 return -EOPNOTSUPP;
3985 if (len < sizeof(int))
3986 return -EINVAL;
3987 len = sizeof(int);
3988 if (put_user(len, optlen))
3989 return -EFAULT;
3990 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3991 return -EFAULT;
3992 return 0;
3995 /* Helper routine to branch off an association to a new socket. */
3996 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3997 struct socket **sockp)
3999 struct sock *sk = asoc->base.sk;
4000 struct socket *sock;
4001 struct sctp_af *af;
4002 int err = 0;
4004 /* An association cannot be branched off from an already peeled-off
4005 * socket, nor is this supported for tcp style sockets.
4007 if (!sctp_style(sk, UDP))
4008 return -EINVAL;
4010 /* Create a new socket. */
4011 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4012 if (err < 0)
4013 return err;
4015 sctp_copy_sock(sock->sk, sk, asoc);
4017 /* Make peeled-off sockets more like 1-1 accepted sockets.
4018 * Set the daddr and initialize id to something more random
4020 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4021 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4023 /* Populate the fields of the newsk from the oldsk and migrate the
4024 * asoc to the newsk.
4026 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4028 *sockp = sock;
4030 return err;
4033 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4035 sctp_peeloff_arg_t peeloff;
4036 struct socket *newsock;
4037 int retval = 0;
4038 struct sctp_association *asoc;
4040 if (len < sizeof(sctp_peeloff_arg_t))
4041 return -EINVAL;
4042 len = sizeof(sctp_peeloff_arg_t);
4043 if (copy_from_user(&peeloff, optval, len))
4044 return -EFAULT;
4046 asoc = sctp_id2assoc(sk, peeloff.associd);
4047 if (!asoc) {
4048 retval = -EINVAL;
4049 goto out;
4052 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4054 retval = sctp_do_peeloff(asoc, &newsock);
4055 if (retval < 0)
4056 goto out;
4058 /* Map the socket to an unused fd that can be returned to the user. */
4059 retval = sock_map_fd(newsock, 0);
4060 if (retval < 0) {
4061 sock_release(newsock);
4062 goto out;
4065 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4066 __func__, sk, asoc, newsock->sk, retval);
4068 /* Return the fd mapped to the new socket. */
4069 peeloff.sd = retval;
4070 if (put_user(len, optlen))
4071 return -EFAULT;
4072 if (copy_to_user(optval, &peeloff, len))
4073 retval = -EFAULT;
4075 out:
4076 return retval;
4079 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4081 * Applications can enable or disable heartbeats for any peer address of
4082 * an association, modify an address's heartbeat interval, force a
4083 * heartbeat to be sent immediately, and adjust the address's maximum
4084 * number of retransmissions sent before an address is considered
4085 * unreachable. The following structure is used to access and modify an
4086 * address's parameters:
4088 * struct sctp_paddrparams {
4089 * sctp_assoc_t spp_assoc_id;
4090 * struct sockaddr_storage spp_address;
4091 * uint32_t spp_hbinterval;
4092 * uint16_t spp_pathmaxrxt;
4093 * uint32_t spp_pathmtu;
4094 * uint32_t spp_sackdelay;
4095 * uint32_t spp_flags;
4096 * };
4098 * spp_assoc_id - (one-to-many style socket) This is filled in the
4099 * application, and identifies the association for
4100 * this query.
4101 * spp_address - This specifies which address is of interest.
4102 * spp_hbinterval - This contains the value of the heartbeat interval,
4103 * in milliseconds. If a value of zero
4104 * is present in this field then no changes are to
4105 * be made to this parameter.
4106 * spp_pathmaxrxt - This contains the maximum number of
4107 * retransmissions before this address shall be
4108 * considered unreachable. If a value of zero
4109 * is present in this field then no changes are to
4110 * be made to this parameter.
4111 * spp_pathmtu - When Path MTU discovery is disabled the value
4112 * specified here will be the "fixed" path mtu.
4113 * Note that if the spp_address field is empty
4114 * then all associations on this address will
4115 * have this fixed path mtu set upon them.
4117 * spp_sackdelay - When delayed sack is enabled, this value specifies
4118 * the number of milliseconds that sacks will be delayed
4119 * for. This value will apply to all addresses of an
4120 * association if the spp_address field is empty. Note
4121 * also, that if delayed sack is enabled and this
4122 * value is set to 0, no change is made to the last
4123 * recorded delayed sack timer value.
4125 * spp_flags - These flags are used to control various features
4126 * on an association. The flag field may contain
4127 * zero or more of the following options.
4129 * SPP_HB_ENABLE - Enable heartbeats on the
4130 * specified address. Note that if the address
4131 * field is empty all addresses for the association
4132 * have heartbeats enabled upon them.
4134 * SPP_HB_DISABLE - Disable heartbeats on the
4135 * speicifed address. Note that if the address
4136 * field is empty all addresses for the association
4137 * will have their heartbeats disabled. Note also
4138 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4139 * mutually exclusive, only one of these two should
4140 * be specified. Enabling both fields will have
4141 * undetermined results.
4143 * SPP_HB_DEMAND - Request a user initiated heartbeat
4144 * to be made immediately.
4146 * SPP_PMTUD_ENABLE - This field will enable PMTU
4147 * discovery upon the specified address. Note that
4148 * if the address feild is empty then all addresses
4149 * on the association are effected.
4151 * SPP_PMTUD_DISABLE - This field will disable PMTU
4152 * discovery upon the specified address. Note that
4153 * if the address feild is empty then all addresses
4154 * on the association are effected. Not also that
4155 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4156 * exclusive. Enabling both will have undetermined
4157 * results.
4159 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4160 * on delayed sack. The time specified in spp_sackdelay
4161 * is used to specify the sack delay for this address. Note
4162 * that if spp_address is empty then all addresses will
4163 * enable delayed sack and take on the sack delay
4164 * value specified in spp_sackdelay.
4165 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4166 * off delayed sack. If the spp_address field is blank then
4167 * delayed sack is disabled for the entire association. Note
4168 * also that this field is mutually exclusive to
4169 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4170 * results.
4172 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4173 char __user *optval, int __user *optlen)
4175 struct sctp_paddrparams params;
4176 struct sctp_transport *trans = NULL;
4177 struct sctp_association *asoc = NULL;
4178 struct sctp_sock *sp = sctp_sk(sk);
4180 if (len < sizeof(struct sctp_paddrparams))
4181 return -EINVAL;
4182 len = sizeof(struct sctp_paddrparams);
4183 if (copy_from_user(&params, optval, len))
4184 return -EFAULT;
4186 /* If an address other than INADDR_ANY is specified, and
4187 * no transport is found, then the request is invalid.
4189 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4190 trans = sctp_addr_id2transport(sk, &params.spp_address,
4191 params.spp_assoc_id);
4192 if (!trans) {
4193 SCTP_DEBUG_PRINTK("Failed no transport\n");
4194 return -EINVAL;
4198 /* Get association, if assoc_id != 0 and the socket is a one
4199 * to many style socket, and an association was not found, then
4200 * the id was invalid.
4202 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4203 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4204 SCTP_DEBUG_PRINTK("Failed no association\n");
4205 return -EINVAL;
4208 if (trans) {
4209 /* Fetch transport values. */
4210 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4211 params.spp_pathmtu = trans->pathmtu;
4212 params.spp_pathmaxrxt = trans->pathmaxrxt;
4213 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4215 /*draft-11 doesn't say what to return in spp_flags*/
4216 params.spp_flags = trans->param_flags;
4217 } else if (asoc) {
4218 /* Fetch association values. */
4219 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4220 params.spp_pathmtu = asoc->pathmtu;
4221 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4222 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4224 /*draft-11 doesn't say what to return in spp_flags*/
4225 params.spp_flags = asoc->param_flags;
4226 } else {
4227 /* Fetch socket values. */
4228 params.spp_hbinterval = sp->hbinterval;
4229 params.spp_pathmtu = sp->pathmtu;
4230 params.spp_sackdelay = sp->sackdelay;
4231 params.spp_pathmaxrxt = sp->pathmaxrxt;
4233 /*draft-11 doesn't say what to return in spp_flags*/
4234 params.spp_flags = sp->param_flags;
4237 if (copy_to_user(optval, &params, len))
4238 return -EFAULT;
4240 if (put_user(len, optlen))
4241 return -EFAULT;
4243 return 0;
4247 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4249 * This option will effect the way delayed acks are performed. This
4250 * option allows you to get or set the delayed ack time, in
4251 * milliseconds. It also allows changing the delayed ack frequency.
4252 * Changing the frequency to 1 disables the delayed sack algorithm. If
4253 * the assoc_id is 0, then this sets or gets the endpoints default
4254 * values. If the assoc_id field is non-zero, then the set or get
4255 * effects the specified association for the one to many model (the
4256 * assoc_id field is ignored by the one to one model). Note that if
4257 * sack_delay or sack_freq are 0 when setting this option, then the
4258 * current values will remain unchanged.
4260 * struct sctp_sack_info {
4261 * sctp_assoc_t sack_assoc_id;
4262 * uint32_t sack_delay;
4263 * uint32_t sack_freq;
4264 * };
4266 * sack_assoc_id - This parameter, indicates which association the user
4267 * is performing an action upon. Note that if this field's value is
4268 * zero then the endpoints default value is changed (effecting future
4269 * associations only).
4271 * sack_delay - This parameter contains the number of milliseconds that
4272 * the user is requesting the delayed ACK timer be set to. Note that
4273 * this value is defined in the standard to be between 200 and 500
4274 * milliseconds.
4276 * sack_freq - This parameter contains the number of packets that must
4277 * be received before a sack is sent without waiting for the delay
4278 * timer to expire. The default value for this is 2, setting this
4279 * value to 1 will disable the delayed sack algorithm.
4281 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4282 char __user *optval,
4283 int __user *optlen)
4285 struct sctp_sack_info params;
4286 struct sctp_association *asoc = NULL;
4287 struct sctp_sock *sp = sctp_sk(sk);
4289 if (len >= sizeof(struct sctp_sack_info)) {
4290 len = sizeof(struct sctp_sack_info);
4292 if (copy_from_user(&params, optval, len))
4293 return -EFAULT;
4294 } else if (len == sizeof(struct sctp_assoc_value)) {
4295 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4296 "in delayed_ack socket option deprecated\n");
4297 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4298 if (copy_from_user(&params, optval, len))
4299 return -EFAULT;
4300 } else
4301 return - EINVAL;
4303 /* Get association, if sack_assoc_id != 0 and the socket is a one
4304 * to many style socket, and an association was not found, then
4305 * the id was invalid.
4307 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4308 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4309 return -EINVAL;
4311 if (asoc) {
4312 /* Fetch association values. */
4313 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4314 params.sack_delay = jiffies_to_msecs(
4315 asoc->sackdelay);
4316 params.sack_freq = asoc->sackfreq;
4318 } else {
4319 params.sack_delay = 0;
4320 params.sack_freq = 1;
4322 } else {
4323 /* Fetch socket values. */
4324 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4325 params.sack_delay = sp->sackdelay;
4326 params.sack_freq = sp->sackfreq;
4327 } else {
4328 params.sack_delay = 0;
4329 params.sack_freq = 1;
4333 if (copy_to_user(optval, &params, len))
4334 return -EFAULT;
4336 if (put_user(len, optlen))
4337 return -EFAULT;
4339 return 0;
4342 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4344 * Applications can specify protocol parameters for the default association
4345 * initialization. The option name argument to setsockopt() and getsockopt()
4346 * is SCTP_INITMSG.
4348 * Setting initialization parameters is effective only on an unconnected
4349 * socket (for UDP-style sockets only future associations are effected
4350 * by the change). With TCP-style sockets, this option is inherited by
4351 * sockets derived from a listener socket.
4353 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4355 if (len < sizeof(struct sctp_initmsg))
4356 return -EINVAL;
4357 len = sizeof(struct sctp_initmsg);
4358 if (put_user(len, optlen))
4359 return -EFAULT;
4360 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4361 return -EFAULT;
4362 return 0;
4366 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4367 char __user *optval, int __user *optlen)
4369 struct sctp_association *asoc;
4370 int cnt = 0;
4371 struct sctp_getaddrs getaddrs;
4372 struct sctp_transport *from;
4373 void __user *to;
4374 union sctp_addr temp;
4375 struct sctp_sock *sp = sctp_sk(sk);
4376 int addrlen;
4377 size_t space_left;
4378 int bytes_copied;
4380 if (len < sizeof(struct sctp_getaddrs))
4381 return -EINVAL;
4383 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4384 return -EFAULT;
4386 /* For UDP-style sockets, id specifies the association to query. */
4387 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4388 if (!asoc)
4389 return -EINVAL;
4391 to = optval + offsetof(struct sctp_getaddrs,addrs);
4392 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4394 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4395 transports) {
4396 memcpy(&temp, &from->ipaddr, sizeof(temp));
4397 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4398 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4399 if (space_left < addrlen)
4400 return -ENOMEM;
4401 if (copy_to_user(to, &temp, addrlen))
4402 return -EFAULT;
4403 to += addrlen;
4404 cnt++;
4405 space_left -= addrlen;
4408 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4409 return -EFAULT;
4410 bytes_copied = ((char __user *)to) - optval;
4411 if (put_user(bytes_copied, optlen))
4412 return -EFAULT;
4414 return 0;
4417 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4418 size_t space_left, int *bytes_copied)
4420 struct sctp_sockaddr_entry *addr;
4421 union sctp_addr temp;
4422 int cnt = 0;
4423 int addrlen;
4425 rcu_read_lock();
4426 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4427 if (!addr->valid)
4428 continue;
4430 if ((PF_INET == sk->sk_family) &&
4431 (AF_INET6 == addr->a.sa.sa_family))
4432 continue;
4433 if ((PF_INET6 == sk->sk_family) &&
4434 inet_v6_ipv6only(sk) &&
4435 (AF_INET == addr->a.sa.sa_family))
4436 continue;
4437 memcpy(&temp, &addr->a, sizeof(temp));
4438 if (!temp.v4.sin_port)
4439 temp.v4.sin_port = htons(port);
4441 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4442 &temp);
4443 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4444 if (space_left < addrlen) {
4445 cnt = -ENOMEM;
4446 break;
4448 memcpy(to, &temp, addrlen);
4450 to += addrlen;
4451 cnt ++;
4452 space_left -= addrlen;
4453 *bytes_copied += addrlen;
4455 rcu_read_unlock();
4457 return cnt;
4461 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4462 char __user *optval, int __user *optlen)
4464 struct sctp_bind_addr *bp;
4465 struct sctp_association *asoc;
4466 int cnt = 0;
4467 struct sctp_getaddrs getaddrs;
4468 struct sctp_sockaddr_entry *addr;
4469 void __user *to;
4470 union sctp_addr temp;
4471 struct sctp_sock *sp = sctp_sk(sk);
4472 int addrlen;
4473 int err = 0;
4474 size_t space_left;
4475 int bytes_copied = 0;
4476 void *addrs;
4477 void *buf;
4479 if (len < sizeof(struct sctp_getaddrs))
4480 return -EINVAL;
4482 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4483 return -EFAULT;
4486 * For UDP-style sockets, id specifies the association to query.
4487 * If the id field is set to the value '0' then the locally bound
4488 * addresses are returned without regard to any particular
4489 * association.
4491 if (0 == getaddrs.assoc_id) {
4492 bp = &sctp_sk(sk)->ep->base.bind_addr;
4493 } else {
4494 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4495 if (!asoc)
4496 return -EINVAL;
4497 bp = &asoc->base.bind_addr;
4500 to = optval + offsetof(struct sctp_getaddrs,addrs);
4501 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4503 addrs = kmalloc(space_left, GFP_KERNEL);
4504 if (!addrs)
4505 return -ENOMEM;
4507 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4508 * addresses from the global local address list.
4510 if (sctp_list_single_entry(&bp->address_list)) {
4511 addr = list_entry(bp->address_list.next,
4512 struct sctp_sockaddr_entry, list);
4513 if (sctp_is_any(sk, &addr->a)) {
4514 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4515 space_left, &bytes_copied);
4516 if (cnt < 0) {
4517 err = cnt;
4518 goto out;
4520 goto copy_getaddrs;
4524 buf = addrs;
4525 /* Protection on the bound address list is not needed since
4526 * in the socket option context we hold a socket lock and
4527 * thus the bound address list can't change.
4529 list_for_each_entry(addr, &bp->address_list, list) {
4530 memcpy(&temp, &addr->a, sizeof(temp));
4531 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4532 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4533 if (space_left < addrlen) {
4534 err = -ENOMEM; /*fixme: right error?*/
4535 goto out;
4537 memcpy(buf, &temp, addrlen);
4538 buf += addrlen;
4539 bytes_copied += addrlen;
4540 cnt ++;
4541 space_left -= addrlen;
4544 copy_getaddrs:
4545 if (copy_to_user(to, addrs, bytes_copied)) {
4546 err = -EFAULT;
4547 goto out;
4549 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4550 err = -EFAULT;
4551 goto out;
4553 if (put_user(bytes_copied, optlen))
4554 err = -EFAULT;
4555 out:
4556 kfree(addrs);
4557 return err;
4560 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4562 * Requests that the local SCTP stack use the enclosed peer address as
4563 * the association primary. The enclosed address must be one of the
4564 * association peer's addresses.
4566 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4567 char __user *optval, int __user *optlen)
4569 struct sctp_prim prim;
4570 struct sctp_association *asoc;
4571 struct sctp_sock *sp = sctp_sk(sk);
4573 if (len < sizeof(struct sctp_prim))
4574 return -EINVAL;
4576 len = sizeof(struct sctp_prim);
4578 if (copy_from_user(&prim, optval, len))
4579 return -EFAULT;
4581 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4582 if (!asoc)
4583 return -EINVAL;
4585 if (!asoc->peer.primary_path)
4586 return -ENOTCONN;
4588 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4589 asoc->peer.primary_path->af_specific->sockaddr_len);
4591 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4592 (union sctp_addr *)&prim.ssp_addr);
4594 if (put_user(len, optlen))
4595 return -EFAULT;
4596 if (copy_to_user(optval, &prim, len))
4597 return -EFAULT;
4599 return 0;
4603 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4605 * Requests that the local endpoint set the specified Adaptation Layer
4606 * Indication parameter for all future INIT and INIT-ACK exchanges.
4608 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4609 char __user *optval, int __user *optlen)
4611 struct sctp_setadaptation adaptation;
4613 if (len < sizeof(struct sctp_setadaptation))
4614 return -EINVAL;
4616 len = sizeof(struct sctp_setadaptation);
4618 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4620 if (put_user(len, optlen))
4621 return -EFAULT;
4622 if (copy_to_user(optval, &adaptation, len))
4623 return -EFAULT;
4625 return 0;
4630 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4632 * Applications that wish to use the sendto() system call may wish to
4633 * specify a default set of parameters that would normally be supplied
4634 * through the inclusion of ancillary data. This socket option allows
4635 * such an application to set the default sctp_sndrcvinfo structure.
4638 * The application that wishes to use this socket option simply passes
4639 * in to this call the sctp_sndrcvinfo structure defined in Section
4640 * 5.2.2) The input parameters accepted by this call include
4641 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4642 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4643 * to this call if the caller is using the UDP model.
4645 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4647 static int sctp_getsockopt_default_send_param(struct sock *sk,
4648 int len, char __user *optval,
4649 int __user *optlen)
4651 struct sctp_sndrcvinfo info;
4652 struct sctp_association *asoc;
4653 struct sctp_sock *sp = sctp_sk(sk);
4655 if (len < sizeof(struct sctp_sndrcvinfo))
4656 return -EINVAL;
4658 len = sizeof(struct sctp_sndrcvinfo);
4660 if (copy_from_user(&info, optval, len))
4661 return -EFAULT;
4663 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4664 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4665 return -EINVAL;
4667 if (asoc) {
4668 info.sinfo_stream = asoc->default_stream;
4669 info.sinfo_flags = asoc->default_flags;
4670 info.sinfo_ppid = asoc->default_ppid;
4671 info.sinfo_context = asoc->default_context;
4672 info.sinfo_timetolive = asoc->default_timetolive;
4673 } else {
4674 info.sinfo_stream = sp->default_stream;
4675 info.sinfo_flags = sp->default_flags;
4676 info.sinfo_ppid = sp->default_ppid;
4677 info.sinfo_context = sp->default_context;
4678 info.sinfo_timetolive = sp->default_timetolive;
4681 if (put_user(len, optlen))
4682 return -EFAULT;
4683 if (copy_to_user(optval, &info, len))
4684 return -EFAULT;
4686 return 0;
4691 * 7.1.5 SCTP_NODELAY
4693 * Turn on/off any Nagle-like algorithm. This means that packets are
4694 * generally sent as soon as possible and no unnecessary delays are
4695 * introduced, at the cost of more packets in the network. Expects an
4696 * integer boolean flag.
4699 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4700 char __user *optval, int __user *optlen)
4702 int val;
4704 if (len < sizeof(int))
4705 return -EINVAL;
4707 len = sizeof(int);
4708 val = (sctp_sk(sk)->nodelay == 1);
4709 if (put_user(len, optlen))
4710 return -EFAULT;
4711 if (copy_to_user(optval, &val, len))
4712 return -EFAULT;
4713 return 0;
4718 * 7.1.1 SCTP_RTOINFO
4720 * The protocol parameters used to initialize and bound retransmission
4721 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4722 * and modify these parameters.
4723 * All parameters are time values, in milliseconds. A value of 0, when
4724 * modifying the parameters, indicates that the current value should not
4725 * be changed.
4728 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4729 char __user *optval,
4730 int __user *optlen) {
4731 struct sctp_rtoinfo rtoinfo;
4732 struct sctp_association *asoc;
4734 if (len < sizeof (struct sctp_rtoinfo))
4735 return -EINVAL;
4737 len = sizeof(struct sctp_rtoinfo);
4739 if (copy_from_user(&rtoinfo, optval, len))
4740 return -EFAULT;
4742 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4744 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4745 return -EINVAL;
4747 /* Values corresponding to the specific association. */
4748 if (asoc) {
4749 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4750 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4751 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4752 } else {
4753 /* Values corresponding to the endpoint. */
4754 struct sctp_sock *sp = sctp_sk(sk);
4756 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4757 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4758 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4761 if (put_user(len, optlen))
4762 return -EFAULT;
4764 if (copy_to_user(optval, &rtoinfo, len))
4765 return -EFAULT;
4767 return 0;
4772 * 7.1.2 SCTP_ASSOCINFO
4774 * This option is used to tune the maximum retransmission attempts
4775 * of the association.
4776 * Returns an error if the new association retransmission value is
4777 * greater than the sum of the retransmission value of the peer.
4778 * See [SCTP] for more information.
4781 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4782 char __user *optval,
4783 int __user *optlen)
4786 struct sctp_assocparams assocparams;
4787 struct sctp_association *asoc;
4788 struct list_head *pos;
4789 int cnt = 0;
4791 if (len < sizeof (struct sctp_assocparams))
4792 return -EINVAL;
4794 len = sizeof(struct sctp_assocparams);
4796 if (copy_from_user(&assocparams, optval, len))
4797 return -EFAULT;
4799 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4801 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4802 return -EINVAL;
4804 /* Values correspoinding to the specific association */
4805 if (asoc) {
4806 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4807 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4808 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4809 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4810 * 1000) +
4811 (asoc->cookie_life.tv_usec
4812 / 1000);
4814 list_for_each(pos, &asoc->peer.transport_addr_list) {
4815 cnt ++;
4818 assocparams.sasoc_number_peer_destinations = cnt;
4819 } else {
4820 /* Values corresponding to the endpoint */
4821 struct sctp_sock *sp = sctp_sk(sk);
4823 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4824 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4825 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4826 assocparams.sasoc_cookie_life =
4827 sp->assocparams.sasoc_cookie_life;
4828 assocparams.sasoc_number_peer_destinations =
4829 sp->assocparams.
4830 sasoc_number_peer_destinations;
4833 if (put_user(len, optlen))
4834 return -EFAULT;
4836 if (copy_to_user(optval, &assocparams, len))
4837 return -EFAULT;
4839 return 0;
4843 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4845 * This socket option is a boolean flag which turns on or off mapped V4
4846 * addresses. If this option is turned on and the socket is type
4847 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4848 * If this option is turned off, then no mapping will be done of V4
4849 * addresses and a user will receive both PF_INET6 and PF_INET type
4850 * addresses on the socket.
4852 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4853 char __user *optval, int __user *optlen)
4855 int val;
4856 struct sctp_sock *sp = sctp_sk(sk);
4858 if (len < sizeof(int))
4859 return -EINVAL;
4861 len = sizeof(int);
4862 val = sp->v4mapped;
4863 if (put_user(len, optlen))
4864 return -EFAULT;
4865 if (copy_to_user(optval, &val, len))
4866 return -EFAULT;
4868 return 0;
4872 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4873 * (chapter and verse is quoted at sctp_setsockopt_context())
4875 static int sctp_getsockopt_context(struct sock *sk, int len,
4876 char __user *optval, int __user *optlen)
4878 struct sctp_assoc_value params;
4879 struct sctp_sock *sp;
4880 struct sctp_association *asoc;
4882 if (len < sizeof(struct sctp_assoc_value))
4883 return -EINVAL;
4885 len = sizeof(struct sctp_assoc_value);
4887 if (copy_from_user(&params, optval, len))
4888 return -EFAULT;
4890 sp = sctp_sk(sk);
4892 if (params.assoc_id != 0) {
4893 asoc = sctp_id2assoc(sk, params.assoc_id);
4894 if (!asoc)
4895 return -EINVAL;
4896 params.assoc_value = asoc->default_rcv_context;
4897 } else {
4898 params.assoc_value = sp->default_rcv_context;
4901 if (put_user(len, optlen))
4902 return -EFAULT;
4903 if (copy_to_user(optval, &params, len))
4904 return -EFAULT;
4906 return 0;
4910 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4911 * This option will get or set the maximum size to put in any outgoing
4912 * SCTP DATA chunk. If a message is larger than this size it will be
4913 * fragmented by SCTP into the specified size. Note that the underlying
4914 * SCTP implementation may fragment into smaller sized chunks when the
4915 * PMTU of the underlying association is smaller than the value set by
4916 * the user. The default value for this option is '0' which indicates
4917 * the user is NOT limiting fragmentation and only the PMTU will effect
4918 * SCTP's choice of DATA chunk size. Note also that values set larger
4919 * than the maximum size of an IP datagram will effectively let SCTP
4920 * control fragmentation (i.e. the same as setting this option to 0).
4922 * The following structure is used to access and modify this parameter:
4924 * struct sctp_assoc_value {
4925 * sctp_assoc_t assoc_id;
4926 * uint32_t assoc_value;
4927 * };
4929 * assoc_id: This parameter is ignored for one-to-one style sockets.
4930 * For one-to-many style sockets this parameter indicates which
4931 * association the user is performing an action upon. Note that if
4932 * this field's value is zero then the endpoints default value is
4933 * changed (effecting future associations only).
4934 * assoc_value: This parameter specifies the maximum size in bytes.
4936 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4937 char __user *optval, int __user *optlen)
4939 struct sctp_assoc_value params;
4940 struct sctp_association *asoc;
4942 if (len == sizeof(int)) {
4943 printk(KERN_WARNING
4944 "SCTP: Use of int in maxseg socket option deprecated\n");
4945 printk(KERN_WARNING
4946 "SCTP: Use struct sctp_assoc_value instead\n");
4947 params.assoc_id = 0;
4948 } else if (len >= sizeof(struct sctp_assoc_value)) {
4949 len = sizeof(struct sctp_assoc_value);
4950 if (copy_from_user(&params, optval, sizeof(params)))
4951 return -EFAULT;
4952 } else
4953 return -EINVAL;
4955 asoc = sctp_id2assoc(sk, params.assoc_id);
4956 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4957 return -EINVAL;
4959 if (asoc)
4960 params.assoc_value = asoc->frag_point;
4961 else
4962 params.assoc_value = sctp_sk(sk)->user_frag;
4964 if (put_user(len, optlen))
4965 return -EFAULT;
4966 if (len == sizeof(int)) {
4967 if (copy_to_user(optval, &params.assoc_value, len))
4968 return -EFAULT;
4969 } else {
4970 if (copy_to_user(optval, &params, len))
4971 return -EFAULT;
4974 return 0;
4978 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4979 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4981 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4982 char __user *optval, int __user *optlen)
4984 int val;
4986 if (len < sizeof(int))
4987 return -EINVAL;
4989 len = sizeof(int);
4991 val = sctp_sk(sk)->frag_interleave;
4992 if (put_user(len, optlen))
4993 return -EFAULT;
4994 if (copy_to_user(optval, &val, len))
4995 return -EFAULT;
4997 return 0;
5001 * 7.1.25. Set or Get the sctp partial delivery point
5002 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5004 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5005 char __user *optval,
5006 int __user *optlen)
5008 u32 val;
5010 if (len < sizeof(u32))
5011 return -EINVAL;
5013 len = sizeof(u32);
5015 val = sctp_sk(sk)->pd_point;
5016 if (put_user(len, optlen))
5017 return -EFAULT;
5018 if (copy_to_user(optval, &val, len))
5019 return -EFAULT;
5021 return -ENOTSUPP;
5025 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5026 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5028 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5029 char __user *optval,
5030 int __user *optlen)
5032 struct sctp_assoc_value params;
5033 struct sctp_sock *sp;
5034 struct sctp_association *asoc;
5036 if (len == sizeof(int)) {
5037 printk(KERN_WARNING
5038 "SCTP: Use of int in max_burst socket option deprecated\n");
5039 printk(KERN_WARNING
5040 "SCTP: Use struct sctp_assoc_value instead\n");
5041 params.assoc_id = 0;
5042 } else if (len >= sizeof(struct sctp_assoc_value)) {
5043 len = sizeof(struct sctp_assoc_value);
5044 if (copy_from_user(&params, optval, len))
5045 return -EFAULT;
5046 } else
5047 return -EINVAL;
5049 sp = sctp_sk(sk);
5051 if (params.assoc_id != 0) {
5052 asoc = sctp_id2assoc(sk, params.assoc_id);
5053 if (!asoc)
5054 return -EINVAL;
5055 params.assoc_value = asoc->max_burst;
5056 } else
5057 params.assoc_value = sp->max_burst;
5059 if (len == sizeof(int)) {
5060 if (copy_to_user(optval, &params.assoc_value, len))
5061 return -EFAULT;
5062 } else {
5063 if (copy_to_user(optval, &params, len))
5064 return -EFAULT;
5067 return 0;
5071 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5072 char __user *optval, int __user *optlen)
5074 struct sctp_hmacalgo __user *p = (void __user *)optval;
5075 struct sctp_hmac_algo_param *hmacs;
5076 __u16 data_len = 0;
5077 u32 num_idents;
5079 if (!sctp_auth_enable)
5080 return -EACCES;
5082 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5083 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5085 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5086 return -EINVAL;
5088 len = sizeof(struct sctp_hmacalgo) + data_len;
5089 num_idents = data_len / sizeof(u16);
5091 if (put_user(len, optlen))
5092 return -EFAULT;
5093 if (put_user(num_idents, &p->shmac_num_idents))
5094 return -EFAULT;
5095 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5096 return -EFAULT;
5097 return 0;
5100 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5101 char __user *optval, int __user *optlen)
5103 struct sctp_authkeyid val;
5104 struct sctp_association *asoc;
5106 if (!sctp_auth_enable)
5107 return -EACCES;
5109 if (len < sizeof(struct sctp_authkeyid))
5110 return -EINVAL;
5111 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5112 return -EFAULT;
5114 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5115 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5116 return -EINVAL;
5118 if (asoc)
5119 val.scact_keynumber = asoc->active_key_id;
5120 else
5121 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5123 len = sizeof(struct sctp_authkeyid);
5124 if (put_user(len, optlen))
5125 return -EFAULT;
5126 if (copy_to_user(optval, &val, len))
5127 return -EFAULT;
5129 return 0;
5132 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5133 char __user *optval, int __user *optlen)
5135 struct sctp_authchunks __user *p = (void __user *)optval;
5136 struct sctp_authchunks val;
5137 struct sctp_association *asoc;
5138 struct sctp_chunks_param *ch;
5139 u32 num_chunks = 0;
5140 char __user *to;
5142 if (!sctp_auth_enable)
5143 return -EACCES;
5145 if (len < sizeof(struct sctp_authchunks))
5146 return -EINVAL;
5148 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5149 return -EFAULT;
5151 to = p->gauth_chunks;
5152 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5153 if (!asoc)
5154 return -EINVAL;
5156 ch = asoc->peer.peer_chunks;
5157 if (!ch)
5158 goto num;
5160 /* See if the user provided enough room for all the data */
5161 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5162 if (len < num_chunks)
5163 return -EINVAL;
5165 if (copy_to_user(to, ch->chunks, num_chunks))
5166 return -EFAULT;
5167 num:
5168 len = sizeof(struct sctp_authchunks) + num_chunks;
5169 if (put_user(len, optlen)) return -EFAULT;
5170 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5171 return -EFAULT;
5172 return 0;
5175 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5176 char __user *optval, int __user *optlen)
5178 struct sctp_authchunks __user *p = (void __user *)optval;
5179 struct sctp_authchunks val;
5180 struct sctp_association *asoc;
5181 struct sctp_chunks_param *ch;
5182 u32 num_chunks = 0;
5183 char __user *to;
5185 if (!sctp_auth_enable)
5186 return -EACCES;
5188 if (len < sizeof(struct sctp_authchunks))
5189 return -EINVAL;
5191 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5192 return -EFAULT;
5194 to = p->gauth_chunks;
5195 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5196 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5197 return -EINVAL;
5199 if (asoc)
5200 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5201 else
5202 ch = sctp_sk(sk)->ep->auth_chunk_list;
5204 if (!ch)
5205 goto num;
5207 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5208 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5209 return -EINVAL;
5211 if (copy_to_user(to, ch->chunks, num_chunks))
5212 return -EFAULT;
5213 num:
5214 len = sizeof(struct sctp_authchunks) + num_chunks;
5215 if (put_user(len, optlen))
5216 return -EFAULT;
5217 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5218 return -EFAULT;
5220 return 0;
5224 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5225 * This option gets the current number of associations that are attached
5226 * to a one-to-many style socket. The option value is an uint32_t.
5228 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5229 char __user *optval, int __user *optlen)
5231 struct sctp_sock *sp = sctp_sk(sk);
5232 struct sctp_association *asoc;
5233 u32 val = 0;
5235 if (sctp_style(sk, TCP))
5236 return -EOPNOTSUPP;
5238 if (len < sizeof(u32))
5239 return -EINVAL;
5241 len = sizeof(u32);
5243 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5244 val++;
5247 if (put_user(len, optlen))
5248 return -EFAULT;
5249 if (copy_to_user(optval, &val, len))
5250 return -EFAULT;
5252 return 0;
5255 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5256 char __user *optval, int __user *optlen)
5258 int retval = 0;
5259 int len;
5261 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5262 sk, optname);
5264 /* I can hardly begin to describe how wrong this is. This is
5265 * so broken as to be worse than useless. The API draft
5266 * REALLY is NOT helpful here... I am not convinced that the
5267 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5268 * are at all well-founded.
5270 if (level != SOL_SCTP) {
5271 struct sctp_af *af = sctp_sk(sk)->pf->af;
5273 retval = af->getsockopt(sk, level, optname, optval, optlen);
5274 return retval;
5277 if (get_user(len, optlen))
5278 return -EFAULT;
5280 sctp_lock_sock(sk);
5282 switch (optname) {
5283 case SCTP_STATUS:
5284 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5285 break;
5286 case SCTP_DISABLE_FRAGMENTS:
5287 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5288 optlen);
5289 break;
5290 case SCTP_EVENTS:
5291 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5292 break;
5293 case SCTP_AUTOCLOSE:
5294 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5295 break;
5296 case SCTP_SOCKOPT_PEELOFF:
5297 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5298 break;
5299 case SCTP_PEER_ADDR_PARAMS:
5300 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5301 optlen);
5302 break;
5303 case SCTP_DELAYED_ACK:
5304 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5305 optlen);
5306 break;
5307 case SCTP_INITMSG:
5308 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5309 break;
5310 case SCTP_GET_PEER_ADDRS:
5311 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5312 optlen);
5313 break;
5314 case SCTP_GET_LOCAL_ADDRS:
5315 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5316 optlen);
5317 break;
5318 case SCTP_SOCKOPT_CONNECTX3:
5319 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5320 break;
5321 case SCTP_DEFAULT_SEND_PARAM:
5322 retval = sctp_getsockopt_default_send_param(sk, len,
5323 optval, optlen);
5324 break;
5325 case SCTP_PRIMARY_ADDR:
5326 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5327 break;
5328 case SCTP_NODELAY:
5329 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5330 break;
5331 case SCTP_RTOINFO:
5332 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5333 break;
5334 case SCTP_ASSOCINFO:
5335 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5336 break;
5337 case SCTP_I_WANT_MAPPED_V4_ADDR:
5338 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5339 break;
5340 case SCTP_MAXSEG:
5341 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5342 break;
5343 case SCTP_GET_PEER_ADDR_INFO:
5344 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5345 optlen);
5346 break;
5347 case SCTP_ADAPTATION_LAYER:
5348 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5349 optlen);
5350 break;
5351 case SCTP_CONTEXT:
5352 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5353 break;
5354 case SCTP_FRAGMENT_INTERLEAVE:
5355 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5356 optlen);
5357 break;
5358 case SCTP_PARTIAL_DELIVERY_POINT:
5359 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5360 optlen);
5361 break;
5362 case SCTP_MAX_BURST:
5363 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5364 break;
5365 case SCTP_AUTH_KEY:
5366 case SCTP_AUTH_CHUNK:
5367 case SCTP_AUTH_DELETE_KEY:
5368 retval = -EOPNOTSUPP;
5369 break;
5370 case SCTP_HMAC_IDENT:
5371 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5372 break;
5373 case SCTP_AUTH_ACTIVE_KEY:
5374 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5375 break;
5376 case SCTP_PEER_AUTH_CHUNKS:
5377 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5378 optlen);
5379 break;
5380 case SCTP_LOCAL_AUTH_CHUNKS:
5381 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5382 optlen);
5383 break;
5384 case SCTP_GET_ASSOC_NUMBER:
5385 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5386 break;
5387 default:
5388 retval = -ENOPROTOOPT;
5389 break;
5392 sctp_release_sock(sk);
5393 return retval;
5396 static void sctp_hash(struct sock *sk)
5398 /* STUB */
5401 static void sctp_unhash(struct sock *sk)
5403 /* STUB */
5406 /* Check if port is acceptable. Possibly find first available port.
5408 * The port hash table (contained in the 'global' SCTP protocol storage
5409 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5410 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5411 * list (the list number is the port number hashed out, so as you
5412 * would expect from a hash function, all the ports in a given list have
5413 * such a number that hashes out to the same list number; you were
5414 * expecting that, right?); so each list has a set of ports, with a
5415 * link to the socket (struct sock) that uses it, the port number and
5416 * a fastreuse flag (FIXME: NPI ipg).
5418 static struct sctp_bind_bucket *sctp_bucket_create(
5419 struct sctp_bind_hashbucket *head, unsigned short snum);
5421 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5423 struct sctp_bind_hashbucket *head; /* hash list */
5424 struct sctp_bind_bucket *pp; /* hash list port iterator */
5425 struct hlist_node *node;
5426 unsigned short snum;
5427 int ret;
5429 snum = ntohs(addr->v4.sin_port);
5431 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5432 sctp_local_bh_disable();
5434 if (snum == 0) {
5435 /* Search for an available port. */
5436 int low, high, remaining, index;
5437 unsigned int rover;
5439 inet_get_local_port_range(&low, &high);
5440 remaining = (high - low) + 1;
5441 rover = net_random() % remaining + low;
5443 do {
5444 rover++;
5445 if ((rover < low) || (rover > high))
5446 rover = low;
5447 if (inet_is_reserved_local_port(rover))
5448 continue;
5449 index = sctp_phashfn(rover);
5450 head = &sctp_port_hashtable[index];
5451 sctp_spin_lock(&head->lock);
5452 sctp_for_each_hentry(pp, node, &head->chain)
5453 if (pp->port == rover)
5454 goto next;
5455 break;
5456 next:
5457 sctp_spin_unlock(&head->lock);
5458 } while (--remaining > 0);
5460 /* Exhausted local port range during search? */
5461 ret = 1;
5462 if (remaining <= 0)
5463 goto fail;
5465 /* OK, here is the one we will use. HEAD (the port
5466 * hash table list entry) is non-NULL and we hold it's
5467 * mutex.
5469 snum = rover;
5470 } else {
5471 /* We are given an specific port number; we verify
5472 * that it is not being used. If it is used, we will
5473 * exahust the search in the hash list corresponding
5474 * to the port number (snum) - we detect that with the
5475 * port iterator, pp being NULL.
5477 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5478 sctp_spin_lock(&head->lock);
5479 sctp_for_each_hentry(pp, node, &head->chain) {
5480 if (pp->port == snum)
5481 goto pp_found;
5484 pp = NULL;
5485 goto pp_not_found;
5486 pp_found:
5487 if (!hlist_empty(&pp->owner)) {
5488 /* We had a port hash table hit - there is an
5489 * available port (pp != NULL) and it is being
5490 * used by other socket (pp->owner not empty); that other
5491 * socket is going to be sk2.
5493 int reuse = sk->sk_reuse;
5494 struct sock *sk2;
5496 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5497 if (pp->fastreuse && sk->sk_reuse &&
5498 sk->sk_state != SCTP_SS_LISTENING)
5499 goto success;
5501 /* Run through the list of sockets bound to the port
5502 * (pp->port) [via the pointers bind_next and
5503 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5504 * we get the endpoint they describe and run through
5505 * the endpoint's list of IP (v4 or v6) addresses,
5506 * comparing each of the addresses with the address of
5507 * the socket sk. If we find a match, then that means
5508 * that this port/socket (sk) combination are already
5509 * in an endpoint.
5511 sk_for_each_bound(sk2, node, &pp->owner) {
5512 struct sctp_endpoint *ep2;
5513 ep2 = sctp_sk(sk2)->ep;
5515 if (sk == sk2 ||
5516 (reuse && sk2->sk_reuse &&
5517 sk2->sk_state != SCTP_SS_LISTENING))
5518 continue;
5520 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5521 sctp_sk(sk2), sctp_sk(sk))) {
5522 ret = (long)sk2;
5523 goto fail_unlock;
5526 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5528 pp_not_found:
5529 /* If there was a hash table miss, create a new port. */
5530 ret = 1;
5531 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5532 goto fail_unlock;
5534 /* In either case (hit or miss), make sure fastreuse is 1 only
5535 * if sk->sk_reuse is too (that is, if the caller requested
5536 * SO_REUSEADDR on this socket -sk-).
5538 if (hlist_empty(&pp->owner)) {
5539 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5540 pp->fastreuse = 1;
5541 else
5542 pp->fastreuse = 0;
5543 } else if (pp->fastreuse &&
5544 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5545 pp->fastreuse = 0;
5547 /* We are set, so fill up all the data in the hash table
5548 * entry, tie the socket list information with the rest of the
5549 * sockets FIXME: Blurry, NPI (ipg).
5551 success:
5552 if (!sctp_sk(sk)->bind_hash) {
5553 inet_sk(sk)->inet_num = snum;
5554 sk_add_bind_node(sk, &pp->owner);
5555 sctp_sk(sk)->bind_hash = pp;
5557 ret = 0;
5559 fail_unlock:
5560 sctp_spin_unlock(&head->lock);
5562 fail:
5563 sctp_local_bh_enable();
5564 return ret;
5567 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5568 * port is requested.
5570 static int sctp_get_port(struct sock *sk, unsigned short snum)
5572 long ret;
5573 union sctp_addr addr;
5574 struct sctp_af *af = sctp_sk(sk)->pf->af;
5576 /* Set up a dummy address struct from the sk. */
5577 af->from_sk(&addr, sk);
5578 addr.v4.sin_port = htons(snum);
5580 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5581 ret = sctp_get_port_local(sk, &addr);
5583 return (ret ? 1 : 0);
5587 * Move a socket to LISTENING state.
5589 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5591 struct sctp_sock *sp = sctp_sk(sk);
5592 struct sctp_endpoint *ep = sp->ep;
5593 struct crypto_hash *tfm = NULL;
5595 /* Allocate HMAC for generating cookie. */
5596 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5597 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5598 if (IS_ERR(tfm)) {
5599 if (net_ratelimit()) {
5600 printk(KERN_INFO
5601 "SCTP: failed to load transform for %s: %ld\n",
5602 sctp_hmac_alg, PTR_ERR(tfm));
5604 return -ENOSYS;
5606 sctp_sk(sk)->hmac = tfm;
5610 * If a bind() or sctp_bindx() is not called prior to a listen()
5611 * call that allows new associations to be accepted, the system
5612 * picks an ephemeral port and will choose an address set equivalent
5613 * to binding with a wildcard address.
5615 * This is not currently spelled out in the SCTP sockets
5616 * extensions draft, but follows the practice as seen in TCP
5617 * sockets.
5620 sk->sk_state = SCTP_SS_LISTENING;
5621 if (!ep->base.bind_addr.port) {
5622 if (sctp_autobind(sk))
5623 return -EAGAIN;
5624 } else {
5625 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5626 sk->sk_state = SCTP_SS_CLOSED;
5627 return -EADDRINUSE;
5631 sk->sk_max_ack_backlog = backlog;
5632 sctp_hash_endpoint(ep);
5633 return 0;
5637 * 4.1.3 / 5.1.3 listen()
5639 * By default, new associations are not accepted for UDP style sockets.
5640 * An application uses listen() to mark a socket as being able to
5641 * accept new associations.
5643 * On TCP style sockets, applications use listen() to ready the SCTP
5644 * endpoint for accepting inbound associations.
5646 * On both types of endpoints a backlog of '0' disables listening.
5648 * Move a socket to LISTENING state.
5650 int sctp_inet_listen(struct socket *sock, int backlog)
5652 struct sock *sk = sock->sk;
5653 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5654 int err = -EINVAL;
5656 if (unlikely(backlog < 0))
5657 return err;
5659 sctp_lock_sock(sk);
5661 /* Peeled-off sockets are not allowed to listen(). */
5662 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5663 goto out;
5665 if (sock->state != SS_UNCONNECTED)
5666 goto out;
5668 /* If backlog is zero, disable listening. */
5669 if (!backlog) {
5670 if (sctp_sstate(sk, CLOSED))
5671 goto out;
5673 err = 0;
5674 sctp_unhash_endpoint(ep);
5675 sk->sk_state = SCTP_SS_CLOSED;
5676 if (sk->sk_reuse)
5677 sctp_sk(sk)->bind_hash->fastreuse = 1;
5678 goto out;
5681 /* If we are already listening, just update the backlog */
5682 if (sctp_sstate(sk, LISTENING))
5683 sk->sk_max_ack_backlog = backlog;
5684 else {
5685 err = sctp_listen_start(sk, backlog);
5686 if (err)
5687 goto out;
5690 err = 0;
5691 out:
5692 sctp_release_sock(sk);
5693 return err;
5697 * This function is done by modeling the current datagram_poll() and the
5698 * tcp_poll(). Note that, based on these implementations, we don't
5699 * lock the socket in this function, even though it seems that,
5700 * ideally, locking or some other mechanisms can be used to ensure
5701 * the integrity of the counters (sndbuf and wmem_alloc) used
5702 * in this place. We assume that we don't need locks either until proven
5703 * otherwise.
5705 * Another thing to note is that we include the Async I/O support
5706 * here, again, by modeling the current TCP/UDP code. We don't have
5707 * a good way to test with it yet.
5709 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5711 struct sock *sk = sock->sk;
5712 struct sctp_sock *sp = sctp_sk(sk);
5713 unsigned int mask;
5715 poll_wait(file, sk_sleep(sk), wait);
5717 /* A TCP-style listening socket becomes readable when the accept queue
5718 * is not empty.
5720 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5721 return (!list_empty(&sp->ep->asocs)) ?
5722 (POLLIN | POLLRDNORM) : 0;
5724 mask = 0;
5726 /* Is there any exceptional events? */
5727 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5728 mask |= POLLERR;
5729 if (sk->sk_shutdown & RCV_SHUTDOWN)
5730 mask |= POLLRDHUP;
5731 if (sk->sk_shutdown == SHUTDOWN_MASK)
5732 mask |= POLLHUP;
5734 /* Is it readable? Reconsider this code with TCP-style support. */
5735 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5736 (sk->sk_shutdown & RCV_SHUTDOWN))
5737 mask |= POLLIN | POLLRDNORM;
5739 /* The association is either gone or not ready. */
5740 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5741 return mask;
5743 /* Is it writable? */
5744 if (sctp_writeable(sk)) {
5745 mask |= POLLOUT | POLLWRNORM;
5746 } else {
5747 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5749 * Since the socket is not locked, the buffer
5750 * might be made available after the writeable check and
5751 * before the bit is set. This could cause a lost I/O
5752 * signal. tcp_poll() has a race breaker for this race
5753 * condition. Based on their implementation, we put
5754 * in the following code to cover it as well.
5756 if (sctp_writeable(sk))
5757 mask |= POLLOUT | POLLWRNORM;
5759 return mask;
5762 /********************************************************************
5763 * 2nd Level Abstractions
5764 ********************************************************************/
5766 static struct sctp_bind_bucket *sctp_bucket_create(
5767 struct sctp_bind_hashbucket *head, unsigned short snum)
5769 struct sctp_bind_bucket *pp;
5771 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5772 if (pp) {
5773 SCTP_DBG_OBJCNT_INC(bind_bucket);
5774 pp->port = snum;
5775 pp->fastreuse = 0;
5776 INIT_HLIST_HEAD(&pp->owner);
5777 hlist_add_head(&pp->node, &head->chain);
5779 return pp;
5782 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5783 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5785 if (pp && hlist_empty(&pp->owner)) {
5786 __hlist_del(&pp->node);
5787 kmem_cache_free(sctp_bucket_cachep, pp);
5788 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5792 /* Release this socket's reference to a local port. */
5793 static inline void __sctp_put_port(struct sock *sk)
5795 struct sctp_bind_hashbucket *head =
5796 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5797 struct sctp_bind_bucket *pp;
5799 sctp_spin_lock(&head->lock);
5800 pp = sctp_sk(sk)->bind_hash;
5801 __sk_del_bind_node(sk);
5802 sctp_sk(sk)->bind_hash = NULL;
5803 inet_sk(sk)->inet_num = 0;
5804 sctp_bucket_destroy(pp);
5805 sctp_spin_unlock(&head->lock);
5808 void sctp_put_port(struct sock *sk)
5810 sctp_local_bh_disable();
5811 __sctp_put_port(sk);
5812 sctp_local_bh_enable();
5816 * The system picks an ephemeral port and choose an address set equivalent
5817 * to binding with a wildcard address.
5818 * One of those addresses will be the primary address for the association.
5819 * This automatically enables the multihoming capability of SCTP.
5821 static int sctp_autobind(struct sock *sk)
5823 union sctp_addr autoaddr;
5824 struct sctp_af *af;
5825 __be16 port;
5827 /* Initialize a local sockaddr structure to INADDR_ANY. */
5828 af = sctp_sk(sk)->pf->af;
5830 port = htons(inet_sk(sk)->inet_num);
5831 af->inaddr_any(&autoaddr, port);
5833 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5836 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5838 * From RFC 2292
5839 * 4.2 The cmsghdr Structure *
5841 * When ancillary data is sent or received, any number of ancillary data
5842 * objects can be specified by the msg_control and msg_controllen members of
5843 * the msghdr structure, because each object is preceded by
5844 * a cmsghdr structure defining the object's length (the cmsg_len member).
5845 * Historically Berkeley-derived implementations have passed only one object
5846 * at a time, but this API allows multiple objects to be
5847 * passed in a single call to sendmsg() or recvmsg(). The following example
5848 * shows two ancillary data objects in a control buffer.
5850 * |<--------------------------- msg_controllen -------------------------->|
5851 * | |
5853 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5855 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5856 * | | |
5858 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5860 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5861 * | | | | |
5863 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5864 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5866 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5868 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5872 * msg_control
5873 * points here
5875 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5876 sctp_cmsgs_t *cmsgs)
5878 struct cmsghdr *cmsg;
5879 struct msghdr *my_msg = (struct msghdr *)msg;
5881 for (cmsg = CMSG_FIRSTHDR(msg);
5882 cmsg != NULL;
5883 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5884 if (!CMSG_OK(my_msg, cmsg))
5885 return -EINVAL;
5887 /* Should we parse this header or ignore? */
5888 if (cmsg->cmsg_level != IPPROTO_SCTP)
5889 continue;
5891 /* Strictly check lengths following example in SCM code. */
5892 switch (cmsg->cmsg_type) {
5893 case SCTP_INIT:
5894 /* SCTP Socket API Extension
5895 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5897 * This cmsghdr structure provides information for
5898 * initializing new SCTP associations with sendmsg().
5899 * The SCTP_INITMSG socket option uses this same data
5900 * structure. This structure is not used for
5901 * recvmsg().
5903 * cmsg_level cmsg_type cmsg_data[]
5904 * ------------ ------------ ----------------------
5905 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5907 if (cmsg->cmsg_len !=
5908 CMSG_LEN(sizeof(struct sctp_initmsg)))
5909 return -EINVAL;
5910 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5911 break;
5913 case SCTP_SNDRCV:
5914 /* SCTP Socket API Extension
5915 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5917 * This cmsghdr structure specifies SCTP options for
5918 * sendmsg() and describes SCTP header information
5919 * about a received message through recvmsg().
5921 * cmsg_level cmsg_type cmsg_data[]
5922 * ------------ ------------ ----------------------
5923 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5925 if (cmsg->cmsg_len !=
5926 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5927 return -EINVAL;
5929 cmsgs->info =
5930 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5932 /* Minimally, validate the sinfo_flags. */
5933 if (cmsgs->info->sinfo_flags &
5934 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5935 SCTP_ABORT | SCTP_EOF))
5936 return -EINVAL;
5937 break;
5939 default:
5940 return -EINVAL;
5943 return 0;
5947 * Wait for a packet..
5948 * Note: This function is the same function as in core/datagram.c
5949 * with a few modifications to make lksctp work.
5951 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5953 int error;
5954 DEFINE_WAIT(wait);
5956 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
5958 /* Socket errors? */
5959 error = sock_error(sk);
5960 if (error)
5961 goto out;
5963 if (!skb_queue_empty(&sk->sk_receive_queue))
5964 goto ready;
5966 /* Socket shut down? */
5967 if (sk->sk_shutdown & RCV_SHUTDOWN)
5968 goto out;
5970 /* Sequenced packets can come disconnected. If so we report the
5971 * problem.
5973 error = -ENOTCONN;
5975 /* Is there a good reason to think that we may receive some data? */
5976 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5977 goto out;
5979 /* Handle signals. */
5980 if (signal_pending(current))
5981 goto interrupted;
5983 /* Let another process have a go. Since we are going to sleep
5984 * anyway. Note: This may cause odd behaviors if the message
5985 * does not fit in the user's buffer, but this seems to be the
5986 * only way to honor MSG_DONTWAIT realistically.
5988 sctp_release_sock(sk);
5989 *timeo_p = schedule_timeout(*timeo_p);
5990 sctp_lock_sock(sk);
5992 ready:
5993 finish_wait(sk_sleep(sk), &wait);
5994 return 0;
5996 interrupted:
5997 error = sock_intr_errno(*timeo_p);
5999 out:
6000 finish_wait(sk_sleep(sk), &wait);
6001 *err = error;
6002 return error;
6005 /* Receive a datagram.
6006 * Note: This is pretty much the same routine as in core/datagram.c
6007 * with a few changes to make lksctp work.
6009 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6010 int noblock, int *err)
6012 int error;
6013 struct sk_buff *skb;
6014 long timeo;
6016 timeo = sock_rcvtimeo(sk, noblock);
6018 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6019 timeo, MAX_SCHEDULE_TIMEOUT);
6021 do {
6022 /* Again only user level code calls this function,
6023 * so nothing interrupt level
6024 * will suddenly eat the receive_queue.
6026 * Look at current nfs client by the way...
6027 * However, this function was corrent in any case. 8)
6029 if (flags & MSG_PEEK) {
6030 spin_lock_bh(&sk->sk_receive_queue.lock);
6031 skb = skb_peek(&sk->sk_receive_queue);
6032 if (skb)
6033 atomic_inc(&skb->users);
6034 spin_unlock_bh(&sk->sk_receive_queue.lock);
6035 } else {
6036 skb = skb_dequeue(&sk->sk_receive_queue);
6039 if (skb)
6040 return skb;
6042 /* Caller is allowed not to check sk->sk_err before calling. */
6043 error = sock_error(sk);
6044 if (error)
6045 goto no_packet;
6047 if (sk->sk_shutdown & RCV_SHUTDOWN)
6048 break;
6050 /* User doesn't want to wait. */
6051 error = -EAGAIN;
6052 if (!timeo)
6053 goto no_packet;
6054 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6056 return NULL;
6058 no_packet:
6059 *err = error;
6060 return NULL;
6063 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6064 static void __sctp_write_space(struct sctp_association *asoc)
6066 struct sock *sk = asoc->base.sk;
6067 struct socket *sock = sk->sk_socket;
6069 if ((sctp_wspace(asoc) > 0) && sock) {
6070 if (waitqueue_active(&asoc->wait))
6071 wake_up_interruptible(&asoc->wait);
6073 if (sctp_writeable(sk)) {
6074 if (sk_sleep(sk) && waitqueue_active(sk_sleep(sk)))
6075 wake_up_interruptible(sk_sleep(sk));
6077 /* Note that we try to include the Async I/O support
6078 * here by modeling from the current TCP/UDP code.
6079 * We have not tested with it yet.
6081 if (sock->wq->fasync_list &&
6082 !(sk->sk_shutdown & SEND_SHUTDOWN))
6083 sock_wake_async(sock,
6084 SOCK_WAKE_SPACE, POLL_OUT);
6089 /* Do accounting for the sndbuf space.
6090 * Decrement the used sndbuf space of the corresponding association by the
6091 * data size which was just transmitted(freed).
6093 static void sctp_wfree(struct sk_buff *skb)
6095 struct sctp_association *asoc;
6096 struct sctp_chunk *chunk;
6097 struct sock *sk;
6099 /* Get the saved chunk pointer. */
6100 chunk = *((struct sctp_chunk **)(skb->cb));
6101 asoc = chunk->asoc;
6102 sk = asoc->base.sk;
6103 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6104 sizeof(struct sk_buff) +
6105 sizeof(struct sctp_chunk);
6107 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6110 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6112 sk->sk_wmem_queued -= skb->truesize;
6113 sk_mem_uncharge(sk, skb->truesize);
6115 sock_wfree(skb);
6116 __sctp_write_space(asoc);
6118 sctp_association_put(asoc);
6121 /* Do accounting for the receive space on the socket.
6122 * Accounting for the association is done in ulpevent.c
6123 * We set this as a destructor for the cloned data skbs so that
6124 * accounting is done at the correct time.
6126 void sctp_sock_rfree(struct sk_buff *skb)
6128 struct sock *sk = skb->sk;
6129 struct sctp_ulpevent *event = sctp_skb2event(skb);
6131 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6134 * Mimic the behavior of sock_rfree
6136 sk_mem_uncharge(sk, event->rmem_len);
6140 /* Helper function to wait for space in the sndbuf. */
6141 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6142 size_t msg_len)
6144 struct sock *sk = asoc->base.sk;
6145 int err = 0;
6146 long current_timeo = *timeo_p;
6147 DEFINE_WAIT(wait);
6149 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6150 asoc, (long)(*timeo_p), msg_len);
6152 /* Increment the association's refcnt. */
6153 sctp_association_hold(asoc);
6155 /* Wait on the association specific sndbuf space. */
6156 for (;;) {
6157 prepare_to_wait_exclusive(&asoc->wait, &wait,
6158 TASK_INTERRUPTIBLE);
6159 if (!*timeo_p)
6160 goto do_nonblock;
6161 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6162 asoc->base.dead)
6163 goto do_error;
6164 if (signal_pending(current))
6165 goto do_interrupted;
6166 if (msg_len <= sctp_wspace(asoc))
6167 break;
6169 /* Let another process have a go. Since we are going
6170 * to sleep anyway.
6172 sctp_release_sock(sk);
6173 current_timeo = schedule_timeout(current_timeo);
6174 BUG_ON(sk != asoc->base.sk);
6175 sctp_lock_sock(sk);
6177 *timeo_p = current_timeo;
6180 out:
6181 finish_wait(&asoc->wait, &wait);
6183 /* Release the association's refcnt. */
6184 sctp_association_put(asoc);
6186 return err;
6188 do_error:
6189 err = -EPIPE;
6190 goto out;
6192 do_interrupted:
6193 err = sock_intr_errno(*timeo_p);
6194 goto out;
6196 do_nonblock:
6197 err = -EAGAIN;
6198 goto out;
6201 void sctp_data_ready(struct sock *sk, int len)
6203 struct socket_wq *wq;
6205 rcu_read_lock();
6206 wq = rcu_dereference(sk->sk_wq);
6207 if (wq_has_sleeper(wq))
6208 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6209 POLLRDNORM | POLLRDBAND);
6210 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6211 rcu_read_unlock();
6214 /* If socket sndbuf has changed, wake up all per association waiters. */
6215 void sctp_write_space(struct sock *sk)
6217 struct sctp_association *asoc;
6219 /* Wake up the tasks in each wait queue. */
6220 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6221 __sctp_write_space(asoc);
6225 /* Is there any sndbuf space available on the socket?
6227 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6228 * associations on the same socket. For a UDP-style socket with
6229 * multiple associations, it is possible for it to be "unwriteable"
6230 * prematurely. I assume that this is acceptable because
6231 * a premature "unwriteable" is better than an accidental "writeable" which
6232 * would cause an unwanted block under certain circumstances. For the 1-1
6233 * UDP-style sockets or TCP-style sockets, this code should work.
6234 * - Daisy
6236 static int sctp_writeable(struct sock *sk)
6238 int amt = 0;
6240 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6241 if (amt < 0)
6242 amt = 0;
6243 return amt;
6246 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6247 * returns immediately with EINPROGRESS.
6249 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6251 struct sock *sk = asoc->base.sk;
6252 int err = 0;
6253 long current_timeo = *timeo_p;
6254 DEFINE_WAIT(wait);
6256 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6257 (long)(*timeo_p));
6259 /* Increment the association's refcnt. */
6260 sctp_association_hold(asoc);
6262 for (;;) {
6263 prepare_to_wait_exclusive(&asoc->wait, &wait,
6264 TASK_INTERRUPTIBLE);
6265 if (!*timeo_p)
6266 goto do_nonblock;
6267 if (sk->sk_shutdown & RCV_SHUTDOWN)
6268 break;
6269 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6270 asoc->base.dead)
6271 goto do_error;
6272 if (signal_pending(current))
6273 goto do_interrupted;
6275 if (sctp_state(asoc, ESTABLISHED))
6276 break;
6278 /* Let another process have a go. Since we are going
6279 * to sleep anyway.
6281 sctp_release_sock(sk);
6282 current_timeo = schedule_timeout(current_timeo);
6283 sctp_lock_sock(sk);
6285 *timeo_p = current_timeo;
6288 out:
6289 finish_wait(&asoc->wait, &wait);
6291 /* Release the association's refcnt. */
6292 sctp_association_put(asoc);
6294 return err;
6296 do_error:
6297 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6298 err = -ETIMEDOUT;
6299 else
6300 err = -ECONNREFUSED;
6301 goto out;
6303 do_interrupted:
6304 err = sock_intr_errno(*timeo_p);
6305 goto out;
6307 do_nonblock:
6308 err = -EINPROGRESS;
6309 goto out;
6312 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6314 struct sctp_endpoint *ep;
6315 int err = 0;
6316 DEFINE_WAIT(wait);
6318 ep = sctp_sk(sk)->ep;
6321 for (;;) {
6322 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6323 TASK_INTERRUPTIBLE);
6325 if (list_empty(&ep->asocs)) {
6326 sctp_release_sock(sk);
6327 timeo = schedule_timeout(timeo);
6328 sctp_lock_sock(sk);
6331 err = -EINVAL;
6332 if (!sctp_sstate(sk, LISTENING))
6333 break;
6335 err = 0;
6336 if (!list_empty(&ep->asocs))
6337 break;
6339 err = sock_intr_errno(timeo);
6340 if (signal_pending(current))
6341 break;
6343 err = -EAGAIN;
6344 if (!timeo)
6345 break;
6348 finish_wait(sk_sleep(sk), &wait);
6350 return err;
6353 static void sctp_wait_for_close(struct sock *sk, long timeout)
6355 DEFINE_WAIT(wait);
6357 do {
6358 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6359 if (list_empty(&sctp_sk(sk)->ep->asocs))
6360 break;
6361 sctp_release_sock(sk);
6362 timeout = schedule_timeout(timeout);
6363 sctp_lock_sock(sk);
6364 } while (!signal_pending(current) && timeout);
6366 finish_wait(sk_sleep(sk), &wait);
6369 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6371 struct sk_buff *frag;
6373 if (!skb->data_len)
6374 goto done;
6376 /* Don't forget the fragments. */
6377 skb_walk_frags(skb, frag)
6378 sctp_skb_set_owner_r_frag(frag, sk);
6380 done:
6381 sctp_skb_set_owner_r(skb, sk);
6384 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6385 struct sctp_association *asoc)
6387 struct inet_sock *inet = inet_sk(sk);
6388 struct inet_sock *newinet;
6390 newsk->sk_type = sk->sk_type;
6391 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6392 newsk->sk_flags = sk->sk_flags;
6393 newsk->sk_no_check = sk->sk_no_check;
6394 newsk->sk_reuse = sk->sk_reuse;
6396 newsk->sk_shutdown = sk->sk_shutdown;
6397 newsk->sk_destruct = inet_sock_destruct;
6398 newsk->sk_family = sk->sk_family;
6399 newsk->sk_protocol = IPPROTO_SCTP;
6400 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6401 newsk->sk_sndbuf = sk->sk_sndbuf;
6402 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6403 newsk->sk_lingertime = sk->sk_lingertime;
6404 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6405 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6407 newinet = inet_sk(newsk);
6409 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6410 * getsockname() and getpeername()
6412 newinet->inet_sport = inet->inet_sport;
6413 newinet->inet_saddr = inet->inet_saddr;
6414 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6415 newinet->inet_dport = htons(asoc->peer.port);
6416 newinet->pmtudisc = inet->pmtudisc;
6417 newinet->inet_id = asoc->next_tsn ^ jiffies;
6419 newinet->uc_ttl = inet->uc_ttl;
6420 newinet->mc_loop = 1;
6421 newinet->mc_ttl = 1;
6422 newinet->mc_index = 0;
6423 newinet->mc_list = NULL;
6426 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6427 * and its messages to the newsk.
6429 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6430 struct sctp_association *assoc,
6431 sctp_socket_type_t type)
6433 struct sctp_sock *oldsp = sctp_sk(oldsk);
6434 struct sctp_sock *newsp = sctp_sk(newsk);
6435 struct sctp_bind_bucket *pp; /* hash list port iterator */
6436 struct sctp_endpoint *newep = newsp->ep;
6437 struct sk_buff *skb, *tmp;
6438 struct sctp_ulpevent *event;
6439 struct sctp_bind_hashbucket *head;
6441 /* Migrate socket buffer sizes and all the socket level options to the
6442 * new socket.
6444 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6445 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6446 /* Brute force copy old sctp opt. */
6447 inet_sk_copy_descendant(newsk, oldsk);
6449 /* Restore the ep value that was overwritten with the above structure
6450 * copy.
6452 newsp->ep = newep;
6453 newsp->hmac = NULL;
6455 /* Hook this new socket in to the bind_hash list. */
6456 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6457 sctp_local_bh_disable();
6458 sctp_spin_lock(&head->lock);
6459 pp = sctp_sk(oldsk)->bind_hash;
6460 sk_add_bind_node(newsk, &pp->owner);
6461 sctp_sk(newsk)->bind_hash = pp;
6462 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6463 sctp_spin_unlock(&head->lock);
6464 sctp_local_bh_enable();
6466 /* Copy the bind_addr list from the original endpoint to the new
6467 * endpoint so that we can handle restarts properly
6469 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6470 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6472 /* Move any messages in the old socket's receive queue that are for the
6473 * peeled off association to the new socket's receive queue.
6475 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6476 event = sctp_skb2event(skb);
6477 if (event->asoc == assoc) {
6478 __skb_unlink(skb, &oldsk->sk_receive_queue);
6479 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6480 sctp_skb_set_owner_r_frag(skb, newsk);
6484 /* Clean up any messages pending delivery due to partial
6485 * delivery. Three cases:
6486 * 1) No partial deliver; no work.
6487 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6488 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6490 skb_queue_head_init(&newsp->pd_lobby);
6491 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6493 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6494 struct sk_buff_head *queue;
6496 /* Decide which queue to move pd_lobby skbs to. */
6497 if (assoc->ulpq.pd_mode) {
6498 queue = &newsp->pd_lobby;
6499 } else
6500 queue = &newsk->sk_receive_queue;
6502 /* Walk through the pd_lobby, looking for skbs that
6503 * need moved to the new socket.
6505 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6506 event = sctp_skb2event(skb);
6507 if (event->asoc == assoc) {
6508 __skb_unlink(skb, &oldsp->pd_lobby);
6509 __skb_queue_tail(queue, skb);
6510 sctp_skb_set_owner_r_frag(skb, newsk);
6514 /* Clear up any skbs waiting for the partial
6515 * delivery to finish.
6517 if (assoc->ulpq.pd_mode)
6518 sctp_clear_pd(oldsk, NULL);
6522 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6523 sctp_skb_set_owner_r_frag(skb, newsk);
6525 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6526 sctp_skb_set_owner_r_frag(skb, newsk);
6528 /* Set the type of socket to indicate that it is peeled off from the
6529 * original UDP-style socket or created with the accept() call on a
6530 * TCP-style socket..
6532 newsp->type = type;
6534 /* Mark the new socket "in-use" by the user so that any packets
6535 * that may arrive on the association after we've moved it are
6536 * queued to the backlog. This prevents a potential race between
6537 * backlog processing on the old socket and new-packet processing
6538 * on the new socket.
6540 * The caller has just allocated newsk so we can guarantee that other
6541 * paths won't try to lock it and then oldsk.
6543 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6544 sctp_assoc_migrate(assoc, newsk);
6546 /* If the association on the newsk is already closed before accept()
6547 * is called, set RCV_SHUTDOWN flag.
6549 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6550 newsk->sk_shutdown |= RCV_SHUTDOWN;
6552 newsk->sk_state = SCTP_SS_ESTABLISHED;
6553 sctp_release_sock(newsk);
6557 /* This proto struct describes the ULP interface for SCTP. */
6558 struct proto sctp_prot = {
6559 .name = "SCTP",
6560 .owner = THIS_MODULE,
6561 .close = sctp_close,
6562 .connect = sctp_connect,
6563 .disconnect = sctp_disconnect,
6564 .accept = sctp_accept,
6565 .ioctl = sctp_ioctl,
6566 .init = sctp_init_sock,
6567 .destroy = sctp_destroy_sock,
6568 .shutdown = sctp_shutdown,
6569 .setsockopt = sctp_setsockopt,
6570 .getsockopt = sctp_getsockopt,
6571 .sendmsg = sctp_sendmsg,
6572 .recvmsg = sctp_recvmsg,
6573 .bind = sctp_bind,
6574 .backlog_rcv = sctp_backlog_rcv,
6575 .hash = sctp_hash,
6576 .unhash = sctp_unhash,
6577 .get_port = sctp_get_port,
6578 .obj_size = sizeof(struct sctp_sock),
6579 .sysctl_mem = sysctl_sctp_mem,
6580 .sysctl_rmem = sysctl_sctp_rmem,
6581 .sysctl_wmem = sysctl_sctp_wmem,
6582 .memory_pressure = &sctp_memory_pressure,
6583 .enter_memory_pressure = sctp_enter_memory_pressure,
6584 .memory_allocated = &sctp_memory_allocated,
6585 .sockets_allocated = &sctp_sockets_allocated,
6588 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6590 struct proto sctpv6_prot = {
6591 .name = "SCTPv6",
6592 .owner = THIS_MODULE,
6593 .close = sctp_close,
6594 .connect = sctp_connect,
6595 .disconnect = sctp_disconnect,
6596 .accept = sctp_accept,
6597 .ioctl = sctp_ioctl,
6598 .init = sctp_init_sock,
6599 .destroy = sctp_destroy_sock,
6600 .shutdown = sctp_shutdown,
6601 .setsockopt = sctp_setsockopt,
6602 .getsockopt = sctp_getsockopt,
6603 .sendmsg = sctp_sendmsg,
6604 .recvmsg = sctp_recvmsg,
6605 .bind = sctp_bind,
6606 .backlog_rcv = sctp_backlog_rcv,
6607 .hash = sctp_hash,
6608 .unhash = sctp_unhash,
6609 .get_port = sctp_get_port,
6610 .obj_size = sizeof(struct sctp6_sock),
6611 .sysctl_mem = sysctl_sctp_mem,
6612 .sysctl_rmem = sysctl_sctp_rmem,
6613 .sysctl_wmem = sysctl_sctp_wmem,
6614 .memory_pressure = &sctp_memory_pressure,
6615 .enter_memory_pressure = sctp_enter_memory_pressure,
6616 .memory_allocated = &sctp_memory_allocated,
6617 .sockets_allocated = &sctp_sockets_allocated,
6619 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */