[MIPS] Misc fixes for plat_irq_dispatch functions
[linux-2.6/mini2440.git] / net / sctp / socket.c
blob536298c2eda26034d3f0afab142ad95cbcee2e7d
1 /* SCTP kernel reference 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 reference 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 * The SCTP reference 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 * The SCTP reference 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>
71 #include <net/ip.h>
72 #include <net/icmp.h>
73 #include <net/route.h>
74 #include <net/ipv6.h>
75 #include <net/inet_common.h>
77 #include <linux/socket.h> /* for sa_family_t */
78 #include <net/sock.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* WARNING: Please do not remove the SCTP_STATIC attribute to
83 * any of the functions below as they are used to export functions
84 * used by a project regression testsuite.
87 /* Forward declarations for internal helper functions. */
88 static int sctp_writeable(struct sock *sk);
89 static void sctp_wfree(struct sk_buff *skb);
90 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
91 size_t msg_len);
92 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
93 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
94 static int sctp_wait_for_accept(struct sock *sk, long timeo);
95 static void sctp_wait_for_close(struct sock *sk, long timeo);
96 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
97 union sctp_addr *addr, int len);
98 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
99 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
102 static int sctp_send_asconf(struct sctp_association *asoc,
103 struct sctp_chunk *chunk);
104 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
105 static int sctp_autobind(struct sock *sk);
106 static void sctp_sock_migrate(struct sock *, struct sock *,
107 struct sctp_association *, sctp_socket_type_t);
108 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
110 extern struct kmem_cache *sctp_bucket_cachep;
112 /* Get the sndbuf space available at the time on the association. */
113 static inline int sctp_wspace(struct sctp_association *asoc)
115 struct sock *sk = asoc->base.sk;
116 int amt = 0;
118 if (asoc->ep->sndbuf_policy) {
119 /* make sure that no association uses more than sk_sndbuf */
120 amt = sk->sk_sndbuf - asoc->sndbuf_used;
121 } else {
122 /* do socket level accounting */
123 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
126 if (amt < 0)
127 amt = 0;
129 return amt;
132 /* Increment the used sndbuf space count of the corresponding association by
133 * the size of the outgoing data chunk.
134 * Also, set the skb destructor for sndbuf accounting later.
136 * Since it is always 1-1 between chunk and skb, and also a new skb is always
137 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
138 * destructor in the data chunk skb for the purpose of the sndbuf space
139 * tracking.
141 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
143 struct sctp_association *asoc = chunk->asoc;
144 struct sock *sk = asoc->base.sk;
146 /* The sndbuf space is tracked per association. */
147 sctp_association_hold(asoc);
149 skb_set_owner_w(chunk->skb, sk);
151 chunk->skb->destructor = sctp_wfree;
152 /* Save the chunk pointer in skb for sctp_wfree to use later. */
153 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
155 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
156 sizeof(struct sk_buff) +
157 sizeof(struct sctp_chunk);
159 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
162 /* Verify that this is a valid address. */
163 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
164 int len)
166 struct sctp_af *af;
168 /* Verify basic sockaddr. */
169 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
170 if (!af)
171 return -EINVAL;
173 /* Is this a valid SCTP address? */
174 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
175 return -EINVAL;
177 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
178 return -EINVAL;
180 return 0;
183 /* Look up the association by its id. If this is not a UDP-style
184 * socket, the ID field is always ignored.
186 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
188 struct sctp_association *asoc = NULL;
190 /* If this is not a UDP-style socket, assoc id should be ignored. */
191 if (!sctp_style(sk, UDP)) {
192 /* Return NULL if the socket state is not ESTABLISHED. It
193 * could be a TCP-style listening socket or a socket which
194 * hasn't yet called connect() to establish an association.
196 if (!sctp_sstate(sk, ESTABLISHED))
197 return NULL;
199 /* Get the first and the only association from the list. */
200 if (!list_empty(&sctp_sk(sk)->ep->asocs))
201 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
202 struct sctp_association, asocs);
203 return asoc;
206 /* Otherwise this is a UDP-style socket. */
207 if (!id || (id == (sctp_assoc_t)-1))
208 return NULL;
210 spin_lock_bh(&sctp_assocs_id_lock);
211 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
212 spin_unlock_bh(&sctp_assocs_id_lock);
214 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
215 return NULL;
217 return asoc;
220 /* Look up the transport from an address and an assoc id. If both address and
221 * id are specified, the associations matching the address and the id should be
222 * the same.
224 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
225 struct sockaddr_storage *addr,
226 sctp_assoc_t id)
228 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
229 struct sctp_transport *transport;
230 union sctp_addr *laddr = (union sctp_addr *)addr;
232 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
233 laddr,
234 &transport);
236 if (!addr_asoc)
237 return NULL;
239 id_asoc = sctp_id2assoc(sk, id);
240 if (id_asoc && (id_asoc != addr_asoc))
241 return NULL;
243 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
244 (union sctp_addr *)addr);
246 return transport;
249 /* API 3.1.2 bind() - UDP Style Syntax
250 * The syntax of bind() is,
252 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
254 * sd - the socket descriptor returned by socket().
255 * addr - the address structure (struct sockaddr_in or struct
256 * sockaddr_in6 [RFC 2553]),
257 * addr_len - the size of the address structure.
259 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
261 int retval = 0;
263 sctp_lock_sock(sk);
265 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
266 sk, addr, addr_len);
268 /* Disallow binding twice. */
269 if (!sctp_sk(sk)->ep->base.bind_addr.port)
270 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
271 addr_len);
272 else
273 retval = -EINVAL;
275 sctp_release_sock(sk);
277 return retval;
280 static long sctp_get_port_local(struct sock *, union sctp_addr *);
282 /* Verify this is a valid sockaddr. */
283 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
284 union sctp_addr *addr, int len)
286 struct sctp_af *af;
288 /* Check minimum size. */
289 if (len < sizeof (struct sockaddr))
290 return NULL;
292 /* Does this PF support this AF? */
293 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
294 return NULL;
296 /* If we get this far, af is valid. */
297 af = sctp_get_af_specific(addr->sa.sa_family);
299 if (len < af->sockaddr_len)
300 return NULL;
302 return af;
305 /* Bind a local address either to an endpoint or to an association. */
306 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
308 struct sctp_sock *sp = sctp_sk(sk);
309 struct sctp_endpoint *ep = sp->ep;
310 struct sctp_bind_addr *bp = &ep->base.bind_addr;
311 struct sctp_af *af;
312 unsigned short snum;
313 int ret = 0;
315 /* Common sockaddr verification. */
316 af = sctp_sockaddr_af(sp, addr, len);
317 if (!af) {
318 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
319 sk, addr, len);
320 return -EINVAL;
323 snum = ntohs(addr->v4.sin_port);
325 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
326 ", port: %d, new port: %d, len: %d)\n",
328 addr,
329 bp->port, snum,
330 len);
332 /* PF specific bind() address verification. */
333 if (!sp->pf->bind_verify(sp, addr))
334 return -EADDRNOTAVAIL;
336 /* We must either be unbound, or bind to the same port. */
337 if (bp->port && (snum != bp->port)) {
338 SCTP_DEBUG_PRINTK("sctp_do_bind:"
339 " New port %d does not match existing port "
340 "%d.\n", snum, bp->port);
341 return -EINVAL;
344 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
345 return -EACCES;
347 /* Make sure we are allowed to bind here.
348 * The function sctp_get_port_local() does duplicate address
349 * detection.
351 if ((ret = sctp_get_port_local(sk, addr))) {
352 if (ret == (long) sk) {
353 /* This endpoint has a conflicting address. */
354 return -EINVAL;
355 } else {
356 return -EADDRINUSE;
360 /* Refresh ephemeral port. */
361 if (!bp->port)
362 bp->port = inet_sk(sk)->num;
364 /* Add the address to the bind address list. */
365 sctp_local_bh_disable();
366 sctp_write_lock(&ep->base.addr_lock);
368 /* Use GFP_ATOMIC since BHs are disabled. */
369 ret = sctp_add_bind_addr(bp, addr, 1, GFP_ATOMIC);
370 sctp_write_unlock(&ep->base.addr_lock);
371 sctp_local_bh_enable();
373 /* Copy back into socket for getsockname() use. */
374 if (!ret) {
375 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
376 af->to_sk_saddr(addr, sk);
379 return ret;
382 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
384 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
385 * at any one time. If a sender, after sending an ASCONF chunk, decides
386 * it needs to transfer another ASCONF Chunk, it MUST wait until the
387 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
388 * subsequent ASCONF. Note this restriction binds each side, so at any
389 * time two ASCONF may be in-transit on any given association (one sent
390 * from each endpoint).
392 static int sctp_send_asconf(struct sctp_association *asoc,
393 struct sctp_chunk *chunk)
395 int retval = 0;
397 /* If there is an outstanding ASCONF chunk, queue it for later
398 * transmission.
400 if (asoc->addip_last_asconf) {
401 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
402 goto out;
405 /* Hold the chunk until an ASCONF_ACK is received. */
406 sctp_chunk_hold(chunk);
407 retval = sctp_primitive_ASCONF(asoc, chunk);
408 if (retval)
409 sctp_chunk_free(chunk);
410 else
411 asoc->addip_last_asconf = chunk;
413 out:
414 return retval;
417 /* Add a list of addresses as bind addresses to local endpoint or
418 * association.
420 * Basically run through each address specified in the addrs/addrcnt
421 * array/length pair, determine if it is IPv6 or IPv4 and call
422 * sctp_do_bind() on it.
424 * If any of them fails, then the operation will be reversed and the
425 * ones that were added will be removed.
427 * Only sctp_setsockopt_bindx() is supposed to call this function.
429 int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
431 int cnt;
432 int retval = 0;
433 void *addr_buf;
434 struct sockaddr *sa_addr;
435 struct sctp_af *af;
437 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
438 sk, addrs, addrcnt);
440 addr_buf = addrs;
441 for (cnt = 0; cnt < addrcnt; cnt++) {
442 /* The list may contain either IPv4 or IPv6 address;
443 * determine the address length for walking thru the list.
445 sa_addr = (struct sockaddr *)addr_buf;
446 af = sctp_get_af_specific(sa_addr->sa_family);
447 if (!af) {
448 retval = -EINVAL;
449 goto err_bindx_add;
452 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
453 af->sockaddr_len);
455 addr_buf += af->sockaddr_len;
457 err_bindx_add:
458 if (retval < 0) {
459 /* Failed. Cleanup the ones that have been added */
460 if (cnt > 0)
461 sctp_bindx_rem(sk, addrs, cnt);
462 return retval;
466 return retval;
469 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
470 * associations that are part of the endpoint indicating that a list of local
471 * addresses are added to the endpoint.
473 * If any of the addresses is already in the bind address list of the
474 * association, we do not send the chunk for that association. But it will not
475 * affect other associations.
477 * Only sctp_setsockopt_bindx() is supposed to call this function.
479 static int sctp_send_asconf_add_ip(struct sock *sk,
480 struct sockaddr *addrs,
481 int addrcnt)
483 struct sctp_sock *sp;
484 struct sctp_endpoint *ep;
485 struct sctp_association *asoc;
486 struct sctp_bind_addr *bp;
487 struct sctp_chunk *chunk;
488 struct sctp_sockaddr_entry *laddr;
489 union sctp_addr *addr;
490 union sctp_addr saveaddr;
491 void *addr_buf;
492 struct sctp_af *af;
493 struct list_head *pos;
494 struct list_head *p;
495 int i;
496 int retval = 0;
498 if (!sctp_addip_enable)
499 return retval;
501 sp = sctp_sk(sk);
502 ep = sp->ep;
504 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
505 __FUNCTION__, sk, addrs, addrcnt);
507 list_for_each(pos, &ep->asocs) {
508 asoc = list_entry(pos, struct sctp_association, asocs);
510 if (!asoc->peer.asconf_capable)
511 continue;
513 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
514 continue;
516 if (!sctp_state(asoc, ESTABLISHED))
517 continue;
519 /* Check if any address in the packed array of addresses is
520 * in the bind address list of the association. If so,
521 * do not send the asconf chunk to its peer, but continue with
522 * other associations.
524 addr_buf = addrs;
525 for (i = 0; i < addrcnt; i++) {
526 addr = (union sctp_addr *)addr_buf;
527 af = sctp_get_af_specific(addr->v4.sin_family);
528 if (!af) {
529 retval = -EINVAL;
530 goto out;
533 if (sctp_assoc_lookup_laddr(asoc, addr))
534 break;
536 addr_buf += af->sockaddr_len;
538 if (i < addrcnt)
539 continue;
541 /* Use the first address in bind addr list of association as
542 * Address Parameter of ASCONF CHUNK.
544 sctp_read_lock(&asoc->base.addr_lock);
545 bp = &asoc->base.bind_addr;
546 p = bp->address_list.next;
547 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
548 sctp_read_unlock(&asoc->base.addr_lock);
550 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
551 addrcnt, SCTP_PARAM_ADD_IP);
552 if (!chunk) {
553 retval = -ENOMEM;
554 goto out;
557 retval = sctp_send_asconf(asoc, chunk);
558 if (retval)
559 goto out;
561 /* Add the new addresses to the bind address list with
562 * use_as_src set to 0.
564 sctp_local_bh_disable();
565 sctp_write_lock(&asoc->base.addr_lock);
566 addr_buf = addrs;
567 for (i = 0; i < addrcnt; i++) {
568 addr = (union sctp_addr *)addr_buf;
569 af = sctp_get_af_specific(addr->v4.sin_family);
570 memcpy(&saveaddr, addr, af->sockaddr_len);
571 retval = sctp_add_bind_addr(bp, &saveaddr, 0,
572 GFP_ATOMIC);
573 addr_buf += af->sockaddr_len;
575 sctp_write_unlock(&asoc->base.addr_lock);
576 sctp_local_bh_enable();
579 out:
580 return retval;
583 /* Remove a list of addresses from bind addresses list. Do not remove the
584 * last address.
586 * Basically run through each address specified in the addrs/addrcnt
587 * array/length pair, determine if it is IPv6 or IPv4 and call
588 * sctp_del_bind() on it.
590 * If any of them fails, then the operation will be reversed and the
591 * ones that were removed will be added back.
593 * At least one address has to be left; if only one address is
594 * available, the operation will return -EBUSY.
596 * Only sctp_setsockopt_bindx() is supposed to call this function.
598 int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
600 struct sctp_sock *sp = sctp_sk(sk);
601 struct sctp_endpoint *ep = sp->ep;
602 int cnt;
603 struct sctp_bind_addr *bp = &ep->base.bind_addr;
604 int retval = 0;
605 void *addr_buf;
606 union sctp_addr *sa_addr;
607 struct sctp_af *af;
609 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
610 sk, addrs, addrcnt);
612 addr_buf = addrs;
613 for (cnt = 0; cnt < addrcnt; cnt++) {
614 /* If the bind address list is empty or if there is only one
615 * bind address, there is nothing more to be removed (we need
616 * at least one address here).
618 if (list_empty(&bp->address_list) ||
619 (sctp_list_single_entry(&bp->address_list))) {
620 retval = -EBUSY;
621 goto err_bindx_rem;
624 sa_addr = (union sctp_addr *)addr_buf;
625 af = sctp_get_af_specific(sa_addr->sa.sa_family);
626 if (!af) {
627 retval = -EINVAL;
628 goto err_bindx_rem;
630 if (sa_addr->v4.sin_port != htons(bp->port)) {
631 retval = -EINVAL;
632 goto err_bindx_rem;
635 /* FIXME - There is probably a need to check if sk->sk_saddr and
636 * sk->sk_rcv_addr are currently set to one of the addresses to
637 * be removed. This is something which needs to be looked into
638 * when we are fixing the outstanding issues with multi-homing
639 * socket routing and failover schemes. Refer to comments in
640 * sctp_do_bind(). -daisy
642 sctp_local_bh_disable();
643 sctp_write_lock(&ep->base.addr_lock);
645 retval = sctp_del_bind_addr(bp, sa_addr);
647 sctp_write_unlock(&ep->base.addr_lock);
648 sctp_local_bh_enable();
650 addr_buf += af->sockaddr_len;
651 err_bindx_rem:
652 if (retval < 0) {
653 /* Failed. Add the ones that has been removed back */
654 if (cnt > 0)
655 sctp_bindx_add(sk, addrs, cnt);
656 return retval;
660 return retval;
663 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
664 * the associations that are part of the endpoint indicating that a list of
665 * local addresses are removed from the endpoint.
667 * If any of the addresses is already in the bind address list of the
668 * association, we do not send the chunk for that association. But it will not
669 * affect other associations.
671 * Only sctp_setsockopt_bindx() is supposed to call this function.
673 static int sctp_send_asconf_del_ip(struct sock *sk,
674 struct sockaddr *addrs,
675 int addrcnt)
677 struct sctp_sock *sp;
678 struct sctp_endpoint *ep;
679 struct sctp_association *asoc;
680 struct sctp_transport *transport;
681 struct sctp_bind_addr *bp;
682 struct sctp_chunk *chunk;
683 union sctp_addr *laddr;
684 void *addr_buf;
685 struct sctp_af *af;
686 struct list_head *pos, *pos1;
687 struct sctp_sockaddr_entry *saddr;
688 int i;
689 int retval = 0;
691 if (!sctp_addip_enable)
692 return retval;
694 sp = sctp_sk(sk);
695 ep = sp->ep;
697 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
698 __FUNCTION__, sk, addrs, addrcnt);
700 list_for_each(pos, &ep->asocs) {
701 asoc = list_entry(pos, struct sctp_association, asocs);
703 if (!asoc->peer.asconf_capable)
704 continue;
706 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
707 continue;
709 if (!sctp_state(asoc, ESTABLISHED))
710 continue;
712 /* Check if any address in the packed array of addresses is
713 * not present in the bind address list of the association.
714 * If so, do not send the asconf chunk to its peer, but
715 * continue with other associations.
717 addr_buf = addrs;
718 for (i = 0; i < addrcnt; i++) {
719 laddr = (union sctp_addr *)addr_buf;
720 af = sctp_get_af_specific(laddr->v4.sin_family);
721 if (!af) {
722 retval = -EINVAL;
723 goto out;
726 if (!sctp_assoc_lookup_laddr(asoc, laddr))
727 break;
729 addr_buf += af->sockaddr_len;
731 if (i < addrcnt)
732 continue;
734 /* Find one address in the association's bind address list
735 * that is not in the packed array of addresses. This is to
736 * make sure that we do not delete all the addresses in the
737 * association.
739 sctp_read_lock(&asoc->base.addr_lock);
740 bp = &asoc->base.bind_addr;
741 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
742 addrcnt, sp);
743 sctp_read_unlock(&asoc->base.addr_lock);
744 if (!laddr)
745 continue;
747 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
748 SCTP_PARAM_DEL_IP);
749 if (!chunk) {
750 retval = -ENOMEM;
751 goto out;
754 /* Reset use_as_src flag for the addresses in the bind address
755 * list that are to be deleted.
757 sctp_local_bh_disable();
758 sctp_write_lock(&asoc->base.addr_lock);
759 addr_buf = addrs;
760 for (i = 0; i < addrcnt; i++) {
761 laddr = (union sctp_addr *)addr_buf;
762 af = sctp_get_af_specific(laddr->v4.sin_family);
763 list_for_each(pos1, &bp->address_list) {
764 saddr = list_entry(pos1,
765 struct sctp_sockaddr_entry,
766 list);
767 if (sctp_cmp_addr_exact(&saddr->a, laddr))
768 saddr->use_as_src = 0;
770 addr_buf += af->sockaddr_len;
772 sctp_write_unlock(&asoc->base.addr_lock);
773 sctp_local_bh_enable();
775 /* Update the route and saddr entries for all the transports
776 * as some of the addresses in the bind address list are
777 * about to be deleted and cannot be used as source addresses.
779 list_for_each(pos1, &asoc->peer.transport_addr_list) {
780 transport = list_entry(pos1, struct sctp_transport,
781 transports);
782 dst_release(transport->dst);
783 sctp_transport_route(transport, NULL,
784 sctp_sk(asoc->base.sk));
787 retval = sctp_send_asconf(asoc, chunk);
789 out:
790 return retval;
793 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
795 * API 8.1
796 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
797 * int flags);
799 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
800 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
801 * or IPv6 addresses.
803 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
804 * Section 3.1.2 for this usage.
806 * addrs is a pointer to an array of one or more socket addresses. Each
807 * address is contained in its appropriate structure (i.e. struct
808 * sockaddr_in or struct sockaddr_in6) the family of the address type
809 * must be used to distinguish the address length (note that this
810 * representation is termed a "packed array" of addresses). The caller
811 * specifies the number of addresses in the array with addrcnt.
813 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
814 * -1, and sets errno to the appropriate error code.
816 * For SCTP, the port given in each socket address must be the same, or
817 * sctp_bindx() will fail, setting errno to EINVAL.
819 * The flags parameter is formed from the bitwise OR of zero or more of
820 * the following currently defined flags:
822 * SCTP_BINDX_ADD_ADDR
824 * SCTP_BINDX_REM_ADDR
826 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
827 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
828 * addresses from the association. The two flags are mutually exclusive;
829 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
830 * not remove all addresses from an association; sctp_bindx() will
831 * reject such an attempt with EINVAL.
833 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
834 * additional addresses with an endpoint after calling bind(). Or use
835 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
836 * socket is associated with so that no new association accepted will be
837 * associated with those addresses. If the endpoint supports dynamic
838 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
839 * endpoint to send the appropriate message to the peer to change the
840 * peers address lists.
842 * Adding and removing addresses from a connected association is
843 * optional functionality. Implementations that do not support this
844 * functionality should return EOPNOTSUPP.
846 * Basically do nothing but copying the addresses from user to kernel
847 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
848 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
849 * from userspace.
851 * We don't use copy_from_user() for optimization: we first do the
852 * sanity checks (buffer size -fast- and access check-healthy
853 * pointer); if all of those succeed, then we can alloc the memory
854 * (expensive operation) needed to copy the data to kernel. Then we do
855 * the copying without checking the user space area
856 * (__copy_from_user()).
858 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
859 * it.
861 * sk The sk of the socket
862 * addrs The pointer to the addresses in user land
863 * addrssize Size of the addrs buffer
864 * op Operation to perform (add or remove, see the flags of
865 * sctp_bindx)
867 * Returns 0 if ok, <0 errno code on error.
869 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
870 struct sockaddr __user *addrs,
871 int addrs_size, int op)
873 struct sockaddr *kaddrs;
874 int err;
875 int addrcnt = 0;
876 int walk_size = 0;
877 struct sockaddr *sa_addr;
878 void *addr_buf;
879 struct sctp_af *af;
881 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
882 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
884 if (unlikely(addrs_size <= 0))
885 return -EINVAL;
887 /* Check the user passed a healthy pointer. */
888 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
889 return -EFAULT;
891 /* Alloc space for the address array in kernel memory. */
892 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
893 if (unlikely(!kaddrs))
894 return -ENOMEM;
896 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
897 kfree(kaddrs);
898 return -EFAULT;
901 /* Walk through the addrs buffer and count the number of addresses. */
902 addr_buf = kaddrs;
903 while (walk_size < addrs_size) {
904 sa_addr = (struct sockaddr *)addr_buf;
905 af = sctp_get_af_specific(sa_addr->sa_family);
907 /* If the address family is not supported or if this address
908 * causes the address buffer to overflow return EINVAL.
910 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
911 kfree(kaddrs);
912 return -EINVAL;
914 addrcnt++;
915 addr_buf += af->sockaddr_len;
916 walk_size += af->sockaddr_len;
919 /* Do the work. */
920 switch (op) {
921 case SCTP_BINDX_ADD_ADDR:
922 err = sctp_bindx_add(sk, kaddrs, addrcnt);
923 if (err)
924 goto out;
925 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
926 break;
928 case SCTP_BINDX_REM_ADDR:
929 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
930 if (err)
931 goto out;
932 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
933 break;
935 default:
936 err = -EINVAL;
937 break;
940 out:
941 kfree(kaddrs);
943 return err;
946 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
948 * Common routine for handling connect() and sctp_connectx().
949 * Connect will come in with just a single address.
951 static int __sctp_connect(struct sock* sk,
952 struct sockaddr *kaddrs,
953 int addrs_size)
955 struct sctp_sock *sp;
956 struct sctp_endpoint *ep;
957 struct sctp_association *asoc = NULL;
958 struct sctp_association *asoc2;
959 struct sctp_transport *transport;
960 union sctp_addr to;
961 struct sctp_af *af;
962 sctp_scope_t scope;
963 long timeo;
964 int err = 0;
965 int addrcnt = 0;
966 int walk_size = 0;
967 union sctp_addr *sa_addr;
968 void *addr_buf;
970 sp = sctp_sk(sk);
971 ep = sp->ep;
973 /* connect() cannot be done on a socket that is already in ESTABLISHED
974 * state - UDP-style peeled off socket or a TCP-style socket that
975 * is already connected.
976 * It cannot be done even on a TCP-style listening socket.
978 if (sctp_sstate(sk, ESTABLISHED) ||
979 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
980 err = -EISCONN;
981 goto out_free;
984 /* Walk through the addrs buffer and count the number of addresses. */
985 addr_buf = kaddrs;
986 while (walk_size < addrs_size) {
987 sa_addr = (union sctp_addr *)addr_buf;
988 af = sctp_get_af_specific(sa_addr->sa.sa_family);
990 /* If the address family is not supported or if this address
991 * causes the address buffer to overflow return EINVAL.
993 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
994 err = -EINVAL;
995 goto out_free;
998 err = sctp_verify_addr(sk, sa_addr, af->sockaddr_len);
999 if (err)
1000 goto out_free;
1002 memcpy(&to, sa_addr, af->sockaddr_len);
1004 /* Check if there already is a matching association on the
1005 * endpoint (other than the one created here).
1007 asoc2 = sctp_endpoint_lookup_assoc(ep, sa_addr, &transport);
1008 if (asoc2 && asoc2 != asoc) {
1009 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1010 err = -EISCONN;
1011 else
1012 err = -EALREADY;
1013 goto out_free;
1016 /* If we could not find a matching association on the endpoint,
1017 * make sure that there is no peeled-off association matching
1018 * the peer address even on another socket.
1020 if (sctp_endpoint_is_peeled_off(ep, sa_addr)) {
1021 err = -EADDRNOTAVAIL;
1022 goto out_free;
1025 if (!asoc) {
1026 /* If a bind() or sctp_bindx() is not called prior to
1027 * an sctp_connectx() call, the system picks an
1028 * ephemeral port and will choose an address set
1029 * equivalent to binding with a wildcard address.
1031 if (!ep->base.bind_addr.port) {
1032 if (sctp_autobind(sk)) {
1033 err = -EAGAIN;
1034 goto out_free;
1036 } else {
1038 * If an unprivileged user inherits a 1-many
1039 * style socket with open associations on a
1040 * privileged port, it MAY be permitted to
1041 * accept new associations, but it SHOULD NOT
1042 * be permitted to open new associations.
1044 if (ep->base.bind_addr.port < PROT_SOCK &&
1045 !capable(CAP_NET_BIND_SERVICE)) {
1046 err = -EACCES;
1047 goto out_free;
1051 scope = sctp_scope(sa_addr);
1052 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1053 if (!asoc) {
1054 err = -ENOMEM;
1055 goto out_free;
1059 /* Prime the peer's transport structures. */
1060 transport = sctp_assoc_add_peer(asoc, sa_addr, GFP_KERNEL,
1061 SCTP_UNKNOWN);
1062 if (!transport) {
1063 err = -ENOMEM;
1064 goto out_free;
1067 addrcnt++;
1068 addr_buf += af->sockaddr_len;
1069 walk_size += af->sockaddr_len;
1072 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1073 if (err < 0) {
1074 goto out_free;
1077 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1078 if (err < 0) {
1079 goto out_free;
1082 /* Initialize sk's dport and daddr for getpeername() */
1083 inet_sk(sk)->dport = htons(asoc->peer.port);
1084 af = sctp_get_af_specific(to.sa.sa_family);
1085 af->to_sk_daddr(&to, sk);
1086 sk->sk_err = 0;
1088 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1089 err = sctp_wait_for_connect(asoc, &timeo);
1091 /* Don't free association on exit. */
1092 asoc = NULL;
1094 out_free:
1096 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1097 " kaddrs: %p err: %d\n",
1098 asoc, kaddrs, err);
1099 if (asoc)
1100 sctp_association_free(asoc);
1101 return err;
1104 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1106 * API 8.9
1107 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1109 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1110 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1111 * or IPv6 addresses.
1113 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1114 * Section 3.1.2 for this usage.
1116 * addrs is a pointer to an array of one or more socket addresses. Each
1117 * address is contained in its appropriate structure (i.e. struct
1118 * sockaddr_in or struct sockaddr_in6) the family of the address type
1119 * must be used to distengish the address length (note that this
1120 * representation is termed a "packed array" of addresses). The caller
1121 * specifies the number of addresses in the array with addrcnt.
1123 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1124 * -1, and sets errno to the appropriate error code.
1126 * For SCTP, the port given in each socket address must be the same, or
1127 * sctp_connectx() will fail, setting errno to EINVAL.
1129 * An application can use sctp_connectx to initiate an association with
1130 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1131 * allows a caller to specify multiple addresses at which a peer can be
1132 * reached. The way the SCTP stack uses the list of addresses to set up
1133 * the association is implementation dependant. This function only
1134 * specifies that the stack will try to make use of all the addresses in
1135 * the list when needed.
1137 * Note that the list of addresses passed in is only used for setting up
1138 * the association. It does not necessarily equal the set of addresses
1139 * the peer uses for the resulting association. If the caller wants to
1140 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1141 * retrieve them after the association has been set up.
1143 * Basically do nothing but copying the addresses from user to kernel
1144 * land and invoking either sctp_connectx(). This is used for tunneling
1145 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1147 * We don't use copy_from_user() for optimization: we first do the
1148 * sanity checks (buffer size -fast- and access check-healthy
1149 * pointer); if all of those succeed, then we can alloc the memory
1150 * (expensive operation) needed to copy the data to kernel. Then we do
1151 * the copying without checking the user space area
1152 * (__copy_from_user()).
1154 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1155 * it.
1157 * sk The sk of the socket
1158 * addrs The pointer to the addresses in user land
1159 * addrssize Size of the addrs buffer
1161 * Returns 0 if ok, <0 errno code on error.
1163 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1164 struct sockaddr __user *addrs,
1165 int addrs_size)
1167 int err = 0;
1168 struct sockaddr *kaddrs;
1170 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1171 __FUNCTION__, sk, addrs, addrs_size);
1173 if (unlikely(addrs_size <= 0))
1174 return -EINVAL;
1176 /* Check the user passed a healthy pointer. */
1177 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1178 return -EFAULT;
1180 /* Alloc space for the address array in kernel memory. */
1181 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1182 if (unlikely(!kaddrs))
1183 return -ENOMEM;
1185 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1186 err = -EFAULT;
1187 } else {
1188 err = __sctp_connect(sk, kaddrs, addrs_size);
1191 kfree(kaddrs);
1192 return err;
1195 /* API 3.1.4 close() - UDP Style Syntax
1196 * Applications use close() to perform graceful shutdown (as described in
1197 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1198 * by a UDP-style socket.
1200 * The syntax is
1202 * ret = close(int sd);
1204 * sd - the socket descriptor of the associations to be closed.
1206 * To gracefully shutdown a specific association represented by the
1207 * UDP-style socket, an application should use the sendmsg() call,
1208 * passing no user data, but including the appropriate flag in the
1209 * ancillary data (see Section xxxx).
1211 * If sd in the close() call is a branched-off socket representing only
1212 * one association, the shutdown is performed on that association only.
1214 * 4.1.6 close() - TCP Style Syntax
1216 * Applications use close() to gracefully close down an association.
1218 * The syntax is:
1220 * int close(int sd);
1222 * sd - the socket descriptor of the association to be closed.
1224 * After an application calls close() on a socket descriptor, no further
1225 * socket operations will succeed on that descriptor.
1227 * API 7.1.4 SO_LINGER
1229 * An application using the TCP-style socket can use this option to
1230 * perform the SCTP ABORT primitive. The linger option structure is:
1232 * struct linger {
1233 * int l_onoff; // option on/off
1234 * int l_linger; // linger time
1235 * };
1237 * To enable the option, set l_onoff to 1. If the l_linger value is set
1238 * to 0, calling close() is the same as the ABORT primitive. If the
1239 * value is set to a negative value, the setsockopt() call will return
1240 * an error. If the value is set to a positive value linger_time, the
1241 * close() can be blocked for at most linger_time ms. If the graceful
1242 * shutdown phase does not finish during this period, close() will
1243 * return but the graceful shutdown phase continues in the system.
1245 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1247 struct sctp_endpoint *ep;
1248 struct sctp_association *asoc;
1249 struct list_head *pos, *temp;
1251 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1253 sctp_lock_sock(sk);
1254 sk->sk_shutdown = SHUTDOWN_MASK;
1256 ep = sctp_sk(sk)->ep;
1258 /* Walk all associations on an endpoint. */
1259 list_for_each_safe(pos, temp, &ep->asocs) {
1260 asoc = list_entry(pos, struct sctp_association, asocs);
1262 if (sctp_style(sk, TCP)) {
1263 /* A closed association can still be in the list if
1264 * it belongs to a TCP-style listening socket that is
1265 * not yet accepted. If so, free it. If not, send an
1266 * ABORT or SHUTDOWN based on the linger options.
1268 if (sctp_state(asoc, CLOSED)) {
1269 sctp_unhash_established(asoc);
1270 sctp_association_free(asoc);
1271 continue;
1275 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1276 struct sctp_chunk *chunk;
1278 chunk = sctp_make_abort_user(asoc, NULL, 0);
1279 if (chunk)
1280 sctp_primitive_ABORT(asoc, chunk);
1281 } else
1282 sctp_primitive_SHUTDOWN(asoc, NULL);
1285 /* Clean up any skbs sitting on the receive queue. */
1286 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1287 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1289 /* On a TCP-style socket, block for at most linger_time if set. */
1290 if (sctp_style(sk, TCP) && timeout)
1291 sctp_wait_for_close(sk, timeout);
1293 /* This will run the backlog queue. */
1294 sctp_release_sock(sk);
1296 /* Supposedly, no process has access to the socket, but
1297 * the net layers still may.
1299 sctp_local_bh_disable();
1300 sctp_bh_lock_sock(sk);
1302 /* Hold the sock, since sk_common_release() will put sock_put()
1303 * and we have just a little more cleanup.
1305 sock_hold(sk);
1306 sk_common_release(sk);
1308 sctp_bh_unlock_sock(sk);
1309 sctp_local_bh_enable();
1311 sock_put(sk);
1313 SCTP_DBG_OBJCNT_DEC(sock);
1316 /* Handle EPIPE error. */
1317 static int sctp_error(struct sock *sk, int flags, int err)
1319 if (err == -EPIPE)
1320 err = sock_error(sk) ? : -EPIPE;
1321 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1322 send_sig(SIGPIPE, current, 0);
1323 return err;
1326 /* API 3.1.3 sendmsg() - UDP Style Syntax
1328 * An application uses sendmsg() and recvmsg() calls to transmit data to
1329 * and receive data from its peer.
1331 * ssize_t sendmsg(int socket, const struct msghdr *message,
1332 * int flags);
1334 * socket - the socket descriptor of the endpoint.
1335 * message - pointer to the msghdr structure which contains a single
1336 * user message and possibly some ancillary data.
1338 * See Section 5 for complete description of the data
1339 * structures.
1341 * flags - flags sent or received with the user message, see Section
1342 * 5 for complete description of the flags.
1344 * Note: This function could use a rewrite especially when explicit
1345 * connect support comes in.
1347 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1349 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1351 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1352 struct msghdr *msg, size_t msg_len)
1354 struct sctp_sock *sp;
1355 struct sctp_endpoint *ep;
1356 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1357 struct sctp_transport *transport, *chunk_tp;
1358 struct sctp_chunk *chunk;
1359 union sctp_addr to;
1360 struct sockaddr *msg_name = NULL;
1361 struct sctp_sndrcvinfo default_sinfo = { 0 };
1362 struct sctp_sndrcvinfo *sinfo;
1363 struct sctp_initmsg *sinit;
1364 sctp_assoc_t associd = 0;
1365 sctp_cmsgs_t cmsgs = { NULL };
1366 int err;
1367 sctp_scope_t scope;
1368 long timeo;
1369 __u16 sinfo_flags = 0;
1370 struct sctp_datamsg *datamsg;
1371 struct list_head *pos;
1372 int msg_flags = msg->msg_flags;
1374 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1375 sk, msg, msg_len);
1377 err = 0;
1378 sp = sctp_sk(sk);
1379 ep = sp->ep;
1381 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1383 /* We cannot send a message over a TCP-style listening socket. */
1384 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1385 err = -EPIPE;
1386 goto out_nounlock;
1389 /* Parse out the SCTP CMSGs. */
1390 err = sctp_msghdr_parse(msg, &cmsgs);
1392 if (err) {
1393 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1394 goto out_nounlock;
1397 /* Fetch the destination address for this packet. This
1398 * address only selects the association--it is not necessarily
1399 * the address we will send to.
1400 * For a peeled-off socket, msg_name is ignored.
1402 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1403 int msg_namelen = msg->msg_namelen;
1405 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1406 msg_namelen);
1407 if (err)
1408 return err;
1410 if (msg_namelen > sizeof(to))
1411 msg_namelen = sizeof(to);
1412 memcpy(&to, msg->msg_name, msg_namelen);
1413 msg_name = msg->msg_name;
1416 sinfo = cmsgs.info;
1417 sinit = cmsgs.init;
1419 /* Did the user specify SNDRCVINFO? */
1420 if (sinfo) {
1421 sinfo_flags = sinfo->sinfo_flags;
1422 associd = sinfo->sinfo_assoc_id;
1425 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1426 msg_len, sinfo_flags);
1428 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1429 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1430 err = -EINVAL;
1431 goto out_nounlock;
1434 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1435 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1436 * If SCTP_ABORT is set, the message length could be non zero with
1437 * the msg_iov set to the user abort reason.
1439 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1440 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1441 err = -EINVAL;
1442 goto out_nounlock;
1445 /* If SCTP_ADDR_OVER is set, there must be an address
1446 * specified in msg_name.
1448 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1449 err = -EINVAL;
1450 goto out_nounlock;
1453 transport = NULL;
1455 SCTP_DEBUG_PRINTK("About to look up association.\n");
1457 sctp_lock_sock(sk);
1459 /* If a msg_name has been specified, assume this is to be used. */
1460 if (msg_name) {
1461 /* Look for a matching association on the endpoint. */
1462 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1463 if (!asoc) {
1464 /* If we could not find a matching association on the
1465 * endpoint, make sure that it is not a TCP-style
1466 * socket that already has an association or there is
1467 * no peeled-off association on another socket.
1469 if ((sctp_style(sk, TCP) &&
1470 sctp_sstate(sk, ESTABLISHED)) ||
1471 sctp_endpoint_is_peeled_off(ep, &to)) {
1472 err = -EADDRNOTAVAIL;
1473 goto out_unlock;
1476 } else {
1477 asoc = sctp_id2assoc(sk, associd);
1478 if (!asoc) {
1479 err = -EPIPE;
1480 goto out_unlock;
1484 if (asoc) {
1485 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1487 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1488 * socket that has an association in CLOSED state. This can
1489 * happen when an accepted socket has an association that is
1490 * already CLOSED.
1492 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1493 err = -EPIPE;
1494 goto out_unlock;
1497 if (sinfo_flags & SCTP_EOF) {
1498 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1499 asoc);
1500 sctp_primitive_SHUTDOWN(asoc, NULL);
1501 err = 0;
1502 goto out_unlock;
1504 if (sinfo_flags & SCTP_ABORT) {
1505 struct sctp_chunk *chunk;
1507 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1508 if (!chunk) {
1509 err = -ENOMEM;
1510 goto out_unlock;
1513 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1514 sctp_primitive_ABORT(asoc, chunk);
1515 err = 0;
1516 goto out_unlock;
1520 /* Do we need to create the association? */
1521 if (!asoc) {
1522 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1524 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1525 err = -EINVAL;
1526 goto out_unlock;
1529 /* Check for invalid stream against the stream counts,
1530 * either the default or the user specified stream counts.
1532 if (sinfo) {
1533 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1534 /* Check against the defaults. */
1535 if (sinfo->sinfo_stream >=
1536 sp->initmsg.sinit_num_ostreams) {
1537 err = -EINVAL;
1538 goto out_unlock;
1540 } else {
1541 /* Check against the requested. */
1542 if (sinfo->sinfo_stream >=
1543 sinit->sinit_num_ostreams) {
1544 err = -EINVAL;
1545 goto out_unlock;
1551 * API 3.1.2 bind() - UDP Style Syntax
1552 * If a bind() or sctp_bindx() is not called prior to a
1553 * sendmsg() call that initiates a new association, the
1554 * system picks an ephemeral port and will choose an address
1555 * set equivalent to binding with a wildcard address.
1557 if (!ep->base.bind_addr.port) {
1558 if (sctp_autobind(sk)) {
1559 err = -EAGAIN;
1560 goto out_unlock;
1562 } else {
1564 * If an unprivileged user inherits a one-to-many
1565 * style socket with open associations on a privileged
1566 * port, it MAY be permitted to accept new associations,
1567 * but it SHOULD NOT be permitted to open new
1568 * associations.
1570 if (ep->base.bind_addr.port < PROT_SOCK &&
1571 !capable(CAP_NET_BIND_SERVICE)) {
1572 err = -EACCES;
1573 goto out_unlock;
1577 scope = sctp_scope(&to);
1578 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1579 if (!new_asoc) {
1580 err = -ENOMEM;
1581 goto out_unlock;
1583 asoc = new_asoc;
1585 /* If the SCTP_INIT ancillary data is specified, set all
1586 * the association init values accordingly.
1588 if (sinit) {
1589 if (sinit->sinit_num_ostreams) {
1590 asoc->c.sinit_num_ostreams =
1591 sinit->sinit_num_ostreams;
1593 if (sinit->sinit_max_instreams) {
1594 asoc->c.sinit_max_instreams =
1595 sinit->sinit_max_instreams;
1597 if (sinit->sinit_max_attempts) {
1598 asoc->max_init_attempts
1599 = sinit->sinit_max_attempts;
1601 if (sinit->sinit_max_init_timeo) {
1602 asoc->max_init_timeo =
1603 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1607 /* Prime the peer's transport structures. */
1608 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1609 if (!transport) {
1610 err = -ENOMEM;
1611 goto out_free;
1613 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1614 if (err < 0) {
1615 err = -ENOMEM;
1616 goto out_free;
1620 /* ASSERT: we have a valid association at this point. */
1621 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1623 if (!sinfo) {
1624 /* If the user didn't specify SNDRCVINFO, make up one with
1625 * some defaults.
1627 default_sinfo.sinfo_stream = asoc->default_stream;
1628 default_sinfo.sinfo_flags = asoc->default_flags;
1629 default_sinfo.sinfo_ppid = asoc->default_ppid;
1630 default_sinfo.sinfo_context = asoc->default_context;
1631 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1632 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1633 sinfo = &default_sinfo;
1636 /* API 7.1.7, the sndbuf size per association bounds the
1637 * maximum size of data that can be sent in a single send call.
1639 if (msg_len > sk->sk_sndbuf) {
1640 err = -EMSGSIZE;
1641 goto out_free;
1644 /* If fragmentation is disabled and the message length exceeds the
1645 * association fragmentation point, return EMSGSIZE. The I-D
1646 * does not specify what this error is, but this looks like
1647 * a great fit.
1649 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1650 err = -EMSGSIZE;
1651 goto out_free;
1654 if (sinfo) {
1655 /* Check for invalid stream. */
1656 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1657 err = -EINVAL;
1658 goto out_free;
1662 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1663 if (!sctp_wspace(asoc)) {
1664 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1665 if (err)
1666 goto out_free;
1669 /* If an address is passed with the sendto/sendmsg call, it is used
1670 * to override the primary destination address in the TCP model, or
1671 * when SCTP_ADDR_OVER flag is set in the UDP model.
1673 if ((sctp_style(sk, TCP) && msg_name) ||
1674 (sinfo_flags & SCTP_ADDR_OVER)) {
1675 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1676 if (!chunk_tp) {
1677 err = -EINVAL;
1678 goto out_free;
1680 } else
1681 chunk_tp = NULL;
1683 /* Auto-connect, if we aren't connected already. */
1684 if (sctp_state(asoc, CLOSED)) {
1685 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1686 if (err < 0)
1687 goto out_free;
1688 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1691 /* Break the message into multiple chunks of maximum size. */
1692 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1693 if (!datamsg) {
1694 err = -ENOMEM;
1695 goto out_free;
1698 /* Now send the (possibly) fragmented message. */
1699 list_for_each(pos, &datamsg->chunks) {
1700 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1701 sctp_datamsg_track(chunk);
1703 /* Do accounting for the write space. */
1704 sctp_set_owner_w(chunk);
1706 chunk->transport = chunk_tp;
1708 /* Send it to the lower layers. Note: all chunks
1709 * must either fail or succeed. The lower layer
1710 * works that way today. Keep it that way or this
1711 * breaks.
1713 err = sctp_primitive_SEND(asoc, chunk);
1714 /* Did the lower layer accept the chunk? */
1715 if (err)
1716 sctp_chunk_free(chunk);
1717 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1720 sctp_datamsg_free(datamsg);
1721 if (err)
1722 goto out_free;
1723 else
1724 err = msg_len;
1726 /* If we are already past ASSOCIATE, the lower
1727 * layers are responsible for association cleanup.
1729 goto out_unlock;
1731 out_free:
1732 if (new_asoc)
1733 sctp_association_free(asoc);
1734 out_unlock:
1735 sctp_release_sock(sk);
1737 out_nounlock:
1738 return sctp_error(sk, msg_flags, err);
1740 #if 0
1741 do_sock_err:
1742 if (msg_len)
1743 err = msg_len;
1744 else
1745 err = sock_error(sk);
1746 goto out;
1748 do_interrupted:
1749 if (msg_len)
1750 err = msg_len;
1751 goto out;
1752 #endif /* 0 */
1755 /* This is an extended version of skb_pull() that removes the data from the
1756 * start of a skb even when data is spread across the list of skb's in the
1757 * frag_list. len specifies the total amount of data that needs to be removed.
1758 * when 'len' bytes could be removed from the skb, it returns 0.
1759 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1760 * could not be removed.
1762 static int sctp_skb_pull(struct sk_buff *skb, int len)
1764 struct sk_buff *list;
1765 int skb_len = skb_headlen(skb);
1766 int rlen;
1768 if (len <= skb_len) {
1769 __skb_pull(skb, len);
1770 return 0;
1772 len -= skb_len;
1773 __skb_pull(skb, skb_len);
1775 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1776 rlen = sctp_skb_pull(list, len);
1777 skb->len -= (len-rlen);
1778 skb->data_len -= (len-rlen);
1780 if (!rlen)
1781 return 0;
1783 len = rlen;
1786 return len;
1789 /* API 3.1.3 recvmsg() - UDP Style Syntax
1791 * ssize_t recvmsg(int socket, struct msghdr *message,
1792 * int flags);
1794 * socket - the socket descriptor of the endpoint.
1795 * message - pointer to the msghdr structure which contains a single
1796 * user message and possibly some ancillary data.
1798 * See Section 5 for complete description of the data
1799 * structures.
1801 * flags - flags sent or received with the user message, see Section
1802 * 5 for complete description of the flags.
1804 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1806 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1807 struct msghdr *msg, size_t len, int noblock,
1808 int flags, int *addr_len)
1810 struct sctp_ulpevent *event = NULL;
1811 struct sctp_sock *sp = sctp_sk(sk);
1812 struct sk_buff *skb;
1813 int copied;
1814 int err = 0;
1815 int skb_len;
1817 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1818 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1819 "len", len, "knoblauch", noblock,
1820 "flags", flags, "addr_len", addr_len);
1822 sctp_lock_sock(sk);
1824 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1825 err = -ENOTCONN;
1826 goto out;
1829 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1830 if (!skb)
1831 goto out;
1833 /* Get the total length of the skb including any skb's in the
1834 * frag_list.
1836 skb_len = skb->len;
1838 copied = skb_len;
1839 if (copied > len)
1840 copied = len;
1842 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1844 event = sctp_skb2event(skb);
1846 if (err)
1847 goto out_free;
1849 sock_recv_timestamp(msg, sk, skb);
1850 if (sctp_ulpevent_is_notification(event)) {
1851 msg->msg_flags |= MSG_NOTIFICATION;
1852 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1853 } else {
1854 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1857 /* Check if we allow SCTP_SNDRCVINFO. */
1858 if (sp->subscribe.sctp_data_io_event)
1859 sctp_ulpevent_read_sndrcvinfo(event, msg);
1860 #if 0
1861 /* FIXME: we should be calling IP/IPv6 layers. */
1862 if (sk->sk_protinfo.af_inet.cmsg_flags)
1863 ip_cmsg_recv(msg, skb);
1864 #endif
1866 err = copied;
1868 /* If skb's length exceeds the user's buffer, update the skb and
1869 * push it back to the receive_queue so that the next call to
1870 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1872 if (skb_len > copied) {
1873 msg->msg_flags &= ~MSG_EOR;
1874 if (flags & MSG_PEEK)
1875 goto out_free;
1876 sctp_skb_pull(skb, copied);
1877 skb_queue_head(&sk->sk_receive_queue, skb);
1879 /* When only partial message is copied to the user, increase
1880 * rwnd by that amount. If all the data in the skb is read,
1881 * rwnd is updated when the event is freed.
1883 sctp_assoc_rwnd_increase(event->asoc, copied);
1884 goto out;
1885 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1886 (event->msg_flags & MSG_EOR))
1887 msg->msg_flags |= MSG_EOR;
1888 else
1889 msg->msg_flags &= ~MSG_EOR;
1891 out_free:
1892 if (flags & MSG_PEEK) {
1893 /* Release the skb reference acquired after peeking the skb in
1894 * sctp_skb_recv_datagram().
1896 kfree_skb(skb);
1897 } else {
1898 /* Free the event which includes releasing the reference to
1899 * the owner of the skb, freeing the skb and updating the
1900 * rwnd.
1902 sctp_ulpevent_free(event);
1904 out:
1905 sctp_release_sock(sk);
1906 return err;
1909 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1911 * This option is a on/off flag. If enabled no SCTP message
1912 * fragmentation will be performed. Instead if a message being sent
1913 * exceeds the current PMTU size, the message will NOT be sent and
1914 * instead a error will be indicated to the user.
1916 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1917 char __user *optval, int optlen)
1919 int val;
1921 if (optlen < sizeof(int))
1922 return -EINVAL;
1924 if (get_user(val, (int __user *)optval))
1925 return -EFAULT;
1927 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1929 return 0;
1932 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1933 int optlen)
1935 if (optlen != sizeof(struct sctp_event_subscribe))
1936 return -EINVAL;
1937 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1938 return -EFAULT;
1939 return 0;
1942 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1944 * This socket option is applicable to the UDP-style socket only. When
1945 * set it will cause associations that are idle for more than the
1946 * specified number of seconds to automatically close. An association
1947 * being idle is defined an association that has NOT sent or received
1948 * user data. The special value of '0' indicates that no automatic
1949 * close of any associations should be performed. The option expects an
1950 * integer defining the number of seconds of idle time before an
1951 * association is closed.
1953 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1954 int optlen)
1956 struct sctp_sock *sp = sctp_sk(sk);
1958 /* Applicable to UDP-style socket only */
1959 if (sctp_style(sk, TCP))
1960 return -EOPNOTSUPP;
1961 if (optlen != sizeof(int))
1962 return -EINVAL;
1963 if (copy_from_user(&sp->autoclose, optval, optlen))
1964 return -EFAULT;
1966 return 0;
1969 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1971 * Applications can enable or disable heartbeats for any peer address of
1972 * an association, modify an address's heartbeat interval, force a
1973 * heartbeat to be sent immediately, and adjust the address's maximum
1974 * number of retransmissions sent before an address is considered
1975 * unreachable. The following structure is used to access and modify an
1976 * address's parameters:
1978 * struct sctp_paddrparams {
1979 * sctp_assoc_t spp_assoc_id;
1980 * struct sockaddr_storage spp_address;
1981 * uint32_t spp_hbinterval;
1982 * uint16_t spp_pathmaxrxt;
1983 * uint32_t spp_pathmtu;
1984 * uint32_t spp_sackdelay;
1985 * uint32_t spp_flags;
1986 * };
1988 * spp_assoc_id - (one-to-many style socket) This is filled in the
1989 * application, and identifies the association for
1990 * this query.
1991 * spp_address - This specifies which address is of interest.
1992 * spp_hbinterval - This contains the value of the heartbeat interval,
1993 * in milliseconds. If a value of zero
1994 * is present in this field then no changes are to
1995 * be made to this parameter.
1996 * spp_pathmaxrxt - This contains the maximum number of
1997 * retransmissions before this address shall be
1998 * considered unreachable. If a value of zero
1999 * is present in this field then no changes are to
2000 * be made to this parameter.
2001 * spp_pathmtu - When Path MTU discovery is disabled the value
2002 * specified here will be the "fixed" path mtu.
2003 * Note that if the spp_address field is empty
2004 * then all associations on this address will
2005 * have this fixed path mtu set upon them.
2007 * spp_sackdelay - When delayed sack is enabled, this value specifies
2008 * the number of milliseconds that sacks will be delayed
2009 * for. This value will apply to all addresses of an
2010 * association if the spp_address field is empty. Note
2011 * also, that if delayed sack is enabled and this
2012 * value is set to 0, no change is made to the last
2013 * recorded delayed sack timer value.
2015 * spp_flags - These flags are used to control various features
2016 * on an association. The flag field may contain
2017 * zero or more of the following options.
2019 * SPP_HB_ENABLE - Enable heartbeats on the
2020 * specified address. Note that if the address
2021 * field is empty all addresses for the association
2022 * have heartbeats enabled upon them.
2024 * SPP_HB_DISABLE - Disable heartbeats on the
2025 * speicifed address. Note that if the address
2026 * field is empty all addresses for the association
2027 * will have their heartbeats disabled. Note also
2028 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2029 * mutually exclusive, only one of these two should
2030 * be specified. Enabling both fields will have
2031 * undetermined results.
2033 * SPP_HB_DEMAND - Request a user initiated heartbeat
2034 * to be made immediately.
2036 * SPP_PMTUD_ENABLE - This field will enable PMTU
2037 * discovery upon the specified address. Note that
2038 * if the address feild is empty then all addresses
2039 * on the association are effected.
2041 * SPP_PMTUD_DISABLE - This field will disable PMTU
2042 * discovery upon the specified address. Note that
2043 * if the address feild is empty then all addresses
2044 * on the association are effected. Not also that
2045 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2046 * exclusive. Enabling both will have undetermined
2047 * results.
2049 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2050 * on delayed sack. The time specified in spp_sackdelay
2051 * is used to specify the sack delay for this address. Note
2052 * that if spp_address is empty then all addresses will
2053 * enable delayed sack and take on the sack delay
2054 * value specified in spp_sackdelay.
2055 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2056 * off delayed sack. If the spp_address field is blank then
2057 * delayed sack is disabled for the entire association. Note
2058 * also that this field is mutually exclusive to
2059 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2060 * results.
2062 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2063 struct sctp_transport *trans,
2064 struct sctp_association *asoc,
2065 struct sctp_sock *sp,
2066 int hb_change,
2067 int pmtud_change,
2068 int sackdelay_change)
2070 int error;
2072 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2073 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2074 if (error)
2075 return error;
2078 if (params->spp_hbinterval) {
2079 if (trans) {
2080 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2081 } else if (asoc) {
2082 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2083 } else {
2084 sp->hbinterval = params->spp_hbinterval;
2088 if (hb_change) {
2089 if (trans) {
2090 trans->param_flags =
2091 (trans->param_flags & ~SPP_HB) | hb_change;
2092 } else if (asoc) {
2093 asoc->param_flags =
2094 (asoc->param_flags & ~SPP_HB) | hb_change;
2095 } else {
2096 sp->param_flags =
2097 (sp->param_flags & ~SPP_HB) | hb_change;
2101 if (params->spp_pathmtu) {
2102 if (trans) {
2103 trans->pathmtu = params->spp_pathmtu;
2104 sctp_assoc_sync_pmtu(asoc);
2105 } else if (asoc) {
2106 asoc->pathmtu = params->spp_pathmtu;
2107 sctp_frag_point(sp, params->spp_pathmtu);
2108 } else {
2109 sp->pathmtu = params->spp_pathmtu;
2113 if (pmtud_change) {
2114 if (trans) {
2115 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2116 (params->spp_flags & SPP_PMTUD_ENABLE);
2117 trans->param_flags =
2118 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2119 if (update) {
2120 sctp_transport_pmtu(trans);
2121 sctp_assoc_sync_pmtu(asoc);
2123 } else if (asoc) {
2124 asoc->param_flags =
2125 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2126 } else {
2127 sp->param_flags =
2128 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2132 if (params->spp_sackdelay) {
2133 if (trans) {
2134 trans->sackdelay =
2135 msecs_to_jiffies(params->spp_sackdelay);
2136 } else if (asoc) {
2137 asoc->sackdelay =
2138 msecs_to_jiffies(params->spp_sackdelay);
2139 } else {
2140 sp->sackdelay = params->spp_sackdelay;
2144 if (sackdelay_change) {
2145 if (trans) {
2146 trans->param_flags =
2147 (trans->param_flags & ~SPP_SACKDELAY) |
2148 sackdelay_change;
2149 } else if (asoc) {
2150 asoc->param_flags =
2151 (asoc->param_flags & ~SPP_SACKDELAY) |
2152 sackdelay_change;
2153 } else {
2154 sp->param_flags =
2155 (sp->param_flags & ~SPP_SACKDELAY) |
2156 sackdelay_change;
2160 if (params->spp_pathmaxrxt) {
2161 if (trans) {
2162 trans->pathmaxrxt = params->spp_pathmaxrxt;
2163 } else if (asoc) {
2164 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2165 } else {
2166 sp->pathmaxrxt = params->spp_pathmaxrxt;
2170 return 0;
2173 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2174 char __user *optval, int optlen)
2176 struct sctp_paddrparams params;
2177 struct sctp_transport *trans = NULL;
2178 struct sctp_association *asoc = NULL;
2179 struct sctp_sock *sp = sctp_sk(sk);
2180 int error;
2181 int hb_change, pmtud_change, sackdelay_change;
2183 if (optlen != sizeof(struct sctp_paddrparams))
2184 return - EINVAL;
2186 if (copy_from_user(&params, optval, optlen))
2187 return -EFAULT;
2189 /* Validate flags and value parameters. */
2190 hb_change = params.spp_flags & SPP_HB;
2191 pmtud_change = params.spp_flags & SPP_PMTUD;
2192 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2194 if (hb_change == SPP_HB ||
2195 pmtud_change == SPP_PMTUD ||
2196 sackdelay_change == SPP_SACKDELAY ||
2197 params.spp_sackdelay > 500 ||
2198 (params.spp_pathmtu
2199 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2200 return -EINVAL;
2202 /* If an address other than INADDR_ANY is specified, and
2203 * no transport is found, then the request is invalid.
2205 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2206 trans = sctp_addr_id2transport(sk, &params.spp_address,
2207 params.spp_assoc_id);
2208 if (!trans)
2209 return -EINVAL;
2212 /* Get association, if assoc_id != 0 and the socket is a one
2213 * to many style socket, and an association was not found, then
2214 * the id was invalid.
2216 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2217 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2218 return -EINVAL;
2220 /* Heartbeat demand can only be sent on a transport or
2221 * association, but not a socket.
2223 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2224 return -EINVAL;
2226 /* Process parameters. */
2227 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2228 hb_change, pmtud_change,
2229 sackdelay_change);
2231 if (error)
2232 return error;
2234 /* If changes are for association, also apply parameters to each
2235 * transport.
2237 if (!trans && asoc) {
2238 struct list_head *pos;
2240 list_for_each(pos, &asoc->peer.transport_addr_list) {
2241 trans = list_entry(pos, struct sctp_transport,
2242 transports);
2243 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2244 hb_change, pmtud_change,
2245 sackdelay_change);
2249 return 0;
2252 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2254 * This options will get or set the delayed ack timer. The time is set
2255 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2256 * endpoints default delayed ack timer value. If the assoc_id field is
2257 * non-zero, then the set or get effects the specified association.
2259 * struct sctp_assoc_value {
2260 * sctp_assoc_t assoc_id;
2261 * uint32_t assoc_value;
2262 * };
2264 * assoc_id - This parameter, indicates which association the
2265 * user is preforming an action upon. Note that if
2266 * this field's value is zero then the endpoints
2267 * default value is changed (effecting future
2268 * associations only).
2270 * assoc_value - This parameter contains the number of milliseconds
2271 * that the user is requesting the delayed ACK timer
2272 * be set to. Note that this value is defined in
2273 * the standard to be between 200 and 500 milliseconds.
2275 * Note: a value of zero will leave the value alone,
2276 * but disable SACK delay. A non-zero value will also
2277 * enable SACK delay.
2280 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2281 char __user *optval, int optlen)
2283 struct sctp_assoc_value params;
2284 struct sctp_transport *trans = NULL;
2285 struct sctp_association *asoc = NULL;
2286 struct sctp_sock *sp = sctp_sk(sk);
2288 if (optlen != sizeof(struct sctp_assoc_value))
2289 return - EINVAL;
2291 if (copy_from_user(&params, optval, optlen))
2292 return -EFAULT;
2294 /* Validate value parameter. */
2295 if (params.assoc_value > 500)
2296 return -EINVAL;
2298 /* Get association, if assoc_id != 0 and the socket is a one
2299 * to many style socket, and an association was not found, then
2300 * the id was invalid.
2302 asoc = sctp_id2assoc(sk, params.assoc_id);
2303 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2304 return -EINVAL;
2306 if (params.assoc_value) {
2307 if (asoc) {
2308 asoc->sackdelay =
2309 msecs_to_jiffies(params.assoc_value);
2310 asoc->param_flags =
2311 (asoc->param_flags & ~SPP_SACKDELAY) |
2312 SPP_SACKDELAY_ENABLE;
2313 } else {
2314 sp->sackdelay = params.assoc_value;
2315 sp->param_flags =
2316 (sp->param_flags & ~SPP_SACKDELAY) |
2317 SPP_SACKDELAY_ENABLE;
2319 } else {
2320 if (asoc) {
2321 asoc->param_flags =
2322 (asoc->param_flags & ~SPP_SACKDELAY) |
2323 SPP_SACKDELAY_DISABLE;
2324 } else {
2325 sp->param_flags =
2326 (sp->param_flags & ~SPP_SACKDELAY) |
2327 SPP_SACKDELAY_DISABLE;
2331 /* If change is for association, also apply to each transport. */
2332 if (asoc) {
2333 struct list_head *pos;
2335 list_for_each(pos, &asoc->peer.transport_addr_list) {
2336 trans = list_entry(pos, struct sctp_transport,
2337 transports);
2338 if (params.assoc_value) {
2339 trans->sackdelay =
2340 msecs_to_jiffies(params.assoc_value);
2341 trans->param_flags =
2342 (trans->param_flags & ~SPP_SACKDELAY) |
2343 SPP_SACKDELAY_ENABLE;
2344 } else {
2345 trans->param_flags =
2346 (trans->param_flags & ~SPP_SACKDELAY) |
2347 SPP_SACKDELAY_DISABLE;
2352 return 0;
2355 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2357 * Applications can specify protocol parameters for the default association
2358 * initialization. The option name argument to setsockopt() and getsockopt()
2359 * is SCTP_INITMSG.
2361 * Setting initialization parameters is effective only on an unconnected
2362 * socket (for UDP-style sockets only future associations are effected
2363 * by the change). With TCP-style sockets, this option is inherited by
2364 * sockets derived from a listener socket.
2366 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2368 struct sctp_initmsg sinit;
2369 struct sctp_sock *sp = sctp_sk(sk);
2371 if (optlen != sizeof(struct sctp_initmsg))
2372 return -EINVAL;
2373 if (copy_from_user(&sinit, optval, optlen))
2374 return -EFAULT;
2376 if (sinit.sinit_num_ostreams)
2377 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2378 if (sinit.sinit_max_instreams)
2379 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2380 if (sinit.sinit_max_attempts)
2381 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2382 if (sinit.sinit_max_init_timeo)
2383 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2385 return 0;
2389 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2391 * Applications that wish to use the sendto() system call may wish to
2392 * specify a default set of parameters that would normally be supplied
2393 * through the inclusion of ancillary data. This socket option allows
2394 * such an application to set the default sctp_sndrcvinfo structure.
2395 * The application that wishes to use this socket option simply passes
2396 * in to this call the sctp_sndrcvinfo structure defined in Section
2397 * 5.2.2) The input parameters accepted by this call include
2398 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2399 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2400 * to this call if the caller is using the UDP model.
2402 static int sctp_setsockopt_default_send_param(struct sock *sk,
2403 char __user *optval, int optlen)
2405 struct sctp_sndrcvinfo info;
2406 struct sctp_association *asoc;
2407 struct sctp_sock *sp = sctp_sk(sk);
2409 if (optlen != sizeof(struct sctp_sndrcvinfo))
2410 return -EINVAL;
2411 if (copy_from_user(&info, optval, optlen))
2412 return -EFAULT;
2414 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2415 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2416 return -EINVAL;
2418 if (asoc) {
2419 asoc->default_stream = info.sinfo_stream;
2420 asoc->default_flags = info.sinfo_flags;
2421 asoc->default_ppid = info.sinfo_ppid;
2422 asoc->default_context = info.sinfo_context;
2423 asoc->default_timetolive = info.sinfo_timetolive;
2424 } else {
2425 sp->default_stream = info.sinfo_stream;
2426 sp->default_flags = info.sinfo_flags;
2427 sp->default_ppid = info.sinfo_ppid;
2428 sp->default_context = info.sinfo_context;
2429 sp->default_timetolive = info.sinfo_timetolive;
2432 return 0;
2435 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2437 * Requests that the local SCTP stack use the enclosed peer address as
2438 * the association primary. The enclosed address must be one of the
2439 * association peer's addresses.
2441 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2442 int optlen)
2444 struct sctp_prim prim;
2445 struct sctp_transport *trans;
2447 if (optlen != sizeof(struct sctp_prim))
2448 return -EINVAL;
2450 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2451 return -EFAULT;
2453 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2454 if (!trans)
2455 return -EINVAL;
2457 sctp_assoc_set_primary(trans->asoc, trans);
2459 return 0;
2463 * 7.1.5 SCTP_NODELAY
2465 * Turn on/off any Nagle-like algorithm. This means that packets are
2466 * generally sent as soon as possible and no unnecessary delays are
2467 * introduced, at the cost of more packets in the network. Expects an
2468 * integer boolean flag.
2470 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2471 int optlen)
2473 int val;
2475 if (optlen < sizeof(int))
2476 return -EINVAL;
2477 if (get_user(val, (int __user *)optval))
2478 return -EFAULT;
2480 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2481 return 0;
2486 * 7.1.1 SCTP_RTOINFO
2488 * The protocol parameters used to initialize and bound retransmission
2489 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2490 * and modify these parameters.
2491 * All parameters are time values, in milliseconds. A value of 0, when
2492 * modifying the parameters, indicates that the current value should not
2493 * be changed.
2496 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2497 struct sctp_rtoinfo rtoinfo;
2498 struct sctp_association *asoc;
2500 if (optlen != sizeof (struct sctp_rtoinfo))
2501 return -EINVAL;
2503 if (copy_from_user(&rtoinfo, optval, optlen))
2504 return -EFAULT;
2506 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2508 /* Set the values to the specific association */
2509 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2510 return -EINVAL;
2512 if (asoc) {
2513 if (rtoinfo.srto_initial != 0)
2514 asoc->rto_initial =
2515 msecs_to_jiffies(rtoinfo.srto_initial);
2516 if (rtoinfo.srto_max != 0)
2517 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2518 if (rtoinfo.srto_min != 0)
2519 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2520 } else {
2521 /* If there is no association or the association-id = 0
2522 * set the values to the endpoint.
2524 struct sctp_sock *sp = sctp_sk(sk);
2526 if (rtoinfo.srto_initial != 0)
2527 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2528 if (rtoinfo.srto_max != 0)
2529 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2530 if (rtoinfo.srto_min != 0)
2531 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2534 return 0;
2539 * 7.1.2 SCTP_ASSOCINFO
2541 * This option is used to tune the the maximum retransmission attempts
2542 * of the association.
2543 * Returns an error if the new association retransmission value is
2544 * greater than the sum of the retransmission value of the peer.
2545 * See [SCTP] for more information.
2548 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2551 struct sctp_assocparams assocparams;
2552 struct sctp_association *asoc;
2554 if (optlen != sizeof(struct sctp_assocparams))
2555 return -EINVAL;
2556 if (copy_from_user(&assocparams, optval, optlen))
2557 return -EFAULT;
2559 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2561 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2562 return -EINVAL;
2564 /* Set the values to the specific association */
2565 if (asoc) {
2566 if (assocparams.sasoc_asocmaxrxt != 0) {
2567 __u32 path_sum = 0;
2568 int paths = 0;
2569 struct list_head *pos;
2570 struct sctp_transport *peer_addr;
2572 list_for_each(pos, &asoc->peer.transport_addr_list) {
2573 peer_addr = list_entry(pos,
2574 struct sctp_transport,
2575 transports);
2576 path_sum += peer_addr->pathmaxrxt;
2577 paths++;
2580 /* Only validate asocmaxrxt if we have more then
2581 * one path/transport. We do this because path
2582 * retransmissions are only counted when we have more
2583 * then one path.
2585 if (paths > 1 &&
2586 assocparams.sasoc_asocmaxrxt > path_sum)
2587 return -EINVAL;
2589 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2592 if (assocparams.sasoc_cookie_life != 0) {
2593 asoc->cookie_life.tv_sec =
2594 assocparams.sasoc_cookie_life / 1000;
2595 asoc->cookie_life.tv_usec =
2596 (assocparams.sasoc_cookie_life % 1000)
2597 * 1000;
2599 } else {
2600 /* Set the values to the endpoint */
2601 struct sctp_sock *sp = sctp_sk(sk);
2603 if (assocparams.sasoc_asocmaxrxt != 0)
2604 sp->assocparams.sasoc_asocmaxrxt =
2605 assocparams.sasoc_asocmaxrxt;
2606 if (assocparams.sasoc_cookie_life != 0)
2607 sp->assocparams.sasoc_cookie_life =
2608 assocparams.sasoc_cookie_life;
2610 return 0;
2614 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2616 * This socket option is a boolean flag which turns on or off mapped V4
2617 * addresses. If this option is turned on and the socket is type
2618 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2619 * If this option is turned off, then no mapping will be done of V4
2620 * addresses and a user will receive both PF_INET6 and PF_INET type
2621 * addresses on the socket.
2623 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2625 int val;
2626 struct sctp_sock *sp = sctp_sk(sk);
2628 if (optlen < sizeof(int))
2629 return -EINVAL;
2630 if (get_user(val, (int __user *)optval))
2631 return -EFAULT;
2632 if (val)
2633 sp->v4mapped = 1;
2634 else
2635 sp->v4mapped = 0;
2637 return 0;
2641 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2643 * This socket option specifies the maximum size to put in any outgoing
2644 * SCTP chunk. If a message is larger than this size it will be
2645 * fragmented by SCTP into the specified size. Note that the underlying
2646 * SCTP implementation may fragment into smaller sized chunks when the
2647 * PMTU of the underlying association is smaller than the value set by
2648 * the user.
2650 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2652 struct sctp_association *asoc;
2653 struct list_head *pos;
2654 struct sctp_sock *sp = sctp_sk(sk);
2655 int val;
2657 if (optlen < sizeof(int))
2658 return -EINVAL;
2659 if (get_user(val, (int __user *)optval))
2660 return -EFAULT;
2661 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2662 return -EINVAL;
2663 sp->user_frag = val;
2665 /* Update the frag_point of the existing associations. */
2666 list_for_each(pos, &(sp->ep->asocs)) {
2667 asoc = list_entry(pos, struct sctp_association, asocs);
2668 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2671 return 0;
2676 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2678 * Requests that the peer mark the enclosed address as the association
2679 * primary. The enclosed address must be one of the association's
2680 * locally bound addresses. The following structure is used to make a
2681 * set primary request:
2683 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2684 int optlen)
2686 struct sctp_sock *sp;
2687 struct sctp_endpoint *ep;
2688 struct sctp_association *asoc = NULL;
2689 struct sctp_setpeerprim prim;
2690 struct sctp_chunk *chunk;
2691 int err;
2693 sp = sctp_sk(sk);
2694 ep = sp->ep;
2696 if (!sctp_addip_enable)
2697 return -EPERM;
2699 if (optlen != sizeof(struct sctp_setpeerprim))
2700 return -EINVAL;
2702 if (copy_from_user(&prim, optval, optlen))
2703 return -EFAULT;
2705 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2706 if (!asoc)
2707 return -EINVAL;
2709 if (!asoc->peer.asconf_capable)
2710 return -EPERM;
2712 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2713 return -EPERM;
2715 if (!sctp_state(asoc, ESTABLISHED))
2716 return -ENOTCONN;
2718 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2719 return -EADDRNOTAVAIL;
2721 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2722 chunk = sctp_make_asconf_set_prim(asoc,
2723 (union sctp_addr *)&prim.sspp_addr);
2724 if (!chunk)
2725 return -ENOMEM;
2727 err = sctp_send_asconf(asoc, chunk);
2729 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2731 return err;
2734 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2735 int optlen)
2737 struct sctp_setadaptation adaptation;
2739 if (optlen != sizeof(struct sctp_setadaptation))
2740 return -EINVAL;
2741 if (copy_from_user(&adaptation, optval, optlen))
2742 return -EFAULT;
2744 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2746 return 0;
2750 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2752 * The context field in the sctp_sndrcvinfo structure is normally only
2753 * used when a failed message is retrieved holding the value that was
2754 * sent down on the actual send call. This option allows the setting of
2755 * a default context on an association basis that will be received on
2756 * reading messages from the peer. This is especially helpful in the
2757 * one-2-many model for an application to keep some reference to an
2758 * internal state machine that is processing messages on the
2759 * association. Note that the setting of this value only effects
2760 * received messages from the peer and does not effect the value that is
2761 * saved with outbound messages.
2763 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2764 int optlen)
2766 struct sctp_assoc_value params;
2767 struct sctp_sock *sp;
2768 struct sctp_association *asoc;
2770 if (optlen != sizeof(struct sctp_assoc_value))
2771 return -EINVAL;
2772 if (copy_from_user(&params, optval, optlen))
2773 return -EFAULT;
2775 sp = sctp_sk(sk);
2777 if (params.assoc_id != 0) {
2778 asoc = sctp_id2assoc(sk, params.assoc_id);
2779 if (!asoc)
2780 return -EINVAL;
2781 asoc->default_rcv_context = params.assoc_value;
2782 } else {
2783 sp->default_rcv_context = params.assoc_value;
2786 return 0;
2789 /* API 6.2 setsockopt(), getsockopt()
2791 * Applications use setsockopt() and getsockopt() to set or retrieve
2792 * socket options. Socket options are used to change the default
2793 * behavior of sockets calls. They are described in Section 7.
2795 * The syntax is:
2797 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2798 * int __user *optlen);
2799 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2800 * int optlen);
2802 * sd - the socket descript.
2803 * level - set to IPPROTO_SCTP for all SCTP options.
2804 * optname - the option name.
2805 * optval - the buffer to store the value of the option.
2806 * optlen - the size of the buffer.
2808 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2809 char __user *optval, int optlen)
2811 int retval = 0;
2813 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2814 sk, optname);
2816 /* I can hardly begin to describe how wrong this is. This is
2817 * so broken as to be worse than useless. The API draft
2818 * REALLY is NOT helpful here... I am not convinced that the
2819 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2820 * are at all well-founded.
2822 if (level != SOL_SCTP) {
2823 struct sctp_af *af = sctp_sk(sk)->pf->af;
2824 retval = af->setsockopt(sk, level, optname, optval, optlen);
2825 goto out_nounlock;
2828 sctp_lock_sock(sk);
2830 switch (optname) {
2831 case SCTP_SOCKOPT_BINDX_ADD:
2832 /* 'optlen' is the size of the addresses buffer. */
2833 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2834 optlen, SCTP_BINDX_ADD_ADDR);
2835 break;
2837 case SCTP_SOCKOPT_BINDX_REM:
2838 /* 'optlen' is the size of the addresses buffer. */
2839 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2840 optlen, SCTP_BINDX_REM_ADDR);
2841 break;
2843 case SCTP_SOCKOPT_CONNECTX:
2844 /* 'optlen' is the size of the addresses buffer. */
2845 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2846 optlen);
2847 break;
2849 case SCTP_DISABLE_FRAGMENTS:
2850 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2851 break;
2853 case SCTP_EVENTS:
2854 retval = sctp_setsockopt_events(sk, optval, optlen);
2855 break;
2857 case SCTP_AUTOCLOSE:
2858 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2859 break;
2861 case SCTP_PEER_ADDR_PARAMS:
2862 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2863 break;
2865 case SCTP_DELAYED_ACK_TIME:
2866 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2867 break;
2869 case SCTP_INITMSG:
2870 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2871 break;
2872 case SCTP_DEFAULT_SEND_PARAM:
2873 retval = sctp_setsockopt_default_send_param(sk, optval,
2874 optlen);
2875 break;
2876 case SCTP_PRIMARY_ADDR:
2877 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2878 break;
2879 case SCTP_SET_PEER_PRIMARY_ADDR:
2880 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2881 break;
2882 case SCTP_NODELAY:
2883 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2884 break;
2885 case SCTP_RTOINFO:
2886 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2887 break;
2888 case SCTP_ASSOCINFO:
2889 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2890 break;
2891 case SCTP_I_WANT_MAPPED_V4_ADDR:
2892 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2893 break;
2894 case SCTP_MAXSEG:
2895 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2896 break;
2897 case SCTP_ADAPTATION_LAYER:
2898 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
2899 break;
2900 case SCTP_CONTEXT:
2901 retval = sctp_setsockopt_context(sk, optval, optlen);
2902 break;
2904 default:
2905 retval = -ENOPROTOOPT;
2906 break;
2909 sctp_release_sock(sk);
2911 out_nounlock:
2912 return retval;
2915 /* API 3.1.6 connect() - UDP Style Syntax
2917 * An application may use the connect() call in the UDP model to initiate an
2918 * association without sending data.
2920 * The syntax is:
2922 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2924 * sd: the socket descriptor to have a new association added to.
2926 * nam: the address structure (either struct sockaddr_in or struct
2927 * sockaddr_in6 defined in RFC2553 [7]).
2929 * len: the size of the address.
2931 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
2932 int addr_len)
2934 int err = 0;
2935 struct sctp_af *af;
2937 sctp_lock_sock(sk);
2939 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2940 __FUNCTION__, sk, addr, addr_len);
2942 /* Validate addr_len before calling common connect/connectx routine. */
2943 af = sctp_get_af_specific(addr->sa_family);
2944 if (!af || addr_len < af->sockaddr_len) {
2945 err = -EINVAL;
2946 } else {
2947 /* Pass correct addr len to common routine (so it knows there
2948 * is only one address being passed.
2950 err = __sctp_connect(sk, addr, af->sockaddr_len);
2953 sctp_release_sock(sk);
2954 return err;
2957 /* FIXME: Write comments. */
2958 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
2960 return -EOPNOTSUPP; /* STUB */
2963 /* 4.1.4 accept() - TCP Style Syntax
2965 * Applications use accept() call to remove an established SCTP
2966 * association from the accept queue of the endpoint. A new socket
2967 * descriptor will be returned from accept() to represent the newly
2968 * formed association.
2970 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
2972 struct sctp_sock *sp;
2973 struct sctp_endpoint *ep;
2974 struct sock *newsk = NULL;
2975 struct sctp_association *asoc;
2976 long timeo;
2977 int error = 0;
2979 sctp_lock_sock(sk);
2981 sp = sctp_sk(sk);
2982 ep = sp->ep;
2984 if (!sctp_style(sk, TCP)) {
2985 error = -EOPNOTSUPP;
2986 goto out;
2989 if (!sctp_sstate(sk, LISTENING)) {
2990 error = -EINVAL;
2991 goto out;
2994 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2996 error = sctp_wait_for_accept(sk, timeo);
2997 if (error)
2998 goto out;
3000 /* We treat the list of associations on the endpoint as the accept
3001 * queue and pick the first association on the list.
3003 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3005 newsk = sp->pf->create_accept_sk(sk, asoc);
3006 if (!newsk) {
3007 error = -ENOMEM;
3008 goto out;
3011 /* Populate the fields of the newsk from the oldsk and migrate the
3012 * asoc to the newsk.
3014 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3016 out:
3017 sctp_release_sock(sk);
3018 *err = error;
3019 return newsk;
3022 /* The SCTP ioctl handler. */
3023 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3025 return -ENOIOCTLCMD;
3028 /* This is the function which gets called during socket creation to
3029 * initialized the SCTP-specific portion of the sock.
3030 * The sock structure should already be zero-filled memory.
3032 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3034 struct sctp_endpoint *ep;
3035 struct sctp_sock *sp;
3037 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3039 sp = sctp_sk(sk);
3041 /* Initialize the SCTP per socket area. */
3042 switch (sk->sk_type) {
3043 case SOCK_SEQPACKET:
3044 sp->type = SCTP_SOCKET_UDP;
3045 break;
3046 case SOCK_STREAM:
3047 sp->type = SCTP_SOCKET_TCP;
3048 break;
3049 default:
3050 return -ESOCKTNOSUPPORT;
3053 /* Initialize default send parameters. These parameters can be
3054 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3056 sp->default_stream = 0;
3057 sp->default_ppid = 0;
3058 sp->default_flags = 0;
3059 sp->default_context = 0;
3060 sp->default_timetolive = 0;
3062 sp->default_rcv_context = 0;
3064 /* Initialize default setup parameters. These parameters
3065 * can be modified with the SCTP_INITMSG socket option or
3066 * overridden by the SCTP_INIT CMSG.
3068 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3069 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3070 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3071 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3073 /* Initialize default RTO related parameters. These parameters can
3074 * be modified for with the SCTP_RTOINFO socket option.
3076 sp->rtoinfo.srto_initial = sctp_rto_initial;
3077 sp->rtoinfo.srto_max = sctp_rto_max;
3078 sp->rtoinfo.srto_min = sctp_rto_min;
3080 /* Initialize default association related parameters. These parameters
3081 * can be modified with the SCTP_ASSOCINFO socket option.
3083 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3084 sp->assocparams.sasoc_number_peer_destinations = 0;
3085 sp->assocparams.sasoc_peer_rwnd = 0;
3086 sp->assocparams.sasoc_local_rwnd = 0;
3087 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3089 /* Initialize default event subscriptions. By default, all the
3090 * options are off.
3092 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3094 /* Default Peer Address Parameters. These defaults can
3095 * be modified via SCTP_PEER_ADDR_PARAMS
3097 sp->hbinterval = sctp_hb_interval;
3098 sp->pathmaxrxt = sctp_max_retrans_path;
3099 sp->pathmtu = 0; // allow default discovery
3100 sp->sackdelay = sctp_sack_timeout;
3101 sp->param_flags = SPP_HB_ENABLE |
3102 SPP_PMTUD_ENABLE |
3103 SPP_SACKDELAY_ENABLE;
3105 /* If enabled no SCTP message fragmentation will be performed.
3106 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3108 sp->disable_fragments = 0;
3110 /* Enable Nagle algorithm by default. */
3111 sp->nodelay = 0;
3113 /* Enable by default. */
3114 sp->v4mapped = 1;
3116 /* Auto-close idle associations after the configured
3117 * number of seconds. A value of 0 disables this
3118 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3119 * for UDP-style sockets only.
3121 sp->autoclose = 0;
3123 /* User specified fragmentation limit. */
3124 sp->user_frag = 0;
3126 sp->adaptation_ind = 0;
3128 sp->pf = sctp_get_pf_specific(sk->sk_family);
3130 /* Control variables for partial data delivery. */
3131 sp->pd_mode = 0;
3132 skb_queue_head_init(&sp->pd_lobby);
3134 /* Create a per socket endpoint structure. Even if we
3135 * change the data structure relationships, this may still
3136 * be useful for storing pre-connect address information.
3138 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3139 if (!ep)
3140 return -ENOMEM;
3142 sp->ep = ep;
3143 sp->hmac = NULL;
3145 SCTP_DBG_OBJCNT_INC(sock);
3146 return 0;
3149 /* Cleanup any SCTP per socket resources. */
3150 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3152 struct sctp_endpoint *ep;
3154 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3156 /* Release our hold on the endpoint. */
3157 ep = sctp_sk(sk)->ep;
3158 sctp_endpoint_free(ep);
3160 return 0;
3163 /* API 4.1.7 shutdown() - TCP Style Syntax
3164 * int shutdown(int socket, int how);
3166 * sd - the socket descriptor of the association to be closed.
3167 * how - Specifies the type of shutdown. The values are
3168 * as follows:
3169 * SHUT_RD
3170 * Disables further receive operations. No SCTP
3171 * protocol action is taken.
3172 * SHUT_WR
3173 * Disables further send operations, and initiates
3174 * the SCTP shutdown sequence.
3175 * SHUT_RDWR
3176 * Disables further send and receive operations
3177 * and initiates the SCTP shutdown sequence.
3179 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3181 struct sctp_endpoint *ep;
3182 struct sctp_association *asoc;
3184 if (!sctp_style(sk, TCP))
3185 return;
3187 if (how & SEND_SHUTDOWN) {
3188 ep = sctp_sk(sk)->ep;
3189 if (!list_empty(&ep->asocs)) {
3190 asoc = list_entry(ep->asocs.next,
3191 struct sctp_association, asocs);
3192 sctp_primitive_SHUTDOWN(asoc, NULL);
3197 /* 7.2.1 Association Status (SCTP_STATUS)
3199 * Applications can retrieve current status information about an
3200 * association, including association state, peer receiver window size,
3201 * number of unacked data chunks, and number of data chunks pending
3202 * receipt. This information is read-only.
3204 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3205 char __user *optval,
3206 int __user *optlen)
3208 struct sctp_status status;
3209 struct sctp_association *asoc = NULL;
3210 struct sctp_transport *transport;
3211 sctp_assoc_t associd;
3212 int retval = 0;
3214 if (len != sizeof(status)) {
3215 retval = -EINVAL;
3216 goto out;
3219 if (copy_from_user(&status, optval, sizeof(status))) {
3220 retval = -EFAULT;
3221 goto out;
3224 associd = status.sstat_assoc_id;
3225 asoc = sctp_id2assoc(sk, associd);
3226 if (!asoc) {
3227 retval = -EINVAL;
3228 goto out;
3231 transport = asoc->peer.primary_path;
3233 status.sstat_assoc_id = sctp_assoc2id(asoc);
3234 status.sstat_state = asoc->state;
3235 status.sstat_rwnd = asoc->peer.rwnd;
3236 status.sstat_unackdata = asoc->unack_data;
3238 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3239 status.sstat_instrms = asoc->c.sinit_max_instreams;
3240 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3241 status.sstat_fragmentation_point = asoc->frag_point;
3242 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3243 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3244 transport->af_specific->sockaddr_len);
3245 /* Map ipv4 address into v4-mapped-on-v6 address. */
3246 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3247 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3248 status.sstat_primary.spinfo_state = transport->state;
3249 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3250 status.sstat_primary.spinfo_srtt = transport->srtt;
3251 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3252 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3254 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3255 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3257 if (put_user(len, optlen)) {
3258 retval = -EFAULT;
3259 goto out;
3262 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3263 len, status.sstat_state, status.sstat_rwnd,
3264 status.sstat_assoc_id);
3266 if (copy_to_user(optval, &status, len)) {
3267 retval = -EFAULT;
3268 goto out;
3271 out:
3272 return (retval);
3276 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3278 * Applications can retrieve information about a specific peer address
3279 * of an association, including its reachability state, congestion
3280 * window, and retransmission timer values. This information is
3281 * read-only.
3283 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3284 char __user *optval,
3285 int __user *optlen)
3287 struct sctp_paddrinfo pinfo;
3288 struct sctp_transport *transport;
3289 int retval = 0;
3291 if (len != sizeof(pinfo)) {
3292 retval = -EINVAL;
3293 goto out;
3296 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3297 retval = -EFAULT;
3298 goto out;
3301 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3302 pinfo.spinfo_assoc_id);
3303 if (!transport)
3304 return -EINVAL;
3306 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3307 pinfo.spinfo_state = transport->state;
3308 pinfo.spinfo_cwnd = transport->cwnd;
3309 pinfo.spinfo_srtt = transport->srtt;
3310 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3311 pinfo.spinfo_mtu = transport->pathmtu;
3313 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3314 pinfo.spinfo_state = SCTP_ACTIVE;
3316 if (put_user(len, optlen)) {
3317 retval = -EFAULT;
3318 goto out;
3321 if (copy_to_user(optval, &pinfo, len)) {
3322 retval = -EFAULT;
3323 goto out;
3326 out:
3327 return (retval);
3330 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3332 * This option is a on/off flag. If enabled no SCTP message
3333 * fragmentation will be performed. Instead if a message being sent
3334 * exceeds the current PMTU size, the message will NOT be sent and
3335 * instead a error will be indicated to the user.
3337 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3338 char __user *optval, int __user *optlen)
3340 int val;
3342 if (len < sizeof(int))
3343 return -EINVAL;
3345 len = sizeof(int);
3346 val = (sctp_sk(sk)->disable_fragments == 1);
3347 if (put_user(len, optlen))
3348 return -EFAULT;
3349 if (copy_to_user(optval, &val, len))
3350 return -EFAULT;
3351 return 0;
3354 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3356 * This socket option is used to specify various notifications and
3357 * ancillary data the user wishes to receive.
3359 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3360 int __user *optlen)
3362 if (len != sizeof(struct sctp_event_subscribe))
3363 return -EINVAL;
3364 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3365 return -EFAULT;
3366 return 0;
3369 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3371 * This socket option is applicable to the UDP-style socket only. When
3372 * set it will cause associations that are idle for more than the
3373 * specified number of seconds to automatically close. An association
3374 * being idle is defined an association that has NOT sent or received
3375 * user data. The special value of '0' indicates that no automatic
3376 * close of any associations should be performed. The option expects an
3377 * integer defining the number of seconds of idle time before an
3378 * association is closed.
3380 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3382 /* Applicable to UDP-style socket only */
3383 if (sctp_style(sk, TCP))
3384 return -EOPNOTSUPP;
3385 if (len != sizeof(int))
3386 return -EINVAL;
3387 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3388 return -EFAULT;
3389 return 0;
3392 /* Helper routine to branch off an association to a new socket. */
3393 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3394 struct socket **sockp)
3396 struct sock *sk = asoc->base.sk;
3397 struct socket *sock;
3398 struct inet_sock *inetsk;
3399 int err = 0;
3401 /* An association cannot be branched off from an already peeled-off
3402 * socket, nor is this supported for tcp style sockets.
3404 if (!sctp_style(sk, UDP))
3405 return -EINVAL;
3407 /* Create a new socket. */
3408 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3409 if (err < 0)
3410 return err;
3412 /* Populate the fields of the newsk from the oldsk and migrate the
3413 * asoc to the newsk.
3415 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3417 /* Make peeled-off sockets more like 1-1 accepted sockets.
3418 * Set the daddr and initialize id to something more random
3420 inetsk = inet_sk(sock->sk);
3421 inetsk->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
3422 inetsk->id = asoc->next_tsn ^ jiffies;
3424 *sockp = sock;
3426 return err;
3429 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3431 sctp_peeloff_arg_t peeloff;
3432 struct socket *newsock;
3433 int retval = 0;
3434 struct sctp_association *asoc;
3436 if (len != sizeof(sctp_peeloff_arg_t))
3437 return -EINVAL;
3438 if (copy_from_user(&peeloff, optval, len))
3439 return -EFAULT;
3441 asoc = sctp_id2assoc(sk, peeloff.associd);
3442 if (!asoc) {
3443 retval = -EINVAL;
3444 goto out;
3447 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3449 retval = sctp_do_peeloff(asoc, &newsock);
3450 if (retval < 0)
3451 goto out;
3453 /* Map the socket to an unused fd that can be returned to the user. */
3454 retval = sock_map_fd(newsock);
3455 if (retval < 0) {
3456 sock_release(newsock);
3457 goto out;
3460 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3461 __FUNCTION__, sk, asoc, newsock->sk, retval);
3463 /* Return the fd mapped to the new socket. */
3464 peeloff.sd = retval;
3465 if (copy_to_user(optval, &peeloff, len))
3466 retval = -EFAULT;
3468 out:
3469 return retval;
3472 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3474 * Applications can enable or disable heartbeats for any peer address of
3475 * an association, modify an address's heartbeat interval, force a
3476 * heartbeat to be sent immediately, and adjust the address's maximum
3477 * number of retransmissions sent before an address is considered
3478 * unreachable. The following structure is used to access and modify an
3479 * address's parameters:
3481 * struct sctp_paddrparams {
3482 * sctp_assoc_t spp_assoc_id;
3483 * struct sockaddr_storage spp_address;
3484 * uint32_t spp_hbinterval;
3485 * uint16_t spp_pathmaxrxt;
3486 * uint32_t spp_pathmtu;
3487 * uint32_t spp_sackdelay;
3488 * uint32_t spp_flags;
3489 * };
3491 * spp_assoc_id - (one-to-many style socket) This is filled in the
3492 * application, and identifies the association for
3493 * this query.
3494 * spp_address - This specifies which address is of interest.
3495 * spp_hbinterval - This contains the value of the heartbeat interval,
3496 * in milliseconds. If a value of zero
3497 * is present in this field then no changes are to
3498 * be made to this parameter.
3499 * spp_pathmaxrxt - This contains the maximum number of
3500 * retransmissions before this address shall be
3501 * considered unreachable. If a value of zero
3502 * is present in this field then no changes are to
3503 * be made to this parameter.
3504 * spp_pathmtu - When Path MTU discovery is disabled the value
3505 * specified here will be the "fixed" path mtu.
3506 * Note that if the spp_address field is empty
3507 * then all associations on this address will
3508 * have this fixed path mtu set upon them.
3510 * spp_sackdelay - When delayed sack is enabled, this value specifies
3511 * the number of milliseconds that sacks will be delayed
3512 * for. This value will apply to all addresses of an
3513 * association if the spp_address field is empty. Note
3514 * also, that if delayed sack is enabled and this
3515 * value is set to 0, no change is made to the last
3516 * recorded delayed sack timer value.
3518 * spp_flags - These flags are used to control various features
3519 * on an association. The flag field may contain
3520 * zero or more of the following options.
3522 * SPP_HB_ENABLE - Enable heartbeats on the
3523 * specified address. Note that if the address
3524 * field is empty all addresses for the association
3525 * have heartbeats enabled upon them.
3527 * SPP_HB_DISABLE - Disable heartbeats on the
3528 * speicifed address. Note that if the address
3529 * field is empty all addresses for the association
3530 * will have their heartbeats disabled. Note also
3531 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3532 * mutually exclusive, only one of these two should
3533 * be specified. Enabling both fields will have
3534 * undetermined results.
3536 * SPP_HB_DEMAND - Request a user initiated heartbeat
3537 * to be made immediately.
3539 * SPP_PMTUD_ENABLE - This field will enable PMTU
3540 * discovery upon the specified address. Note that
3541 * if the address feild is empty then all addresses
3542 * on the association are effected.
3544 * SPP_PMTUD_DISABLE - This field will disable PMTU
3545 * discovery upon the specified address. Note that
3546 * if the address feild is empty then all addresses
3547 * on the association are effected. Not also that
3548 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3549 * exclusive. Enabling both will have undetermined
3550 * results.
3552 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3553 * on delayed sack. The time specified in spp_sackdelay
3554 * is used to specify the sack delay for this address. Note
3555 * that if spp_address is empty then all addresses will
3556 * enable delayed sack and take on the sack delay
3557 * value specified in spp_sackdelay.
3558 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3559 * off delayed sack. If the spp_address field is blank then
3560 * delayed sack is disabled for the entire association. Note
3561 * also that this field is mutually exclusive to
3562 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3563 * results.
3565 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3566 char __user *optval, int __user *optlen)
3568 struct sctp_paddrparams params;
3569 struct sctp_transport *trans = NULL;
3570 struct sctp_association *asoc = NULL;
3571 struct sctp_sock *sp = sctp_sk(sk);
3573 if (len != sizeof(struct sctp_paddrparams))
3574 return -EINVAL;
3576 if (copy_from_user(&params, optval, len))
3577 return -EFAULT;
3579 /* If an address other than INADDR_ANY is specified, and
3580 * no transport is found, then the request is invalid.
3582 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3583 trans = sctp_addr_id2transport(sk, &params.spp_address,
3584 params.spp_assoc_id);
3585 if (!trans) {
3586 SCTP_DEBUG_PRINTK("Failed no transport\n");
3587 return -EINVAL;
3591 /* Get association, if assoc_id != 0 and the socket is a one
3592 * to many style socket, and an association was not found, then
3593 * the id was invalid.
3595 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3596 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3597 SCTP_DEBUG_PRINTK("Failed no association\n");
3598 return -EINVAL;
3601 if (trans) {
3602 /* Fetch transport values. */
3603 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3604 params.spp_pathmtu = trans->pathmtu;
3605 params.spp_pathmaxrxt = trans->pathmaxrxt;
3606 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3608 /*draft-11 doesn't say what to return in spp_flags*/
3609 params.spp_flags = trans->param_flags;
3610 } else if (asoc) {
3611 /* Fetch association values. */
3612 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3613 params.spp_pathmtu = asoc->pathmtu;
3614 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3615 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3617 /*draft-11 doesn't say what to return in spp_flags*/
3618 params.spp_flags = asoc->param_flags;
3619 } else {
3620 /* Fetch socket values. */
3621 params.spp_hbinterval = sp->hbinterval;
3622 params.spp_pathmtu = sp->pathmtu;
3623 params.spp_sackdelay = sp->sackdelay;
3624 params.spp_pathmaxrxt = sp->pathmaxrxt;
3626 /*draft-11 doesn't say what to return in spp_flags*/
3627 params.spp_flags = sp->param_flags;
3630 if (copy_to_user(optval, &params, len))
3631 return -EFAULT;
3633 if (put_user(len, optlen))
3634 return -EFAULT;
3636 return 0;
3639 /* 7.1.24. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3641 * This options will get or set the delayed ack timer. The time is set
3642 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3643 * endpoints default delayed ack timer value. If the assoc_id field is
3644 * non-zero, then the set or get effects the specified association.
3646 * struct sctp_assoc_value {
3647 * sctp_assoc_t assoc_id;
3648 * uint32_t assoc_value;
3649 * };
3651 * assoc_id - This parameter, indicates which association the
3652 * user is preforming an action upon. Note that if
3653 * this field's value is zero then the endpoints
3654 * default value is changed (effecting future
3655 * associations only).
3657 * assoc_value - This parameter contains the number of milliseconds
3658 * that the user is requesting the delayed ACK timer
3659 * be set to. Note that this value is defined in
3660 * the standard to be between 200 and 500 milliseconds.
3662 * Note: a value of zero will leave the value alone,
3663 * but disable SACK delay. A non-zero value will also
3664 * enable SACK delay.
3666 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3667 char __user *optval,
3668 int __user *optlen)
3670 struct sctp_assoc_value params;
3671 struct sctp_association *asoc = NULL;
3672 struct sctp_sock *sp = sctp_sk(sk);
3674 if (len != sizeof(struct sctp_assoc_value))
3675 return - EINVAL;
3677 if (copy_from_user(&params, optval, len))
3678 return -EFAULT;
3680 /* Get association, if assoc_id != 0 and the socket is a one
3681 * to many style socket, and an association was not found, then
3682 * the id was invalid.
3684 asoc = sctp_id2assoc(sk, params.assoc_id);
3685 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3686 return -EINVAL;
3688 if (asoc) {
3689 /* Fetch association values. */
3690 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3691 params.assoc_value = jiffies_to_msecs(
3692 asoc->sackdelay);
3693 else
3694 params.assoc_value = 0;
3695 } else {
3696 /* Fetch socket values. */
3697 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3698 params.assoc_value = sp->sackdelay;
3699 else
3700 params.assoc_value = 0;
3703 if (copy_to_user(optval, &params, len))
3704 return -EFAULT;
3706 if (put_user(len, optlen))
3707 return -EFAULT;
3709 return 0;
3712 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3714 * Applications can specify protocol parameters for the default association
3715 * initialization. The option name argument to setsockopt() and getsockopt()
3716 * is SCTP_INITMSG.
3718 * Setting initialization parameters is effective only on an unconnected
3719 * socket (for UDP-style sockets only future associations are effected
3720 * by the change). With TCP-style sockets, this option is inherited by
3721 * sockets derived from a listener socket.
3723 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3725 if (len != sizeof(struct sctp_initmsg))
3726 return -EINVAL;
3727 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3728 return -EFAULT;
3729 return 0;
3732 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3733 char __user *optval,
3734 int __user *optlen)
3736 sctp_assoc_t id;
3737 struct sctp_association *asoc;
3738 struct list_head *pos;
3739 int cnt = 0;
3741 if (len != sizeof(sctp_assoc_t))
3742 return -EINVAL;
3744 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3745 return -EFAULT;
3747 /* For UDP-style sockets, id specifies the association to query. */
3748 asoc = sctp_id2assoc(sk, id);
3749 if (!asoc)
3750 return -EINVAL;
3752 list_for_each(pos, &asoc->peer.transport_addr_list) {
3753 cnt ++;
3756 return cnt;
3760 * Old API for getting list of peer addresses. Does not work for 32-bit
3761 * programs running on a 64-bit kernel
3763 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3764 char __user *optval,
3765 int __user *optlen)
3767 struct sctp_association *asoc;
3768 struct list_head *pos;
3769 int cnt = 0;
3770 struct sctp_getaddrs_old getaddrs;
3771 struct sctp_transport *from;
3772 void __user *to;
3773 union sctp_addr temp;
3774 struct sctp_sock *sp = sctp_sk(sk);
3775 int addrlen;
3777 if (len != sizeof(struct sctp_getaddrs_old))
3778 return -EINVAL;
3780 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3781 return -EFAULT;
3783 if (getaddrs.addr_num <= 0) return -EINVAL;
3785 /* For UDP-style sockets, id specifies the association to query. */
3786 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3787 if (!asoc)
3788 return -EINVAL;
3790 to = (void __user *)getaddrs.addrs;
3791 list_for_each(pos, &asoc->peer.transport_addr_list) {
3792 from = list_entry(pos, struct sctp_transport, transports);
3793 memcpy(&temp, &from->ipaddr, sizeof(temp));
3794 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3795 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3796 if (copy_to_user(to, &temp, addrlen))
3797 return -EFAULT;
3798 to += addrlen ;
3799 cnt ++;
3800 if (cnt >= getaddrs.addr_num) break;
3802 getaddrs.addr_num = cnt;
3803 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3804 return -EFAULT;
3806 return 0;
3809 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3810 char __user *optval, int __user *optlen)
3812 struct sctp_association *asoc;
3813 struct list_head *pos;
3814 int cnt = 0;
3815 struct sctp_getaddrs getaddrs;
3816 struct sctp_transport *from;
3817 void __user *to;
3818 union sctp_addr temp;
3819 struct sctp_sock *sp = sctp_sk(sk);
3820 int addrlen;
3821 size_t space_left;
3822 int bytes_copied;
3824 if (len < sizeof(struct sctp_getaddrs))
3825 return -EINVAL;
3827 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3828 return -EFAULT;
3830 /* For UDP-style sockets, id specifies the association to query. */
3831 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3832 if (!asoc)
3833 return -EINVAL;
3835 to = optval + offsetof(struct sctp_getaddrs,addrs);
3836 space_left = len - sizeof(struct sctp_getaddrs) -
3837 offsetof(struct sctp_getaddrs,addrs);
3839 list_for_each(pos, &asoc->peer.transport_addr_list) {
3840 from = list_entry(pos, struct sctp_transport, transports);
3841 memcpy(&temp, &from->ipaddr, sizeof(temp));
3842 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3843 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3844 if(space_left < addrlen)
3845 return -ENOMEM;
3846 if (copy_to_user(to, &temp, addrlen))
3847 return -EFAULT;
3848 to += addrlen;
3849 cnt++;
3850 space_left -= addrlen;
3853 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3854 return -EFAULT;
3855 bytes_copied = ((char __user *)to) - optval;
3856 if (put_user(bytes_copied, optlen))
3857 return -EFAULT;
3859 return 0;
3862 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3863 char __user *optval,
3864 int __user *optlen)
3866 sctp_assoc_t id;
3867 struct sctp_bind_addr *bp;
3868 struct sctp_association *asoc;
3869 struct list_head *pos, *temp;
3870 struct sctp_sockaddr_entry *addr;
3871 rwlock_t *addr_lock;
3872 int cnt = 0;
3874 if (len != sizeof(sctp_assoc_t))
3875 return -EINVAL;
3877 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3878 return -EFAULT;
3881 * For UDP-style sockets, id specifies the association to query.
3882 * If the id field is set to the value '0' then the locally bound
3883 * addresses are returned without regard to any particular
3884 * association.
3886 if (0 == id) {
3887 bp = &sctp_sk(sk)->ep->base.bind_addr;
3888 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3889 } else {
3890 asoc = sctp_id2assoc(sk, id);
3891 if (!asoc)
3892 return -EINVAL;
3893 bp = &asoc->base.bind_addr;
3894 addr_lock = &asoc->base.addr_lock;
3897 sctp_read_lock(addr_lock);
3899 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3900 * addresses from the global local address list.
3902 if (sctp_list_single_entry(&bp->address_list)) {
3903 addr = list_entry(bp->address_list.next,
3904 struct sctp_sockaddr_entry, list);
3905 if (sctp_is_any(&addr->a)) {
3906 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
3907 addr = list_entry(pos,
3908 struct sctp_sockaddr_entry,
3909 list);
3910 if ((PF_INET == sk->sk_family) &&
3911 (AF_INET6 == addr->a.sa.sa_family))
3912 continue;
3913 cnt++;
3915 } else {
3916 cnt = 1;
3918 goto done;
3921 list_for_each(pos, &bp->address_list) {
3922 cnt ++;
3925 done:
3926 sctp_read_unlock(addr_lock);
3927 return cnt;
3930 /* Helper function that copies local addresses to user and returns the number
3931 * of addresses copied.
3933 static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3934 void __user *to)
3936 struct list_head *pos, *next;
3937 struct sctp_sockaddr_entry *addr;
3938 union sctp_addr temp;
3939 int cnt = 0;
3940 int addrlen;
3942 list_for_each_safe(pos, next, &sctp_local_addr_list) {
3943 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3944 if ((PF_INET == sk->sk_family) &&
3945 (AF_INET6 == addr->a.sa.sa_family))
3946 continue;
3947 memcpy(&temp, &addr->a, sizeof(temp));
3948 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3949 &temp);
3950 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3951 if (copy_to_user(to, &temp, addrlen))
3952 return -EFAULT;
3954 to += addrlen;
3955 cnt ++;
3956 if (cnt >= max_addrs) break;
3959 return cnt;
3962 static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
3963 void __user **to, size_t space_left)
3965 struct list_head *pos, *next;
3966 struct sctp_sockaddr_entry *addr;
3967 union sctp_addr temp;
3968 int cnt = 0;
3969 int addrlen;
3971 list_for_each_safe(pos, next, &sctp_local_addr_list) {
3972 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3973 if ((PF_INET == sk->sk_family) &&
3974 (AF_INET6 == addr->a.sa.sa_family))
3975 continue;
3976 memcpy(&temp, &addr->a, sizeof(temp));
3977 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3978 &temp);
3979 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
3980 if(space_left<addrlen)
3981 return -ENOMEM;
3982 if (copy_to_user(*to, &temp, addrlen))
3983 return -EFAULT;
3985 *to += addrlen;
3986 cnt ++;
3987 space_left -= addrlen;
3990 return cnt;
3993 /* Old API for getting list of local addresses. Does not work for 32-bit
3994 * programs running on a 64-bit kernel
3996 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
3997 char __user *optval, int __user *optlen)
3999 struct sctp_bind_addr *bp;
4000 struct sctp_association *asoc;
4001 struct list_head *pos;
4002 int cnt = 0;
4003 struct sctp_getaddrs_old getaddrs;
4004 struct sctp_sockaddr_entry *addr;
4005 void __user *to;
4006 union sctp_addr temp;
4007 struct sctp_sock *sp = sctp_sk(sk);
4008 int addrlen;
4009 rwlock_t *addr_lock;
4010 int err = 0;
4012 if (len != sizeof(struct sctp_getaddrs_old))
4013 return -EINVAL;
4015 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
4016 return -EFAULT;
4018 if (getaddrs.addr_num <= 0) return -EINVAL;
4020 * For UDP-style sockets, id specifies the association to query.
4021 * If the id field is set to the value '0' then the locally bound
4022 * addresses are returned without regard to any particular
4023 * association.
4025 if (0 == getaddrs.assoc_id) {
4026 bp = &sctp_sk(sk)->ep->base.bind_addr;
4027 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4028 } else {
4029 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4030 if (!asoc)
4031 return -EINVAL;
4032 bp = &asoc->base.bind_addr;
4033 addr_lock = &asoc->base.addr_lock;
4036 to = getaddrs.addrs;
4038 sctp_read_lock(addr_lock);
4040 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4041 * addresses from the global local address list.
4043 if (sctp_list_single_entry(&bp->address_list)) {
4044 addr = list_entry(bp->address_list.next,
4045 struct sctp_sockaddr_entry, list);
4046 if (sctp_is_any(&addr->a)) {
4047 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
4048 getaddrs.addr_num,
4049 to);
4050 if (cnt < 0) {
4051 err = cnt;
4052 goto unlock;
4054 goto copy_getaddrs;
4058 list_for_each(pos, &bp->address_list) {
4059 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4060 memcpy(&temp, &addr->a, sizeof(temp));
4061 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4062 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4063 if (copy_to_user(to, &temp, addrlen)) {
4064 err = -EFAULT;
4065 goto unlock;
4067 to += addrlen;
4068 cnt ++;
4069 if (cnt >= getaddrs.addr_num) break;
4072 copy_getaddrs:
4073 getaddrs.addr_num = cnt;
4074 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
4075 err = -EFAULT;
4077 unlock:
4078 sctp_read_unlock(addr_lock);
4079 return err;
4082 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4083 char __user *optval, int __user *optlen)
4085 struct sctp_bind_addr *bp;
4086 struct sctp_association *asoc;
4087 struct list_head *pos;
4088 int cnt = 0;
4089 struct sctp_getaddrs getaddrs;
4090 struct sctp_sockaddr_entry *addr;
4091 void __user *to;
4092 union sctp_addr temp;
4093 struct sctp_sock *sp = sctp_sk(sk);
4094 int addrlen;
4095 rwlock_t *addr_lock;
4096 int err = 0;
4097 size_t space_left;
4098 int bytes_copied;
4100 if (len <= sizeof(struct sctp_getaddrs))
4101 return -EINVAL;
4103 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4104 return -EFAULT;
4107 * For UDP-style sockets, id specifies the association to query.
4108 * If the id field is set to the value '0' then the locally bound
4109 * addresses are returned without regard to any particular
4110 * association.
4112 if (0 == getaddrs.assoc_id) {
4113 bp = &sctp_sk(sk)->ep->base.bind_addr;
4114 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4115 } else {
4116 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4117 if (!asoc)
4118 return -EINVAL;
4119 bp = &asoc->base.bind_addr;
4120 addr_lock = &asoc->base.addr_lock;
4123 to = optval + offsetof(struct sctp_getaddrs,addrs);
4124 space_left = len - sizeof(struct sctp_getaddrs) -
4125 offsetof(struct sctp_getaddrs,addrs);
4127 sctp_read_lock(addr_lock);
4129 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4130 * addresses from the global local address list.
4132 if (sctp_list_single_entry(&bp->address_list)) {
4133 addr = list_entry(bp->address_list.next,
4134 struct sctp_sockaddr_entry, list);
4135 if (sctp_is_any(&addr->a)) {
4136 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4137 &to, space_left);
4138 if (cnt < 0) {
4139 err = cnt;
4140 goto unlock;
4142 goto copy_getaddrs;
4146 list_for_each(pos, &bp->address_list) {
4147 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4148 memcpy(&temp, &addr->a, sizeof(temp));
4149 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4150 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4151 if(space_left < addrlen)
4152 return -ENOMEM; /*fixme: right error?*/
4153 if (copy_to_user(to, &temp, addrlen)) {
4154 err = -EFAULT;
4155 goto unlock;
4157 to += addrlen;
4158 cnt ++;
4159 space_left -= addrlen;
4162 copy_getaddrs:
4163 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4164 return -EFAULT;
4165 bytes_copied = ((char __user *)to) - optval;
4166 if (put_user(bytes_copied, optlen))
4167 return -EFAULT;
4169 unlock:
4170 sctp_read_unlock(addr_lock);
4171 return err;
4174 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4176 * Requests that the local SCTP stack use the enclosed peer address as
4177 * the association primary. The enclosed address must be one of the
4178 * association peer's addresses.
4180 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4181 char __user *optval, int __user *optlen)
4183 struct sctp_prim prim;
4184 struct sctp_association *asoc;
4185 struct sctp_sock *sp = sctp_sk(sk);
4187 if (len != sizeof(struct sctp_prim))
4188 return -EINVAL;
4190 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4191 return -EFAULT;
4193 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4194 if (!asoc)
4195 return -EINVAL;
4197 if (!asoc->peer.primary_path)
4198 return -ENOTCONN;
4200 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4201 asoc->peer.primary_path->af_specific->sockaddr_len);
4203 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4204 (union sctp_addr *)&prim.ssp_addr);
4206 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4207 return -EFAULT;
4209 return 0;
4213 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4215 * Requests that the local endpoint set the specified Adaptation Layer
4216 * Indication parameter for all future INIT and INIT-ACK exchanges.
4218 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4219 char __user *optval, int __user *optlen)
4221 struct sctp_setadaptation adaptation;
4223 if (len != sizeof(struct sctp_setadaptation))
4224 return -EINVAL;
4226 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4227 if (copy_to_user(optval, &adaptation, len))
4228 return -EFAULT;
4230 return 0;
4235 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4237 * Applications that wish to use the sendto() system call may wish to
4238 * specify a default set of parameters that would normally be supplied
4239 * through the inclusion of ancillary data. This socket option allows
4240 * such an application to set the default sctp_sndrcvinfo structure.
4243 * The application that wishes to use this socket option simply passes
4244 * in to this call the sctp_sndrcvinfo structure defined in Section
4245 * 5.2.2) The input parameters accepted by this call include
4246 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4247 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4248 * to this call if the caller is using the UDP model.
4250 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4252 static int sctp_getsockopt_default_send_param(struct sock *sk,
4253 int len, char __user *optval,
4254 int __user *optlen)
4256 struct sctp_sndrcvinfo info;
4257 struct sctp_association *asoc;
4258 struct sctp_sock *sp = sctp_sk(sk);
4260 if (len != sizeof(struct sctp_sndrcvinfo))
4261 return -EINVAL;
4262 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4263 return -EFAULT;
4265 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4266 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4267 return -EINVAL;
4269 if (asoc) {
4270 info.sinfo_stream = asoc->default_stream;
4271 info.sinfo_flags = asoc->default_flags;
4272 info.sinfo_ppid = asoc->default_ppid;
4273 info.sinfo_context = asoc->default_context;
4274 info.sinfo_timetolive = asoc->default_timetolive;
4275 } else {
4276 info.sinfo_stream = sp->default_stream;
4277 info.sinfo_flags = sp->default_flags;
4278 info.sinfo_ppid = sp->default_ppid;
4279 info.sinfo_context = sp->default_context;
4280 info.sinfo_timetolive = sp->default_timetolive;
4283 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4284 return -EFAULT;
4286 return 0;
4291 * 7.1.5 SCTP_NODELAY
4293 * Turn on/off any Nagle-like algorithm. This means that packets are
4294 * generally sent as soon as possible and no unnecessary delays are
4295 * introduced, at the cost of more packets in the network. Expects an
4296 * integer boolean flag.
4299 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4300 char __user *optval, int __user *optlen)
4302 int val;
4304 if (len < sizeof(int))
4305 return -EINVAL;
4307 len = sizeof(int);
4308 val = (sctp_sk(sk)->nodelay == 1);
4309 if (put_user(len, optlen))
4310 return -EFAULT;
4311 if (copy_to_user(optval, &val, len))
4312 return -EFAULT;
4313 return 0;
4318 * 7.1.1 SCTP_RTOINFO
4320 * The protocol parameters used to initialize and bound retransmission
4321 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4322 * and modify these parameters.
4323 * All parameters are time values, in milliseconds. A value of 0, when
4324 * modifying the parameters, indicates that the current value should not
4325 * be changed.
4328 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4329 char __user *optval,
4330 int __user *optlen) {
4331 struct sctp_rtoinfo rtoinfo;
4332 struct sctp_association *asoc;
4334 if (len != sizeof (struct sctp_rtoinfo))
4335 return -EINVAL;
4337 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4338 return -EFAULT;
4340 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4342 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4343 return -EINVAL;
4345 /* Values corresponding to the specific association. */
4346 if (asoc) {
4347 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4348 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4349 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4350 } else {
4351 /* Values corresponding to the endpoint. */
4352 struct sctp_sock *sp = sctp_sk(sk);
4354 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4355 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4356 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4359 if (put_user(len, optlen))
4360 return -EFAULT;
4362 if (copy_to_user(optval, &rtoinfo, len))
4363 return -EFAULT;
4365 return 0;
4370 * 7.1.2 SCTP_ASSOCINFO
4372 * This option is used to tune the the maximum retransmission attempts
4373 * of the association.
4374 * Returns an error if the new association retransmission value is
4375 * greater than the sum of the retransmission value of the peer.
4376 * See [SCTP] for more information.
4379 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4380 char __user *optval,
4381 int __user *optlen)
4384 struct sctp_assocparams assocparams;
4385 struct sctp_association *asoc;
4386 struct list_head *pos;
4387 int cnt = 0;
4389 if (len != sizeof (struct sctp_assocparams))
4390 return -EINVAL;
4392 if (copy_from_user(&assocparams, optval,
4393 sizeof (struct sctp_assocparams)))
4394 return -EFAULT;
4396 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4398 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4399 return -EINVAL;
4401 /* Values correspoinding to the specific association */
4402 if (asoc) {
4403 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4404 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4405 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4406 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4407 * 1000) +
4408 (asoc->cookie_life.tv_usec
4409 / 1000);
4411 list_for_each(pos, &asoc->peer.transport_addr_list) {
4412 cnt ++;
4415 assocparams.sasoc_number_peer_destinations = cnt;
4416 } else {
4417 /* Values corresponding to the endpoint */
4418 struct sctp_sock *sp = sctp_sk(sk);
4420 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4421 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4422 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4423 assocparams.sasoc_cookie_life =
4424 sp->assocparams.sasoc_cookie_life;
4425 assocparams.sasoc_number_peer_destinations =
4426 sp->assocparams.
4427 sasoc_number_peer_destinations;
4430 if (put_user(len, optlen))
4431 return -EFAULT;
4433 if (copy_to_user(optval, &assocparams, len))
4434 return -EFAULT;
4436 return 0;
4440 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4442 * This socket option is a boolean flag which turns on or off mapped V4
4443 * addresses. If this option is turned on and the socket is type
4444 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4445 * If this option is turned off, then no mapping will be done of V4
4446 * addresses and a user will receive both PF_INET6 and PF_INET type
4447 * addresses on the socket.
4449 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4450 char __user *optval, int __user *optlen)
4452 int val;
4453 struct sctp_sock *sp = sctp_sk(sk);
4455 if (len < sizeof(int))
4456 return -EINVAL;
4458 len = sizeof(int);
4459 val = sp->v4mapped;
4460 if (put_user(len, optlen))
4461 return -EFAULT;
4462 if (copy_to_user(optval, &val, len))
4463 return -EFAULT;
4465 return 0;
4469 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4470 * (chapter and verse is quoted at sctp_setsockopt_context())
4472 static int sctp_getsockopt_context(struct sock *sk, int len,
4473 char __user *optval, int __user *optlen)
4475 struct sctp_assoc_value params;
4476 struct sctp_sock *sp;
4477 struct sctp_association *asoc;
4479 if (len != sizeof(struct sctp_assoc_value))
4480 return -EINVAL;
4482 if (copy_from_user(&params, optval, len))
4483 return -EFAULT;
4485 sp = sctp_sk(sk);
4487 if (params.assoc_id != 0) {
4488 asoc = sctp_id2assoc(sk, params.assoc_id);
4489 if (!asoc)
4490 return -EINVAL;
4491 params.assoc_value = asoc->default_rcv_context;
4492 } else {
4493 params.assoc_value = sp->default_rcv_context;
4496 if (put_user(len, optlen))
4497 return -EFAULT;
4498 if (copy_to_user(optval, &params, len))
4499 return -EFAULT;
4501 return 0;
4505 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4507 * This socket option specifies the maximum size to put in any outgoing
4508 * SCTP chunk. If a message is larger than this size it will be
4509 * fragmented by SCTP into the specified size. Note that the underlying
4510 * SCTP implementation may fragment into smaller sized chunks when the
4511 * PMTU of the underlying association is smaller than the value set by
4512 * the user.
4514 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4515 char __user *optval, int __user *optlen)
4517 int val;
4519 if (len < sizeof(int))
4520 return -EINVAL;
4522 len = sizeof(int);
4524 val = sctp_sk(sk)->user_frag;
4525 if (put_user(len, optlen))
4526 return -EFAULT;
4527 if (copy_to_user(optval, &val, len))
4528 return -EFAULT;
4530 return 0;
4533 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4534 char __user *optval, int __user *optlen)
4536 int retval = 0;
4537 int len;
4539 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4540 sk, optname);
4542 /* I can hardly begin to describe how wrong this is. This is
4543 * so broken as to be worse than useless. The API draft
4544 * REALLY is NOT helpful here... I am not convinced that the
4545 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4546 * are at all well-founded.
4548 if (level != SOL_SCTP) {
4549 struct sctp_af *af = sctp_sk(sk)->pf->af;
4551 retval = af->getsockopt(sk, level, optname, optval, optlen);
4552 return retval;
4555 if (get_user(len, optlen))
4556 return -EFAULT;
4558 sctp_lock_sock(sk);
4560 switch (optname) {
4561 case SCTP_STATUS:
4562 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4563 break;
4564 case SCTP_DISABLE_FRAGMENTS:
4565 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4566 optlen);
4567 break;
4568 case SCTP_EVENTS:
4569 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4570 break;
4571 case SCTP_AUTOCLOSE:
4572 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4573 break;
4574 case SCTP_SOCKOPT_PEELOFF:
4575 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4576 break;
4577 case SCTP_PEER_ADDR_PARAMS:
4578 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4579 optlen);
4580 break;
4581 case SCTP_DELAYED_ACK_TIME:
4582 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4583 optlen);
4584 break;
4585 case SCTP_INITMSG:
4586 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4587 break;
4588 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4589 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4590 optlen);
4591 break;
4592 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4593 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4594 optlen);
4595 break;
4596 case SCTP_GET_PEER_ADDRS_OLD:
4597 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4598 optlen);
4599 break;
4600 case SCTP_GET_LOCAL_ADDRS_OLD:
4601 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4602 optlen);
4603 break;
4604 case SCTP_GET_PEER_ADDRS:
4605 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4606 optlen);
4607 break;
4608 case SCTP_GET_LOCAL_ADDRS:
4609 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4610 optlen);
4611 break;
4612 case SCTP_DEFAULT_SEND_PARAM:
4613 retval = sctp_getsockopt_default_send_param(sk, len,
4614 optval, optlen);
4615 break;
4616 case SCTP_PRIMARY_ADDR:
4617 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4618 break;
4619 case SCTP_NODELAY:
4620 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4621 break;
4622 case SCTP_RTOINFO:
4623 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4624 break;
4625 case SCTP_ASSOCINFO:
4626 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4627 break;
4628 case SCTP_I_WANT_MAPPED_V4_ADDR:
4629 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4630 break;
4631 case SCTP_MAXSEG:
4632 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4633 break;
4634 case SCTP_GET_PEER_ADDR_INFO:
4635 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4636 optlen);
4637 break;
4638 case SCTP_ADAPTATION_LAYER:
4639 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
4640 optlen);
4641 break;
4642 case SCTP_CONTEXT:
4643 retval = sctp_getsockopt_context(sk, len, optval, optlen);
4644 break;
4645 default:
4646 retval = -ENOPROTOOPT;
4647 break;
4650 sctp_release_sock(sk);
4651 return retval;
4654 static void sctp_hash(struct sock *sk)
4656 /* STUB */
4659 static void sctp_unhash(struct sock *sk)
4661 /* STUB */
4664 /* Check if port is acceptable. Possibly find first available port.
4666 * The port hash table (contained in the 'global' SCTP protocol storage
4667 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4668 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4669 * list (the list number is the port number hashed out, so as you
4670 * would expect from a hash function, all the ports in a given list have
4671 * such a number that hashes out to the same list number; you were
4672 * expecting that, right?); so each list has a set of ports, with a
4673 * link to the socket (struct sock) that uses it, the port number and
4674 * a fastreuse flag (FIXME: NPI ipg).
4676 static struct sctp_bind_bucket *sctp_bucket_create(
4677 struct sctp_bind_hashbucket *head, unsigned short snum);
4679 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4681 struct sctp_bind_hashbucket *head; /* hash list */
4682 struct sctp_bind_bucket *pp; /* hash list port iterator */
4683 unsigned short snum;
4684 int ret;
4686 snum = ntohs(addr->v4.sin_port);
4688 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4689 sctp_local_bh_disable();
4691 if (snum == 0) {
4692 /* Search for an available port.
4694 * 'sctp_port_rover' was the last port assigned, so
4695 * we start to search from 'sctp_port_rover +
4696 * 1'. What we do is first check if port 'rover' is
4697 * already in the hash table; if not, we use that; if
4698 * it is, we try next.
4700 int low = sysctl_local_port_range[0];
4701 int high = sysctl_local_port_range[1];
4702 int remaining = (high - low) + 1;
4703 int rover;
4704 int index;
4706 sctp_spin_lock(&sctp_port_alloc_lock);
4707 rover = sctp_port_rover;
4708 do {
4709 rover++;
4710 if ((rover < low) || (rover > high))
4711 rover = low;
4712 index = sctp_phashfn(rover);
4713 head = &sctp_port_hashtable[index];
4714 sctp_spin_lock(&head->lock);
4715 for (pp = head->chain; pp; pp = pp->next)
4716 if (pp->port == rover)
4717 goto next;
4718 break;
4719 next:
4720 sctp_spin_unlock(&head->lock);
4721 } while (--remaining > 0);
4722 sctp_port_rover = rover;
4723 sctp_spin_unlock(&sctp_port_alloc_lock);
4725 /* Exhausted local port range during search? */
4726 ret = 1;
4727 if (remaining <= 0)
4728 goto fail;
4730 /* OK, here is the one we will use. HEAD (the port
4731 * hash table list entry) is non-NULL and we hold it's
4732 * mutex.
4734 snum = rover;
4735 } else {
4736 /* We are given an specific port number; we verify
4737 * that it is not being used. If it is used, we will
4738 * exahust the search in the hash list corresponding
4739 * to the port number (snum) - we detect that with the
4740 * port iterator, pp being NULL.
4742 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4743 sctp_spin_lock(&head->lock);
4744 for (pp = head->chain; pp; pp = pp->next) {
4745 if (pp->port == snum)
4746 goto pp_found;
4749 pp = NULL;
4750 goto pp_not_found;
4751 pp_found:
4752 if (!hlist_empty(&pp->owner)) {
4753 /* We had a port hash table hit - there is an
4754 * available port (pp != NULL) and it is being
4755 * used by other socket (pp->owner not empty); that other
4756 * socket is going to be sk2.
4758 int reuse = sk->sk_reuse;
4759 struct sock *sk2;
4760 struct hlist_node *node;
4762 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4763 if (pp->fastreuse && sk->sk_reuse)
4764 goto success;
4766 /* Run through the list of sockets bound to the port
4767 * (pp->port) [via the pointers bind_next and
4768 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4769 * we get the endpoint they describe and run through
4770 * the endpoint's list of IP (v4 or v6) addresses,
4771 * comparing each of the addresses with the address of
4772 * the socket sk. If we find a match, then that means
4773 * that this port/socket (sk) combination are already
4774 * in an endpoint.
4776 sk_for_each_bound(sk2, node, &pp->owner) {
4777 struct sctp_endpoint *ep2;
4778 ep2 = sctp_sk(sk2)->ep;
4780 if (reuse && sk2->sk_reuse)
4781 continue;
4783 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4784 sctp_sk(sk))) {
4785 ret = (long)sk2;
4786 goto fail_unlock;
4789 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4791 pp_not_found:
4792 /* If there was a hash table miss, create a new port. */
4793 ret = 1;
4794 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4795 goto fail_unlock;
4797 /* In either case (hit or miss), make sure fastreuse is 1 only
4798 * if sk->sk_reuse is too (that is, if the caller requested
4799 * SO_REUSEADDR on this socket -sk-).
4801 if (hlist_empty(&pp->owner))
4802 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4803 else if (pp->fastreuse && !sk->sk_reuse)
4804 pp->fastreuse = 0;
4806 /* We are set, so fill up all the data in the hash table
4807 * entry, tie the socket list information with the rest of the
4808 * sockets FIXME: Blurry, NPI (ipg).
4810 success:
4811 inet_sk(sk)->num = snum;
4812 if (!sctp_sk(sk)->bind_hash) {
4813 sk_add_bind_node(sk, &pp->owner);
4814 sctp_sk(sk)->bind_hash = pp;
4816 ret = 0;
4818 fail_unlock:
4819 sctp_spin_unlock(&head->lock);
4821 fail:
4822 sctp_local_bh_enable();
4823 return ret;
4826 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4827 * port is requested.
4829 static int sctp_get_port(struct sock *sk, unsigned short snum)
4831 long ret;
4832 union sctp_addr addr;
4833 struct sctp_af *af = sctp_sk(sk)->pf->af;
4835 /* Set up a dummy address struct from the sk. */
4836 af->from_sk(&addr, sk);
4837 addr.v4.sin_port = htons(snum);
4839 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4840 ret = sctp_get_port_local(sk, &addr);
4842 return (ret ? 1 : 0);
4846 * 3.1.3 listen() - UDP Style Syntax
4848 * By default, new associations are not accepted for UDP style sockets.
4849 * An application uses listen() to mark a socket as being able to
4850 * accept new associations.
4852 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4854 struct sctp_sock *sp = sctp_sk(sk);
4855 struct sctp_endpoint *ep = sp->ep;
4857 /* Only UDP style sockets that are not peeled off are allowed to
4858 * listen().
4860 if (!sctp_style(sk, UDP))
4861 return -EINVAL;
4863 /* If backlog is zero, disable listening. */
4864 if (!backlog) {
4865 if (sctp_sstate(sk, CLOSED))
4866 return 0;
4868 sctp_unhash_endpoint(ep);
4869 sk->sk_state = SCTP_SS_CLOSED;
4872 /* Return if we are already listening. */
4873 if (sctp_sstate(sk, LISTENING))
4874 return 0;
4877 * If a bind() or sctp_bindx() is not called prior to a listen()
4878 * call that allows new associations to be accepted, the system
4879 * picks an ephemeral port and will choose an address set equivalent
4880 * to binding with a wildcard address.
4882 * This is not currently spelled out in the SCTP sockets
4883 * extensions draft, but follows the practice as seen in TCP
4884 * sockets.
4886 if (!ep->base.bind_addr.port) {
4887 if (sctp_autobind(sk))
4888 return -EAGAIN;
4890 sk->sk_state = SCTP_SS_LISTENING;
4891 sctp_hash_endpoint(ep);
4892 return 0;
4896 * 4.1.3 listen() - TCP Style Syntax
4898 * Applications uses listen() to ready the SCTP endpoint for accepting
4899 * inbound associations.
4901 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4903 struct sctp_sock *sp = sctp_sk(sk);
4904 struct sctp_endpoint *ep = sp->ep;
4906 /* If backlog is zero, disable listening. */
4907 if (!backlog) {
4908 if (sctp_sstate(sk, CLOSED))
4909 return 0;
4911 sctp_unhash_endpoint(ep);
4912 sk->sk_state = SCTP_SS_CLOSED;
4915 if (sctp_sstate(sk, LISTENING))
4916 return 0;
4919 * If a bind() or sctp_bindx() is not called prior to a listen()
4920 * call that allows new associations to be accepted, the system
4921 * picks an ephemeral port and will choose an address set equivalent
4922 * to binding with a wildcard address.
4924 * This is not currently spelled out in the SCTP sockets
4925 * extensions draft, but follows the practice as seen in TCP
4926 * sockets.
4928 if (!ep->base.bind_addr.port) {
4929 if (sctp_autobind(sk))
4930 return -EAGAIN;
4932 sk->sk_state = SCTP_SS_LISTENING;
4933 sk->sk_max_ack_backlog = backlog;
4934 sctp_hash_endpoint(ep);
4935 return 0;
4939 * Move a socket to LISTENING state.
4941 int sctp_inet_listen(struct socket *sock, int backlog)
4943 struct sock *sk = sock->sk;
4944 struct crypto_hash *tfm = NULL;
4945 int err = -EINVAL;
4947 if (unlikely(backlog < 0))
4948 goto out;
4950 sctp_lock_sock(sk);
4952 if (sock->state != SS_UNCONNECTED)
4953 goto out;
4955 /* Allocate HMAC for generating cookie. */
4956 if (sctp_hmac_alg) {
4957 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
4958 if (!tfm) {
4959 err = -ENOSYS;
4960 goto out;
4964 switch (sock->type) {
4965 case SOCK_SEQPACKET:
4966 err = sctp_seqpacket_listen(sk, backlog);
4967 break;
4968 case SOCK_STREAM:
4969 err = sctp_stream_listen(sk, backlog);
4970 break;
4971 default:
4972 break;
4974 if (err)
4975 goto cleanup;
4977 /* Store away the transform reference. */
4978 sctp_sk(sk)->hmac = tfm;
4979 out:
4980 sctp_release_sock(sk);
4981 return err;
4982 cleanup:
4983 crypto_free_hash(tfm);
4984 goto out;
4988 * This function is done by modeling the current datagram_poll() and the
4989 * tcp_poll(). Note that, based on these implementations, we don't
4990 * lock the socket in this function, even though it seems that,
4991 * ideally, locking or some other mechanisms can be used to ensure
4992 * the integrity of the counters (sndbuf and wmem_alloc) used
4993 * in this place. We assume that we don't need locks either until proven
4994 * otherwise.
4996 * Another thing to note is that we include the Async I/O support
4997 * here, again, by modeling the current TCP/UDP code. We don't have
4998 * a good way to test with it yet.
5000 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5002 struct sock *sk = sock->sk;
5003 struct sctp_sock *sp = sctp_sk(sk);
5004 unsigned int mask;
5006 poll_wait(file, sk->sk_sleep, wait);
5008 /* A TCP-style listening socket becomes readable when the accept queue
5009 * is not empty.
5011 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5012 return (!list_empty(&sp->ep->asocs)) ?
5013 (POLLIN | POLLRDNORM) : 0;
5015 mask = 0;
5017 /* Is there any exceptional events? */
5018 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5019 mask |= POLLERR;
5020 if (sk->sk_shutdown & RCV_SHUTDOWN)
5021 mask |= POLLRDHUP;
5022 if (sk->sk_shutdown == SHUTDOWN_MASK)
5023 mask |= POLLHUP;
5025 /* Is it readable? Reconsider this code with TCP-style support. */
5026 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5027 (sk->sk_shutdown & RCV_SHUTDOWN))
5028 mask |= POLLIN | POLLRDNORM;
5030 /* The association is either gone or not ready. */
5031 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5032 return mask;
5034 /* Is it writable? */
5035 if (sctp_writeable(sk)) {
5036 mask |= POLLOUT | POLLWRNORM;
5037 } else {
5038 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5040 * Since the socket is not locked, the buffer
5041 * might be made available after the writeable check and
5042 * before the bit is set. This could cause a lost I/O
5043 * signal. tcp_poll() has a race breaker for this race
5044 * condition. Based on their implementation, we put
5045 * in the following code to cover it as well.
5047 if (sctp_writeable(sk))
5048 mask |= POLLOUT | POLLWRNORM;
5050 return mask;
5053 /********************************************************************
5054 * 2nd Level Abstractions
5055 ********************************************************************/
5057 static struct sctp_bind_bucket *sctp_bucket_create(
5058 struct sctp_bind_hashbucket *head, unsigned short snum)
5060 struct sctp_bind_bucket *pp;
5062 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5063 SCTP_DBG_OBJCNT_INC(bind_bucket);
5064 if (pp) {
5065 pp->port = snum;
5066 pp->fastreuse = 0;
5067 INIT_HLIST_HEAD(&pp->owner);
5068 if ((pp->next = head->chain) != NULL)
5069 pp->next->pprev = &pp->next;
5070 head->chain = pp;
5071 pp->pprev = &head->chain;
5073 return pp;
5076 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5077 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5079 if (pp && hlist_empty(&pp->owner)) {
5080 if (pp->next)
5081 pp->next->pprev = pp->pprev;
5082 *(pp->pprev) = pp->next;
5083 kmem_cache_free(sctp_bucket_cachep, pp);
5084 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5088 /* Release this socket's reference to a local port. */
5089 static inline void __sctp_put_port(struct sock *sk)
5091 struct sctp_bind_hashbucket *head =
5092 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5093 struct sctp_bind_bucket *pp;
5095 sctp_spin_lock(&head->lock);
5096 pp = sctp_sk(sk)->bind_hash;
5097 __sk_del_bind_node(sk);
5098 sctp_sk(sk)->bind_hash = NULL;
5099 inet_sk(sk)->num = 0;
5100 sctp_bucket_destroy(pp);
5101 sctp_spin_unlock(&head->lock);
5104 void sctp_put_port(struct sock *sk)
5106 sctp_local_bh_disable();
5107 __sctp_put_port(sk);
5108 sctp_local_bh_enable();
5112 * The system picks an ephemeral port and choose an address set equivalent
5113 * to binding with a wildcard address.
5114 * One of those addresses will be the primary address for the association.
5115 * This automatically enables the multihoming capability of SCTP.
5117 static int sctp_autobind(struct sock *sk)
5119 union sctp_addr autoaddr;
5120 struct sctp_af *af;
5121 __be16 port;
5123 /* Initialize a local sockaddr structure to INADDR_ANY. */
5124 af = sctp_sk(sk)->pf->af;
5126 port = htons(inet_sk(sk)->num);
5127 af->inaddr_any(&autoaddr, port);
5129 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5132 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5134 * From RFC 2292
5135 * 4.2 The cmsghdr Structure *
5137 * When ancillary data is sent or received, any number of ancillary data
5138 * objects can be specified by the msg_control and msg_controllen members of
5139 * the msghdr structure, because each object is preceded by
5140 * a cmsghdr structure defining the object's length (the cmsg_len member).
5141 * Historically Berkeley-derived implementations have passed only one object
5142 * at a time, but this API allows multiple objects to be
5143 * passed in a single call to sendmsg() or recvmsg(). The following example
5144 * shows two ancillary data objects in a control buffer.
5146 * |<--------------------------- msg_controllen -------------------------->|
5147 * | |
5149 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5151 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5152 * | | |
5154 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5156 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5157 * | | | | |
5159 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5160 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5162 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5164 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5168 * msg_control
5169 * points here
5171 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5172 sctp_cmsgs_t *cmsgs)
5174 struct cmsghdr *cmsg;
5176 for (cmsg = CMSG_FIRSTHDR(msg);
5177 cmsg != NULL;
5178 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5179 if (!CMSG_OK(msg, cmsg))
5180 return -EINVAL;
5182 /* Should we parse this header or ignore? */
5183 if (cmsg->cmsg_level != IPPROTO_SCTP)
5184 continue;
5186 /* Strictly check lengths following example in SCM code. */
5187 switch (cmsg->cmsg_type) {
5188 case SCTP_INIT:
5189 /* SCTP Socket API Extension
5190 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5192 * This cmsghdr structure provides information for
5193 * initializing new SCTP associations with sendmsg().
5194 * The SCTP_INITMSG socket option uses this same data
5195 * structure. This structure is not used for
5196 * recvmsg().
5198 * cmsg_level cmsg_type cmsg_data[]
5199 * ------------ ------------ ----------------------
5200 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5202 if (cmsg->cmsg_len !=
5203 CMSG_LEN(sizeof(struct sctp_initmsg)))
5204 return -EINVAL;
5205 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5206 break;
5208 case SCTP_SNDRCV:
5209 /* SCTP Socket API Extension
5210 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5212 * This cmsghdr structure specifies SCTP options for
5213 * sendmsg() and describes SCTP header information
5214 * about a received message through recvmsg().
5216 * cmsg_level cmsg_type cmsg_data[]
5217 * ------------ ------------ ----------------------
5218 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5220 if (cmsg->cmsg_len !=
5221 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5222 return -EINVAL;
5224 cmsgs->info =
5225 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5227 /* Minimally, validate the sinfo_flags. */
5228 if (cmsgs->info->sinfo_flags &
5229 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5230 SCTP_ABORT | SCTP_EOF))
5231 return -EINVAL;
5232 break;
5234 default:
5235 return -EINVAL;
5238 return 0;
5242 * Wait for a packet..
5243 * Note: This function is the same function as in core/datagram.c
5244 * with a few modifications to make lksctp work.
5246 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5248 int error;
5249 DEFINE_WAIT(wait);
5251 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5253 /* Socket errors? */
5254 error = sock_error(sk);
5255 if (error)
5256 goto out;
5258 if (!skb_queue_empty(&sk->sk_receive_queue))
5259 goto ready;
5261 /* Socket shut down? */
5262 if (sk->sk_shutdown & RCV_SHUTDOWN)
5263 goto out;
5265 /* Sequenced packets can come disconnected. If so we report the
5266 * problem.
5268 error = -ENOTCONN;
5270 /* Is there a good reason to think that we may receive some data? */
5271 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5272 goto out;
5274 /* Handle signals. */
5275 if (signal_pending(current))
5276 goto interrupted;
5278 /* Let another process have a go. Since we are going to sleep
5279 * anyway. Note: This may cause odd behaviors if the message
5280 * does not fit in the user's buffer, but this seems to be the
5281 * only way to honor MSG_DONTWAIT realistically.
5283 sctp_release_sock(sk);
5284 *timeo_p = schedule_timeout(*timeo_p);
5285 sctp_lock_sock(sk);
5287 ready:
5288 finish_wait(sk->sk_sleep, &wait);
5289 return 0;
5291 interrupted:
5292 error = sock_intr_errno(*timeo_p);
5294 out:
5295 finish_wait(sk->sk_sleep, &wait);
5296 *err = error;
5297 return error;
5300 /* Receive a datagram.
5301 * Note: This is pretty much the same routine as in core/datagram.c
5302 * with a few changes to make lksctp work.
5304 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5305 int noblock, int *err)
5307 int error;
5308 struct sk_buff *skb;
5309 long timeo;
5311 timeo = sock_rcvtimeo(sk, noblock);
5313 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5314 timeo, MAX_SCHEDULE_TIMEOUT);
5316 do {
5317 /* Again only user level code calls this function,
5318 * so nothing interrupt level
5319 * will suddenly eat the receive_queue.
5321 * Look at current nfs client by the way...
5322 * However, this function was corrent in any case. 8)
5324 if (flags & MSG_PEEK) {
5325 spin_lock_bh(&sk->sk_receive_queue.lock);
5326 skb = skb_peek(&sk->sk_receive_queue);
5327 if (skb)
5328 atomic_inc(&skb->users);
5329 spin_unlock_bh(&sk->sk_receive_queue.lock);
5330 } else {
5331 skb = skb_dequeue(&sk->sk_receive_queue);
5334 if (skb)
5335 return skb;
5337 /* Caller is allowed not to check sk->sk_err before calling. */
5338 error = sock_error(sk);
5339 if (error)
5340 goto no_packet;
5342 if (sk->sk_shutdown & RCV_SHUTDOWN)
5343 break;
5345 /* User doesn't want to wait. */
5346 error = -EAGAIN;
5347 if (!timeo)
5348 goto no_packet;
5349 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5351 return NULL;
5353 no_packet:
5354 *err = error;
5355 return NULL;
5358 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5359 static void __sctp_write_space(struct sctp_association *asoc)
5361 struct sock *sk = asoc->base.sk;
5362 struct socket *sock = sk->sk_socket;
5364 if ((sctp_wspace(asoc) > 0) && sock) {
5365 if (waitqueue_active(&asoc->wait))
5366 wake_up_interruptible(&asoc->wait);
5368 if (sctp_writeable(sk)) {
5369 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5370 wake_up_interruptible(sk->sk_sleep);
5372 /* Note that we try to include the Async I/O support
5373 * here by modeling from the current TCP/UDP code.
5374 * We have not tested with it yet.
5376 if (sock->fasync_list &&
5377 !(sk->sk_shutdown & SEND_SHUTDOWN))
5378 sock_wake_async(sock, 2, POLL_OUT);
5383 /* Do accounting for the sndbuf space.
5384 * Decrement the used sndbuf space of the corresponding association by the
5385 * data size which was just transmitted(freed).
5387 static void sctp_wfree(struct sk_buff *skb)
5389 struct sctp_association *asoc;
5390 struct sctp_chunk *chunk;
5391 struct sock *sk;
5393 /* Get the saved chunk pointer. */
5394 chunk = *((struct sctp_chunk **)(skb->cb));
5395 asoc = chunk->asoc;
5396 sk = asoc->base.sk;
5397 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5398 sizeof(struct sk_buff) +
5399 sizeof(struct sctp_chunk);
5401 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5403 sock_wfree(skb);
5404 __sctp_write_space(asoc);
5406 sctp_association_put(asoc);
5409 /* Do accounting for the receive space on the socket.
5410 * Accounting for the association is done in ulpevent.c
5411 * We set this as a destructor for the cloned data skbs so that
5412 * accounting is done at the correct time.
5414 void sctp_sock_rfree(struct sk_buff *skb)
5416 struct sock *sk = skb->sk;
5417 struct sctp_ulpevent *event = sctp_skb2event(skb);
5419 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
5423 /* Helper function to wait for space in the sndbuf. */
5424 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5425 size_t msg_len)
5427 struct sock *sk = asoc->base.sk;
5428 int err = 0;
5429 long current_timeo = *timeo_p;
5430 DEFINE_WAIT(wait);
5432 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5433 asoc, (long)(*timeo_p), msg_len);
5435 /* Increment the association's refcnt. */
5436 sctp_association_hold(asoc);
5438 /* Wait on the association specific sndbuf space. */
5439 for (;;) {
5440 prepare_to_wait_exclusive(&asoc->wait, &wait,
5441 TASK_INTERRUPTIBLE);
5442 if (!*timeo_p)
5443 goto do_nonblock;
5444 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5445 asoc->base.dead)
5446 goto do_error;
5447 if (signal_pending(current))
5448 goto do_interrupted;
5449 if (msg_len <= sctp_wspace(asoc))
5450 break;
5452 /* Let another process have a go. Since we are going
5453 * to sleep anyway.
5455 sctp_release_sock(sk);
5456 current_timeo = schedule_timeout(current_timeo);
5457 BUG_ON(sk != asoc->base.sk);
5458 sctp_lock_sock(sk);
5460 *timeo_p = current_timeo;
5463 out:
5464 finish_wait(&asoc->wait, &wait);
5466 /* Release the association's refcnt. */
5467 sctp_association_put(asoc);
5469 return err;
5471 do_error:
5472 err = -EPIPE;
5473 goto out;
5475 do_interrupted:
5476 err = sock_intr_errno(*timeo_p);
5477 goto out;
5479 do_nonblock:
5480 err = -EAGAIN;
5481 goto out;
5484 /* If socket sndbuf has changed, wake up all per association waiters. */
5485 void sctp_write_space(struct sock *sk)
5487 struct sctp_association *asoc;
5488 struct list_head *pos;
5490 /* Wake up the tasks in each wait queue. */
5491 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5492 asoc = list_entry(pos, struct sctp_association, asocs);
5493 __sctp_write_space(asoc);
5497 /* Is there any sndbuf space available on the socket?
5499 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5500 * associations on the same socket. For a UDP-style socket with
5501 * multiple associations, it is possible for it to be "unwriteable"
5502 * prematurely. I assume that this is acceptable because
5503 * a premature "unwriteable" is better than an accidental "writeable" which
5504 * would cause an unwanted block under certain circumstances. For the 1-1
5505 * UDP-style sockets or TCP-style sockets, this code should work.
5506 * - Daisy
5508 static int sctp_writeable(struct sock *sk)
5510 int amt = 0;
5512 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5513 if (amt < 0)
5514 amt = 0;
5515 return amt;
5518 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5519 * returns immediately with EINPROGRESS.
5521 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5523 struct sock *sk = asoc->base.sk;
5524 int err = 0;
5525 long current_timeo = *timeo_p;
5526 DEFINE_WAIT(wait);
5528 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5529 (long)(*timeo_p));
5531 /* Increment the association's refcnt. */
5532 sctp_association_hold(asoc);
5534 for (;;) {
5535 prepare_to_wait_exclusive(&asoc->wait, &wait,
5536 TASK_INTERRUPTIBLE);
5537 if (!*timeo_p)
5538 goto do_nonblock;
5539 if (sk->sk_shutdown & RCV_SHUTDOWN)
5540 break;
5541 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5542 asoc->base.dead)
5543 goto do_error;
5544 if (signal_pending(current))
5545 goto do_interrupted;
5547 if (sctp_state(asoc, ESTABLISHED))
5548 break;
5550 /* Let another process have a go. Since we are going
5551 * to sleep anyway.
5553 sctp_release_sock(sk);
5554 current_timeo = schedule_timeout(current_timeo);
5555 sctp_lock_sock(sk);
5557 *timeo_p = current_timeo;
5560 out:
5561 finish_wait(&asoc->wait, &wait);
5563 /* Release the association's refcnt. */
5564 sctp_association_put(asoc);
5566 return err;
5568 do_error:
5569 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5570 err = -ETIMEDOUT;
5571 else
5572 err = -ECONNREFUSED;
5573 goto out;
5575 do_interrupted:
5576 err = sock_intr_errno(*timeo_p);
5577 goto out;
5579 do_nonblock:
5580 err = -EINPROGRESS;
5581 goto out;
5584 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5586 struct sctp_endpoint *ep;
5587 int err = 0;
5588 DEFINE_WAIT(wait);
5590 ep = sctp_sk(sk)->ep;
5593 for (;;) {
5594 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5595 TASK_INTERRUPTIBLE);
5597 if (list_empty(&ep->asocs)) {
5598 sctp_release_sock(sk);
5599 timeo = schedule_timeout(timeo);
5600 sctp_lock_sock(sk);
5603 err = -EINVAL;
5604 if (!sctp_sstate(sk, LISTENING))
5605 break;
5607 err = 0;
5608 if (!list_empty(&ep->asocs))
5609 break;
5611 err = sock_intr_errno(timeo);
5612 if (signal_pending(current))
5613 break;
5615 err = -EAGAIN;
5616 if (!timeo)
5617 break;
5620 finish_wait(sk->sk_sleep, &wait);
5622 return err;
5625 void sctp_wait_for_close(struct sock *sk, long timeout)
5627 DEFINE_WAIT(wait);
5629 do {
5630 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5631 if (list_empty(&sctp_sk(sk)->ep->asocs))
5632 break;
5633 sctp_release_sock(sk);
5634 timeout = schedule_timeout(timeout);
5635 sctp_lock_sock(sk);
5636 } while (!signal_pending(current) && timeout);
5638 finish_wait(sk->sk_sleep, &wait);
5641 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5642 * and its messages to the newsk.
5644 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5645 struct sctp_association *assoc,
5646 sctp_socket_type_t type)
5648 struct sctp_sock *oldsp = sctp_sk(oldsk);
5649 struct sctp_sock *newsp = sctp_sk(newsk);
5650 struct sctp_bind_bucket *pp; /* hash list port iterator */
5651 struct sctp_endpoint *newep = newsp->ep;
5652 struct sk_buff *skb, *tmp;
5653 struct sctp_ulpevent *event;
5654 int flags = 0;
5656 /* Migrate socket buffer sizes and all the socket level options to the
5657 * new socket.
5659 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5660 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5661 /* Brute force copy old sctp opt. */
5662 inet_sk_copy_descendant(newsk, oldsk);
5664 /* Restore the ep value that was overwritten with the above structure
5665 * copy.
5667 newsp->ep = newep;
5668 newsp->hmac = NULL;
5670 /* Hook this new socket in to the bind_hash list. */
5671 pp = sctp_sk(oldsk)->bind_hash;
5672 sk_add_bind_node(newsk, &pp->owner);
5673 sctp_sk(newsk)->bind_hash = pp;
5674 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5676 /* Copy the bind_addr list from the original endpoint to the new
5677 * endpoint so that we can handle restarts properly
5679 if (PF_INET6 == assoc->base.sk->sk_family)
5680 flags = SCTP_ADDR6_ALLOWED;
5681 if (assoc->peer.ipv4_address)
5682 flags |= SCTP_ADDR4_PEERSUPP;
5683 if (assoc->peer.ipv6_address)
5684 flags |= SCTP_ADDR6_PEERSUPP;
5685 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5686 &oldsp->ep->base.bind_addr,
5687 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5689 /* Move any messages in the old socket's receive queue that are for the
5690 * peeled off association to the new socket's receive queue.
5692 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5693 event = sctp_skb2event(skb);
5694 if (event->asoc == assoc) {
5695 sctp_sock_rfree(skb);
5696 __skb_unlink(skb, &oldsk->sk_receive_queue);
5697 __skb_queue_tail(&newsk->sk_receive_queue, skb);
5698 sctp_skb_set_owner_r(skb, newsk);
5702 /* Clean up any messages pending delivery due to partial
5703 * delivery. Three cases:
5704 * 1) No partial deliver; no work.
5705 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5706 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5708 skb_queue_head_init(&newsp->pd_lobby);
5709 sctp_sk(newsk)->pd_mode = assoc->ulpq.pd_mode;
5711 if (sctp_sk(oldsk)->pd_mode) {
5712 struct sk_buff_head *queue;
5714 /* Decide which queue to move pd_lobby skbs to. */
5715 if (assoc->ulpq.pd_mode) {
5716 queue = &newsp->pd_lobby;
5717 } else
5718 queue = &newsk->sk_receive_queue;
5720 /* Walk through the pd_lobby, looking for skbs that
5721 * need moved to the new socket.
5723 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5724 event = sctp_skb2event(skb);
5725 if (event->asoc == assoc) {
5726 sctp_sock_rfree(skb);
5727 __skb_unlink(skb, &oldsp->pd_lobby);
5728 __skb_queue_tail(queue, skb);
5729 sctp_skb_set_owner_r(skb, newsk);
5733 /* Clear up any skbs waiting for the partial
5734 * delivery to finish.
5736 if (assoc->ulpq.pd_mode)
5737 sctp_clear_pd(oldsk);
5741 /* Set the type of socket to indicate that it is peeled off from the
5742 * original UDP-style socket or created with the accept() call on a
5743 * TCP-style socket..
5745 newsp->type = type;
5747 /* Mark the new socket "in-use" by the user so that any packets
5748 * that may arrive on the association after we've moved it are
5749 * queued to the backlog. This prevents a potential race between
5750 * backlog processing on the old socket and new-packet processing
5751 * on the new socket.
5753 sctp_lock_sock(newsk);
5754 sctp_assoc_migrate(assoc, newsk);
5756 /* If the association on the newsk is already closed before accept()
5757 * is called, set RCV_SHUTDOWN flag.
5759 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5760 newsk->sk_shutdown |= RCV_SHUTDOWN;
5762 newsk->sk_state = SCTP_SS_ESTABLISHED;
5763 sctp_release_sock(newsk);
5766 /* This proto struct describes the ULP interface for SCTP. */
5767 struct proto sctp_prot = {
5768 .name = "SCTP",
5769 .owner = THIS_MODULE,
5770 .close = sctp_close,
5771 .connect = sctp_connect,
5772 .disconnect = sctp_disconnect,
5773 .accept = sctp_accept,
5774 .ioctl = sctp_ioctl,
5775 .init = sctp_init_sock,
5776 .destroy = sctp_destroy_sock,
5777 .shutdown = sctp_shutdown,
5778 .setsockopt = sctp_setsockopt,
5779 .getsockopt = sctp_getsockopt,
5780 .sendmsg = sctp_sendmsg,
5781 .recvmsg = sctp_recvmsg,
5782 .bind = sctp_bind,
5783 .backlog_rcv = sctp_backlog_rcv,
5784 .hash = sctp_hash,
5785 .unhash = sctp_unhash,
5786 .get_port = sctp_get_port,
5787 .obj_size = sizeof(struct sctp_sock),
5790 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5791 struct proto sctpv6_prot = {
5792 .name = "SCTPv6",
5793 .owner = THIS_MODULE,
5794 .close = sctp_close,
5795 .connect = sctp_connect,
5796 .disconnect = sctp_disconnect,
5797 .accept = sctp_accept,
5798 .ioctl = sctp_ioctl,
5799 .init = sctp_init_sock,
5800 .destroy = sctp_destroy_sock,
5801 .shutdown = sctp_shutdown,
5802 .setsockopt = sctp_setsockopt,
5803 .getsockopt = sctp_getsockopt,
5804 .sendmsg = sctp_sendmsg,
5805 .recvmsg = sctp_recvmsg,
5806 .bind = sctp_bind,
5807 .backlog_rcv = sctp_backlog_rcv,
5808 .hash = sctp_hash,
5809 .unhash = sctp_unhash,
5810 .get_port = sctp_get_port,
5811 .obj_size = sizeof(struct sctp6_sock),
5813 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */