[SCTP]: Implement SCTP_FRAGMENT_INTERLEAVE socket option
[firewire-audio.git] / net / sctp / socket.c
blobb4be473c68b05aee9cc4148df428720fc969283f
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;
631 if (!af->addr_valid(sa_addr, sp, NULL)) {
632 retval = -EADDRNOTAVAIL;
633 goto err_bindx_rem;
636 if (sa_addr->v4.sin_port != htons(bp->port)) {
637 retval = -EINVAL;
638 goto err_bindx_rem;
641 /* FIXME - There is probably a need to check if sk->sk_saddr and
642 * sk->sk_rcv_addr are currently set to one of the addresses to
643 * be removed. This is something which needs to be looked into
644 * when we are fixing the outstanding issues with multi-homing
645 * socket routing and failover schemes. Refer to comments in
646 * sctp_do_bind(). -daisy
648 sctp_local_bh_disable();
649 sctp_write_lock(&ep->base.addr_lock);
651 retval = sctp_del_bind_addr(bp, sa_addr);
653 sctp_write_unlock(&ep->base.addr_lock);
654 sctp_local_bh_enable();
656 addr_buf += af->sockaddr_len;
657 err_bindx_rem:
658 if (retval < 0) {
659 /* Failed. Add the ones that has been removed back */
660 if (cnt > 0)
661 sctp_bindx_add(sk, addrs, cnt);
662 return retval;
666 return retval;
669 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
670 * the associations that are part of the endpoint indicating that a list of
671 * local addresses are removed from the endpoint.
673 * If any of the addresses is already in the bind address list of the
674 * association, we do not send the chunk for that association. But it will not
675 * affect other associations.
677 * Only sctp_setsockopt_bindx() is supposed to call this function.
679 static int sctp_send_asconf_del_ip(struct sock *sk,
680 struct sockaddr *addrs,
681 int addrcnt)
683 struct sctp_sock *sp;
684 struct sctp_endpoint *ep;
685 struct sctp_association *asoc;
686 struct sctp_transport *transport;
687 struct sctp_bind_addr *bp;
688 struct sctp_chunk *chunk;
689 union sctp_addr *laddr;
690 void *addr_buf;
691 struct sctp_af *af;
692 struct list_head *pos, *pos1;
693 struct sctp_sockaddr_entry *saddr;
694 int i;
695 int retval = 0;
697 if (!sctp_addip_enable)
698 return retval;
700 sp = sctp_sk(sk);
701 ep = sp->ep;
703 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
704 __FUNCTION__, sk, addrs, addrcnt);
706 list_for_each(pos, &ep->asocs) {
707 asoc = list_entry(pos, struct sctp_association, asocs);
709 if (!asoc->peer.asconf_capable)
710 continue;
712 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
713 continue;
715 if (!sctp_state(asoc, ESTABLISHED))
716 continue;
718 /* Check if any address in the packed array of addresses is
719 * not present in the bind address list of the association.
720 * If so, do not send the asconf chunk to its peer, but
721 * continue with other associations.
723 addr_buf = addrs;
724 for (i = 0; i < addrcnt; i++) {
725 laddr = (union sctp_addr *)addr_buf;
726 af = sctp_get_af_specific(laddr->v4.sin_family);
727 if (!af) {
728 retval = -EINVAL;
729 goto out;
732 if (!sctp_assoc_lookup_laddr(asoc, laddr))
733 break;
735 addr_buf += af->sockaddr_len;
737 if (i < addrcnt)
738 continue;
740 /* Find one address in the association's bind address list
741 * that is not in the packed array of addresses. This is to
742 * make sure that we do not delete all the addresses in the
743 * association.
745 sctp_read_lock(&asoc->base.addr_lock);
746 bp = &asoc->base.bind_addr;
747 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
748 addrcnt, sp);
749 sctp_read_unlock(&asoc->base.addr_lock);
750 if (!laddr)
751 continue;
753 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
754 SCTP_PARAM_DEL_IP);
755 if (!chunk) {
756 retval = -ENOMEM;
757 goto out;
760 /* Reset use_as_src flag for the addresses in the bind address
761 * list that are to be deleted.
763 sctp_local_bh_disable();
764 sctp_write_lock(&asoc->base.addr_lock);
765 addr_buf = addrs;
766 for (i = 0; i < addrcnt; i++) {
767 laddr = (union sctp_addr *)addr_buf;
768 af = sctp_get_af_specific(laddr->v4.sin_family);
769 list_for_each(pos1, &bp->address_list) {
770 saddr = list_entry(pos1,
771 struct sctp_sockaddr_entry,
772 list);
773 if (sctp_cmp_addr_exact(&saddr->a, laddr))
774 saddr->use_as_src = 0;
776 addr_buf += af->sockaddr_len;
778 sctp_write_unlock(&asoc->base.addr_lock);
779 sctp_local_bh_enable();
781 /* Update the route and saddr entries for all the transports
782 * as some of the addresses in the bind address list are
783 * about to be deleted and cannot be used as source addresses.
785 list_for_each(pos1, &asoc->peer.transport_addr_list) {
786 transport = list_entry(pos1, struct sctp_transport,
787 transports);
788 dst_release(transport->dst);
789 sctp_transport_route(transport, NULL,
790 sctp_sk(asoc->base.sk));
793 retval = sctp_send_asconf(asoc, chunk);
795 out:
796 return retval;
799 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
801 * API 8.1
802 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
803 * int flags);
805 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
806 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
807 * or IPv6 addresses.
809 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
810 * Section 3.1.2 for this usage.
812 * addrs is a pointer to an array of one or more socket addresses. Each
813 * address is contained in its appropriate structure (i.e. struct
814 * sockaddr_in or struct sockaddr_in6) the family of the address type
815 * must be used to distinguish the address length (note that this
816 * representation is termed a "packed array" of addresses). The caller
817 * specifies the number of addresses in the array with addrcnt.
819 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
820 * -1, and sets errno to the appropriate error code.
822 * For SCTP, the port given in each socket address must be the same, or
823 * sctp_bindx() will fail, setting errno to EINVAL.
825 * The flags parameter is formed from the bitwise OR of zero or more of
826 * the following currently defined flags:
828 * SCTP_BINDX_ADD_ADDR
830 * SCTP_BINDX_REM_ADDR
832 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
833 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
834 * addresses from the association. The two flags are mutually exclusive;
835 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
836 * not remove all addresses from an association; sctp_bindx() will
837 * reject such an attempt with EINVAL.
839 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
840 * additional addresses with an endpoint after calling bind(). Or use
841 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
842 * socket is associated with so that no new association accepted will be
843 * associated with those addresses. If the endpoint supports dynamic
844 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
845 * endpoint to send the appropriate message to the peer to change the
846 * peers address lists.
848 * Adding and removing addresses from a connected association is
849 * optional functionality. Implementations that do not support this
850 * functionality should return EOPNOTSUPP.
852 * Basically do nothing but copying the addresses from user to kernel
853 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
854 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
855 * from userspace.
857 * We don't use copy_from_user() for optimization: we first do the
858 * sanity checks (buffer size -fast- and access check-healthy
859 * pointer); if all of those succeed, then we can alloc the memory
860 * (expensive operation) needed to copy the data to kernel. Then we do
861 * the copying without checking the user space area
862 * (__copy_from_user()).
864 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
865 * it.
867 * sk The sk of the socket
868 * addrs The pointer to the addresses in user land
869 * addrssize Size of the addrs buffer
870 * op Operation to perform (add or remove, see the flags of
871 * sctp_bindx)
873 * Returns 0 if ok, <0 errno code on error.
875 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
876 struct sockaddr __user *addrs,
877 int addrs_size, int op)
879 struct sockaddr *kaddrs;
880 int err;
881 int addrcnt = 0;
882 int walk_size = 0;
883 struct sockaddr *sa_addr;
884 void *addr_buf;
885 struct sctp_af *af;
887 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
888 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
890 if (unlikely(addrs_size <= 0))
891 return -EINVAL;
893 /* Check the user passed a healthy pointer. */
894 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
895 return -EFAULT;
897 /* Alloc space for the address array in kernel memory. */
898 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
899 if (unlikely(!kaddrs))
900 return -ENOMEM;
902 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
903 kfree(kaddrs);
904 return -EFAULT;
907 /* Walk through the addrs buffer and count the number of addresses. */
908 addr_buf = kaddrs;
909 while (walk_size < addrs_size) {
910 sa_addr = (struct sockaddr *)addr_buf;
911 af = sctp_get_af_specific(sa_addr->sa_family);
913 /* If the address family is not supported or if this address
914 * causes the address buffer to overflow return EINVAL.
916 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
917 kfree(kaddrs);
918 return -EINVAL;
920 addrcnt++;
921 addr_buf += af->sockaddr_len;
922 walk_size += af->sockaddr_len;
925 /* Do the work. */
926 switch (op) {
927 case SCTP_BINDX_ADD_ADDR:
928 err = sctp_bindx_add(sk, kaddrs, addrcnt);
929 if (err)
930 goto out;
931 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
932 break;
934 case SCTP_BINDX_REM_ADDR:
935 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
936 if (err)
937 goto out;
938 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
939 break;
941 default:
942 err = -EINVAL;
943 break;
946 out:
947 kfree(kaddrs);
949 return err;
952 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
954 * Common routine for handling connect() and sctp_connectx().
955 * Connect will come in with just a single address.
957 static int __sctp_connect(struct sock* sk,
958 struct sockaddr *kaddrs,
959 int addrs_size)
961 struct sctp_sock *sp;
962 struct sctp_endpoint *ep;
963 struct sctp_association *asoc = NULL;
964 struct sctp_association *asoc2;
965 struct sctp_transport *transport;
966 union sctp_addr to;
967 struct sctp_af *af;
968 sctp_scope_t scope;
969 long timeo;
970 int err = 0;
971 int addrcnt = 0;
972 int walk_size = 0;
973 union sctp_addr *sa_addr;
974 void *addr_buf;
976 sp = sctp_sk(sk);
977 ep = sp->ep;
979 /* connect() cannot be done on a socket that is already in ESTABLISHED
980 * state - UDP-style peeled off socket or a TCP-style socket that
981 * is already connected.
982 * It cannot be done even on a TCP-style listening socket.
984 if (sctp_sstate(sk, ESTABLISHED) ||
985 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
986 err = -EISCONN;
987 goto out_free;
990 /* Walk through the addrs buffer and count the number of addresses. */
991 addr_buf = kaddrs;
992 while (walk_size < addrs_size) {
993 sa_addr = (union sctp_addr *)addr_buf;
994 af = sctp_get_af_specific(sa_addr->sa.sa_family);
996 /* If the address family is not supported or if this address
997 * causes the address buffer to overflow return EINVAL.
999 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1000 err = -EINVAL;
1001 goto out_free;
1004 err = sctp_verify_addr(sk, sa_addr, af->sockaddr_len);
1005 if (err)
1006 goto out_free;
1008 memcpy(&to, sa_addr, af->sockaddr_len);
1010 /* Check if there already is a matching association on the
1011 * endpoint (other than the one created here).
1013 asoc2 = sctp_endpoint_lookup_assoc(ep, sa_addr, &transport);
1014 if (asoc2 && asoc2 != asoc) {
1015 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1016 err = -EISCONN;
1017 else
1018 err = -EALREADY;
1019 goto out_free;
1022 /* If we could not find a matching association on the endpoint,
1023 * make sure that there is no peeled-off association matching
1024 * the peer address even on another socket.
1026 if (sctp_endpoint_is_peeled_off(ep, sa_addr)) {
1027 err = -EADDRNOTAVAIL;
1028 goto out_free;
1031 if (!asoc) {
1032 /* If a bind() or sctp_bindx() is not called prior to
1033 * an sctp_connectx() call, the system picks an
1034 * ephemeral port and will choose an address set
1035 * equivalent to binding with a wildcard address.
1037 if (!ep->base.bind_addr.port) {
1038 if (sctp_autobind(sk)) {
1039 err = -EAGAIN;
1040 goto out_free;
1042 } else {
1044 * If an unprivileged user inherits a 1-many
1045 * style socket with open associations on a
1046 * privileged port, it MAY be permitted to
1047 * accept new associations, but it SHOULD NOT
1048 * be permitted to open new associations.
1050 if (ep->base.bind_addr.port < PROT_SOCK &&
1051 !capable(CAP_NET_BIND_SERVICE)) {
1052 err = -EACCES;
1053 goto out_free;
1057 scope = sctp_scope(sa_addr);
1058 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1059 if (!asoc) {
1060 err = -ENOMEM;
1061 goto out_free;
1065 /* Prime the peer's transport structures. */
1066 transport = sctp_assoc_add_peer(asoc, sa_addr, GFP_KERNEL,
1067 SCTP_UNKNOWN);
1068 if (!transport) {
1069 err = -ENOMEM;
1070 goto out_free;
1073 addrcnt++;
1074 addr_buf += af->sockaddr_len;
1075 walk_size += af->sockaddr_len;
1078 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1079 if (err < 0) {
1080 goto out_free;
1083 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1084 if (err < 0) {
1085 goto out_free;
1088 /* Initialize sk's dport and daddr for getpeername() */
1089 inet_sk(sk)->dport = htons(asoc->peer.port);
1090 af = sctp_get_af_specific(to.sa.sa_family);
1091 af->to_sk_daddr(&to, sk);
1092 sk->sk_err = 0;
1094 timeo = sock_sndtimeo(sk, sk->sk_socket->file->f_flags & O_NONBLOCK);
1095 err = sctp_wait_for_connect(asoc, &timeo);
1097 /* Don't free association on exit. */
1098 asoc = NULL;
1100 out_free:
1102 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1103 " kaddrs: %p err: %d\n",
1104 asoc, kaddrs, err);
1105 if (asoc)
1106 sctp_association_free(asoc);
1107 return err;
1110 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1112 * API 8.9
1113 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1115 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1116 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1117 * or IPv6 addresses.
1119 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1120 * Section 3.1.2 for this usage.
1122 * addrs is a pointer to an array of one or more socket addresses. Each
1123 * address is contained in its appropriate structure (i.e. struct
1124 * sockaddr_in or struct sockaddr_in6) the family of the address type
1125 * must be used to distengish the address length (note that this
1126 * representation is termed a "packed array" of addresses). The caller
1127 * specifies the number of addresses in the array with addrcnt.
1129 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1130 * -1, and sets errno to the appropriate error code.
1132 * For SCTP, the port given in each socket address must be the same, or
1133 * sctp_connectx() will fail, setting errno to EINVAL.
1135 * An application can use sctp_connectx to initiate an association with
1136 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1137 * allows a caller to specify multiple addresses at which a peer can be
1138 * reached. The way the SCTP stack uses the list of addresses to set up
1139 * the association is implementation dependant. This function only
1140 * specifies that the stack will try to make use of all the addresses in
1141 * the list when needed.
1143 * Note that the list of addresses passed in is only used for setting up
1144 * the association. It does not necessarily equal the set of addresses
1145 * the peer uses for the resulting association. If the caller wants to
1146 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1147 * retrieve them after the association has been set up.
1149 * Basically do nothing but copying the addresses from user to kernel
1150 * land and invoking either sctp_connectx(). This is used for tunneling
1151 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1153 * We don't use copy_from_user() for optimization: we first do the
1154 * sanity checks (buffer size -fast- and access check-healthy
1155 * pointer); if all of those succeed, then we can alloc the memory
1156 * (expensive operation) needed to copy the data to kernel. Then we do
1157 * the copying without checking the user space area
1158 * (__copy_from_user()).
1160 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1161 * it.
1163 * sk The sk of the socket
1164 * addrs The pointer to the addresses in user land
1165 * addrssize Size of the addrs buffer
1167 * Returns 0 if ok, <0 errno code on error.
1169 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1170 struct sockaddr __user *addrs,
1171 int addrs_size)
1173 int err = 0;
1174 struct sockaddr *kaddrs;
1176 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1177 __FUNCTION__, sk, addrs, addrs_size);
1179 if (unlikely(addrs_size <= 0))
1180 return -EINVAL;
1182 /* Check the user passed a healthy pointer. */
1183 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1184 return -EFAULT;
1186 /* Alloc space for the address array in kernel memory. */
1187 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1188 if (unlikely(!kaddrs))
1189 return -ENOMEM;
1191 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1192 err = -EFAULT;
1193 } else {
1194 err = __sctp_connect(sk, kaddrs, addrs_size);
1197 kfree(kaddrs);
1198 return err;
1201 /* API 3.1.4 close() - UDP Style Syntax
1202 * Applications use close() to perform graceful shutdown (as described in
1203 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1204 * by a UDP-style socket.
1206 * The syntax is
1208 * ret = close(int sd);
1210 * sd - the socket descriptor of the associations to be closed.
1212 * To gracefully shutdown a specific association represented by the
1213 * UDP-style socket, an application should use the sendmsg() call,
1214 * passing no user data, but including the appropriate flag in the
1215 * ancillary data (see Section xxxx).
1217 * If sd in the close() call is a branched-off socket representing only
1218 * one association, the shutdown is performed on that association only.
1220 * 4.1.6 close() - TCP Style Syntax
1222 * Applications use close() to gracefully close down an association.
1224 * The syntax is:
1226 * int close(int sd);
1228 * sd - the socket descriptor of the association to be closed.
1230 * After an application calls close() on a socket descriptor, no further
1231 * socket operations will succeed on that descriptor.
1233 * API 7.1.4 SO_LINGER
1235 * An application using the TCP-style socket can use this option to
1236 * perform the SCTP ABORT primitive. The linger option structure is:
1238 * struct linger {
1239 * int l_onoff; // option on/off
1240 * int l_linger; // linger time
1241 * };
1243 * To enable the option, set l_onoff to 1. If the l_linger value is set
1244 * to 0, calling close() is the same as the ABORT primitive. If the
1245 * value is set to a negative value, the setsockopt() call will return
1246 * an error. If the value is set to a positive value linger_time, the
1247 * close() can be blocked for at most linger_time ms. If the graceful
1248 * shutdown phase does not finish during this period, close() will
1249 * return but the graceful shutdown phase continues in the system.
1251 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1253 struct sctp_endpoint *ep;
1254 struct sctp_association *asoc;
1255 struct list_head *pos, *temp;
1257 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1259 sctp_lock_sock(sk);
1260 sk->sk_shutdown = SHUTDOWN_MASK;
1262 ep = sctp_sk(sk)->ep;
1264 /* Walk all associations on an endpoint. */
1265 list_for_each_safe(pos, temp, &ep->asocs) {
1266 asoc = list_entry(pos, struct sctp_association, asocs);
1268 if (sctp_style(sk, TCP)) {
1269 /* A closed association can still be in the list if
1270 * it belongs to a TCP-style listening socket that is
1271 * not yet accepted. If so, free it. If not, send an
1272 * ABORT or SHUTDOWN based on the linger options.
1274 if (sctp_state(asoc, CLOSED)) {
1275 sctp_unhash_established(asoc);
1276 sctp_association_free(asoc);
1277 continue;
1281 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1282 struct sctp_chunk *chunk;
1284 chunk = sctp_make_abort_user(asoc, NULL, 0);
1285 if (chunk)
1286 sctp_primitive_ABORT(asoc, chunk);
1287 } else
1288 sctp_primitive_SHUTDOWN(asoc, NULL);
1291 /* Clean up any skbs sitting on the receive queue. */
1292 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1293 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1295 /* On a TCP-style socket, block for at most linger_time if set. */
1296 if (sctp_style(sk, TCP) && timeout)
1297 sctp_wait_for_close(sk, timeout);
1299 /* This will run the backlog queue. */
1300 sctp_release_sock(sk);
1302 /* Supposedly, no process has access to the socket, but
1303 * the net layers still may.
1305 sctp_local_bh_disable();
1306 sctp_bh_lock_sock(sk);
1308 /* Hold the sock, since sk_common_release() will put sock_put()
1309 * and we have just a little more cleanup.
1311 sock_hold(sk);
1312 sk_common_release(sk);
1314 sctp_bh_unlock_sock(sk);
1315 sctp_local_bh_enable();
1317 sock_put(sk);
1319 SCTP_DBG_OBJCNT_DEC(sock);
1322 /* Handle EPIPE error. */
1323 static int sctp_error(struct sock *sk, int flags, int err)
1325 if (err == -EPIPE)
1326 err = sock_error(sk) ? : -EPIPE;
1327 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1328 send_sig(SIGPIPE, current, 0);
1329 return err;
1332 /* API 3.1.3 sendmsg() - UDP Style Syntax
1334 * An application uses sendmsg() and recvmsg() calls to transmit data to
1335 * and receive data from its peer.
1337 * ssize_t sendmsg(int socket, const struct msghdr *message,
1338 * int flags);
1340 * socket - the socket descriptor of the endpoint.
1341 * message - pointer to the msghdr structure which contains a single
1342 * user message and possibly some ancillary data.
1344 * See Section 5 for complete description of the data
1345 * structures.
1347 * flags - flags sent or received with the user message, see Section
1348 * 5 for complete description of the flags.
1350 * Note: This function could use a rewrite especially when explicit
1351 * connect support comes in.
1353 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1355 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1357 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1358 struct msghdr *msg, size_t msg_len)
1360 struct sctp_sock *sp;
1361 struct sctp_endpoint *ep;
1362 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1363 struct sctp_transport *transport, *chunk_tp;
1364 struct sctp_chunk *chunk;
1365 union sctp_addr to;
1366 struct sockaddr *msg_name = NULL;
1367 struct sctp_sndrcvinfo default_sinfo = { 0 };
1368 struct sctp_sndrcvinfo *sinfo;
1369 struct sctp_initmsg *sinit;
1370 sctp_assoc_t associd = 0;
1371 sctp_cmsgs_t cmsgs = { NULL };
1372 int err;
1373 sctp_scope_t scope;
1374 long timeo;
1375 __u16 sinfo_flags = 0;
1376 struct sctp_datamsg *datamsg;
1377 struct list_head *pos;
1378 int msg_flags = msg->msg_flags;
1380 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1381 sk, msg, msg_len);
1383 err = 0;
1384 sp = sctp_sk(sk);
1385 ep = sp->ep;
1387 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1389 /* We cannot send a message over a TCP-style listening socket. */
1390 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1391 err = -EPIPE;
1392 goto out_nounlock;
1395 /* Parse out the SCTP CMSGs. */
1396 err = sctp_msghdr_parse(msg, &cmsgs);
1398 if (err) {
1399 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1400 goto out_nounlock;
1403 /* Fetch the destination address for this packet. This
1404 * address only selects the association--it is not necessarily
1405 * the address we will send to.
1406 * For a peeled-off socket, msg_name is ignored.
1408 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1409 int msg_namelen = msg->msg_namelen;
1411 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1412 msg_namelen);
1413 if (err)
1414 return err;
1416 if (msg_namelen > sizeof(to))
1417 msg_namelen = sizeof(to);
1418 memcpy(&to, msg->msg_name, msg_namelen);
1419 msg_name = msg->msg_name;
1422 sinfo = cmsgs.info;
1423 sinit = cmsgs.init;
1425 /* Did the user specify SNDRCVINFO? */
1426 if (sinfo) {
1427 sinfo_flags = sinfo->sinfo_flags;
1428 associd = sinfo->sinfo_assoc_id;
1431 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1432 msg_len, sinfo_flags);
1434 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1435 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1436 err = -EINVAL;
1437 goto out_nounlock;
1440 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1441 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1442 * If SCTP_ABORT is set, the message length could be non zero with
1443 * the msg_iov set to the user abort reason.
1445 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1446 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1447 err = -EINVAL;
1448 goto out_nounlock;
1451 /* If SCTP_ADDR_OVER is set, there must be an address
1452 * specified in msg_name.
1454 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1455 err = -EINVAL;
1456 goto out_nounlock;
1459 transport = NULL;
1461 SCTP_DEBUG_PRINTK("About to look up association.\n");
1463 sctp_lock_sock(sk);
1465 /* If a msg_name has been specified, assume this is to be used. */
1466 if (msg_name) {
1467 /* Look for a matching association on the endpoint. */
1468 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1469 if (!asoc) {
1470 /* If we could not find a matching association on the
1471 * endpoint, make sure that it is not a TCP-style
1472 * socket that already has an association or there is
1473 * no peeled-off association on another socket.
1475 if ((sctp_style(sk, TCP) &&
1476 sctp_sstate(sk, ESTABLISHED)) ||
1477 sctp_endpoint_is_peeled_off(ep, &to)) {
1478 err = -EADDRNOTAVAIL;
1479 goto out_unlock;
1482 } else {
1483 asoc = sctp_id2assoc(sk, associd);
1484 if (!asoc) {
1485 err = -EPIPE;
1486 goto out_unlock;
1490 if (asoc) {
1491 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1493 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1494 * socket that has an association in CLOSED state. This can
1495 * happen when an accepted socket has an association that is
1496 * already CLOSED.
1498 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1499 err = -EPIPE;
1500 goto out_unlock;
1503 if (sinfo_flags & SCTP_EOF) {
1504 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1505 asoc);
1506 sctp_primitive_SHUTDOWN(asoc, NULL);
1507 err = 0;
1508 goto out_unlock;
1510 if (sinfo_flags & SCTP_ABORT) {
1511 struct sctp_chunk *chunk;
1513 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1514 if (!chunk) {
1515 err = -ENOMEM;
1516 goto out_unlock;
1519 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1520 sctp_primitive_ABORT(asoc, chunk);
1521 err = 0;
1522 goto out_unlock;
1526 /* Do we need to create the association? */
1527 if (!asoc) {
1528 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1530 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1531 err = -EINVAL;
1532 goto out_unlock;
1535 /* Check for invalid stream against the stream counts,
1536 * either the default or the user specified stream counts.
1538 if (sinfo) {
1539 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1540 /* Check against the defaults. */
1541 if (sinfo->sinfo_stream >=
1542 sp->initmsg.sinit_num_ostreams) {
1543 err = -EINVAL;
1544 goto out_unlock;
1546 } else {
1547 /* Check against the requested. */
1548 if (sinfo->sinfo_stream >=
1549 sinit->sinit_num_ostreams) {
1550 err = -EINVAL;
1551 goto out_unlock;
1557 * API 3.1.2 bind() - UDP Style Syntax
1558 * If a bind() or sctp_bindx() is not called prior to a
1559 * sendmsg() call that initiates a new association, the
1560 * system picks an ephemeral port and will choose an address
1561 * set equivalent to binding with a wildcard address.
1563 if (!ep->base.bind_addr.port) {
1564 if (sctp_autobind(sk)) {
1565 err = -EAGAIN;
1566 goto out_unlock;
1568 } else {
1570 * If an unprivileged user inherits a one-to-many
1571 * style socket with open associations on a privileged
1572 * port, it MAY be permitted to accept new associations,
1573 * but it SHOULD NOT be permitted to open new
1574 * associations.
1576 if (ep->base.bind_addr.port < PROT_SOCK &&
1577 !capable(CAP_NET_BIND_SERVICE)) {
1578 err = -EACCES;
1579 goto out_unlock;
1583 scope = sctp_scope(&to);
1584 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1585 if (!new_asoc) {
1586 err = -ENOMEM;
1587 goto out_unlock;
1589 asoc = new_asoc;
1591 /* If the SCTP_INIT ancillary data is specified, set all
1592 * the association init values accordingly.
1594 if (sinit) {
1595 if (sinit->sinit_num_ostreams) {
1596 asoc->c.sinit_num_ostreams =
1597 sinit->sinit_num_ostreams;
1599 if (sinit->sinit_max_instreams) {
1600 asoc->c.sinit_max_instreams =
1601 sinit->sinit_max_instreams;
1603 if (sinit->sinit_max_attempts) {
1604 asoc->max_init_attempts
1605 = sinit->sinit_max_attempts;
1607 if (sinit->sinit_max_init_timeo) {
1608 asoc->max_init_timeo =
1609 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1613 /* Prime the peer's transport structures. */
1614 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1615 if (!transport) {
1616 err = -ENOMEM;
1617 goto out_free;
1619 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1620 if (err < 0) {
1621 err = -ENOMEM;
1622 goto out_free;
1626 /* ASSERT: we have a valid association at this point. */
1627 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1629 if (!sinfo) {
1630 /* If the user didn't specify SNDRCVINFO, make up one with
1631 * some defaults.
1633 default_sinfo.sinfo_stream = asoc->default_stream;
1634 default_sinfo.sinfo_flags = asoc->default_flags;
1635 default_sinfo.sinfo_ppid = asoc->default_ppid;
1636 default_sinfo.sinfo_context = asoc->default_context;
1637 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1638 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1639 sinfo = &default_sinfo;
1642 /* API 7.1.7, the sndbuf size per association bounds the
1643 * maximum size of data that can be sent in a single send call.
1645 if (msg_len > sk->sk_sndbuf) {
1646 err = -EMSGSIZE;
1647 goto out_free;
1650 /* If fragmentation is disabled and the message length exceeds the
1651 * association fragmentation point, return EMSGSIZE. The I-D
1652 * does not specify what this error is, but this looks like
1653 * a great fit.
1655 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1656 err = -EMSGSIZE;
1657 goto out_free;
1660 if (sinfo) {
1661 /* Check for invalid stream. */
1662 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1663 err = -EINVAL;
1664 goto out_free;
1668 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1669 if (!sctp_wspace(asoc)) {
1670 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1671 if (err)
1672 goto out_free;
1675 /* If an address is passed with the sendto/sendmsg call, it is used
1676 * to override the primary destination address in the TCP model, or
1677 * when SCTP_ADDR_OVER flag is set in the UDP model.
1679 if ((sctp_style(sk, TCP) && msg_name) ||
1680 (sinfo_flags & SCTP_ADDR_OVER)) {
1681 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1682 if (!chunk_tp) {
1683 err = -EINVAL;
1684 goto out_free;
1686 } else
1687 chunk_tp = NULL;
1689 /* Auto-connect, if we aren't connected already. */
1690 if (sctp_state(asoc, CLOSED)) {
1691 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1692 if (err < 0)
1693 goto out_free;
1694 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1697 /* Break the message into multiple chunks of maximum size. */
1698 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1699 if (!datamsg) {
1700 err = -ENOMEM;
1701 goto out_free;
1704 /* Now send the (possibly) fragmented message. */
1705 list_for_each(pos, &datamsg->chunks) {
1706 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1707 sctp_datamsg_track(chunk);
1709 /* Do accounting for the write space. */
1710 sctp_set_owner_w(chunk);
1712 chunk->transport = chunk_tp;
1714 /* Send it to the lower layers. Note: all chunks
1715 * must either fail or succeed. The lower layer
1716 * works that way today. Keep it that way or this
1717 * breaks.
1719 err = sctp_primitive_SEND(asoc, chunk);
1720 /* Did the lower layer accept the chunk? */
1721 if (err)
1722 sctp_chunk_free(chunk);
1723 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1726 sctp_datamsg_free(datamsg);
1727 if (err)
1728 goto out_free;
1729 else
1730 err = msg_len;
1732 /* If we are already past ASSOCIATE, the lower
1733 * layers are responsible for association cleanup.
1735 goto out_unlock;
1737 out_free:
1738 if (new_asoc)
1739 sctp_association_free(asoc);
1740 out_unlock:
1741 sctp_release_sock(sk);
1743 out_nounlock:
1744 return sctp_error(sk, msg_flags, err);
1746 #if 0
1747 do_sock_err:
1748 if (msg_len)
1749 err = msg_len;
1750 else
1751 err = sock_error(sk);
1752 goto out;
1754 do_interrupted:
1755 if (msg_len)
1756 err = msg_len;
1757 goto out;
1758 #endif /* 0 */
1761 /* This is an extended version of skb_pull() that removes the data from the
1762 * start of a skb even when data is spread across the list of skb's in the
1763 * frag_list. len specifies the total amount of data that needs to be removed.
1764 * when 'len' bytes could be removed from the skb, it returns 0.
1765 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1766 * could not be removed.
1768 static int sctp_skb_pull(struct sk_buff *skb, int len)
1770 struct sk_buff *list;
1771 int skb_len = skb_headlen(skb);
1772 int rlen;
1774 if (len <= skb_len) {
1775 __skb_pull(skb, len);
1776 return 0;
1778 len -= skb_len;
1779 __skb_pull(skb, skb_len);
1781 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1782 rlen = sctp_skb_pull(list, len);
1783 skb->len -= (len-rlen);
1784 skb->data_len -= (len-rlen);
1786 if (!rlen)
1787 return 0;
1789 len = rlen;
1792 return len;
1795 /* API 3.1.3 recvmsg() - UDP Style Syntax
1797 * ssize_t recvmsg(int socket, struct msghdr *message,
1798 * int flags);
1800 * socket - the socket descriptor of the endpoint.
1801 * message - pointer to the msghdr structure which contains a single
1802 * user message and possibly some ancillary data.
1804 * See Section 5 for complete description of the data
1805 * structures.
1807 * flags - flags sent or received with the user message, see Section
1808 * 5 for complete description of the flags.
1810 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1812 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1813 struct msghdr *msg, size_t len, int noblock,
1814 int flags, int *addr_len)
1816 struct sctp_ulpevent *event = NULL;
1817 struct sctp_sock *sp = sctp_sk(sk);
1818 struct sk_buff *skb;
1819 int copied;
1820 int err = 0;
1821 int skb_len;
1823 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1824 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1825 "len", len, "knoblauch", noblock,
1826 "flags", flags, "addr_len", addr_len);
1828 sctp_lock_sock(sk);
1830 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1831 err = -ENOTCONN;
1832 goto out;
1835 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1836 if (!skb)
1837 goto out;
1839 /* Get the total length of the skb including any skb's in the
1840 * frag_list.
1842 skb_len = skb->len;
1844 copied = skb_len;
1845 if (copied > len)
1846 copied = len;
1848 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1850 event = sctp_skb2event(skb);
1852 if (err)
1853 goto out_free;
1855 sock_recv_timestamp(msg, sk, skb);
1856 if (sctp_ulpevent_is_notification(event)) {
1857 msg->msg_flags |= MSG_NOTIFICATION;
1858 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1859 } else {
1860 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1863 /* Check if we allow SCTP_SNDRCVINFO. */
1864 if (sp->subscribe.sctp_data_io_event)
1865 sctp_ulpevent_read_sndrcvinfo(event, msg);
1866 #if 0
1867 /* FIXME: we should be calling IP/IPv6 layers. */
1868 if (sk->sk_protinfo.af_inet.cmsg_flags)
1869 ip_cmsg_recv(msg, skb);
1870 #endif
1872 err = copied;
1874 /* If skb's length exceeds the user's buffer, update the skb and
1875 * push it back to the receive_queue so that the next call to
1876 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1878 if (skb_len > copied) {
1879 msg->msg_flags &= ~MSG_EOR;
1880 if (flags & MSG_PEEK)
1881 goto out_free;
1882 sctp_skb_pull(skb, copied);
1883 skb_queue_head(&sk->sk_receive_queue, skb);
1885 /* When only partial message is copied to the user, increase
1886 * rwnd by that amount. If all the data in the skb is read,
1887 * rwnd is updated when the event is freed.
1889 sctp_assoc_rwnd_increase(event->asoc, copied);
1890 goto out;
1891 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1892 (event->msg_flags & MSG_EOR))
1893 msg->msg_flags |= MSG_EOR;
1894 else
1895 msg->msg_flags &= ~MSG_EOR;
1897 out_free:
1898 if (flags & MSG_PEEK) {
1899 /* Release the skb reference acquired after peeking the skb in
1900 * sctp_skb_recv_datagram().
1902 kfree_skb(skb);
1903 } else {
1904 /* Free the event which includes releasing the reference to
1905 * the owner of the skb, freeing the skb and updating the
1906 * rwnd.
1908 sctp_ulpevent_free(event);
1910 out:
1911 sctp_release_sock(sk);
1912 return err;
1915 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1917 * This option is a on/off flag. If enabled no SCTP message
1918 * fragmentation will be performed. Instead if a message being sent
1919 * exceeds the current PMTU size, the message will NOT be sent and
1920 * instead a error will be indicated to the user.
1922 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1923 char __user *optval, int optlen)
1925 int val;
1927 if (optlen < sizeof(int))
1928 return -EINVAL;
1930 if (get_user(val, (int __user *)optval))
1931 return -EFAULT;
1933 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1935 return 0;
1938 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1939 int optlen)
1941 if (optlen != sizeof(struct sctp_event_subscribe))
1942 return -EINVAL;
1943 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1944 return -EFAULT;
1945 return 0;
1948 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1950 * This socket option is applicable to the UDP-style socket only. When
1951 * set it will cause associations that are idle for more than the
1952 * specified number of seconds to automatically close. An association
1953 * being idle is defined an association that has NOT sent or received
1954 * user data. The special value of '0' indicates that no automatic
1955 * close of any associations should be performed. The option expects an
1956 * integer defining the number of seconds of idle time before an
1957 * association is closed.
1959 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1960 int optlen)
1962 struct sctp_sock *sp = sctp_sk(sk);
1964 /* Applicable to UDP-style socket only */
1965 if (sctp_style(sk, TCP))
1966 return -EOPNOTSUPP;
1967 if (optlen != sizeof(int))
1968 return -EINVAL;
1969 if (copy_from_user(&sp->autoclose, optval, optlen))
1970 return -EFAULT;
1972 return 0;
1975 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1977 * Applications can enable or disable heartbeats for any peer address of
1978 * an association, modify an address's heartbeat interval, force a
1979 * heartbeat to be sent immediately, and adjust the address's maximum
1980 * number of retransmissions sent before an address is considered
1981 * unreachable. The following structure is used to access and modify an
1982 * address's parameters:
1984 * struct sctp_paddrparams {
1985 * sctp_assoc_t spp_assoc_id;
1986 * struct sockaddr_storage spp_address;
1987 * uint32_t spp_hbinterval;
1988 * uint16_t spp_pathmaxrxt;
1989 * uint32_t spp_pathmtu;
1990 * uint32_t spp_sackdelay;
1991 * uint32_t spp_flags;
1992 * };
1994 * spp_assoc_id - (one-to-many style socket) This is filled in the
1995 * application, and identifies the association for
1996 * this query.
1997 * spp_address - This specifies which address is of interest.
1998 * spp_hbinterval - This contains the value of the heartbeat interval,
1999 * in milliseconds. If a value of zero
2000 * is present in this field then no changes are to
2001 * be made to this parameter.
2002 * spp_pathmaxrxt - This contains the maximum number of
2003 * retransmissions before this address shall be
2004 * considered unreachable. If a value of zero
2005 * is present in this field then no changes are to
2006 * be made to this parameter.
2007 * spp_pathmtu - When Path MTU discovery is disabled the value
2008 * specified here will be the "fixed" path mtu.
2009 * Note that if the spp_address field is empty
2010 * then all associations on this address will
2011 * have this fixed path mtu set upon them.
2013 * spp_sackdelay - When delayed sack is enabled, this value specifies
2014 * the number of milliseconds that sacks will be delayed
2015 * for. This value will apply to all addresses of an
2016 * association if the spp_address field is empty. Note
2017 * also, that if delayed sack is enabled and this
2018 * value is set to 0, no change is made to the last
2019 * recorded delayed sack timer value.
2021 * spp_flags - These flags are used to control various features
2022 * on an association. The flag field may contain
2023 * zero or more of the following options.
2025 * SPP_HB_ENABLE - Enable heartbeats on the
2026 * specified address. Note that if the address
2027 * field is empty all addresses for the association
2028 * have heartbeats enabled upon them.
2030 * SPP_HB_DISABLE - Disable heartbeats on the
2031 * speicifed address. Note that if the address
2032 * field is empty all addresses for the association
2033 * will have their heartbeats disabled. Note also
2034 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2035 * mutually exclusive, only one of these two should
2036 * be specified. Enabling both fields will have
2037 * undetermined results.
2039 * SPP_HB_DEMAND - Request a user initiated heartbeat
2040 * to be made immediately.
2042 * SPP_PMTUD_ENABLE - This field will enable PMTU
2043 * discovery upon the specified address. Note that
2044 * if the address feild is empty then all addresses
2045 * on the association are effected.
2047 * SPP_PMTUD_DISABLE - This field will disable PMTU
2048 * discovery upon the specified address. Note that
2049 * if the address feild is empty then all addresses
2050 * on the association are effected. Not also that
2051 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2052 * exclusive. Enabling both will have undetermined
2053 * results.
2055 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2056 * on delayed sack. The time specified in spp_sackdelay
2057 * is used to specify the sack delay for this address. Note
2058 * that if spp_address is empty then all addresses will
2059 * enable delayed sack and take on the sack delay
2060 * value specified in spp_sackdelay.
2061 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2062 * off delayed sack. If the spp_address field is blank then
2063 * delayed sack is disabled for the entire association. Note
2064 * also that this field is mutually exclusive to
2065 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2066 * results.
2068 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2069 struct sctp_transport *trans,
2070 struct sctp_association *asoc,
2071 struct sctp_sock *sp,
2072 int hb_change,
2073 int pmtud_change,
2074 int sackdelay_change)
2076 int error;
2078 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2079 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2080 if (error)
2081 return error;
2084 if (params->spp_hbinterval) {
2085 if (trans) {
2086 trans->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2087 } else if (asoc) {
2088 asoc->hbinterval = msecs_to_jiffies(params->spp_hbinterval);
2089 } else {
2090 sp->hbinterval = params->spp_hbinterval;
2094 if (hb_change) {
2095 if (trans) {
2096 trans->param_flags =
2097 (trans->param_flags & ~SPP_HB) | hb_change;
2098 } else if (asoc) {
2099 asoc->param_flags =
2100 (asoc->param_flags & ~SPP_HB) | hb_change;
2101 } else {
2102 sp->param_flags =
2103 (sp->param_flags & ~SPP_HB) | hb_change;
2107 if (params->spp_pathmtu) {
2108 if (trans) {
2109 trans->pathmtu = params->spp_pathmtu;
2110 sctp_assoc_sync_pmtu(asoc);
2111 } else if (asoc) {
2112 asoc->pathmtu = params->spp_pathmtu;
2113 sctp_frag_point(sp, params->spp_pathmtu);
2114 } else {
2115 sp->pathmtu = params->spp_pathmtu;
2119 if (pmtud_change) {
2120 if (trans) {
2121 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2122 (params->spp_flags & SPP_PMTUD_ENABLE);
2123 trans->param_flags =
2124 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2125 if (update) {
2126 sctp_transport_pmtu(trans);
2127 sctp_assoc_sync_pmtu(asoc);
2129 } else if (asoc) {
2130 asoc->param_flags =
2131 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2132 } else {
2133 sp->param_flags =
2134 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2138 if (params->spp_sackdelay) {
2139 if (trans) {
2140 trans->sackdelay =
2141 msecs_to_jiffies(params->spp_sackdelay);
2142 } else if (asoc) {
2143 asoc->sackdelay =
2144 msecs_to_jiffies(params->spp_sackdelay);
2145 } else {
2146 sp->sackdelay = params->spp_sackdelay;
2150 if (sackdelay_change) {
2151 if (trans) {
2152 trans->param_flags =
2153 (trans->param_flags & ~SPP_SACKDELAY) |
2154 sackdelay_change;
2155 } else if (asoc) {
2156 asoc->param_flags =
2157 (asoc->param_flags & ~SPP_SACKDELAY) |
2158 sackdelay_change;
2159 } else {
2160 sp->param_flags =
2161 (sp->param_flags & ~SPP_SACKDELAY) |
2162 sackdelay_change;
2166 if (params->spp_pathmaxrxt) {
2167 if (trans) {
2168 trans->pathmaxrxt = params->spp_pathmaxrxt;
2169 } else if (asoc) {
2170 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2171 } else {
2172 sp->pathmaxrxt = params->spp_pathmaxrxt;
2176 return 0;
2179 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2180 char __user *optval, int optlen)
2182 struct sctp_paddrparams params;
2183 struct sctp_transport *trans = NULL;
2184 struct sctp_association *asoc = NULL;
2185 struct sctp_sock *sp = sctp_sk(sk);
2186 int error;
2187 int hb_change, pmtud_change, sackdelay_change;
2189 if (optlen != sizeof(struct sctp_paddrparams))
2190 return - EINVAL;
2192 if (copy_from_user(&params, optval, optlen))
2193 return -EFAULT;
2195 /* Validate flags and value parameters. */
2196 hb_change = params.spp_flags & SPP_HB;
2197 pmtud_change = params.spp_flags & SPP_PMTUD;
2198 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2200 if (hb_change == SPP_HB ||
2201 pmtud_change == SPP_PMTUD ||
2202 sackdelay_change == SPP_SACKDELAY ||
2203 params.spp_sackdelay > 500 ||
2204 (params.spp_pathmtu
2205 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2206 return -EINVAL;
2208 /* If an address other than INADDR_ANY is specified, and
2209 * no transport is found, then the request is invalid.
2211 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2212 trans = sctp_addr_id2transport(sk, &params.spp_address,
2213 params.spp_assoc_id);
2214 if (!trans)
2215 return -EINVAL;
2218 /* Get association, if assoc_id != 0 and the socket is a one
2219 * to many style socket, and an association was not found, then
2220 * the id was invalid.
2222 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2223 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2224 return -EINVAL;
2226 /* Heartbeat demand can only be sent on a transport or
2227 * association, but not a socket.
2229 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2230 return -EINVAL;
2232 /* Process parameters. */
2233 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2234 hb_change, pmtud_change,
2235 sackdelay_change);
2237 if (error)
2238 return error;
2240 /* If changes are for association, also apply parameters to each
2241 * transport.
2243 if (!trans && asoc) {
2244 struct list_head *pos;
2246 list_for_each(pos, &asoc->peer.transport_addr_list) {
2247 trans = list_entry(pos, struct sctp_transport,
2248 transports);
2249 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2250 hb_change, pmtud_change,
2251 sackdelay_change);
2255 return 0;
2258 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2260 * This options will get or set the delayed ack timer. The time is set
2261 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2262 * endpoints default delayed ack timer value. If the assoc_id field is
2263 * non-zero, then the set or get effects the specified association.
2265 * struct sctp_assoc_value {
2266 * sctp_assoc_t assoc_id;
2267 * uint32_t assoc_value;
2268 * };
2270 * assoc_id - This parameter, indicates which association the
2271 * user is preforming an action upon. Note that if
2272 * this field's value is zero then the endpoints
2273 * default value is changed (effecting future
2274 * associations only).
2276 * assoc_value - This parameter contains the number of milliseconds
2277 * that the user is requesting the delayed ACK timer
2278 * be set to. Note that this value is defined in
2279 * the standard to be between 200 and 500 milliseconds.
2281 * Note: a value of zero will leave the value alone,
2282 * but disable SACK delay. A non-zero value will also
2283 * enable SACK delay.
2286 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2287 char __user *optval, int optlen)
2289 struct sctp_assoc_value params;
2290 struct sctp_transport *trans = NULL;
2291 struct sctp_association *asoc = NULL;
2292 struct sctp_sock *sp = sctp_sk(sk);
2294 if (optlen != sizeof(struct sctp_assoc_value))
2295 return - EINVAL;
2297 if (copy_from_user(&params, optval, optlen))
2298 return -EFAULT;
2300 /* Validate value parameter. */
2301 if (params.assoc_value > 500)
2302 return -EINVAL;
2304 /* Get association, if assoc_id != 0 and the socket is a one
2305 * to many style socket, and an association was not found, then
2306 * the id was invalid.
2308 asoc = sctp_id2assoc(sk, params.assoc_id);
2309 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2310 return -EINVAL;
2312 if (params.assoc_value) {
2313 if (asoc) {
2314 asoc->sackdelay =
2315 msecs_to_jiffies(params.assoc_value);
2316 asoc->param_flags =
2317 (asoc->param_flags & ~SPP_SACKDELAY) |
2318 SPP_SACKDELAY_ENABLE;
2319 } else {
2320 sp->sackdelay = params.assoc_value;
2321 sp->param_flags =
2322 (sp->param_flags & ~SPP_SACKDELAY) |
2323 SPP_SACKDELAY_ENABLE;
2325 } else {
2326 if (asoc) {
2327 asoc->param_flags =
2328 (asoc->param_flags & ~SPP_SACKDELAY) |
2329 SPP_SACKDELAY_DISABLE;
2330 } else {
2331 sp->param_flags =
2332 (sp->param_flags & ~SPP_SACKDELAY) |
2333 SPP_SACKDELAY_DISABLE;
2337 /* If change is for association, also apply to each transport. */
2338 if (asoc) {
2339 struct list_head *pos;
2341 list_for_each(pos, &asoc->peer.transport_addr_list) {
2342 trans = list_entry(pos, struct sctp_transport,
2343 transports);
2344 if (params.assoc_value) {
2345 trans->sackdelay =
2346 msecs_to_jiffies(params.assoc_value);
2347 trans->param_flags =
2348 (trans->param_flags & ~SPP_SACKDELAY) |
2349 SPP_SACKDELAY_ENABLE;
2350 } else {
2351 trans->param_flags =
2352 (trans->param_flags & ~SPP_SACKDELAY) |
2353 SPP_SACKDELAY_DISABLE;
2358 return 0;
2361 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2363 * Applications can specify protocol parameters for the default association
2364 * initialization. The option name argument to setsockopt() and getsockopt()
2365 * is SCTP_INITMSG.
2367 * Setting initialization parameters is effective only on an unconnected
2368 * socket (for UDP-style sockets only future associations are effected
2369 * by the change). With TCP-style sockets, this option is inherited by
2370 * sockets derived from a listener socket.
2372 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2374 struct sctp_initmsg sinit;
2375 struct sctp_sock *sp = sctp_sk(sk);
2377 if (optlen != sizeof(struct sctp_initmsg))
2378 return -EINVAL;
2379 if (copy_from_user(&sinit, optval, optlen))
2380 return -EFAULT;
2382 if (sinit.sinit_num_ostreams)
2383 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2384 if (sinit.sinit_max_instreams)
2385 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2386 if (sinit.sinit_max_attempts)
2387 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2388 if (sinit.sinit_max_init_timeo)
2389 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2391 return 0;
2395 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2397 * Applications that wish to use the sendto() system call may wish to
2398 * specify a default set of parameters that would normally be supplied
2399 * through the inclusion of ancillary data. This socket option allows
2400 * such an application to set the default sctp_sndrcvinfo structure.
2401 * The application that wishes to use this socket option simply passes
2402 * in to this call the sctp_sndrcvinfo structure defined in Section
2403 * 5.2.2) The input parameters accepted by this call include
2404 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2405 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2406 * to this call if the caller is using the UDP model.
2408 static int sctp_setsockopt_default_send_param(struct sock *sk,
2409 char __user *optval, int optlen)
2411 struct sctp_sndrcvinfo info;
2412 struct sctp_association *asoc;
2413 struct sctp_sock *sp = sctp_sk(sk);
2415 if (optlen != sizeof(struct sctp_sndrcvinfo))
2416 return -EINVAL;
2417 if (copy_from_user(&info, optval, optlen))
2418 return -EFAULT;
2420 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2421 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2422 return -EINVAL;
2424 if (asoc) {
2425 asoc->default_stream = info.sinfo_stream;
2426 asoc->default_flags = info.sinfo_flags;
2427 asoc->default_ppid = info.sinfo_ppid;
2428 asoc->default_context = info.sinfo_context;
2429 asoc->default_timetolive = info.sinfo_timetolive;
2430 } else {
2431 sp->default_stream = info.sinfo_stream;
2432 sp->default_flags = info.sinfo_flags;
2433 sp->default_ppid = info.sinfo_ppid;
2434 sp->default_context = info.sinfo_context;
2435 sp->default_timetolive = info.sinfo_timetolive;
2438 return 0;
2441 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2443 * Requests that the local SCTP stack use the enclosed peer address as
2444 * the association primary. The enclosed address must be one of the
2445 * association peer's addresses.
2447 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2448 int optlen)
2450 struct sctp_prim prim;
2451 struct sctp_transport *trans;
2453 if (optlen != sizeof(struct sctp_prim))
2454 return -EINVAL;
2456 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2457 return -EFAULT;
2459 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2460 if (!trans)
2461 return -EINVAL;
2463 sctp_assoc_set_primary(trans->asoc, trans);
2465 return 0;
2469 * 7.1.5 SCTP_NODELAY
2471 * Turn on/off any Nagle-like algorithm. This means that packets are
2472 * generally sent as soon as possible and no unnecessary delays are
2473 * introduced, at the cost of more packets in the network. Expects an
2474 * integer boolean flag.
2476 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2477 int optlen)
2479 int val;
2481 if (optlen < sizeof(int))
2482 return -EINVAL;
2483 if (get_user(val, (int __user *)optval))
2484 return -EFAULT;
2486 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2487 return 0;
2492 * 7.1.1 SCTP_RTOINFO
2494 * The protocol parameters used to initialize and bound retransmission
2495 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2496 * and modify these parameters.
2497 * All parameters are time values, in milliseconds. A value of 0, when
2498 * modifying the parameters, indicates that the current value should not
2499 * be changed.
2502 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2503 struct sctp_rtoinfo rtoinfo;
2504 struct sctp_association *asoc;
2506 if (optlen != sizeof (struct sctp_rtoinfo))
2507 return -EINVAL;
2509 if (copy_from_user(&rtoinfo, optval, optlen))
2510 return -EFAULT;
2512 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2514 /* Set the values to the specific association */
2515 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2516 return -EINVAL;
2518 if (asoc) {
2519 if (rtoinfo.srto_initial != 0)
2520 asoc->rto_initial =
2521 msecs_to_jiffies(rtoinfo.srto_initial);
2522 if (rtoinfo.srto_max != 0)
2523 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2524 if (rtoinfo.srto_min != 0)
2525 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2526 } else {
2527 /* If there is no association or the association-id = 0
2528 * set the values to the endpoint.
2530 struct sctp_sock *sp = sctp_sk(sk);
2532 if (rtoinfo.srto_initial != 0)
2533 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2534 if (rtoinfo.srto_max != 0)
2535 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2536 if (rtoinfo.srto_min != 0)
2537 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2540 return 0;
2545 * 7.1.2 SCTP_ASSOCINFO
2547 * This option is used to tune the the maximum retransmission attempts
2548 * of the association.
2549 * Returns an error if the new association retransmission value is
2550 * greater than the sum of the retransmission value of the peer.
2551 * See [SCTP] for more information.
2554 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2557 struct sctp_assocparams assocparams;
2558 struct sctp_association *asoc;
2560 if (optlen != sizeof(struct sctp_assocparams))
2561 return -EINVAL;
2562 if (copy_from_user(&assocparams, optval, optlen))
2563 return -EFAULT;
2565 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2567 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2568 return -EINVAL;
2570 /* Set the values to the specific association */
2571 if (asoc) {
2572 if (assocparams.sasoc_asocmaxrxt != 0) {
2573 __u32 path_sum = 0;
2574 int paths = 0;
2575 struct list_head *pos;
2576 struct sctp_transport *peer_addr;
2578 list_for_each(pos, &asoc->peer.transport_addr_list) {
2579 peer_addr = list_entry(pos,
2580 struct sctp_transport,
2581 transports);
2582 path_sum += peer_addr->pathmaxrxt;
2583 paths++;
2586 /* Only validate asocmaxrxt if we have more then
2587 * one path/transport. We do this because path
2588 * retransmissions are only counted when we have more
2589 * then one path.
2591 if (paths > 1 &&
2592 assocparams.sasoc_asocmaxrxt > path_sum)
2593 return -EINVAL;
2595 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2598 if (assocparams.sasoc_cookie_life != 0) {
2599 asoc->cookie_life.tv_sec =
2600 assocparams.sasoc_cookie_life / 1000;
2601 asoc->cookie_life.tv_usec =
2602 (assocparams.sasoc_cookie_life % 1000)
2603 * 1000;
2605 } else {
2606 /* Set the values to the endpoint */
2607 struct sctp_sock *sp = sctp_sk(sk);
2609 if (assocparams.sasoc_asocmaxrxt != 0)
2610 sp->assocparams.sasoc_asocmaxrxt =
2611 assocparams.sasoc_asocmaxrxt;
2612 if (assocparams.sasoc_cookie_life != 0)
2613 sp->assocparams.sasoc_cookie_life =
2614 assocparams.sasoc_cookie_life;
2616 return 0;
2620 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2622 * This socket option is a boolean flag which turns on or off mapped V4
2623 * addresses. If this option is turned on and the socket is type
2624 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2625 * If this option is turned off, then no mapping will be done of V4
2626 * addresses and a user will receive both PF_INET6 and PF_INET type
2627 * addresses on the socket.
2629 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2631 int val;
2632 struct sctp_sock *sp = sctp_sk(sk);
2634 if (optlen < sizeof(int))
2635 return -EINVAL;
2636 if (get_user(val, (int __user *)optval))
2637 return -EFAULT;
2638 if (val)
2639 sp->v4mapped = 1;
2640 else
2641 sp->v4mapped = 0;
2643 return 0;
2647 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2649 * This socket option specifies the maximum size to put in any outgoing
2650 * SCTP chunk. If a message is larger than this size it will be
2651 * fragmented by SCTP into the specified size. Note that the underlying
2652 * SCTP implementation may fragment into smaller sized chunks when the
2653 * PMTU of the underlying association is smaller than the value set by
2654 * the user.
2656 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2658 struct sctp_association *asoc;
2659 struct list_head *pos;
2660 struct sctp_sock *sp = sctp_sk(sk);
2661 int val;
2663 if (optlen < sizeof(int))
2664 return -EINVAL;
2665 if (get_user(val, (int __user *)optval))
2666 return -EFAULT;
2667 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2668 return -EINVAL;
2669 sp->user_frag = val;
2671 /* Update the frag_point of the existing associations. */
2672 list_for_each(pos, &(sp->ep->asocs)) {
2673 asoc = list_entry(pos, struct sctp_association, asocs);
2674 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2677 return 0;
2682 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2684 * Requests that the peer mark the enclosed address as the association
2685 * primary. The enclosed address must be one of the association's
2686 * locally bound addresses. The following structure is used to make a
2687 * set primary request:
2689 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2690 int optlen)
2692 struct sctp_sock *sp;
2693 struct sctp_endpoint *ep;
2694 struct sctp_association *asoc = NULL;
2695 struct sctp_setpeerprim prim;
2696 struct sctp_chunk *chunk;
2697 int err;
2699 sp = sctp_sk(sk);
2700 ep = sp->ep;
2702 if (!sctp_addip_enable)
2703 return -EPERM;
2705 if (optlen != sizeof(struct sctp_setpeerprim))
2706 return -EINVAL;
2708 if (copy_from_user(&prim, optval, optlen))
2709 return -EFAULT;
2711 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2712 if (!asoc)
2713 return -EINVAL;
2715 if (!asoc->peer.asconf_capable)
2716 return -EPERM;
2718 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2719 return -EPERM;
2721 if (!sctp_state(asoc, ESTABLISHED))
2722 return -ENOTCONN;
2724 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2725 return -EADDRNOTAVAIL;
2727 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2728 chunk = sctp_make_asconf_set_prim(asoc,
2729 (union sctp_addr *)&prim.sspp_addr);
2730 if (!chunk)
2731 return -ENOMEM;
2733 err = sctp_send_asconf(asoc, chunk);
2735 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2737 return err;
2740 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2741 int optlen)
2743 struct sctp_setadaptation adaptation;
2745 if (optlen != sizeof(struct sctp_setadaptation))
2746 return -EINVAL;
2747 if (copy_from_user(&adaptation, optval, optlen))
2748 return -EFAULT;
2750 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2752 return 0;
2756 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2758 * The context field in the sctp_sndrcvinfo structure is normally only
2759 * used when a failed message is retrieved holding the value that was
2760 * sent down on the actual send call. This option allows the setting of
2761 * a default context on an association basis that will be received on
2762 * reading messages from the peer. This is especially helpful in the
2763 * one-2-many model for an application to keep some reference to an
2764 * internal state machine that is processing messages on the
2765 * association. Note that the setting of this value only effects
2766 * received messages from the peer and does not effect the value that is
2767 * saved with outbound messages.
2769 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2770 int optlen)
2772 struct sctp_assoc_value params;
2773 struct sctp_sock *sp;
2774 struct sctp_association *asoc;
2776 if (optlen != sizeof(struct sctp_assoc_value))
2777 return -EINVAL;
2778 if (copy_from_user(&params, optval, optlen))
2779 return -EFAULT;
2781 sp = sctp_sk(sk);
2783 if (params.assoc_id != 0) {
2784 asoc = sctp_id2assoc(sk, params.assoc_id);
2785 if (!asoc)
2786 return -EINVAL;
2787 asoc->default_rcv_context = params.assoc_value;
2788 } else {
2789 sp->default_rcv_context = params.assoc_value;
2792 return 0;
2796 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2798 * This options will at a minimum specify if the implementation is doing
2799 * fragmented interleave. Fragmented interleave, for a one to many
2800 * socket, is when subsequent calls to receive a message may return
2801 * parts of messages from different associations. Some implementations
2802 * may allow you to turn this value on or off. If so, when turned off,
2803 * no fragment interleave will occur (which will cause a head of line
2804 * blocking amongst multiple associations sharing the same one to many
2805 * socket). When this option is turned on, then each receive call may
2806 * come from a different association (thus the user must receive data
2807 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2808 * association each receive belongs to.
2810 * This option takes a boolean value. A non-zero value indicates that
2811 * fragmented interleave is on. A value of zero indicates that
2812 * fragmented interleave is off.
2814 * Note that it is important that an implementation that allows this
2815 * option to be turned on, have it off by default. Otherwise an unaware
2816 * application using the one to many model may become confused and act
2817 * incorrectly.
2819 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2820 char __user *optval,
2821 int optlen)
2823 int val;
2825 if (optlen != sizeof(int))
2826 return -EINVAL;
2827 if (get_user(val, (int __user *)optval))
2828 return -EFAULT;
2830 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2832 return 0;
2835 /* API 6.2 setsockopt(), getsockopt()
2837 * Applications use setsockopt() and getsockopt() to set or retrieve
2838 * socket options. Socket options are used to change the default
2839 * behavior of sockets calls. They are described in Section 7.
2841 * The syntax is:
2843 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2844 * int __user *optlen);
2845 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2846 * int optlen);
2848 * sd - the socket descript.
2849 * level - set to IPPROTO_SCTP for all SCTP options.
2850 * optname - the option name.
2851 * optval - the buffer to store the value of the option.
2852 * optlen - the size of the buffer.
2854 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2855 char __user *optval, int optlen)
2857 int retval = 0;
2859 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2860 sk, optname);
2862 /* I can hardly begin to describe how wrong this is. This is
2863 * so broken as to be worse than useless. The API draft
2864 * REALLY is NOT helpful here... I am not convinced that the
2865 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2866 * are at all well-founded.
2868 if (level != SOL_SCTP) {
2869 struct sctp_af *af = sctp_sk(sk)->pf->af;
2870 retval = af->setsockopt(sk, level, optname, optval, optlen);
2871 goto out_nounlock;
2874 sctp_lock_sock(sk);
2876 switch (optname) {
2877 case SCTP_SOCKOPT_BINDX_ADD:
2878 /* 'optlen' is the size of the addresses buffer. */
2879 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2880 optlen, SCTP_BINDX_ADD_ADDR);
2881 break;
2883 case SCTP_SOCKOPT_BINDX_REM:
2884 /* 'optlen' is the size of the addresses buffer. */
2885 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2886 optlen, SCTP_BINDX_REM_ADDR);
2887 break;
2889 case SCTP_SOCKOPT_CONNECTX:
2890 /* 'optlen' is the size of the addresses buffer. */
2891 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2892 optlen);
2893 break;
2895 case SCTP_DISABLE_FRAGMENTS:
2896 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2897 break;
2899 case SCTP_EVENTS:
2900 retval = sctp_setsockopt_events(sk, optval, optlen);
2901 break;
2903 case SCTP_AUTOCLOSE:
2904 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2905 break;
2907 case SCTP_PEER_ADDR_PARAMS:
2908 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2909 break;
2911 case SCTP_DELAYED_ACK_TIME:
2912 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
2913 break;
2915 case SCTP_INITMSG:
2916 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
2917 break;
2918 case SCTP_DEFAULT_SEND_PARAM:
2919 retval = sctp_setsockopt_default_send_param(sk, optval,
2920 optlen);
2921 break;
2922 case SCTP_PRIMARY_ADDR:
2923 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
2924 break;
2925 case SCTP_SET_PEER_PRIMARY_ADDR:
2926 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
2927 break;
2928 case SCTP_NODELAY:
2929 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
2930 break;
2931 case SCTP_RTOINFO:
2932 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
2933 break;
2934 case SCTP_ASSOCINFO:
2935 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
2936 break;
2937 case SCTP_I_WANT_MAPPED_V4_ADDR:
2938 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
2939 break;
2940 case SCTP_MAXSEG:
2941 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
2942 break;
2943 case SCTP_ADAPTATION_LAYER:
2944 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
2945 break;
2946 case SCTP_CONTEXT:
2947 retval = sctp_setsockopt_context(sk, optval, optlen);
2948 break;
2949 case SCTP_FRAGMENT_INTERLEAVE:
2950 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
2951 break;
2952 default:
2953 retval = -ENOPROTOOPT;
2954 break;
2957 sctp_release_sock(sk);
2959 out_nounlock:
2960 return retval;
2963 /* API 3.1.6 connect() - UDP Style Syntax
2965 * An application may use the connect() call in the UDP model to initiate an
2966 * association without sending data.
2968 * The syntax is:
2970 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
2972 * sd: the socket descriptor to have a new association added to.
2974 * nam: the address structure (either struct sockaddr_in or struct
2975 * sockaddr_in6 defined in RFC2553 [7]).
2977 * len: the size of the address.
2979 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
2980 int addr_len)
2982 int err = 0;
2983 struct sctp_af *af;
2985 sctp_lock_sock(sk);
2987 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
2988 __FUNCTION__, sk, addr, addr_len);
2990 /* Validate addr_len before calling common connect/connectx routine. */
2991 af = sctp_get_af_specific(addr->sa_family);
2992 if (!af || addr_len < af->sockaddr_len) {
2993 err = -EINVAL;
2994 } else {
2995 /* Pass correct addr len to common routine (so it knows there
2996 * is only one address being passed.
2998 err = __sctp_connect(sk, addr, af->sockaddr_len);
3001 sctp_release_sock(sk);
3002 return err;
3005 /* FIXME: Write comments. */
3006 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3008 return -EOPNOTSUPP; /* STUB */
3011 /* 4.1.4 accept() - TCP Style Syntax
3013 * Applications use accept() call to remove an established SCTP
3014 * association from the accept queue of the endpoint. A new socket
3015 * descriptor will be returned from accept() to represent the newly
3016 * formed association.
3018 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3020 struct sctp_sock *sp;
3021 struct sctp_endpoint *ep;
3022 struct sock *newsk = NULL;
3023 struct sctp_association *asoc;
3024 long timeo;
3025 int error = 0;
3027 sctp_lock_sock(sk);
3029 sp = sctp_sk(sk);
3030 ep = sp->ep;
3032 if (!sctp_style(sk, TCP)) {
3033 error = -EOPNOTSUPP;
3034 goto out;
3037 if (!sctp_sstate(sk, LISTENING)) {
3038 error = -EINVAL;
3039 goto out;
3042 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3044 error = sctp_wait_for_accept(sk, timeo);
3045 if (error)
3046 goto out;
3048 /* We treat the list of associations on the endpoint as the accept
3049 * queue and pick the first association on the list.
3051 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3053 newsk = sp->pf->create_accept_sk(sk, asoc);
3054 if (!newsk) {
3055 error = -ENOMEM;
3056 goto out;
3059 /* Populate the fields of the newsk from the oldsk and migrate the
3060 * asoc to the newsk.
3062 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3064 out:
3065 sctp_release_sock(sk);
3066 *err = error;
3067 return newsk;
3070 /* The SCTP ioctl handler. */
3071 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3073 return -ENOIOCTLCMD;
3076 /* This is the function which gets called during socket creation to
3077 * initialized the SCTP-specific portion of the sock.
3078 * The sock structure should already be zero-filled memory.
3080 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3082 struct sctp_endpoint *ep;
3083 struct sctp_sock *sp;
3085 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3087 sp = sctp_sk(sk);
3089 /* Initialize the SCTP per socket area. */
3090 switch (sk->sk_type) {
3091 case SOCK_SEQPACKET:
3092 sp->type = SCTP_SOCKET_UDP;
3093 break;
3094 case SOCK_STREAM:
3095 sp->type = SCTP_SOCKET_TCP;
3096 break;
3097 default:
3098 return -ESOCKTNOSUPPORT;
3101 /* Initialize default send parameters. These parameters can be
3102 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3104 sp->default_stream = 0;
3105 sp->default_ppid = 0;
3106 sp->default_flags = 0;
3107 sp->default_context = 0;
3108 sp->default_timetolive = 0;
3110 sp->default_rcv_context = 0;
3112 /* Initialize default setup parameters. These parameters
3113 * can be modified with the SCTP_INITMSG socket option or
3114 * overridden by the SCTP_INIT CMSG.
3116 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3117 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3118 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3119 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3121 /* Initialize default RTO related parameters. These parameters can
3122 * be modified for with the SCTP_RTOINFO socket option.
3124 sp->rtoinfo.srto_initial = sctp_rto_initial;
3125 sp->rtoinfo.srto_max = sctp_rto_max;
3126 sp->rtoinfo.srto_min = sctp_rto_min;
3128 /* Initialize default association related parameters. These parameters
3129 * can be modified with the SCTP_ASSOCINFO socket option.
3131 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3132 sp->assocparams.sasoc_number_peer_destinations = 0;
3133 sp->assocparams.sasoc_peer_rwnd = 0;
3134 sp->assocparams.sasoc_local_rwnd = 0;
3135 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3137 /* Initialize default event subscriptions. By default, all the
3138 * options are off.
3140 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3142 /* Default Peer Address Parameters. These defaults can
3143 * be modified via SCTP_PEER_ADDR_PARAMS
3145 sp->hbinterval = sctp_hb_interval;
3146 sp->pathmaxrxt = sctp_max_retrans_path;
3147 sp->pathmtu = 0; // allow default discovery
3148 sp->sackdelay = sctp_sack_timeout;
3149 sp->param_flags = SPP_HB_ENABLE |
3150 SPP_PMTUD_ENABLE |
3151 SPP_SACKDELAY_ENABLE;
3153 /* If enabled no SCTP message fragmentation will be performed.
3154 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3156 sp->disable_fragments = 0;
3158 /* Enable Nagle algorithm by default. */
3159 sp->nodelay = 0;
3161 /* Enable by default. */
3162 sp->v4mapped = 1;
3164 /* Auto-close idle associations after the configured
3165 * number of seconds. A value of 0 disables this
3166 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3167 * for UDP-style sockets only.
3169 sp->autoclose = 0;
3171 /* User specified fragmentation limit. */
3172 sp->user_frag = 0;
3174 sp->adaptation_ind = 0;
3176 sp->pf = sctp_get_pf_specific(sk->sk_family);
3178 /* Control variables for partial data delivery. */
3179 atomic_set(&sp->pd_mode, 0);
3180 skb_queue_head_init(&sp->pd_lobby);
3181 sp->frag_interleave = 0;
3183 /* Create a per socket endpoint structure. Even if we
3184 * change the data structure relationships, this may still
3185 * be useful for storing pre-connect address information.
3187 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3188 if (!ep)
3189 return -ENOMEM;
3191 sp->ep = ep;
3192 sp->hmac = NULL;
3194 SCTP_DBG_OBJCNT_INC(sock);
3195 return 0;
3198 /* Cleanup any SCTP per socket resources. */
3199 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3201 struct sctp_endpoint *ep;
3203 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3205 /* Release our hold on the endpoint. */
3206 ep = sctp_sk(sk)->ep;
3207 sctp_endpoint_free(ep);
3209 return 0;
3212 /* API 4.1.7 shutdown() - TCP Style Syntax
3213 * int shutdown(int socket, int how);
3215 * sd - the socket descriptor of the association to be closed.
3216 * how - Specifies the type of shutdown. The values are
3217 * as follows:
3218 * SHUT_RD
3219 * Disables further receive operations. No SCTP
3220 * protocol action is taken.
3221 * SHUT_WR
3222 * Disables further send operations, and initiates
3223 * the SCTP shutdown sequence.
3224 * SHUT_RDWR
3225 * Disables further send and receive operations
3226 * and initiates the SCTP shutdown sequence.
3228 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3230 struct sctp_endpoint *ep;
3231 struct sctp_association *asoc;
3233 if (!sctp_style(sk, TCP))
3234 return;
3236 if (how & SEND_SHUTDOWN) {
3237 ep = sctp_sk(sk)->ep;
3238 if (!list_empty(&ep->asocs)) {
3239 asoc = list_entry(ep->asocs.next,
3240 struct sctp_association, asocs);
3241 sctp_primitive_SHUTDOWN(asoc, NULL);
3246 /* 7.2.1 Association Status (SCTP_STATUS)
3248 * Applications can retrieve current status information about an
3249 * association, including association state, peer receiver window size,
3250 * number of unacked data chunks, and number of data chunks pending
3251 * receipt. This information is read-only.
3253 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3254 char __user *optval,
3255 int __user *optlen)
3257 struct sctp_status status;
3258 struct sctp_association *asoc = NULL;
3259 struct sctp_transport *transport;
3260 sctp_assoc_t associd;
3261 int retval = 0;
3263 if (len != sizeof(status)) {
3264 retval = -EINVAL;
3265 goto out;
3268 if (copy_from_user(&status, optval, sizeof(status))) {
3269 retval = -EFAULT;
3270 goto out;
3273 associd = status.sstat_assoc_id;
3274 asoc = sctp_id2assoc(sk, associd);
3275 if (!asoc) {
3276 retval = -EINVAL;
3277 goto out;
3280 transport = asoc->peer.primary_path;
3282 status.sstat_assoc_id = sctp_assoc2id(asoc);
3283 status.sstat_state = asoc->state;
3284 status.sstat_rwnd = asoc->peer.rwnd;
3285 status.sstat_unackdata = asoc->unack_data;
3287 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3288 status.sstat_instrms = asoc->c.sinit_max_instreams;
3289 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3290 status.sstat_fragmentation_point = asoc->frag_point;
3291 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3292 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3293 transport->af_specific->sockaddr_len);
3294 /* Map ipv4 address into v4-mapped-on-v6 address. */
3295 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3296 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3297 status.sstat_primary.spinfo_state = transport->state;
3298 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3299 status.sstat_primary.spinfo_srtt = transport->srtt;
3300 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3301 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3303 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3304 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3306 if (put_user(len, optlen)) {
3307 retval = -EFAULT;
3308 goto out;
3311 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3312 len, status.sstat_state, status.sstat_rwnd,
3313 status.sstat_assoc_id);
3315 if (copy_to_user(optval, &status, len)) {
3316 retval = -EFAULT;
3317 goto out;
3320 out:
3321 return (retval);
3325 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3327 * Applications can retrieve information about a specific peer address
3328 * of an association, including its reachability state, congestion
3329 * window, and retransmission timer values. This information is
3330 * read-only.
3332 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3333 char __user *optval,
3334 int __user *optlen)
3336 struct sctp_paddrinfo pinfo;
3337 struct sctp_transport *transport;
3338 int retval = 0;
3340 if (len != sizeof(pinfo)) {
3341 retval = -EINVAL;
3342 goto out;
3345 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3346 retval = -EFAULT;
3347 goto out;
3350 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3351 pinfo.spinfo_assoc_id);
3352 if (!transport)
3353 return -EINVAL;
3355 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3356 pinfo.spinfo_state = transport->state;
3357 pinfo.spinfo_cwnd = transport->cwnd;
3358 pinfo.spinfo_srtt = transport->srtt;
3359 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3360 pinfo.spinfo_mtu = transport->pathmtu;
3362 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3363 pinfo.spinfo_state = SCTP_ACTIVE;
3365 if (put_user(len, optlen)) {
3366 retval = -EFAULT;
3367 goto out;
3370 if (copy_to_user(optval, &pinfo, len)) {
3371 retval = -EFAULT;
3372 goto out;
3375 out:
3376 return (retval);
3379 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3381 * This option is a on/off flag. If enabled no SCTP message
3382 * fragmentation will be performed. Instead if a message being sent
3383 * exceeds the current PMTU size, the message will NOT be sent and
3384 * instead a error will be indicated to the user.
3386 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3387 char __user *optval, int __user *optlen)
3389 int val;
3391 if (len < sizeof(int))
3392 return -EINVAL;
3394 len = sizeof(int);
3395 val = (sctp_sk(sk)->disable_fragments == 1);
3396 if (put_user(len, optlen))
3397 return -EFAULT;
3398 if (copy_to_user(optval, &val, len))
3399 return -EFAULT;
3400 return 0;
3403 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3405 * This socket option is used to specify various notifications and
3406 * ancillary data the user wishes to receive.
3408 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3409 int __user *optlen)
3411 if (len != sizeof(struct sctp_event_subscribe))
3412 return -EINVAL;
3413 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3414 return -EFAULT;
3415 return 0;
3418 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3420 * This socket option is applicable to the UDP-style socket only. When
3421 * set it will cause associations that are idle for more than the
3422 * specified number of seconds to automatically close. An association
3423 * being idle is defined an association that has NOT sent or received
3424 * user data. The special value of '0' indicates that no automatic
3425 * close of any associations should be performed. The option expects an
3426 * integer defining the number of seconds of idle time before an
3427 * association is closed.
3429 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3431 /* Applicable to UDP-style socket only */
3432 if (sctp_style(sk, TCP))
3433 return -EOPNOTSUPP;
3434 if (len != sizeof(int))
3435 return -EINVAL;
3436 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3437 return -EFAULT;
3438 return 0;
3441 /* Helper routine to branch off an association to a new socket. */
3442 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3443 struct socket **sockp)
3445 struct sock *sk = asoc->base.sk;
3446 struct socket *sock;
3447 struct inet_sock *inetsk;
3448 int err = 0;
3450 /* An association cannot be branched off from an already peeled-off
3451 * socket, nor is this supported for tcp style sockets.
3453 if (!sctp_style(sk, UDP))
3454 return -EINVAL;
3456 /* Create a new socket. */
3457 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3458 if (err < 0)
3459 return err;
3461 /* Populate the fields of the newsk from the oldsk and migrate the
3462 * asoc to the newsk.
3464 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3466 /* Make peeled-off sockets more like 1-1 accepted sockets.
3467 * Set the daddr and initialize id to something more random
3469 inetsk = inet_sk(sock->sk);
3470 inetsk->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
3471 inetsk->id = asoc->next_tsn ^ jiffies;
3473 *sockp = sock;
3475 return err;
3478 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3480 sctp_peeloff_arg_t peeloff;
3481 struct socket *newsock;
3482 int retval = 0;
3483 struct sctp_association *asoc;
3485 if (len != sizeof(sctp_peeloff_arg_t))
3486 return -EINVAL;
3487 if (copy_from_user(&peeloff, optval, len))
3488 return -EFAULT;
3490 asoc = sctp_id2assoc(sk, peeloff.associd);
3491 if (!asoc) {
3492 retval = -EINVAL;
3493 goto out;
3496 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3498 retval = sctp_do_peeloff(asoc, &newsock);
3499 if (retval < 0)
3500 goto out;
3502 /* Map the socket to an unused fd that can be returned to the user. */
3503 retval = sock_map_fd(newsock);
3504 if (retval < 0) {
3505 sock_release(newsock);
3506 goto out;
3509 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3510 __FUNCTION__, sk, asoc, newsock->sk, retval);
3512 /* Return the fd mapped to the new socket. */
3513 peeloff.sd = retval;
3514 if (copy_to_user(optval, &peeloff, len))
3515 retval = -EFAULT;
3517 out:
3518 return retval;
3521 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3523 * Applications can enable or disable heartbeats for any peer address of
3524 * an association, modify an address's heartbeat interval, force a
3525 * heartbeat to be sent immediately, and adjust the address's maximum
3526 * number of retransmissions sent before an address is considered
3527 * unreachable. The following structure is used to access and modify an
3528 * address's parameters:
3530 * struct sctp_paddrparams {
3531 * sctp_assoc_t spp_assoc_id;
3532 * struct sockaddr_storage spp_address;
3533 * uint32_t spp_hbinterval;
3534 * uint16_t spp_pathmaxrxt;
3535 * uint32_t spp_pathmtu;
3536 * uint32_t spp_sackdelay;
3537 * uint32_t spp_flags;
3538 * };
3540 * spp_assoc_id - (one-to-many style socket) This is filled in the
3541 * application, and identifies the association for
3542 * this query.
3543 * spp_address - This specifies which address is of interest.
3544 * spp_hbinterval - This contains the value of the heartbeat interval,
3545 * in milliseconds. If a value of zero
3546 * is present in this field then no changes are to
3547 * be made to this parameter.
3548 * spp_pathmaxrxt - This contains the maximum number of
3549 * retransmissions before this address shall be
3550 * considered unreachable. If a value of zero
3551 * is present in this field then no changes are to
3552 * be made to this parameter.
3553 * spp_pathmtu - When Path MTU discovery is disabled the value
3554 * specified here will be the "fixed" path mtu.
3555 * Note that if the spp_address field is empty
3556 * then all associations on this address will
3557 * have this fixed path mtu set upon them.
3559 * spp_sackdelay - When delayed sack is enabled, this value specifies
3560 * the number of milliseconds that sacks will be delayed
3561 * for. This value will apply to all addresses of an
3562 * association if the spp_address field is empty. Note
3563 * also, that if delayed sack is enabled and this
3564 * value is set to 0, no change is made to the last
3565 * recorded delayed sack timer value.
3567 * spp_flags - These flags are used to control various features
3568 * on an association. The flag field may contain
3569 * zero or more of the following options.
3571 * SPP_HB_ENABLE - Enable heartbeats on the
3572 * specified address. Note that if the address
3573 * field is empty all addresses for the association
3574 * have heartbeats enabled upon them.
3576 * SPP_HB_DISABLE - Disable heartbeats on the
3577 * speicifed address. Note that if the address
3578 * field is empty all addresses for the association
3579 * will have their heartbeats disabled. Note also
3580 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3581 * mutually exclusive, only one of these two should
3582 * be specified. Enabling both fields will have
3583 * undetermined results.
3585 * SPP_HB_DEMAND - Request a user initiated heartbeat
3586 * to be made immediately.
3588 * SPP_PMTUD_ENABLE - This field will enable PMTU
3589 * discovery upon the specified address. Note that
3590 * if the address feild is empty then all addresses
3591 * on the association are effected.
3593 * SPP_PMTUD_DISABLE - This field will disable PMTU
3594 * discovery upon the specified address. Note that
3595 * if the address feild is empty then all addresses
3596 * on the association are effected. Not also that
3597 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3598 * exclusive. Enabling both will have undetermined
3599 * results.
3601 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3602 * on delayed sack. The time specified in spp_sackdelay
3603 * is used to specify the sack delay for this address. Note
3604 * that if spp_address is empty then all addresses will
3605 * enable delayed sack and take on the sack delay
3606 * value specified in spp_sackdelay.
3607 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3608 * off delayed sack. If the spp_address field is blank then
3609 * delayed sack is disabled for the entire association. Note
3610 * also that this field is mutually exclusive to
3611 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3612 * results.
3614 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3615 char __user *optval, int __user *optlen)
3617 struct sctp_paddrparams params;
3618 struct sctp_transport *trans = NULL;
3619 struct sctp_association *asoc = NULL;
3620 struct sctp_sock *sp = sctp_sk(sk);
3622 if (len != sizeof(struct sctp_paddrparams))
3623 return -EINVAL;
3625 if (copy_from_user(&params, optval, len))
3626 return -EFAULT;
3628 /* If an address other than INADDR_ANY is specified, and
3629 * no transport is found, then the request is invalid.
3631 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3632 trans = sctp_addr_id2transport(sk, &params.spp_address,
3633 params.spp_assoc_id);
3634 if (!trans) {
3635 SCTP_DEBUG_PRINTK("Failed no transport\n");
3636 return -EINVAL;
3640 /* Get association, if assoc_id != 0 and the socket is a one
3641 * to many style socket, and an association was not found, then
3642 * the id was invalid.
3644 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3645 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3646 SCTP_DEBUG_PRINTK("Failed no association\n");
3647 return -EINVAL;
3650 if (trans) {
3651 /* Fetch transport values. */
3652 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3653 params.spp_pathmtu = trans->pathmtu;
3654 params.spp_pathmaxrxt = trans->pathmaxrxt;
3655 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3657 /*draft-11 doesn't say what to return in spp_flags*/
3658 params.spp_flags = trans->param_flags;
3659 } else if (asoc) {
3660 /* Fetch association values. */
3661 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3662 params.spp_pathmtu = asoc->pathmtu;
3663 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3664 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3666 /*draft-11 doesn't say what to return in spp_flags*/
3667 params.spp_flags = asoc->param_flags;
3668 } else {
3669 /* Fetch socket values. */
3670 params.spp_hbinterval = sp->hbinterval;
3671 params.spp_pathmtu = sp->pathmtu;
3672 params.spp_sackdelay = sp->sackdelay;
3673 params.spp_pathmaxrxt = sp->pathmaxrxt;
3675 /*draft-11 doesn't say what to return in spp_flags*/
3676 params.spp_flags = sp->param_flags;
3679 if (copy_to_user(optval, &params, len))
3680 return -EFAULT;
3682 if (put_user(len, optlen))
3683 return -EFAULT;
3685 return 0;
3688 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3690 * This options will get or set the delayed ack timer. The time is set
3691 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3692 * endpoints default delayed ack timer value. If the assoc_id field is
3693 * non-zero, then the set or get effects the specified association.
3695 * struct sctp_assoc_value {
3696 * sctp_assoc_t assoc_id;
3697 * uint32_t assoc_value;
3698 * };
3700 * assoc_id - This parameter, indicates which association the
3701 * user is preforming an action upon. Note that if
3702 * this field's value is zero then the endpoints
3703 * default value is changed (effecting future
3704 * associations only).
3706 * assoc_value - This parameter contains the number of milliseconds
3707 * that the user is requesting the delayed ACK timer
3708 * be set to. Note that this value is defined in
3709 * the standard to be between 200 and 500 milliseconds.
3711 * Note: a value of zero will leave the value alone,
3712 * but disable SACK delay. A non-zero value will also
3713 * enable SACK delay.
3715 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3716 char __user *optval,
3717 int __user *optlen)
3719 struct sctp_assoc_value params;
3720 struct sctp_association *asoc = NULL;
3721 struct sctp_sock *sp = sctp_sk(sk);
3723 if (len != sizeof(struct sctp_assoc_value))
3724 return - EINVAL;
3726 if (copy_from_user(&params, optval, len))
3727 return -EFAULT;
3729 /* Get association, if assoc_id != 0 and the socket is a one
3730 * to many style socket, and an association was not found, then
3731 * the id was invalid.
3733 asoc = sctp_id2assoc(sk, params.assoc_id);
3734 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3735 return -EINVAL;
3737 if (asoc) {
3738 /* Fetch association values. */
3739 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3740 params.assoc_value = jiffies_to_msecs(
3741 asoc->sackdelay);
3742 else
3743 params.assoc_value = 0;
3744 } else {
3745 /* Fetch socket values. */
3746 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3747 params.assoc_value = sp->sackdelay;
3748 else
3749 params.assoc_value = 0;
3752 if (copy_to_user(optval, &params, len))
3753 return -EFAULT;
3755 if (put_user(len, optlen))
3756 return -EFAULT;
3758 return 0;
3761 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3763 * Applications can specify protocol parameters for the default association
3764 * initialization. The option name argument to setsockopt() and getsockopt()
3765 * is SCTP_INITMSG.
3767 * Setting initialization parameters is effective only on an unconnected
3768 * socket (for UDP-style sockets only future associations are effected
3769 * by the change). With TCP-style sockets, this option is inherited by
3770 * sockets derived from a listener socket.
3772 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3774 if (len != sizeof(struct sctp_initmsg))
3775 return -EINVAL;
3776 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3777 return -EFAULT;
3778 return 0;
3781 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3782 char __user *optval,
3783 int __user *optlen)
3785 sctp_assoc_t id;
3786 struct sctp_association *asoc;
3787 struct list_head *pos;
3788 int cnt = 0;
3790 if (len != sizeof(sctp_assoc_t))
3791 return -EINVAL;
3793 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3794 return -EFAULT;
3796 /* For UDP-style sockets, id specifies the association to query. */
3797 asoc = sctp_id2assoc(sk, id);
3798 if (!asoc)
3799 return -EINVAL;
3801 list_for_each(pos, &asoc->peer.transport_addr_list) {
3802 cnt ++;
3805 return cnt;
3809 * Old API for getting list of peer addresses. Does not work for 32-bit
3810 * programs running on a 64-bit kernel
3812 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3813 char __user *optval,
3814 int __user *optlen)
3816 struct sctp_association *asoc;
3817 struct list_head *pos;
3818 int cnt = 0;
3819 struct sctp_getaddrs_old getaddrs;
3820 struct sctp_transport *from;
3821 void __user *to;
3822 union sctp_addr temp;
3823 struct sctp_sock *sp = sctp_sk(sk);
3824 int addrlen;
3826 if (len != sizeof(struct sctp_getaddrs_old))
3827 return -EINVAL;
3829 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3830 return -EFAULT;
3832 if (getaddrs.addr_num <= 0) return -EINVAL;
3834 /* For UDP-style sockets, id specifies the association to query. */
3835 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3836 if (!asoc)
3837 return -EINVAL;
3839 to = (void __user *)getaddrs.addrs;
3840 list_for_each(pos, &asoc->peer.transport_addr_list) {
3841 from = list_entry(pos, struct sctp_transport, transports);
3842 memcpy(&temp, &from->ipaddr, sizeof(temp));
3843 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3844 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3845 if (copy_to_user(to, &temp, addrlen))
3846 return -EFAULT;
3847 to += addrlen ;
3848 cnt ++;
3849 if (cnt >= getaddrs.addr_num) break;
3851 getaddrs.addr_num = cnt;
3852 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3853 return -EFAULT;
3855 return 0;
3858 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3859 char __user *optval, int __user *optlen)
3861 struct sctp_association *asoc;
3862 struct list_head *pos;
3863 int cnt = 0;
3864 struct sctp_getaddrs getaddrs;
3865 struct sctp_transport *from;
3866 void __user *to;
3867 union sctp_addr temp;
3868 struct sctp_sock *sp = sctp_sk(sk);
3869 int addrlen;
3870 size_t space_left;
3871 int bytes_copied;
3873 if (len < sizeof(struct sctp_getaddrs))
3874 return -EINVAL;
3876 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3877 return -EFAULT;
3879 /* For UDP-style sockets, id specifies the association to query. */
3880 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3881 if (!asoc)
3882 return -EINVAL;
3884 to = optval + offsetof(struct sctp_getaddrs,addrs);
3885 space_left = len - sizeof(struct sctp_getaddrs) -
3886 offsetof(struct sctp_getaddrs,addrs);
3888 list_for_each(pos, &asoc->peer.transport_addr_list) {
3889 from = list_entry(pos, struct sctp_transport, transports);
3890 memcpy(&temp, &from->ipaddr, sizeof(temp));
3891 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3892 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3893 if(space_left < addrlen)
3894 return -ENOMEM;
3895 if (copy_to_user(to, &temp, addrlen))
3896 return -EFAULT;
3897 to += addrlen;
3898 cnt++;
3899 space_left -= addrlen;
3902 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
3903 return -EFAULT;
3904 bytes_copied = ((char __user *)to) - optval;
3905 if (put_user(bytes_copied, optlen))
3906 return -EFAULT;
3908 return 0;
3911 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
3912 char __user *optval,
3913 int __user *optlen)
3915 sctp_assoc_t id;
3916 struct sctp_bind_addr *bp;
3917 struct sctp_association *asoc;
3918 struct list_head *pos, *temp;
3919 struct sctp_sockaddr_entry *addr;
3920 rwlock_t *addr_lock;
3921 int cnt = 0;
3923 if (len != sizeof(sctp_assoc_t))
3924 return -EINVAL;
3926 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3927 return -EFAULT;
3930 * For UDP-style sockets, id specifies the association to query.
3931 * If the id field is set to the value '0' then the locally bound
3932 * addresses are returned without regard to any particular
3933 * association.
3935 if (0 == id) {
3936 bp = &sctp_sk(sk)->ep->base.bind_addr;
3937 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
3938 } else {
3939 asoc = sctp_id2assoc(sk, id);
3940 if (!asoc)
3941 return -EINVAL;
3942 bp = &asoc->base.bind_addr;
3943 addr_lock = &asoc->base.addr_lock;
3946 sctp_read_lock(addr_lock);
3948 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
3949 * addresses from the global local address list.
3951 if (sctp_list_single_entry(&bp->address_list)) {
3952 addr = list_entry(bp->address_list.next,
3953 struct sctp_sockaddr_entry, list);
3954 if (sctp_is_any(&addr->a)) {
3955 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
3956 addr = list_entry(pos,
3957 struct sctp_sockaddr_entry,
3958 list);
3959 if ((PF_INET == sk->sk_family) &&
3960 (AF_INET6 == addr->a.sa.sa_family))
3961 continue;
3962 cnt++;
3964 } else {
3965 cnt = 1;
3967 goto done;
3970 list_for_each(pos, &bp->address_list) {
3971 cnt ++;
3974 done:
3975 sctp_read_unlock(addr_lock);
3976 return cnt;
3979 /* Helper function that copies local addresses to user and returns the number
3980 * of addresses copied.
3982 static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
3983 void __user *to)
3985 struct list_head *pos, *next;
3986 struct sctp_sockaddr_entry *addr;
3987 union sctp_addr temp;
3988 int cnt = 0;
3989 int addrlen;
3991 list_for_each_safe(pos, next, &sctp_local_addr_list) {
3992 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
3993 if ((PF_INET == sk->sk_family) &&
3994 (AF_INET6 == addr->a.sa.sa_family))
3995 continue;
3996 memcpy(&temp, &addr->a, sizeof(temp));
3997 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3998 &temp);
3999 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4000 if (copy_to_user(to, &temp, addrlen))
4001 return -EFAULT;
4003 to += addrlen;
4004 cnt ++;
4005 if (cnt >= max_addrs) break;
4008 return cnt;
4011 static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
4012 void __user **to, size_t space_left)
4014 struct list_head *pos, *next;
4015 struct sctp_sockaddr_entry *addr;
4016 union sctp_addr temp;
4017 int cnt = 0;
4018 int addrlen;
4020 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4021 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4022 if ((PF_INET == sk->sk_family) &&
4023 (AF_INET6 == addr->a.sa.sa_family))
4024 continue;
4025 memcpy(&temp, &addr->a, sizeof(temp));
4026 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4027 &temp);
4028 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4029 if(space_left<addrlen)
4030 return -ENOMEM;
4031 if (copy_to_user(*to, &temp, addrlen))
4032 return -EFAULT;
4034 *to += addrlen;
4035 cnt ++;
4036 space_left -= addrlen;
4039 return cnt;
4042 /* Old API for getting list of local addresses. Does not work for 32-bit
4043 * programs running on a 64-bit kernel
4045 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4046 char __user *optval, int __user *optlen)
4048 struct sctp_bind_addr *bp;
4049 struct sctp_association *asoc;
4050 struct list_head *pos;
4051 int cnt = 0;
4052 struct sctp_getaddrs_old getaddrs;
4053 struct sctp_sockaddr_entry *addr;
4054 void __user *to;
4055 union sctp_addr temp;
4056 struct sctp_sock *sp = sctp_sk(sk);
4057 int addrlen;
4058 rwlock_t *addr_lock;
4059 int err = 0;
4061 if (len != sizeof(struct sctp_getaddrs_old))
4062 return -EINVAL;
4064 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
4065 return -EFAULT;
4067 if (getaddrs.addr_num <= 0) return -EINVAL;
4069 * For UDP-style sockets, id specifies the association to query.
4070 * If the id field is set to the value '0' then the locally bound
4071 * addresses are returned without regard to any particular
4072 * association.
4074 if (0 == getaddrs.assoc_id) {
4075 bp = &sctp_sk(sk)->ep->base.bind_addr;
4076 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4077 } else {
4078 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4079 if (!asoc)
4080 return -EINVAL;
4081 bp = &asoc->base.bind_addr;
4082 addr_lock = &asoc->base.addr_lock;
4085 to = getaddrs.addrs;
4087 sctp_read_lock(addr_lock);
4089 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4090 * addresses from the global local address list.
4092 if (sctp_list_single_entry(&bp->address_list)) {
4093 addr = list_entry(bp->address_list.next,
4094 struct sctp_sockaddr_entry, list);
4095 if (sctp_is_any(&addr->a)) {
4096 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
4097 getaddrs.addr_num,
4098 to);
4099 if (cnt < 0) {
4100 err = cnt;
4101 goto unlock;
4103 goto copy_getaddrs;
4107 list_for_each(pos, &bp->address_list) {
4108 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4109 memcpy(&temp, &addr->a, sizeof(temp));
4110 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4111 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4112 if (copy_to_user(to, &temp, addrlen)) {
4113 err = -EFAULT;
4114 goto unlock;
4116 to += addrlen;
4117 cnt ++;
4118 if (cnt >= getaddrs.addr_num) break;
4121 copy_getaddrs:
4122 getaddrs.addr_num = cnt;
4123 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
4124 err = -EFAULT;
4126 unlock:
4127 sctp_read_unlock(addr_lock);
4128 return err;
4131 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4132 char __user *optval, int __user *optlen)
4134 struct sctp_bind_addr *bp;
4135 struct sctp_association *asoc;
4136 struct list_head *pos;
4137 int cnt = 0;
4138 struct sctp_getaddrs getaddrs;
4139 struct sctp_sockaddr_entry *addr;
4140 void __user *to;
4141 union sctp_addr temp;
4142 struct sctp_sock *sp = sctp_sk(sk);
4143 int addrlen;
4144 rwlock_t *addr_lock;
4145 int err = 0;
4146 size_t space_left;
4147 int bytes_copied;
4149 if (len <= sizeof(struct sctp_getaddrs))
4150 return -EINVAL;
4152 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4153 return -EFAULT;
4156 * For UDP-style sockets, id specifies the association to query.
4157 * If the id field is set to the value '0' then the locally bound
4158 * addresses are returned without regard to any particular
4159 * association.
4161 if (0 == getaddrs.assoc_id) {
4162 bp = &sctp_sk(sk)->ep->base.bind_addr;
4163 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4164 } else {
4165 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4166 if (!asoc)
4167 return -EINVAL;
4168 bp = &asoc->base.bind_addr;
4169 addr_lock = &asoc->base.addr_lock;
4172 to = optval + offsetof(struct sctp_getaddrs,addrs);
4173 space_left = len - sizeof(struct sctp_getaddrs) -
4174 offsetof(struct sctp_getaddrs,addrs);
4176 sctp_read_lock(addr_lock);
4178 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4179 * addresses from the global local address list.
4181 if (sctp_list_single_entry(&bp->address_list)) {
4182 addr = list_entry(bp->address_list.next,
4183 struct sctp_sockaddr_entry, list);
4184 if (sctp_is_any(&addr->a)) {
4185 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4186 &to, space_left);
4187 if (cnt < 0) {
4188 err = cnt;
4189 goto unlock;
4191 goto copy_getaddrs;
4195 list_for_each(pos, &bp->address_list) {
4196 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4197 memcpy(&temp, &addr->a, sizeof(temp));
4198 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4199 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4200 if(space_left < addrlen)
4201 return -ENOMEM; /*fixme: right error?*/
4202 if (copy_to_user(to, &temp, addrlen)) {
4203 err = -EFAULT;
4204 goto unlock;
4206 to += addrlen;
4207 cnt ++;
4208 space_left -= addrlen;
4211 copy_getaddrs:
4212 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4213 return -EFAULT;
4214 bytes_copied = ((char __user *)to) - optval;
4215 if (put_user(bytes_copied, optlen))
4216 return -EFAULT;
4218 unlock:
4219 sctp_read_unlock(addr_lock);
4220 return err;
4223 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4225 * Requests that the local SCTP stack use the enclosed peer address as
4226 * the association primary. The enclosed address must be one of the
4227 * association peer's addresses.
4229 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4230 char __user *optval, int __user *optlen)
4232 struct sctp_prim prim;
4233 struct sctp_association *asoc;
4234 struct sctp_sock *sp = sctp_sk(sk);
4236 if (len != sizeof(struct sctp_prim))
4237 return -EINVAL;
4239 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4240 return -EFAULT;
4242 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4243 if (!asoc)
4244 return -EINVAL;
4246 if (!asoc->peer.primary_path)
4247 return -ENOTCONN;
4249 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4250 asoc->peer.primary_path->af_specific->sockaddr_len);
4252 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4253 (union sctp_addr *)&prim.ssp_addr);
4255 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4256 return -EFAULT;
4258 return 0;
4262 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4264 * Requests that the local endpoint set the specified Adaptation Layer
4265 * Indication parameter for all future INIT and INIT-ACK exchanges.
4267 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4268 char __user *optval, int __user *optlen)
4270 struct sctp_setadaptation adaptation;
4272 if (len != sizeof(struct sctp_setadaptation))
4273 return -EINVAL;
4275 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4276 if (copy_to_user(optval, &adaptation, len))
4277 return -EFAULT;
4279 return 0;
4284 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4286 * Applications that wish to use the sendto() system call may wish to
4287 * specify a default set of parameters that would normally be supplied
4288 * through the inclusion of ancillary data. This socket option allows
4289 * such an application to set the default sctp_sndrcvinfo structure.
4292 * The application that wishes to use this socket option simply passes
4293 * in to this call the sctp_sndrcvinfo structure defined in Section
4294 * 5.2.2) The input parameters accepted by this call include
4295 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4296 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4297 * to this call if the caller is using the UDP model.
4299 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4301 static int sctp_getsockopt_default_send_param(struct sock *sk,
4302 int len, char __user *optval,
4303 int __user *optlen)
4305 struct sctp_sndrcvinfo info;
4306 struct sctp_association *asoc;
4307 struct sctp_sock *sp = sctp_sk(sk);
4309 if (len != sizeof(struct sctp_sndrcvinfo))
4310 return -EINVAL;
4311 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4312 return -EFAULT;
4314 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4315 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4316 return -EINVAL;
4318 if (asoc) {
4319 info.sinfo_stream = asoc->default_stream;
4320 info.sinfo_flags = asoc->default_flags;
4321 info.sinfo_ppid = asoc->default_ppid;
4322 info.sinfo_context = asoc->default_context;
4323 info.sinfo_timetolive = asoc->default_timetolive;
4324 } else {
4325 info.sinfo_stream = sp->default_stream;
4326 info.sinfo_flags = sp->default_flags;
4327 info.sinfo_ppid = sp->default_ppid;
4328 info.sinfo_context = sp->default_context;
4329 info.sinfo_timetolive = sp->default_timetolive;
4332 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4333 return -EFAULT;
4335 return 0;
4340 * 7.1.5 SCTP_NODELAY
4342 * Turn on/off any Nagle-like algorithm. This means that packets are
4343 * generally sent as soon as possible and no unnecessary delays are
4344 * introduced, at the cost of more packets in the network. Expects an
4345 * integer boolean flag.
4348 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4349 char __user *optval, int __user *optlen)
4351 int val;
4353 if (len < sizeof(int))
4354 return -EINVAL;
4356 len = sizeof(int);
4357 val = (sctp_sk(sk)->nodelay == 1);
4358 if (put_user(len, optlen))
4359 return -EFAULT;
4360 if (copy_to_user(optval, &val, len))
4361 return -EFAULT;
4362 return 0;
4367 * 7.1.1 SCTP_RTOINFO
4369 * The protocol parameters used to initialize and bound retransmission
4370 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4371 * and modify these parameters.
4372 * All parameters are time values, in milliseconds. A value of 0, when
4373 * modifying the parameters, indicates that the current value should not
4374 * be changed.
4377 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4378 char __user *optval,
4379 int __user *optlen) {
4380 struct sctp_rtoinfo rtoinfo;
4381 struct sctp_association *asoc;
4383 if (len != sizeof (struct sctp_rtoinfo))
4384 return -EINVAL;
4386 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4387 return -EFAULT;
4389 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4391 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4392 return -EINVAL;
4394 /* Values corresponding to the specific association. */
4395 if (asoc) {
4396 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4397 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4398 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4399 } else {
4400 /* Values corresponding to the endpoint. */
4401 struct sctp_sock *sp = sctp_sk(sk);
4403 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4404 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4405 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4408 if (put_user(len, optlen))
4409 return -EFAULT;
4411 if (copy_to_user(optval, &rtoinfo, len))
4412 return -EFAULT;
4414 return 0;
4419 * 7.1.2 SCTP_ASSOCINFO
4421 * This option is used to tune the the maximum retransmission attempts
4422 * of the association.
4423 * Returns an error if the new association retransmission value is
4424 * greater than the sum of the retransmission value of the peer.
4425 * See [SCTP] for more information.
4428 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4429 char __user *optval,
4430 int __user *optlen)
4433 struct sctp_assocparams assocparams;
4434 struct sctp_association *asoc;
4435 struct list_head *pos;
4436 int cnt = 0;
4438 if (len != sizeof (struct sctp_assocparams))
4439 return -EINVAL;
4441 if (copy_from_user(&assocparams, optval,
4442 sizeof (struct sctp_assocparams)))
4443 return -EFAULT;
4445 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4447 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4448 return -EINVAL;
4450 /* Values correspoinding to the specific association */
4451 if (asoc) {
4452 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4453 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4454 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4455 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4456 * 1000) +
4457 (asoc->cookie_life.tv_usec
4458 / 1000);
4460 list_for_each(pos, &asoc->peer.transport_addr_list) {
4461 cnt ++;
4464 assocparams.sasoc_number_peer_destinations = cnt;
4465 } else {
4466 /* Values corresponding to the endpoint */
4467 struct sctp_sock *sp = sctp_sk(sk);
4469 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4470 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4471 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4472 assocparams.sasoc_cookie_life =
4473 sp->assocparams.sasoc_cookie_life;
4474 assocparams.sasoc_number_peer_destinations =
4475 sp->assocparams.
4476 sasoc_number_peer_destinations;
4479 if (put_user(len, optlen))
4480 return -EFAULT;
4482 if (copy_to_user(optval, &assocparams, len))
4483 return -EFAULT;
4485 return 0;
4489 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4491 * This socket option is a boolean flag which turns on or off mapped V4
4492 * addresses. If this option is turned on and the socket is type
4493 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4494 * If this option is turned off, then no mapping will be done of V4
4495 * addresses and a user will receive both PF_INET6 and PF_INET type
4496 * addresses on the socket.
4498 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4499 char __user *optval, int __user *optlen)
4501 int val;
4502 struct sctp_sock *sp = sctp_sk(sk);
4504 if (len < sizeof(int))
4505 return -EINVAL;
4507 len = sizeof(int);
4508 val = sp->v4mapped;
4509 if (put_user(len, optlen))
4510 return -EFAULT;
4511 if (copy_to_user(optval, &val, len))
4512 return -EFAULT;
4514 return 0;
4518 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4519 * (chapter and verse is quoted at sctp_setsockopt_context())
4521 static int sctp_getsockopt_context(struct sock *sk, int len,
4522 char __user *optval, int __user *optlen)
4524 struct sctp_assoc_value params;
4525 struct sctp_sock *sp;
4526 struct sctp_association *asoc;
4528 if (len != sizeof(struct sctp_assoc_value))
4529 return -EINVAL;
4531 if (copy_from_user(&params, optval, len))
4532 return -EFAULT;
4534 sp = sctp_sk(sk);
4536 if (params.assoc_id != 0) {
4537 asoc = sctp_id2assoc(sk, params.assoc_id);
4538 if (!asoc)
4539 return -EINVAL;
4540 params.assoc_value = asoc->default_rcv_context;
4541 } else {
4542 params.assoc_value = sp->default_rcv_context;
4545 if (put_user(len, optlen))
4546 return -EFAULT;
4547 if (copy_to_user(optval, &params, len))
4548 return -EFAULT;
4550 return 0;
4554 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4556 * This socket option specifies the maximum size to put in any outgoing
4557 * SCTP chunk. If a message is larger than this size it will be
4558 * fragmented by SCTP into the specified size. Note that the underlying
4559 * SCTP implementation may fragment into smaller sized chunks when the
4560 * PMTU of the underlying association is smaller than the value set by
4561 * the user.
4563 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4564 char __user *optval, int __user *optlen)
4566 int val;
4568 if (len < sizeof(int))
4569 return -EINVAL;
4571 len = sizeof(int);
4573 val = sctp_sk(sk)->user_frag;
4574 if (put_user(len, optlen))
4575 return -EFAULT;
4576 if (copy_to_user(optval, &val, len))
4577 return -EFAULT;
4579 return 0;
4583 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4584 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4586 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4587 char __user *optval, int __user *optlen)
4589 int val;
4591 if (len < sizeof(int))
4592 return -EINVAL;
4594 len = sizeof(int);
4596 val = sctp_sk(sk)->frag_interleave;
4597 if (put_user(len, optlen))
4598 return -EFAULT;
4599 if (copy_to_user(optval, &val, len))
4600 return -EFAULT;
4602 return 0;
4605 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4606 char __user *optval, int __user *optlen)
4608 int retval = 0;
4609 int len;
4611 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4612 sk, optname);
4614 /* I can hardly begin to describe how wrong this is. This is
4615 * so broken as to be worse than useless. The API draft
4616 * REALLY is NOT helpful here... I am not convinced that the
4617 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4618 * are at all well-founded.
4620 if (level != SOL_SCTP) {
4621 struct sctp_af *af = sctp_sk(sk)->pf->af;
4623 retval = af->getsockopt(sk, level, optname, optval, optlen);
4624 return retval;
4627 if (get_user(len, optlen))
4628 return -EFAULT;
4630 sctp_lock_sock(sk);
4632 switch (optname) {
4633 case SCTP_STATUS:
4634 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4635 break;
4636 case SCTP_DISABLE_FRAGMENTS:
4637 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4638 optlen);
4639 break;
4640 case SCTP_EVENTS:
4641 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4642 break;
4643 case SCTP_AUTOCLOSE:
4644 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4645 break;
4646 case SCTP_SOCKOPT_PEELOFF:
4647 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4648 break;
4649 case SCTP_PEER_ADDR_PARAMS:
4650 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4651 optlen);
4652 break;
4653 case SCTP_DELAYED_ACK_TIME:
4654 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4655 optlen);
4656 break;
4657 case SCTP_INITMSG:
4658 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4659 break;
4660 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4661 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4662 optlen);
4663 break;
4664 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4665 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4666 optlen);
4667 break;
4668 case SCTP_GET_PEER_ADDRS_OLD:
4669 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4670 optlen);
4671 break;
4672 case SCTP_GET_LOCAL_ADDRS_OLD:
4673 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4674 optlen);
4675 break;
4676 case SCTP_GET_PEER_ADDRS:
4677 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4678 optlen);
4679 break;
4680 case SCTP_GET_LOCAL_ADDRS:
4681 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4682 optlen);
4683 break;
4684 case SCTP_DEFAULT_SEND_PARAM:
4685 retval = sctp_getsockopt_default_send_param(sk, len,
4686 optval, optlen);
4687 break;
4688 case SCTP_PRIMARY_ADDR:
4689 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4690 break;
4691 case SCTP_NODELAY:
4692 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4693 break;
4694 case SCTP_RTOINFO:
4695 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4696 break;
4697 case SCTP_ASSOCINFO:
4698 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4699 break;
4700 case SCTP_I_WANT_MAPPED_V4_ADDR:
4701 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4702 break;
4703 case SCTP_MAXSEG:
4704 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4705 break;
4706 case SCTP_GET_PEER_ADDR_INFO:
4707 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4708 optlen);
4709 break;
4710 case SCTP_ADAPTATION_LAYER:
4711 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
4712 optlen);
4713 break;
4714 case SCTP_CONTEXT:
4715 retval = sctp_getsockopt_context(sk, len, optval, optlen);
4716 break;
4717 case SCTP_FRAGMENT_INTERLEAVE:
4718 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
4719 optlen);
4720 break;
4721 default:
4722 retval = -ENOPROTOOPT;
4723 break;
4726 sctp_release_sock(sk);
4727 return retval;
4730 static void sctp_hash(struct sock *sk)
4732 /* STUB */
4735 static void sctp_unhash(struct sock *sk)
4737 /* STUB */
4740 /* Check if port is acceptable. Possibly find first available port.
4742 * The port hash table (contained in the 'global' SCTP protocol storage
4743 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4744 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4745 * list (the list number is the port number hashed out, so as you
4746 * would expect from a hash function, all the ports in a given list have
4747 * such a number that hashes out to the same list number; you were
4748 * expecting that, right?); so each list has a set of ports, with a
4749 * link to the socket (struct sock) that uses it, the port number and
4750 * a fastreuse flag (FIXME: NPI ipg).
4752 static struct sctp_bind_bucket *sctp_bucket_create(
4753 struct sctp_bind_hashbucket *head, unsigned short snum);
4755 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4757 struct sctp_bind_hashbucket *head; /* hash list */
4758 struct sctp_bind_bucket *pp; /* hash list port iterator */
4759 unsigned short snum;
4760 int ret;
4762 snum = ntohs(addr->v4.sin_port);
4764 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4765 sctp_local_bh_disable();
4767 if (snum == 0) {
4768 /* Search for an available port.
4770 * 'sctp_port_rover' was the last port assigned, so
4771 * we start to search from 'sctp_port_rover +
4772 * 1'. What we do is first check if port 'rover' is
4773 * already in the hash table; if not, we use that; if
4774 * it is, we try next.
4776 int low = sysctl_local_port_range[0];
4777 int high = sysctl_local_port_range[1];
4778 int remaining = (high - low) + 1;
4779 int rover;
4780 int index;
4782 sctp_spin_lock(&sctp_port_alloc_lock);
4783 rover = sctp_port_rover;
4784 do {
4785 rover++;
4786 if ((rover < low) || (rover > high))
4787 rover = low;
4788 index = sctp_phashfn(rover);
4789 head = &sctp_port_hashtable[index];
4790 sctp_spin_lock(&head->lock);
4791 for (pp = head->chain; pp; pp = pp->next)
4792 if (pp->port == rover)
4793 goto next;
4794 break;
4795 next:
4796 sctp_spin_unlock(&head->lock);
4797 } while (--remaining > 0);
4798 sctp_port_rover = rover;
4799 sctp_spin_unlock(&sctp_port_alloc_lock);
4801 /* Exhausted local port range during search? */
4802 ret = 1;
4803 if (remaining <= 0)
4804 goto fail;
4806 /* OK, here is the one we will use. HEAD (the port
4807 * hash table list entry) is non-NULL and we hold it's
4808 * mutex.
4810 snum = rover;
4811 } else {
4812 /* We are given an specific port number; we verify
4813 * that it is not being used. If it is used, we will
4814 * exahust the search in the hash list corresponding
4815 * to the port number (snum) - we detect that with the
4816 * port iterator, pp being NULL.
4818 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4819 sctp_spin_lock(&head->lock);
4820 for (pp = head->chain; pp; pp = pp->next) {
4821 if (pp->port == snum)
4822 goto pp_found;
4825 pp = NULL;
4826 goto pp_not_found;
4827 pp_found:
4828 if (!hlist_empty(&pp->owner)) {
4829 /* We had a port hash table hit - there is an
4830 * available port (pp != NULL) and it is being
4831 * used by other socket (pp->owner not empty); that other
4832 * socket is going to be sk2.
4834 int reuse = sk->sk_reuse;
4835 struct sock *sk2;
4836 struct hlist_node *node;
4838 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4839 if (pp->fastreuse && sk->sk_reuse)
4840 goto success;
4842 /* Run through the list of sockets bound to the port
4843 * (pp->port) [via the pointers bind_next and
4844 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4845 * we get the endpoint they describe and run through
4846 * the endpoint's list of IP (v4 or v6) addresses,
4847 * comparing each of the addresses with the address of
4848 * the socket sk. If we find a match, then that means
4849 * that this port/socket (sk) combination are already
4850 * in an endpoint.
4852 sk_for_each_bound(sk2, node, &pp->owner) {
4853 struct sctp_endpoint *ep2;
4854 ep2 = sctp_sk(sk2)->ep;
4856 if (reuse && sk2->sk_reuse)
4857 continue;
4859 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
4860 sctp_sk(sk))) {
4861 ret = (long)sk2;
4862 goto fail_unlock;
4865 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
4867 pp_not_found:
4868 /* If there was a hash table miss, create a new port. */
4869 ret = 1;
4870 if (!pp && !(pp = sctp_bucket_create(head, snum)))
4871 goto fail_unlock;
4873 /* In either case (hit or miss), make sure fastreuse is 1 only
4874 * if sk->sk_reuse is too (that is, if the caller requested
4875 * SO_REUSEADDR on this socket -sk-).
4877 if (hlist_empty(&pp->owner))
4878 pp->fastreuse = sk->sk_reuse ? 1 : 0;
4879 else if (pp->fastreuse && !sk->sk_reuse)
4880 pp->fastreuse = 0;
4882 /* We are set, so fill up all the data in the hash table
4883 * entry, tie the socket list information with the rest of the
4884 * sockets FIXME: Blurry, NPI (ipg).
4886 success:
4887 inet_sk(sk)->num = snum;
4888 if (!sctp_sk(sk)->bind_hash) {
4889 sk_add_bind_node(sk, &pp->owner);
4890 sctp_sk(sk)->bind_hash = pp;
4892 ret = 0;
4894 fail_unlock:
4895 sctp_spin_unlock(&head->lock);
4897 fail:
4898 sctp_local_bh_enable();
4899 return ret;
4902 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
4903 * port is requested.
4905 static int sctp_get_port(struct sock *sk, unsigned short snum)
4907 long ret;
4908 union sctp_addr addr;
4909 struct sctp_af *af = sctp_sk(sk)->pf->af;
4911 /* Set up a dummy address struct from the sk. */
4912 af->from_sk(&addr, sk);
4913 addr.v4.sin_port = htons(snum);
4915 /* Note: sk->sk_num gets filled in if ephemeral port request. */
4916 ret = sctp_get_port_local(sk, &addr);
4918 return (ret ? 1 : 0);
4922 * 3.1.3 listen() - UDP Style Syntax
4924 * By default, new associations are not accepted for UDP style sockets.
4925 * An application uses listen() to mark a socket as being able to
4926 * accept new associations.
4928 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
4930 struct sctp_sock *sp = sctp_sk(sk);
4931 struct sctp_endpoint *ep = sp->ep;
4933 /* Only UDP style sockets that are not peeled off are allowed to
4934 * listen().
4936 if (!sctp_style(sk, UDP))
4937 return -EINVAL;
4939 /* If backlog is zero, disable listening. */
4940 if (!backlog) {
4941 if (sctp_sstate(sk, CLOSED))
4942 return 0;
4944 sctp_unhash_endpoint(ep);
4945 sk->sk_state = SCTP_SS_CLOSED;
4948 /* Return if we are already listening. */
4949 if (sctp_sstate(sk, LISTENING))
4950 return 0;
4953 * If a bind() or sctp_bindx() is not called prior to a listen()
4954 * call that allows new associations to be accepted, the system
4955 * picks an ephemeral port and will choose an address set equivalent
4956 * to binding with a wildcard address.
4958 * This is not currently spelled out in the SCTP sockets
4959 * extensions draft, but follows the practice as seen in TCP
4960 * sockets.
4962 if (!ep->base.bind_addr.port) {
4963 if (sctp_autobind(sk))
4964 return -EAGAIN;
4966 sk->sk_state = SCTP_SS_LISTENING;
4967 sctp_hash_endpoint(ep);
4968 return 0;
4972 * 4.1.3 listen() - TCP Style Syntax
4974 * Applications uses listen() to ready the SCTP endpoint for accepting
4975 * inbound associations.
4977 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
4979 struct sctp_sock *sp = sctp_sk(sk);
4980 struct sctp_endpoint *ep = sp->ep;
4982 /* If backlog is zero, disable listening. */
4983 if (!backlog) {
4984 if (sctp_sstate(sk, CLOSED))
4985 return 0;
4987 sctp_unhash_endpoint(ep);
4988 sk->sk_state = SCTP_SS_CLOSED;
4991 if (sctp_sstate(sk, LISTENING))
4992 return 0;
4995 * If a bind() or sctp_bindx() is not called prior to a listen()
4996 * call that allows new associations to be accepted, the system
4997 * picks an ephemeral port and will choose an address set equivalent
4998 * to binding with a wildcard address.
5000 * This is not currently spelled out in the SCTP sockets
5001 * extensions draft, but follows the practice as seen in TCP
5002 * sockets.
5004 if (!ep->base.bind_addr.port) {
5005 if (sctp_autobind(sk))
5006 return -EAGAIN;
5008 sk->sk_state = SCTP_SS_LISTENING;
5009 sk->sk_max_ack_backlog = backlog;
5010 sctp_hash_endpoint(ep);
5011 return 0;
5015 * Move a socket to LISTENING state.
5017 int sctp_inet_listen(struct socket *sock, int backlog)
5019 struct sock *sk = sock->sk;
5020 struct crypto_hash *tfm = NULL;
5021 int err = -EINVAL;
5023 if (unlikely(backlog < 0))
5024 goto out;
5026 sctp_lock_sock(sk);
5028 if (sock->state != SS_UNCONNECTED)
5029 goto out;
5031 /* Allocate HMAC for generating cookie. */
5032 if (sctp_hmac_alg) {
5033 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5034 if (!tfm) {
5035 err = -ENOSYS;
5036 goto out;
5040 switch (sock->type) {
5041 case SOCK_SEQPACKET:
5042 err = sctp_seqpacket_listen(sk, backlog);
5043 break;
5044 case SOCK_STREAM:
5045 err = sctp_stream_listen(sk, backlog);
5046 break;
5047 default:
5048 break;
5050 if (err)
5051 goto cleanup;
5053 /* Store away the transform reference. */
5054 sctp_sk(sk)->hmac = tfm;
5055 out:
5056 sctp_release_sock(sk);
5057 return err;
5058 cleanup:
5059 crypto_free_hash(tfm);
5060 goto out;
5064 * This function is done by modeling the current datagram_poll() and the
5065 * tcp_poll(). Note that, based on these implementations, we don't
5066 * lock the socket in this function, even though it seems that,
5067 * ideally, locking or some other mechanisms can be used to ensure
5068 * the integrity of the counters (sndbuf and wmem_alloc) used
5069 * in this place. We assume that we don't need locks either until proven
5070 * otherwise.
5072 * Another thing to note is that we include the Async I/O support
5073 * here, again, by modeling the current TCP/UDP code. We don't have
5074 * a good way to test with it yet.
5076 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5078 struct sock *sk = sock->sk;
5079 struct sctp_sock *sp = sctp_sk(sk);
5080 unsigned int mask;
5082 poll_wait(file, sk->sk_sleep, wait);
5084 /* A TCP-style listening socket becomes readable when the accept queue
5085 * is not empty.
5087 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5088 return (!list_empty(&sp->ep->asocs)) ?
5089 (POLLIN | POLLRDNORM) : 0;
5091 mask = 0;
5093 /* Is there any exceptional events? */
5094 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5095 mask |= POLLERR;
5096 if (sk->sk_shutdown & RCV_SHUTDOWN)
5097 mask |= POLLRDHUP;
5098 if (sk->sk_shutdown == SHUTDOWN_MASK)
5099 mask |= POLLHUP;
5101 /* Is it readable? Reconsider this code with TCP-style support. */
5102 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5103 (sk->sk_shutdown & RCV_SHUTDOWN))
5104 mask |= POLLIN | POLLRDNORM;
5106 /* The association is either gone or not ready. */
5107 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5108 return mask;
5110 /* Is it writable? */
5111 if (sctp_writeable(sk)) {
5112 mask |= POLLOUT | POLLWRNORM;
5113 } else {
5114 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5116 * Since the socket is not locked, the buffer
5117 * might be made available after the writeable check and
5118 * before the bit is set. This could cause a lost I/O
5119 * signal. tcp_poll() has a race breaker for this race
5120 * condition. Based on their implementation, we put
5121 * in the following code to cover it as well.
5123 if (sctp_writeable(sk))
5124 mask |= POLLOUT | POLLWRNORM;
5126 return mask;
5129 /********************************************************************
5130 * 2nd Level Abstractions
5131 ********************************************************************/
5133 static struct sctp_bind_bucket *sctp_bucket_create(
5134 struct sctp_bind_hashbucket *head, unsigned short snum)
5136 struct sctp_bind_bucket *pp;
5138 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5139 SCTP_DBG_OBJCNT_INC(bind_bucket);
5140 if (pp) {
5141 pp->port = snum;
5142 pp->fastreuse = 0;
5143 INIT_HLIST_HEAD(&pp->owner);
5144 if ((pp->next = head->chain) != NULL)
5145 pp->next->pprev = &pp->next;
5146 head->chain = pp;
5147 pp->pprev = &head->chain;
5149 return pp;
5152 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5153 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5155 if (pp && hlist_empty(&pp->owner)) {
5156 if (pp->next)
5157 pp->next->pprev = pp->pprev;
5158 *(pp->pprev) = pp->next;
5159 kmem_cache_free(sctp_bucket_cachep, pp);
5160 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5164 /* Release this socket's reference to a local port. */
5165 static inline void __sctp_put_port(struct sock *sk)
5167 struct sctp_bind_hashbucket *head =
5168 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5169 struct sctp_bind_bucket *pp;
5171 sctp_spin_lock(&head->lock);
5172 pp = sctp_sk(sk)->bind_hash;
5173 __sk_del_bind_node(sk);
5174 sctp_sk(sk)->bind_hash = NULL;
5175 inet_sk(sk)->num = 0;
5176 sctp_bucket_destroy(pp);
5177 sctp_spin_unlock(&head->lock);
5180 void sctp_put_port(struct sock *sk)
5182 sctp_local_bh_disable();
5183 __sctp_put_port(sk);
5184 sctp_local_bh_enable();
5188 * The system picks an ephemeral port and choose an address set equivalent
5189 * to binding with a wildcard address.
5190 * One of those addresses will be the primary address for the association.
5191 * This automatically enables the multihoming capability of SCTP.
5193 static int sctp_autobind(struct sock *sk)
5195 union sctp_addr autoaddr;
5196 struct sctp_af *af;
5197 __be16 port;
5199 /* Initialize a local sockaddr structure to INADDR_ANY. */
5200 af = sctp_sk(sk)->pf->af;
5202 port = htons(inet_sk(sk)->num);
5203 af->inaddr_any(&autoaddr, port);
5205 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5208 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5210 * From RFC 2292
5211 * 4.2 The cmsghdr Structure *
5213 * When ancillary data is sent or received, any number of ancillary data
5214 * objects can be specified by the msg_control and msg_controllen members of
5215 * the msghdr structure, because each object is preceded by
5216 * a cmsghdr structure defining the object's length (the cmsg_len member).
5217 * Historically Berkeley-derived implementations have passed only one object
5218 * at a time, but this API allows multiple objects to be
5219 * passed in a single call to sendmsg() or recvmsg(). The following example
5220 * shows two ancillary data objects in a control buffer.
5222 * |<--------------------------- msg_controllen -------------------------->|
5223 * | |
5225 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5227 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5228 * | | |
5230 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5232 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5233 * | | | | |
5235 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5236 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5238 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5240 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5244 * msg_control
5245 * points here
5247 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5248 sctp_cmsgs_t *cmsgs)
5250 struct cmsghdr *cmsg;
5252 for (cmsg = CMSG_FIRSTHDR(msg);
5253 cmsg != NULL;
5254 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5255 if (!CMSG_OK(msg, cmsg))
5256 return -EINVAL;
5258 /* Should we parse this header or ignore? */
5259 if (cmsg->cmsg_level != IPPROTO_SCTP)
5260 continue;
5262 /* Strictly check lengths following example in SCM code. */
5263 switch (cmsg->cmsg_type) {
5264 case SCTP_INIT:
5265 /* SCTP Socket API Extension
5266 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5268 * This cmsghdr structure provides information for
5269 * initializing new SCTP associations with sendmsg().
5270 * The SCTP_INITMSG socket option uses this same data
5271 * structure. This structure is not used for
5272 * recvmsg().
5274 * cmsg_level cmsg_type cmsg_data[]
5275 * ------------ ------------ ----------------------
5276 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5278 if (cmsg->cmsg_len !=
5279 CMSG_LEN(sizeof(struct sctp_initmsg)))
5280 return -EINVAL;
5281 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5282 break;
5284 case SCTP_SNDRCV:
5285 /* SCTP Socket API Extension
5286 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5288 * This cmsghdr structure specifies SCTP options for
5289 * sendmsg() and describes SCTP header information
5290 * about a received message through recvmsg().
5292 * cmsg_level cmsg_type cmsg_data[]
5293 * ------------ ------------ ----------------------
5294 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5296 if (cmsg->cmsg_len !=
5297 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5298 return -EINVAL;
5300 cmsgs->info =
5301 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5303 /* Minimally, validate the sinfo_flags. */
5304 if (cmsgs->info->sinfo_flags &
5305 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5306 SCTP_ABORT | SCTP_EOF))
5307 return -EINVAL;
5308 break;
5310 default:
5311 return -EINVAL;
5314 return 0;
5318 * Wait for a packet..
5319 * Note: This function is the same function as in core/datagram.c
5320 * with a few modifications to make lksctp work.
5322 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5324 int error;
5325 DEFINE_WAIT(wait);
5327 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5329 /* Socket errors? */
5330 error = sock_error(sk);
5331 if (error)
5332 goto out;
5334 if (!skb_queue_empty(&sk->sk_receive_queue))
5335 goto ready;
5337 /* Socket shut down? */
5338 if (sk->sk_shutdown & RCV_SHUTDOWN)
5339 goto out;
5341 /* Sequenced packets can come disconnected. If so we report the
5342 * problem.
5344 error = -ENOTCONN;
5346 /* Is there a good reason to think that we may receive some data? */
5347 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5348 goto out;
5350 /* Handle signals. */
5351 if (signal_pending(current))
5352 goto interrupted;
5354 /* Let another process have a go. Since we are going to sleep
5355 * anyway. Note: This may cause odd behaviors if the message
5356 * does not fit in the user's buffer, but this seems to be the
5357 * only way to honor MSG_DONTWAIT realistically.
5359 sctp_release_sock(sk);
5360 *timeo_p = schedule_timeout(*timeo_p);
5361 sctp_lock_sock(sk);
5363 ready:
5364 finish_wait(sk->sk_sleep, &wait);
5365 return 0;
5367 interrupted:
5368 error = sock_intr_errno(*timeo_p);
5370 out:
5371 finish_wait(sk->sk_sleep, &wait);
5372 *err = error;
5373 return error;
5376 /* Receive a datagram.
5377 * Note: This is pretty much the same routine as in core/datagram.c
5378 * with a few changes to make lksctp work.
5380 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5381 int noblock, int *err)
5383 int error;
5384 struct sk_buff *skb;
5385 long timeo;
5387 timeo = sock_rcvtimeo(sk, noblock);
5389 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5390 timeo, MAX_SCHEDULE_TIMEOUT);
5392 do {
5393 /* Again only user level code calls this function,
5394 * so nothing interrupt level
5395 * will suddenly eat the receive_queue.
5397 * Look at current nfs client by the way...
5398 * However, this function was corrent in any case. 8)
5400 if (flags & MSG_PEEK) {
5401 spin_lock_bh(&sk->sk_receive_queue.lock);
5402 skb = skb_peek(&sk->sk_receive_queue);
5403 if (skb)
5404 atomic_inc(&skb->users);
5405 spin_unlock_bh(&sk->sk_receive_queue.lock);
5406 } else {
5407 skb = skb_dequeue(&sk->sk_receive_queue);
5410 if (skb)
5411 return skb;
5413 /* Caller is allowed not to check sk->sk_err before calling. */
5414 error = sock_error(sk);
5415 if (error)
5416 goto no_packet;
5418 if (sk->sk_shutdown & RCV_SHUTDOWN)
5419 break;
5421 /* User doesn't want to wait. */
5422 error = -EAGAIN;
5423 if (!timeo)
5424 goto no_packet;
5425 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5427 return NULL;
5429 no_packet:
5430 *err = error;
5431 return NULL;
5434 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5435 static void __sctp_write_space(struct sctp_association *asoc)
5437 struct sock *sk = asoc->base.sk;
5438 struct socket *sock = sk->sk_socket;
5440 if ((sctp_wspace(asoc) > 0) && sock) {
5441 if (waitqueue_active(&asoc->wait))
5442 wake_up_interruptible(&asoc->wait);
5444 if (sctp_writeable(sk)) {
5445 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5446 wake_up_interruptible(sk->sk_sleep);
5448 /* Note that we try to include the Async I/O support
5449 * here by modeling from the current TCP/UDP code.
5450 * We have not tested with it yet.
5452 if (sock->fasync_list &&
5453 !(sk->sk_shutdown & SEND_SHUTDOWN))
5454 sock_wake_async(sock, 2, POLL_OUT);
5459 /* Do accounting for the sndbuf space.
5460 * Decrement the used sndbuf space of the corresponding association by the
5461 * data size which was just transmitted(freed).
5463 static void sctp_wfree(struct sk_buff *skb)
5465 struct sctp_association *asoc;
5466 struct sctp_chunk *chunk;
5467 struct sock *sk;
5469 /* Get the saved chunk pointer. */
5470 chunk = *((struct sctp_chunk **)(skb->cb));
5471 asoc = chunk->asoc;
5472 sk = asoc->base.sk;
5473 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5474 sizeof(struct sk_buff) +
5475 sizeof(struct sctp_chunk);
5477 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5479 sock_wfree(skb);
5480 __sctp_write_space(asoc);
5482 sctp_association_put(asoc);
5485 /* Do accounting for the receive space on the socket.
5486 * Accounting for the association is done in ulpevent.c
5487 * We set this as a destructor for the cloned data skbs so that
5488 * accounting is done at the correct time.
5490 void sctp_sock_rfree(struct sk_buff *skb)
5492 struct sock *sk = skb->sk;
5493 struct sctp_ulpevent *event = sctp_skb2event(skb);
5495 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
5499 /* Helper function to wait for space in the sndbuf. */
5500 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5501 size_t msg_len)
5503 struct sock *sk = asoc->base.sk;
5504 int err = 0;
5505 long current_timeo = *timeo_p;
5506 DEFINE_WAIT(wait);
5508 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5509 asoc, (long)(*timeo_p), msg_len);
5511 /* Increment the association's refcnt. */
5512 sctp_association_hold(asoc);
5514 /* Wait on the association specific sndbuf space. */
5515 for (;;) {
5516 prepare_to_wait_exclusive(&asoc->wait, &wait,
5517 TASK_INTERRUPTIBLE);
5518 if (!*timeo_p)
5519 goto do_nonblock;
5520 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5521 asoc->base.dead)
5522 goto do_error;
5523 if (signal_pending(current))
5524 goto do_interrupted;
5525 if (msg_len <= sctp_wspace(asoc))
5526 break;
5528 /* Let another process have a go. Since we are going
5529 * to sleep anyway.
5531 sctp_release_sock(sk);
5532 current_timeo = schedule_timeout(current_timeo);
5533 BUG_ON(sk != asoc->base.sk);
5534 sctp_lock_sock(sk);
5536 *timeo_p = current_timeo;
5539 out:
5540 finish_wait(&asoc->wait, &wait);
5542 /* Release the association's refcnt. */
5543 sctp_association_put(asoc);
5545 return err;
5547 do_error:
5548 err = -EPIPE;
5549 goto out;
5551 do_interrupted:
5552 err = sock_intr_errno(*timeo_p);
5553 goto out;
5555 do_nonblock:
5556 err = -EAGAIN;
5557 goto out;
5560 /* If socket sndbuf has changed, wake up all per association waiters. */
5561 void sctp_write_space(struct sock *sk)
5563 struct sctp_association *asoc;
5564 struct list_head *pos;
5566 /* Wake up the tasks in each wait queue. */
5567 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5568 asoc = list_entry(pos, struct sctp_association, asocs);
5569 __sctp_write_space(asoc);
5573 /* Is there any sndbuf space available on the socket?
5575 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5576 * associations on the same socket. For a UDP-style socket with
5577 * multiple associations, it is possible for it to be "unwriteable"
5578 * prematurely. I assume that this is acceptable because
5579 * a premature "unwriteable" is better than an accidental "writeable" which
5580 * would cause an unwanted block under certain circumstances. For the 1-1
5581 * UDP-style sockets or TCP-style sockets, this code should work.
5582 * - Daisy
5584 static int sctp_writeable(struct sock *sk)
5586 int amt = 0;
5588 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5589 if (amt < 0)
5590 amt = 0;
5591 return amt;
5594 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5595 * returns immediately with EINPROGRESS.
5597 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5599 struct sock *sk = asoc->base.sk;
5600 int err = 0;
5601 long current_timeo = *timeo_p;
5602 DEFINE_WAIT(wait);
5604 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5605 (long)(*timeo_p));
5607 /* Increment the association's refcnt. */
5608 sctp_association_hold(asoc);
5610 for (;;) {
5611 prepare_to_wait_exclusive(&asoc->wait, &wait,
5612 TASK_INTERRUPTIBLE);
5613 if (!*timeo_p)
5614 goto do_nonblock;
5615 if (sk->sk_shutdown & RCV_SHUTDOWN)
5616 break;
5617 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5618 asoc->base.dead)
5619 goto do_error;
5620 if (signal_pending(current))
5621 goto do_interrupted;
5623 if (sctp_state(asoc, ESTABLISHED))
5624 break;
5626 /* Let another process have a go. Since we are going
5627 * to sleep anyway.
5629 sctp_release_sock(sk);
5630 current_timeo = schedule_timeout(current_timeo);
5631 sctp_lock_sock(sk);
5633 *timeo_p = current_timeo;
5636 out:
5637 finish_wait(&asoc->wait, &wait);
5639 /* Release the association's refcnt. */
5640 sctp_association_put(asoc);
5642 return err;
5644 do_error:
5645 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5646 err = -ETIMEDOUT;
5647 else
5648 err = -ECONNREFUSED;
5649 goto out;
5651 do_interrupted:
5652 err = sock_intr_errno(*timeo_p);
5653 goto out;
5655 do_nonblock:
5656 err = -EINPROGRESS;
5657 goto out;
5660 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5662 struct sctp_endpoint *ep;
5663 int err = 0;
5664 DEFINE_WAIT(wait);
5666 ep = sctp_sk(sk)->ep;
5669 for (;;) {
5670 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5671 TASK_INTERRUPTIBLE);
5673 if (list_empty(&ep->asocs)) {
5674 sctp_release_sock(sk);
5675 timeo = schedule_timeout(timeo);
5676 sctp_lock_sock(sk);
5679 err = -EINVAL;
5680 if (!sctp_sstate(sk, LISTENING))
5681 break;
5683 err = 0;
5684 if (!list_empty(&ep->asocs))
5685 break;
5687 err = sock_intr_errno(timeo);
5688 if (signal_pending(current))
5689 break;
5691 err = -EAGAIN;
5692 if (!timeo)
5693 break;
5696 finish_wait(sk->sk_sleep, &wait);
5698 return err;
5701 void sctp_wait_for_close(struct sock *sk, long timeout)
5703 DEFINE_WAIT(wait);
5705 do {
5706 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5707 if (list_empty(&sctp_sk(sk)->ep->asocs))
5708 break;
5709 sctp_release_sock(sk);
5710 timeout = schedule_timeout(timeout);
5711 sctp_lock_sock(sk);
5712 } while (!signal_pending(current) && timeout);
5714 finish_wait(sk->sk_sleep, &wait);
5717 static void sctp_sock_rfree_frag(struct sk_buff *skb)
5719 struct sk_buff *frag;
5721 if (!skb->data_len)
5722 goto done;
5724 /* Don't forget the fragments. */
5725 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5726 sctp_sock_rfree_frag(frag);
5728 done:
5729 sctp_sock_rfree(skb);
5732 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
5734 struct sk_buff *frag;
5736 if (!skb->data_len)
5737 goto done;
5739 /* Don't forget the fragments. */
5740 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5741 sctp_skb_set_owner_r_frag(frag, sk);
5743 done:
5744 sctp_skb_set_owner_r(skb, sk);
5747 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5748 * and its messages to the newsk.
5750 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5751 struct sctp_association *assoc,
5752 sctp_socket_type_t type)
5754 struct sctp_sock *oldsp = sctp_sk(oldsk);
5755 struct sctp_sock *newsp = sctp_sk(newsk);
5756 struct sctp_bind_bucket *pp; /* hash list port iterator */
5757 struct sctp_endpoint *newep = newsp->ep;
5758 struct sk_buff *skb, *tmp;
5759 struct sctp_ulpevent *event;
5760 int flags = 0;
5762 /* Migrate socket buffer sizes and all the socket level options to the
5763 * new socket.
5765 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5766 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5767 /* Brute force copy old sctp opt. */
5768 inet_sk_copy_descendant(newsk, oldsk);
5770 /* Restore the ep value that was overwritten with the above structure
5771 * copy.
5773 newsp->ep = newep;
5774 newsp->hmac = NULL;
5776 /* Hook this new socket in to the bind_hash list. */
5777 pp = sctp_sk(oldsk)->bind_hash;
5778 sk_add_bind_node(newsk, &pp->owner);
5779 sctp_sk(newsk)->bind_hash = pp;
5780 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5782 /* Copy the bind_addr list from the original endpoint to the new
5783 * endpoint so that we can handle restarts properly
5785 if (PF_INET6 == assoc->base.sk->sk_family)
5786 flags = SCTP_ADDR6_ALLOWED;
5787 if (assoc->peer.ipv4_address)
5788 flags |= SCTP_ADDR4_PEERSUPP;
5789 if (assoc->peer.ipv6_address)
5790 flags |= SCTP_ADDR6_PEERSUPP;
5791 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5792 &oldsp->ep->base.bind_addr,
5793 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5795 /* Move any messages in the old socket's receive queue that are for the
5796 * peeled off association to the new socket's receive queue.
5798 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5799 event = sctp_skb2event(skb);
5800 if (event->asoc == assoc) {
5801 sctp_sock_rfree_frag(skb);
5802 __skb_unlink(skb, &oldsk->sk_receive_queue);
5803 __skb_queue_tail(&newsk->sk_receive_queue, skb);
5804 sctp_skb_set_owner_r_frag(skb, newsk);
5808 /* Clean up any messages pending delivery due to partial
5809 * delivery. Three cases:
5810 * 1) No partial deliver; no work.
5811 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5812 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5814 skb_queue_head_init(&newsp->pd_lobby);
5815 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
5817 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
5818 struct sk_buff_head *queue;
5820 /* Decide which queue to move pd_lobby skbs to. */
5821 if (assoc->ulpq.pd_mode) {
5822 queue = &newsp->pd_lobby;
5823 } else
5824 queue = &newsk->sk_receive_queue;
5826 /* Walk through the pd_lobby, looking for skbs that
5827 * need moved to the new socket.
5829 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5830 event = sctp_skb2event(skb);
5831 if (event->asoc == assoc) {
5832 sctp_sock_rfree_frag(skb);
5833 __skb_unlink(skb, &oldsp->pd_lobby);
5834 __skb_queue_tail(queue, skb);
5835 sctp_skb_set_owner_r_frag(skb, newsk);
5839 /* Clear up any skbs waiting for the partial
5840 * delivery to finish.
5842 if (assoc->ulpq.pd_mode)
5843 sctp_clear_pd(oldsk, NULL);
5847 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
5848 sctp_sock_rfree_frag(skb);
5849 sctp_skb_set_owner_r_frag(skb, newsk);
5852 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
5853 sctp_sock_rfree_frag(skb);
5854 sctp_skb_set_owner_r_frag(skb, newsk);
5857 /* Set the type of socket to indicate that it is peeled off from the
5858 * original UDP-style socket or created with the accept() call on a
5859 * TCP-style socket..
5861 newsp->type = type;
5863 /* Mark the new socket "in-use" by the user so that any packets
5864 * that may arrive on the association after we've moved it are
5865 * queued to the backlog. This prevents a potential race between
5866 * backlog processing on the old socket and new-packet processing
5867 * on the new socket.
5869 sctp_lock_sock(newsk);
5870 sctp_assoc_migrate(assoc, newsk);
5872 /* If the association on the newsk is already closed before accept()
5873 * is called, set RCV_SHUTDOWN flag.
5875 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
5876 newsk->sk_shutdown |= RCV_SHUTDOWN;
5878 newsk->sk_state = SCTP_SS_ESTABLISHED;
5879 sctp_release_sock(newsk);
5882 /* This proto struct describes the ULP interface for SCTP. */
5883 struct proto sctp_prot = {
5884 .name = "SCTP",
5885 .owner = THIS_MODULE,
5886 .close = sctp_close,
5887 .connect = sctp_connect,
5888 .disconnect = sctp_disconnect,
5889 .accept = sctp_accept,
5890 .ioctl = sctp_ioctl,
5891 .init = sctp_init_sock,
5892 .destroy = sctp_destroy_sock,
5893 .shutdown = sctp_shutdown,
5894 .setsockopt = sctp_setsockopt,
5895 .getsockopt = sctp_getsockopt,
5896 .sendmsg = sctp_sendmsg,
5897 .recvmsg = sctp_recvmsg,
5898 .bind = sctp_bind,
5899 .backlog_rcv = sctp_backlog_rcv,
5900 .hash = sctp_hash,
5901 .unhash = sctp_unhash,
5902 .get_port = sctp_get_port,
5903 .obj_size = sizeof(struct sctp_sock),
5906 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
5907 struct proto sctpv6_prot = {
5908 .name = "SCTPv6",
5909 .owner = THIS_MODULE,
5910 .close = sctp_close,
5911 .connect = sctp_connect,
5912 .disconnect = sctp_disconnect,
5913 .accept = sctp_accept,
5914 .ioctl = sctp_ioctl,
5915 .init = sctp_init_sock,
5916 .destroy = sctp_destroy_sock,
5917 .shutdown = sctp_shutdown,
5918 .setsockopt = sctp_setsockopt,
5919 .getsockopt = sctp_getsockopt,
5920 .sendmsg = sctp_sendmsg,
5921 .recvmsg = sctp_recvmsg,
5922 .bind = sctp_bind,
5923 .backlog_rcv = sctp_backlog_rcv,
5924 .hash = sctp_hash,
5925 .unhash = sctp_unhash,
5926 .get_port = sctp_get_port,
5927 .obj_size = sizeof(struct sctp6_sock),
5929 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */