[SCSI] esp_scsi.c: Fix compilation.
[linux-2.6/openmoko-kernel.git] / net / sctp / socket.c
blob11938fb20395f45b8538181a85318c086c3668e8
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_HB_TIME_IS_ZERO - Specify's that the time for
2043 * heartbeat delayis to be set to the value of 0
2044 * milliseconds.
2046 * SPP_PMTUD_ENABLE - This field will enable PMTU
2047 * discovery upon the specified address. Note that
2048 * if the address feild is empty then all addresses
2049 * on the association are effected.
2051 * SPP_PMTUD_DISABLE - This field will disable PMTU
2052 * discovery upon the specified address. Note that
2053 * if the address feild is empty then all addresses
2054 * on the association are effected. Not also that
2055 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2056 * exclusive. Enabling both will have undetermined
2057 * results.
2059 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2060 * on delayed sack. The time specified in spp_sackdelay
2061 * is used to specify the sack delay for this address. Note
2062 * that if spp_address is empty then all addresses will
2063 * enable delayed sack and take on the sack delay
2064 * value specified in spp_sackdelay.
2065 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2066 * off delayed sack. If the spp_address field is blank then
2067 * delayed sack is disabled for the entire association. Note
2068 * also that this field is mutually exclusive to
2069 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2070 * results.
2072 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2073 struct sctp_transport *trans,
2074 struct sctp_association *asoc,
2075 struct sctp_sock *sp,
2076 int hb_change,
2077 int pmtud_change,
2078 int sackdelay_change)
2080 int error;
2082 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2083 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2084 if (error)
2085 return error;
2088 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2089 * this field is ignored. Note also that a value of zero indicates
2090 * the current setting should be left unchanged.
2092 if (params->spp_flags & SPP_HB_ENABLE) {
2094 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2095 * set. This lets us use 0 value when this flag
2096 * is set.
2098 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2099 params->spp_hbinterval = 0;
2101 if (params->spp_hbinterval ||
2102 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2103 if (trans) {
2104 trans->hbinterval =
2105 msecs_to_jiffies(params->spp_hbinterval);
2106 } else if (asoc) {
2107 asoc->hbinterval =
2108 msecs_to_jiffies(params->spp_hbinterval);
2109 } else {
2110 sp->hbinterval = params->spp_hbinterval;
2115 if (hb_change) {
2116 if (trans) {
2117 trans->param_flags =
2118 (trans->param_flags & ~SPP_HB) | hb_change;
2119 } else if (asoc) {
2120 asoc->param_flags =
2121 (asoc->param_flags & ~SPP_HB) | hb_change;
2122 } else {
2123 sp->param_flags =
2124 (sp->param_flags & ~SPP_HB) | hb_change;
2128 /* When Path MTU discovery is disabled the value specified here will
2129 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2130 * include the flag SPP_PMTUD_DISABLE for this field to have any
2131 * effect).
2133 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2134 if (trans) {
2135 trans->pathmtu = params->spp_pathmtu;
2136 sctp_assoc_sync_pmtu(asoc);
2137 } else if (asoc) {
2138 asoc->pathmtu = params->spp_pathmtu;
2139 sctp_frag_point(sp, params->spp_pathmtu);
2140 } else {
2141 sp->pathmtu = params->spp_pathmtu;
2145 if (pmtud_change) {
2146 if (trans) {
2147 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2148 (params->spp_flags & SPP_PMTUD_ENABLE);
2149 trans->param_flags =
2150 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2151 if (update) {
2152 sctp_transport_pmtu(trans);
2153 sctp_assoc_sync_pmtu(asoc);
2155 } else if (asoc) {
2156 asoc->param_flags =
2157 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2158 } else {
2159 sp->param_flags =
2160 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2164 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2165 * value of this field is ignored. Note also that a value of zero
2166 * indicates the current setting should be left unchanged.
2168 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2169 if (trans) {
2170 trans->sackdelay =
2171 msecs_to_jiffies(params->spp_sackdelay);
2172 } else if (asoc) {
2173 asoc->sackdelay =
2174 msecs_to_jiffies(params->spp_sackdelay);
2175 } else {
2176 sp->sackdelay = params->spp_sackdelay;
2180 if (sackdelay_change) {
2181 if (trans) {
2182 trans->param_flags =
2183 (trans->param_flags & ~SPP_SACKDELAY) |
2184 sackdelay_change;
2185 } else if (asoc) {
2186 asoc->param_flags =
2187 (asoc->param_flags & ~SPP_SACKDELAY) |
2188 sackdelay_change;
2189 } else {
2190 sp->param_flags =
2191 (sp->param_flags & ~SPP_SACKDELAY) |
2192 sackdelay_change;
2196 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2197 * of this field is ignored. Note also that a value of zero
2198 * indicates the current setting should be left unchanged.
2200 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2201 if (trans) {
2202 trans->pathmaxrxt = params->spp_pathmaxrxt;
2203 } else if (asoc) {
2204 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2205 } else {
2206 sp->pathmaxrxt = params->spp_pathmaxrxt;
2210 return 0;
2213 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2214 char __user *optval, int optlen)
2216 struct sctp_paddrparams params;
2217 struct sctp_transport *trans = NULL;
2218 struct sctp_association *asoc = NULL;
2219 struct sctp_sock *sp = sctp_sk(sk);
2220 int error;
2221 int hb_change, pmtud_change, sackdelay_change;
2223 if (optlen != sizeof(struct sctp_paddrparams))
2224 return - EINVAL;
2226 if (copy_from_user(&params, optval, optlen))
2227 return -EFAULT;
2229 /* Validate flags and value parameters. */
2230 hb_change = params.spp_flags & SPP_HB;
2231 pmtud_change = params.spp_flags & SPP_PMTUD;
2232 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2234 if (hb_change == SPP_HB ||
2235 pmtud_change == SPP_PMTUD ||
2236 sackdelay_change == SPP_SACKDELAY ||
2237 params.spp_sackdelay > 500 ||
2238 (params.spp_pathmtu
2239 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2240 return -EINVAL;
2242 /* If an address other than INADDR_ANY is specified, and
2243 * no transport is found, then the request is invalid.
2245 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2246 trans = sctp_addr_id2transport(sk, &params.spp_address,
2247 params.spp_assoc_id);
2248 if (!trans)
2249 return -EINVAL;
2252 /* Get association, if assoc_id != 0 and the socket is a one
2253 * to many style socket, and an association was not found, then
2254 * the id was invalid.
2256 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2257 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2258 return -EINVAL;
2260 /* Heartbeat demand can only be sent on a transport or
2261 * association, but not a socket.
2263 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2264 return -EINVAL;
2266 /* Process parameters. */
2267 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2268 hb_change, pmtud_change,
2269 sackdelay_change);
2271 if (error)
2272 return error;
2274 /* If changes are for association, also apply parameters to each
2275 * transport.
2277 if (!trans && asoc) {
2278 struct list_head *pos;
2280 list_for_each(pos, &asoc->peer.transport_addr_list) {
2281 trans = list_entry(pos, struct sctp_transport,
2282 transports);
2283 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2284 hb_change, pmtud_change,
2285 sackdelay_change);
2289 return 0;
2292 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2294 * This options will get or set the delayed ack timer. The time is set
2295 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2296 * endpoints default delayed ack timer value. If the assoc_id field is
2297 * non-zero, then the set or get effects the specified association.
2299 * struct sctp_assoc_value {
2300 * sctp_assoc_t assoc_id;
2301 * uint32_t assoc_value;
2302 * };
2304 * assoc_id - This parameter, indicates which association the
2305 * user is preforming an action upon. Note that if
2306 * this field's value is zero then the endpoints
2307 * default value is changed (effecting future
2308 * associations only).
2310 * assoc_value - This parameter contains the number of milliseconds
2311 * that the user is requesting the delayed ACK timer
2312 * be set to. Note that this value is defined in
2313 * the standard to be between 200 and 500 milliseconds.
2315 * Note: a value of zero will leave the value alone,
2316 * but disable SACK delay. A non-zero value will also
2317 * enable SACK delay.
2320 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2321 char __user *optval, int optlen)
2323 struct sctp_assoc_value params;
2324 struct sctp_transport *trans = NULL;
2325 struct sctp_association *asoc = NULL;
2326 struct sctp_sock *sp = sctp_sk(sk);
2328 if (optlen != sizeof(struct sctp_assoc_value))
2329 return - EINVAL;
2331 if (copy_from_user(&params, optval, optlen))
2332 return -EFAULT;
2334 /* Validate value parameter. */
2335 if (params.assoc_value > 500)
2336 return -EINVAL;
2338 /* Get association, if assoc_id != 0 and the socket is a one
2339 * to many style socket, and an association was not found, then
2340 * the id was invalid.
2342 asoc = sctp_id2assoc(sk, params.assoc_id);
2343 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2344 return -EINVAL;
2346 if (params.assoc_value) {
2347 if (asoc) {
2348 asoc->sackdelay =
2349 msecs_to_jiffies(params.assoc_value);
2350 asoc->param_flags =
2351 (asoc->param_flags & ~SPP_SACKDELAY) |
2352 SPP_SACKDELAY_ENABLE;
2353 } else {
2354 sp->sackdelay = params.assoc_value;
2355 sp->param_flags =
2356 (sp->param_flags & ~SPP_SACKDELAY) |
2357 SPP_SACKDELAY_ENABLE;
2359 } else {
2360 if (asoc) {
2361 asoc->param_flags =
2362 (asoc->param_flags & ~SPP_SACKDELAY) |
2363 SPP_SACKDELAY_DISABLE;
2364 } else {
2365 sp->param_flags =
2366 (sp->param_flags & ~SPP_SACKDELAY) |
2367 SPP_SACKDELAY_DISABLE;
2371 /* If change is for association, also apply to each transport. */
2372 if (asoc) {
2373 struct list_head *pos;
2375 list_for_each(pos, &asoc->peer.transport_addr_list) {
2376 trans = list_entry(pos, struct sctp_transport,
2377 transports);
2378 if (params.assoc_value) {
2379 trans->sackdelay =
2380 msecs_to_jiffies(params.assoc_value);
2381 trans->param_flags =
2382 (trans->param_flags & ~SPP_SACKDELAY) |
2383 SPP_SACKDELAY_ENABLE;
2384 } else {
2385 trans->param_flags =
2386 (trans->param_flags & ~SPP_SACKDELAY) |
2387 SPP_SACKDELAY_DISABLE;
2392 return 0;
2395 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2397 * Applications can specify protocol parameters for the default association
2398 * initialization. The option name argument to setsockopt() and getsockopt()
2399 * is SCTP_INITMSG.
2401 * Setting initialization parameters is effective only on an unconnected
2402 * socket (for UDP-style sockets only future associations are effected
2403 * by the change). With TCP-style sockets, this option is inherited by
2404 * sockets derived from a listener socket.
2406 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2408 struct sctp_initmsg sinit;
2409 struct sctp_sock *sp = sctp_sk(sk);
2411 if (optlen != sizeof(struct sctp_initmsg))
2412 return -EINVAL;
2413 if (copy_from_user(&sinit, optval, optlen))
2414 return -EFAULT;
2416 if (sinit.sinit_num_ostreams)
2417 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2418 if (sinit.sinit_max_instreams)
2419 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2420 if (sinit.sinit_max_attempts)
2421 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2422 if (sinit.sinit_max_init_timeo)
2423 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2425 return 0;
2429 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2431 * Applications that wish to use the sendto() system call may wish to
2432 * specify a default set of parameters that would normally be supplied
2433 * through the inclusion of ancillary data. This socket option allows
2434 * such an application to set the default sctp_sndrcvinfo structure.
2435 * The application that wishes to use this socket option simply passes
2436 * in to this call the sctp_sndrcvinfo structure defined in Section
2437 * 5.2.2) The input parameters accepted by this call include
2438 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2439 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2440 * to this call if the caller is using the UDP model.
2442 static int sctp_setsockopt_default_send_param(struct sock *sk,
2443 char __user *optval, int optlen)
2445 struct sctp_sndrcvinfo info;
2446 struct sctp_association *asoc;
2447 struct sctp_sock *sp = sctp_sk(sk);
2449 if (optlen != sizeof(struct sctp_sndrcvinfo))
2450 return -EINVAL;
2451 if (copy_from_user(&info, optval, optlen))
2452 return -EFAULT;
2454 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2455 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2456 return -EINVAL;
2458 if (asoc) {
2459 asoc->default_stream = info.sinfo_stream;
2460 asoc->default_flags = info.sinfo_flags;
2461 asoc->default_ppid = info.sinfo_ppid;
2462 asoc->default_context = info.sinfo_context;
2463 asoc->default_timetolive = info.sinfo_timetolive;
2464 } else {
2465 sp->default_stream = info.sinfo_stream;
2466 sp->default_flags = info.sinfo_flags;
2467 sp->default_ppid = info.sinfo_ppid;
2468 sp->default_context = info.sinfo_context;
2469 sp->default_timetolive = info.sinfo_timetolive;
2472 return 0;
2475 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2477 * Requests that the local SCTP stack use the enclosed peer address as
2478 * the association primary. The enclosed address must be one of the
2479 * association peer's addresses.
2481 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2482 int optlen)
2484 struct sctp_prim prim;
2485 struct sctp_transport *trans;
2487 if (optlen != sizeof(struct sctp_prim))
2488 return -EINVAL;
2490 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2491 return -EFAULT;
2493 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2494 if (!trans)
2495 return -EINVAL;
2497 sctp_assoc_set_primary(trans->asoc, trans);
2499 return 0;
2503 * 7.1.5 SCTP_NODELAY
2505 * Turn on/off any Nagle-like algorithm. This means that packets are
2506 * generally sent as soon as possible and no unnecessary delays are
2507 * introduced, at the cost of more packets in the network. Expects an
2508 * integer boolean flag.
2510 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2511 int optlen)
2513 int val;
2515 if (optlen < sizeof(int))
2516 return -EINVAL;
2517 if (get_user(val, (int __user *)optval))
2518 return -EFAULT;
2520 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2521 return 0;
2526 * 7.1.1 SCTP_RTOINFO
2528 * The protocol parameters used to initialize and bound retransmission
2529 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2530 * and modify these parameters.
2531 * All parameters are time values, in milliseconds. A value of 0, when
2532 * modifying the parameters, indicates that the current value should not
2533 * be changed.
2536 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2537 struct sctp_rtoinfo rtoinfo;
2538 struct sctp_association *asoc;
2540 if (optlen != sizeof (struct sctp_rtoinfo))
2541 return -EINVAL;
2543 if (copy_from_user(&rtoinfo, optval, optlen))
2544 return -EFAULT;
2546 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2548 /* Set the values to the specific association */
2549 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2550 return -EINVAL;
2552 if (asoc) {
2553 if (rtoinfo.srto_initial != 0)
2554 asoc->rto_initial =
2555 msecs_to_jiffies(rtoinfo.srto_initial);
2556 if (rtoinfo.srto_max != 0)
2557 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2558 if (rtoinfo.srto_min != 0)
2559 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2560 } else {
2561 /* If there is no association or the association-id = 0
2562 * set the values to the endpoint.
2564 struct sctp_sock *sp = sctp_sk(sk);
2566 if (rtoinfo.srto_initial != 0)
2567 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2568 if (rtoinfo.srto_max != 0)
2569 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2570 if (rtoinfo.srto_min != 0)
2571 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2574 return 0;
2579 * 7.1.2 SCTP_ASSOCINFO
2581 * This option is used to tune the the maximum retransmission attempts
2582 * of the association.
2583 * Returns an error if the new association retransmission value is
2584 * greater than the sum of the retransmission value of the peer.
2585 * See [SCTP] for more information.
2588 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2591 struct sctp_assocparams assocparams;
2592 struct sctp_association *asoc;
2594 if (optlen != sizeof(struct sctp_assocparams))
2595 return -EINVAL;
2596 if (copy_from_user(&assocparams, optval, optlen))
2597 return -EFAULT;
2599 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2601 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2602 return -EINVAL;
2604 /* Set the values to the specific association */
2605 if (asoc) {
2606 if (assocparams.sasoc_asocmaxrxt != 0) {
2607 __u32 path_sum = 0;
2608 int paths = 0;
2609 struct list_head *pos;
2610 struct sctp_transport *peer_addr;
2612 list_for_each(pos, &asoc->peer.transport_addr_list) {
2613 peer_addr = list_entry(pos,
2614 struct sctp_transport,
2615 transports);
2616 path_sum += peer_addr->pathmaxrxt;
2617 paths++;
2620 /* Only validate asocmaxrxt if we have more then
2621 * one path/transport. We do this because path
2622 * retransmissions are only counted when we have more
2623 * then one path.
2625 if (paths > 1 &&
2626 assocparams.sasoc_asocmaxrxt > path_sum)
2627 return -EINVAL;
2629 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2632 if (assocparams.sasoc_cookie_life != 0) {
2633 asoc->cookie_life.tv_sec =
2634 assocparams.sasoc_cookie_life / 1000;
2635 asoc->cookie_life.tv_usec =
2636 (assocparams.sasoc_cookie_life % 1000)
2637 * 1000;
2639 } else {
2640 /* Set the values to the endpoint */
2641 struct sctp_sock *sp = sctp_sk(sk);
2643 if (assocparams.sasoc_asocmaxrxt != 0)
2644 sp->assocparams.sasoc_asocmaxrxt =
2645 assocparams.sasoc_asocmaxrxt;
2646 if (assocparams.sasoc_cookie_life != 0)
2647 sp->assocparams.sasoc_cookie_life =
2648 assocparams.sasoc_cookie_life;
2650 return 0;
2654 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2656 * This socket option is a boolean flag which turns on or off mapped V4
2657 * addresses. If this option is turned on and the socket is type
2658 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2659 * If this option is turned off, then no mapping will be done of V4
2660 * addresses and a user will receive both PF_INET6 and PF_INET type
2661 * addresses on the socket.
2663 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2665 int val;
2666 struct sctp_sock *sp = sctp_sk(sk);
2668 if (optlen < sizeof(int))
2669 return -EINVAL;
2670 if (get_user(val, (int __user *)optval))
2671 return -EFAULT;
2672 if (val)
2673 sp->v4mapped = 1;
2674 else
2675 sp->v4mapped = 0;
2677 return 0;
2681 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2683 * This socket option specifies the maximum size to put in any outgoing
2684 * SCTP chunk. If a message is larger than this size it will be
2685 * fragmented by SCTP into the specified size. Note that the underlying
2686 * SCTP implementation may fragment into smaller sized chunks when the
2687 * PMTU of the underlying association is smaller than the value set by
2688 * the user.
2690 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2692 struct sctp_association *asoc;
2693 struct list_head *pos;
2694 struct sctp_sock *sp = sctp_sk(sk);
2695 int val;
2697 if (optlen < sizeof(int))
2698 return -EINVAL;
2699 if (get_user(val, (int __user *)optval))
2700 return -EFAULT;
2701 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2702 return -EINVAL;
2703 sp->user_frag = val;
2705 /* Update the frag_point of the existing associations. */
2706 list_for_each(pos, &(sp->ep->asocs)) {
2707 asoc = list_entry(pos, struct sctp_association, asocs);
2708 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2711 return 0;
2716 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2718 * Requests that the peer mark the enclosed address as the association
2719 * primary. The enclosed address must be one of the association's
2720 * locally bound addresses. The following structure is used to make a
2721 * set primary request:
2723 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2724 int optlen)
2726 struct sctp_sock *sp;
2727 struct sctp_endpoint *ep;
2728 struct sctp_association *asoc = NULL;
2729 struct sctp_setpeerprim prim;
2730 struct sctp_chunk *chunk;
2731 int err;
2733 sp = sctp_sk(sk);
2734 ep = sp->ep;
2736 if (!sctp_addip_enable)
2737 return -EPERM;
2739 if (optlen != sizeof(struct sctp_setpeerprim))
2740 return -EINVAL;
2742 if (copy_from_user(&prim, optval, optlen))
2743 return -EFAULT;
2745 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2746 if (!asoc)
2747 return -EINVAL;
2749 if (!asoc->peer.asconf_capable)
2750 return -EPERM;
2752 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2753 return -EPERM;
2755 if (!sctp_state(asoc, ESTABLISHED))
2756 return -ENOTCONN;
2758 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2759 return -EADDRNOTAVAIL;
2761 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2762 chunk = sctp_make_asconf_set_prim(asoc,
2763 (union sctp_addr *)&prim.sspp_addr);
2764 if (!chunk)
2765 return -ENOMEM;
2767 err = sctp_send_asconf(asoc, chunk);
2769 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2771 return err;
2774 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2775 int optlen)
2777 struct sctp_setadaptation adaptation;
2779 if (optlen != sizeof(struct sctp_setadaptation))
2780 return -EINVAL;
2781 if (copy_from_user(&adaptation, optval, optlen))
2782 return -EFAULT;
2784 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2786 return 0;
2790 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2792 * The context field in the sctp_sndrcvinfo structure is normally only
2793 * used when a failed message is retrieved holding the value that was
2794 * sent down on the actual send call. This option allows the setting of
2795 * a default context on an association basis that will be received on
2796 * reading messages from the peer. This is especially helpful in the
2797 * one-2-many model for an application to keep some reference to an
2798 * internal state machine that is processing messages on the
2799 * association. Note that the setting of this value only effects
2800 * received messages from the peer and does not effect the value that is
2801 * saved with outbound messages.
2803 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2804 int optlen)
2806 struct sctp_assoc_value params;
2807 struct sctp_sock *sp;
2808 struct sctp_association *asoc;
2810 if (optlen != sizeof(struct sctp_assoc_value))
2811 return -EINVAL;
2812 if (copy_from_user(&params, optval, optlen))
2813 return -EFAULT;
2815 sp = sctp_sk(sk);
2817 if (params.assoc_id != 0) {
2818 asoc = sctp_id2assoc(sk, params.assoc_id);
2819 if (!asoc)
2820 return -EINVAL;
2821 asoc->default_rcv_context = params.assoc_value;
2822 } else {
2823 sp->default_rcv_context = params.assoc_value;
2826 return 0;
2830 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2832 * This options will at a minimum specify if the implementation is doing
2833 * fragmented interleave. Fragmented interleave, for a one to many
2834 * socket, is when subsequent calls to receive a message may return
2835 * parts of messages from different associations. Some implementations
2836 * may allow you to turn this value on or off. If so, when turned off,
2837 * no fragment interleave will occur (which will cause a head of line
2838 * blocking amongst multiple associations sharing the same one to many
2839 * socket). When this option is turned on, then each receive call may
2840 * come from a different association (thus the user must receive data
2841 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2842 * association each receive belongs to.
2844 * This option takes a boolean value. A non-zero value indicates that
2845 * fragmented interleave is on. A value of zero indicates that
2846 * fragmented interleave is off.
2848 * Note that it is important that an implementation that allows this
2849 * option to be turned on, have it off by default. Otherwise an unaware
2850 * application using the one to many model may become confused and act
2851 * incorrectly.
2853 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2854 char __user *optval,
2855 int optlen)
2857 int val;
2859 if (optlen != sizeof(int))
2860 return -EINVAL;
2861 if (get_user(val, (int __user *)optval))
2862 return -EFAULT;
2864 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2866 return 0;
2870 * 7.1.25. Set or Get the sctp partial delivery point
2871 * (SCTP_PARTIAL_DELIVERY_POINT)
2872 * This option will set or get the SCTP partial delivery point. This
2873 * point is the size of a message where the partial delivery API will be
2874 * invoked to help free up rwnd space for the peer. Setting this to a
2875 * lower value will cause partial delivery's to happen more often. The
2876 * calls argument is an integer that sets or gets the partial delivery
2877 * point.
2879 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2880 char __user *optval,
2881 int optlen)
2883 u32 val;
2885 if (optlen != sizeof(u32))
2886 return -EINVAL;
2887 if (get_user(val, (int __user *)optval))
2888 return -EFAULT;
2890 sctp_sk(sk)->pd_point = val;
2892 return 0; /* is this the right error code? */
2896 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2898 * This option will allow a user to change the maximum burst of packets
2899 * that can be emitted by this association. Note that the default value
2900 * is 4, and some implementations may restrict this setting so that it
2901 * can only be lowered.
2903 * NOTE: This text doesn't seem right. Do this on a socket basis with
2904 * future associations inheriting the socket value.
2906 static int sctp_setsockopt_maxburst(struct sock *sk,
2907 char __user *optval,
2908 int optlen)
2910 int val;
2912 if (optlen != sizeof(int))
2913 return -EINVAL;
2914 if (get_user(val, (int __user *)optval))
2915 return -EFAULT;
2917 if (val < 0)
2918 return -EINVAL;
2920 sctp_sk(sk)->max_burst = val;
2922 return 0;
2925 /* API 6.2 setsockopt(), getsockopt()
2927 * Applications use setsockopt() and getsockopt() to set or retrieve
2928 * socket options. Socket options are used to change the default
2929 * behavior of sockets calls. They are described in Section 7.
2931 * The syntax is:
2933 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
2934 * int __user *optlen);
2935 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
2936 * int optlen);
2938 * sd - the socket descript.
2939 * level - set to IPPROTO_SCTP for all SCTP options.
2940 * optname - the option name.
2941 * optval - the buffer to store the value of the option.
2942 * optlen - the size of the buffer.
2944 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
2945 char __user *optval, int optlen)
2947 int retval = 0;
2949 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
2950 sk, optname);
2952 /* I can hardly begin to describe how wrong this is. This is
2953 * so broken as to be worse than useless. The API draft
2954 * REALLY is NOT helpful here... I am not convinced that the
2955 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
2956 * are at all well-founded.
2958 if (level != SOL_SCTP) {
2959 struct sctp_af *af = sctp_sk(sk)->pf->af;
2960 retval = af->setsockopt(sk, level, optname, optval, optlen);
2961 goto out_nounlock;
2964 sctp_lock_sock(sk);
2966 switch (optname) {
2967 case SCTP_SOCKOPT_BINDX_ADD:
2968 /* 'optlen' is the size of the addresses buffer. */
2969 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2970 optlen, SCTP_BINDX_ADD_ADDR);
2971 break;
2973 case SCTP_SOCKOPT_BINDX_REM:
2974 /* 'optlen' is the size of the addresses buffer. */
2975 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
2976 optlen, SCTP_BINDX_REM_ADDR);
2977 break;
2979 case SCTP_SOCKOPT_CONNECTX:
2980 /* 'optlen' is the size of the addresses buffer. */
2981 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
2982 optlen);
2983 break;
2985 case SCTP_DISABLE_FRAGMENTS:
2986 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
2987 break;
2989 case SCTP_EVENTS:
2990 retval = sctp_setsockopt_events(sk, optval, optlen);
2991 break;
2993 case SCTP_AUTOCLOSE:
2994 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
2995 break;
2997 case SCTP_PEER_ADDR_PARAMS:
2998 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
2999 break;
3001 case SCTP_DELAYED_ACK_TIME:
3002 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
3003 break;
3004 case SCTP_PARTIAL_DELIVERY_POINT:
3005 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3006 break;
3008 case SCTP_INITMSG:
3009 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3010 break;
3011 case SCTP_DEFAULT_SEND_PARAM:
3012 retval = sctp_setsockopt_default_send_param(sk, optval,
3013 optlen);
3014 break;
3015 case SCTP_PRIMARY_ADDR:
3016 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3017 break;
3018 case SCTP_SET_PEER_PRIMARY_ADDR:
3019 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3020 break;
3021 case SCTP_NODELAY:
3022 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3023 break;
3024 case SCTP_RTOINFO:
3025 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3026 break;
3027 case SCTP_ASSOCINFO:
3028 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3029 break;
3030 case SCTP_I_WANT_MAPPED_V4_ADDR:
3031 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3032 break;
3033 case SCTP_MAXSEG:
3034 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3035 break;
3036 case SCTP_ADAPTATION_LAYER:
3037 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3038 break;
3039 case SCTP_CONTEXT:
3040 retval = sctp_setsockopt_context(sk, optval, optlen);
3041 break;
3042 case SCTP_FRAGMENT_INTERLEAVE:
3043 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3044 break;
3045 case SCTP_MAX_BURST:
3046 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3047 break;
3048 default:
3049 retval = -ENOPROTOOPT;
3050 break;
3053 sctp_release_sock(sk);
3055 out_nounlock:
3056 return retval;
3059 /* API 3.1.6 connect() - UDP Style Syntax
3061 * An application may use the connect() call in the UDP model to initiate an
3062 * association without sending data.
3064 * The syntax is:
3066 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3068 * sd: the socket descriptor to have a new association added to.
3070 * nam: the address structure (either struct sockaddr_in or struct
3071 * sockaddr_in6 defined in RFC2553 [7]).
3073 * len: the size of the address.
3075 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3076 int addr_len)
3078 int err = 0;
3079 struct sctp_af *af;
3081 sctp_lock_sock(sk);
3083 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3084 __FUNCTION__, sk, addr, addr_len);
3086 /* Validate addr_len before calling common connect/connectx routine. */
3087 af = sctp_get_af_specific(addr->sa_family);
3088 if (!af || addr_len < af->sockaddr_len) {
3089 err = -EINVAL;
3090 } else {
3091 /* Pass correct addr len to common routine (so it knows there
3092 * is only one address being passed.
3094 err = __sctp_connect(sk, addr, af->sockaddr_len);
3097 sctp_release_sock(sk);
3098 return err;
3101 /* FIXME: Write comments. */
3102 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3104 return -EOPNOTSUPP; /* STUB */
3107 /* 4.1.4 accept() - TCP Style Syntax
3109 * Applications use accept() call to remove an established SCTP
3110 * association from the accept queue of the endpoint. A new socket
3111 * descriptor will be returned from accept() to represent the newly
3112 * formed association.
3114 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3116 struct sctp_sock *sp;
3117 struct sctp_endpoint *ep;
3118 struct sock *newsk = NULL;
3119 struct sctp_association *asoc;
3120 long timeo;
3121 int error = 0;
3123 sctp_lock_sock(sk);
3125 sp = sctp_sk(sk);
3126 ep = sp->ep;
3128 if (!sctp_style(sk, TCP)) {
3129 error = -EOPNOTSUPP;
3130 goto out;
3133 if (!sctp_sstate(sk, LISTENING)) {
3134 error = -EINVAL;
3135 goto out;
3138 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3140 error = sctp_wait_for_accept(sk, timeo);
3141 if (error)
3142 goto out;
3144 /* We treat the list of associations on the endpoint as the accept
3145 * queue and pick the first association on the list.
3147 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3149 newsk = sp->pf->create_accept_sk(sk, asoc);
3150 if (!newsk) {
3151 error = -ENOMEM;
3152 goto out;
3155 /* Populate the fields of the newsk from the oldsk and migrate the
3156 * asoc to the newsk.
3158 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3160 out:
3161 sctp_release_sock(sk);
3162 *err = error;
3163 return newsk;
3166 /* The SCTP ioctl handler. */
3167 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3169 return -ENOIOCTLCMD;
3172 /* This is the function which gets called during socket creation to
3173 * initialized the SCTP-specific portion of the sock.
3174 * The sock structure should already be zero-filled memory.
3176 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3178 struct sctp_endpoint *ep;
3179 struct sctp_sock *sp;
3181 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3183 sp = sctp_sk(sk);
3185 /* Initialize the SCTP per socket area. */
3186 switch (sk->sk_type) {
3187 case SOCK_SEQPACKET:
3188 sp->type = SCTP_SOCKET_UDP;
3189 break;
3190 case SOCK_STREAM:
3191 sp->type = SCTP_SOCKET_TCP;
3192 break;
3193 default:
3194 return -ESOCKTNOSUPPORT;
3197 /* Initialize default send parameters. These parameters can be
3198 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3200 sp->default_stream = 0;
3201 sp->default_ppid = 0;
3202 sp->default_flags = 0;
3203 sp->default_context = 0;
3204 sp->default_timetolive = 0;
3206 sp->default_rcv_context = 0;
3207 sp->max_burst = sctp_max_burst;
3209 /* Initialize default setup parameters. These parameters
3210 * can be modified with the SCTP_INITMSG socket option or
3211 * overridden by the SCTP_INIT CMSG.
3213 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3214 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3215 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3216 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3218 /* Initialize default RTO related parameters. These parameters can
3219 * be modified for with the SCTP_RTOINFO socket option.
3221 sp->rtoinfo.srto_initial = sctp_rto_initial;
3222 sp->rtoinfo.srto_max = sctp_rto_max;
3223 sp->rtoinfo.srto_min = sctp_rto_min;
3225 /* Initialize default association related parameters. These parameters
3226 * can be modified with the SCTP_ASSOCINFO socket option.
3228 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3229 sp->assocparams.sasoc_number_peer_destinations = 0;
3230 sp->assocparams.sasoc_peer_rwnd = 0;
3231 sp->assocparams.sasoc_local_rwnd = 0;
3232 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3234 /* Initialize default event subscriptions. By default, all the
3235 * options are off.
3237 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3239 /* Default Peer Address Parameters. These defaults can
3240 * be modified via SCTP_PEER_ADDR_PARAMS
3242 sp->hbinterval = sctp_hb_interval;
3243 sp->pathmaxrxt = sctp_max_retrans_path;
3244 sp->pathmtu = 0; // allow default discovery
3245 sp->sackdelay = sctp_sack_timeout;
3246 sp->param_flags = SPP_HB_ENABLE |
3247 SPP_PMTUD_ENABLE |
3248 SPP_SACKDELAY_ENABLE;
3250 /* If enabled no SCTP message fragmentation will be performed.
3251 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3253 sp->disable_fragments = 0;
3255 /* Enable Nagle algorithm by default. */
3256 sp->nodelay = 0;
3258 /* Enable by default. */
3259 sp->v4mapped = 1;
3261 /* Auto-close idle associations after the configured
3262 * number of seconds. A value of 0 disables this
3263 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3264 * for UDP-style sockets only.
3266 sp->autoclose = 0;
3268 /* User specified fragmentation limit. */
3269 sp->user_frag = 0;
3271 sp->adaptation_ind = 0;
3273 sp->pf = sctp_get_pf_specific(sk->sk_family);
3275 /* Control variables for partial data delivery. */
3276 atomic_set(&sp->pd_mode, 0);
3277 skb_queue_head_init(&sp->pd_lobby);
3278 sp->frag_interleave = 0;
3280 /* Create a per socket endpoint structure. Even if we
3281 * change the data structure relationships, this may still
3282 * be useful for storing pre-connect address information.
3284 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3285 if (!ep)
3286 return -ENOMEM;
3288 sp->ep = ep;
3289 sp->hmac = NULL;
3291 SCTP_DBG_OBJCNT_INC(sock);
3292 return 0;
3295 /* Cleanup any SCTP per socket resources. */
3296 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3298 struct sctp_endpoint *ep;
3300 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3302 /* Release our hold on the endpoint. */
3303 ep = sctp_sk(sk)->ep;
3304 sctp_endpoint_free(ep);
3306 return 0;
3309 /* API 4.1.7 shutdown() - TCP Style Syntax
3310 * int shutdown(int socket, int how);
3312 * sd - the socket descriptor of the association to be closed.
3313 * how - Specifies the type of shutdown. The values are
3314 * as follows:
3315 * SHUT_RD
3316 * Disables further receive operations. No SCTP
3317 * protocol action is taken.
3318 * SHUT_WR
3319 * Disables further send operations, and initiates
3320 * the SCTP shutdown sequence.
3321 * SHUT_RDWR
3322 * Disables further send and receive operations
3323 * and initiates the SCTP shutdown sequence.
3325 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3327 struct sctp_endpoint *ep;
3328 struct sctp_association *asoc;
3330 if (!sctp_style(sk, TCP))
3331 return;
3333 if (how & SEND_SHUTDOWN) {
3334 ep = sctp_sk(sk)->ep;
3335 if (!list_empty(&ep->asocs)) {
3336 asoc = list_entry(ep->asocs.next,
3337 struct sctp_association, asocs);
3338 sctp_primitive_SHUTDOWN(asoc, NULL);
3343 /* 7.2.1 Association Status (SCTP_STATUS)
3345 * Applications can retrieve current status information about an
3346 * association, including association state, peer receiver window size,
3347 * number of unacked data chunks, and number of data chunks pending
3348 * receipt. This information is read-only.
3350 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3351 char __user *optval,
3352 int __user *optlen)
3354 struct sctp_status status;
3355 struct sctp_association *asoc = NULL;
3356 struct sctp_transport *transport;
3357 sctp_assoc_t associd;
3358 int retval = 0;
3360 if (len != sizeof(status)) {
3361 retval = -EINVAL;
3362 goto out;
3365 if (copy_from_user(&status, optval, sizeof(status))) {
3366 retval = -EFAULT;
3367 goto out;
3370 associd = status.sstat_assoc_id;
3371 asoc = sctp_id2assoc(sk, associd);
3372 if (!asoc) {
3373 retval = -EINVAL;
3374 goto out;
3377 transport = asoc->peer.primary_path;
3379 status.sstat_assoc_id = sctp_assoc2id(asoc);
3380 status.sstat_state = asoc->state;
3381 status.sstat_rwnd = asoc->peer.rwnd;
3382 status.sstat_unackdata = asoc->unack_data;
3384 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3385 status.sstat_instrms = asoc->c.sinit_max_instreams;
3386 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3387 status.sstat_fragmentation_point = asoc->frag_point;
3388 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3389 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3390 transport->af_specific->sockaddr_len);
3391 /* Map ipv4 address into v4-mapped-on-v6 address. */
3392 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3393 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3394 status.sstat_primary.spinfo_state = transport->state;
3395 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3396 status.sstat_primary.spinfo_srtt = transport->srtt;
3397 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3398 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3400 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3401 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3403 if (put_user(len, optlen)) {
3404 retval = -EFAULT;
3405 goto out;
3408 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3409 len, status.sstat_state, status.sstat_rwnd,
3410 status.sstat_assoc_id);
3412 if (copy_to_user(optval, &status, len)) {
3413 retval = -EFAULT;
3414 goto out;
3417 out:
3418 return (retval);
3422 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3424 * Applications can retrieve information about a specific peer address
3425 * of an association, including its reachability state, congestion
3426 * window, and retransmission timer values. This information is
3427 * read-only.
3429 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3430 char __user *optval,
3431 int __user *optlen)
3433 struct sctp_paddrinfo pinfo;
3434 struct sctp_transport *transport;
3435 int retval = 0;
3437 if (len != sizeof(pinfo)) {
3438 retval = -EINVAL;
3439 goto out;
3442 if (copy_from_user(&pinfo, optval, sizeof(pinfo))) {
3443 retval = -EFAULT;
3444 goto out;
3447 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3448 pinfo.spinfo_assoc_id);
3449 if (!transport)
3450 return -EINVAL;
3452 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3453 pinfo.spinfo_state = transport->state;
3454 pinfo.spinfo_cwnd = transport->cwnd;
3455 pinfo.spinfo_srtt = transport->srtt;
3456 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3457 pinfo.spinfo_mtu = transport->pathmtu;
3459 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3460 pinfo.spinfo_state = SCTP_ACTIVE;
3462 if (put_user(len, optlen)) {
3463 retval = -EFAULT;
3464 goto out;
3467 if (copy_to_user(optval, &pinfo, len)) {
3468 retval = -EFAULT;
3469 goto out;
3472 out:
3473 return (retval);
3476 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3478 * This option is a on/off flag. If enabled no SCTP message
3479 * fragmentation will be performed. Instead if a message being sent
3480 * exceeds the current PMTU size, the message will NOT be sent and
3481 * instead a error will be indicated to the user.
3483 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3484 char __user *optval, int __user *optlen)
3486 int val;
3488 if (len < sizeof(int))
3489 return -EINVAL;
3491 len = sizeof(int);
3492 val = (sctp_sk(sk)->disable_fragments == 1);
3493 if (put_user(len, optlen))
3494 return -EFAULT;
3495 if (copy_to_user(optval, &val, len))
3496 return -EFAULT;
3497 return 0;
3500 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3502 * This socket option is used to specify various notifications and
3503 * ancillary data the user wishes to receive.
3505 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3506 int __user *optlen)
3508 if (len != sizeof(struct sctp_event_subscribe))
3509 return -EINVAL;
3510 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3511 return -EFAULT;
3512 return 0;
3515 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3517 * This socket option is applicable to the UDP-style socket only. When
3518 * set it will cause associations that are idle for more than the
3519 * specified number of seconds to automatically close. An association
3520 * being idle is defined an association that has NOT sent or received
3521 * user data. The special value of '0' indicates that no automatic
3522 * close of any associations should be performed. The option expects an
3523 * integer defining the number of seconds of idle time before an
3524 * association is closed.
3526 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3528 /* Applicable to UDP-style socket only */
3529 if (sctp_style(sk, TCP))
3530 return -EOPNOTSUPP;
3531 if (len != sizeof(int))
3532 return -EINVAL;
3533 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
3534 return -EFAULT;
3535 return 0;
3538 /* Helper routine to branch off an association to a new socket. */
3539 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3540 struct socket **sockp)
3542 struct sock *sk = asoc->base.sk;
3543 struct socket *sock;
3544 struct inet_sock *inetsk;
3545 int err = 0;
3547 /* An association cannot be branched off from an already peeled-off
3548 * socket, nor is this supported for tcp style sockets.
3550 if (!sctp_style(sk, UDP))
3551 return -EINVAL;
3553 /* Create a new socket. */
3554 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3555 if (err < 0)
3556 return err;
3558 /* Populate the fields of the newsk from the oldsk and migrate the
3559 * asoc to the newsk.
3561 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3563 /* Make peeled-off sockets more like 1-1 accepted sockets.
3564 * Set the daddr and initialize id to something more random
3566 inetsk = inet_sk(sock->sk);
3567 inetsk->daddr = asoc->peer.primary_addr.v4.sin_addr.s_addr;
3568 inetsk->id = asoc->next_tsn ^ jiffies;
3570 *sockp = sock;
3572 return err;
3575 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3577 sctp_peeloff_arg_t peeloff;
3578 struct socket *newsock;
3579 int retval = 0;
3580 struct sctp_association *asoc;
3582 if (len != sizeof(sctp_peeloff_arg_t))
3583 return -EINVAL;
3584 if (copy_from_user(&peeloff, optval, len))
3585 return -EFAULT;
3587 asoc = sctp_id2assoc(sk, peeloff.associd);
3588 if (!asoc) {
3589 retval = -EINVAL;
3590 goto out;
3593 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3595 retval = sctp_do_peeloff(asoc, &newsock);
3596 if (retval < 0)
3597 goto out;
3599 /* Map the socket to an unused fd that can be returned to the user. */
3600 retval = sock_map_fd(newsock);
3601 if (retval < 0) {
3602 sock_release(newsock);
3603 goto out;
3606 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3607 __FUNCTION__, sk, asoc, newsock->sk, retval);
3609 /* Return the fd mapped to the new socket. */
3610 peeloff.sd = retval;
3611 if (copy_to_user(optval, &peeloff, len))
3612 retval = -EFAULT;
3614 out:
3615 return retval;
3618 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3620 * Applications can enable or disable heartbeats for any peer address of
3621 * an association, modify an address's heartbeat interval, force a
3622 * heartbeat to be sent immediately, and adjust the address's maximum
3623 * number of retransmissions sent before an address is considered
3624 * unreachable. The following structure is used to access and modify an
3625 * address's parameters:
3627 * struct sctp_paddrparams {
3628 * sctp_assoc_t spp_assoc_id;
3629 * struct sockaddr_storage spp_address;
3630 * uint32_t spp_hbinterval;
3631 * uint16_t spp_pathmaxrxt;
3632 * uint32_t spp_pathmtu;
3633 * uint32_t spp_sackdelay;
3634 * uint32_t spp_flags;
3635 * };
3637 * spp_assoc_id - (one-to-many style socket) This is filled in the
3638 * application, and identifies the association for
3639 * this query.
3640 * spp_address - This specifies which address is of interest.
3641 * spp_hbinterval - This contains the value of the heartbeat interval,
3642 * in milliseconds. If a value of zero
3643 * is present in this field then no changes are to
3644 * be made to this parameter.
3645 * spp_pathmaxrxt - This contains the maximum number of
3646 * retransmissions before this address shall be
3647 * considered unreachable. If a value of zero
3648 * is present in this field then no changes are to
3649 * be made to this parameter.
3650 * spp_pathmtu - When Path MTU discovery is disabled the value
3651 * specified here will be the "fixed" path mtu.
3652 * Note that if the spp_address field is empty
3653 * then all associations on this address will
3654 * have this fixed path mtu set upon them.
3656 * spp_sackdelay - When delayed sack is enabled, this value specifies
3657 * the number of milliseconds that sacks will be delayed
3658 * for. This value will apply to all addresses of an
3659 * association if the spp_address field is empty. Note
3660 * also, that if delayed sack is enabled and this
3661 * value is set to 0, no change is made to the last
3662 * recorded delayed sack timer value.
3664 * spp_flags - These flags are used to control various features
3665 * on an association. The flag field may contain
3666 * zero or more of the following options.
3668 * SPP_HB_ENABLE - Enable heartbeats on the
3669 * specified address. Note that if the address
3670 * field is empty all addresses for the association
3671 * have heartbeats enabled upon them.
3673 * SPP_HB_DISABLE - Disable heartbeats on the
3674 * speicifed address. Note that if the address
3675 * field is empty all addresses for the association
3676 * will have their heartbeats disabled. Note also
3677 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3678 * mutually exclusive, only one of these two should
3679 * be specified. Enabling both fields will have
3680 * undetermined results.
3682 * SPP_HB_DEMAND - Request a user initiated heartbeat
3683 * to be made immediately.
3685 * SPP_PMTUD_ENABLE - This field will enable PMTU
3686 * discovery upon the specified address. Note that
3687 * if the address feild is empty then all addresses
3688 * on the association are effected.
3690 * SPP_PMTUD_DISABLE - This field will disable PMTU
3691 * discovery upon the specified address. Note that
3692 * if the address feild is empty then all addresses
3693 * on the association are effected. Not also that
3694 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3695 * exclusive. Enabling both will have undetermined
3696 * results.
3698 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3699 * on delayed sack. The time specified in spp_sackdelay
3700 * is used to specify the sack delay for this address. Note
3701 * that if spp_address is empty then all addresses will
3702 * enable delayed sack and take on the sack delay
3703 * value specified in spp_sackdelay.
3704 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3705 * off delayed sack. If the spp_address field is blank then
3706 * delayed sack is disabled for the entire association. Note
3707 * also that this field is mutually exclusive to
3708 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3709 * results.
3711 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3712 char __user *optval, int __user *optlen)
3714 struct sctp_paddrparams params;
3715 struct sctp_transport *trans = NULL;
3716 struct sctp_association *asoc = NULL;
3717 struct sctp_sock *sp = sctp_sk(sk);
3719 if (len != sizeof(struct sctp_paddrparams))
3720 return -EINVAL;
3722 if (copy_from_user(&params, optval, len))
3723 return -EFAULT;
3725 /* If an address other than INADDR_ANY is specified, and
3726 * no transport is found, then the request is invalid.
3728 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3729 trans = sctp_addr_id2transport(sk, &params.spp_address,
3730 params.spp_assoc_id);
3731 if (!trans) {
3732 SCTP_DEBUG_PRINTK("Failed no transport\n");
3733 return -EINVAL;
3737 /* Get association, if assoc_id != 0 and the socket is a one
3738 * to many style socket, and an association was not found, then
3739 * the id was invalid.
3741 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3742 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3743 SCTP_DEBUG_PRINTK("Failed no association\n");
3744 return -EINVAL;
3747 if (trans) {
3748 /* Fetch transport values. */
3749 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3750 params.spp_pathmtu = trans->pathmtu;
3751 params.spp_pathmaxrxt = trans->pathmaxrxt;
3752 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3754 /*draft-11 doesn't say what to return in spp_flags*/
3755 params.spp_flags = trans->param_flags;
3756 } else if (asoc) {
3757 /* Fetch association values. */
3758 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3759 params.spp_pathmtu = asoc->pathmtu;
3760 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3761 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3763 /*draft-11 doesn't say what to return in spp_flags*/
3764 params.spp_flags = asoc->param_flags;
3765 } else {
3766 /* Fetch socket values. */
3767 params.spp_hbinterval = sp->hbinterval;
3768 params.spp_pathmtu = sp->pathmtu;
3769 params.spp_sackdelay = sp->sackdelay;
3770 params.spp_pathmaxrxt = sp->pathmaxrxt;
3772 /*draft-11 doesn't say what to return in spp_flags*/
3773 params.spp_flags = sp->param_flags;
3776 if (copy_to_user(optval, &params, len))
3777 return -EFAULT;
3779 if (put_user(len, optlen))
3780 return -EFAULT;
3782 return 0;
3785 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
3787 * This options will get or set the delayed ack timer. The time is set
3788 * in milliseconds. If the assoc_id is 0, then this sets or gets the
3789 * endpoints default delayed ack timer value. If the assoc_id field is
3790 * non-zero, then the set or get effects the specified association.
3792 * struct sctp_assoc_value {
3793 * sctp_assoc_t assoc_id;
3794 * uint32_t assoc_value;
3795 * };
3797 * assoc_id - This parameter, indicates which association the
3798 * user is preforming an action upon. Note that if
3799 * this field's value is zero then the endpoints
3800 * default value is changed (effecting future
3801 * associations only).
3803 * assoc_value - This parameter contains the number of milliseconds
3804 * that the user is requesting the delayed ACK timer
3805 * be set to. Note that this value is defined in
3806 * the standard to be between 200 and 500 milliseconds.
3808 * Note: a value of zero will leave the value alone,
3809 * but disable SACK delay. A non-zero value will also
3810 * enable SACK delay.
3812 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
3813 char __user *optval,
3814 int __user *optlen)
3816 struct sctp_assoc_value params;
3817 struct sctp_association *asoc = NULL;
3818 struct sctp_sock *sp = sctp_sk(sk);
3820 if (len != sizeof(struct sctp_assoc_value))
3821 return - EINVAL;
3823 if (copy_from_user(&params, optval, len))
3824 return -EFAULT;
3826 /* Get association, if assoc_id != 0 and the socket is a one
3827 * to many style socket, and an association was not found, then
3828 * the id was invalid.
3830 asoc = sctp_id2assoc(sk, params.assoc_id);
3831 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3832 return -EINVAL;
3834 if (asoc) {
3835 /* Fetch association values. */
3836 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
3837 params.assoc_value = jiffies_to_msecs(
3838 asoc->sackdelay);
3839 else
3840 params.assoc_value = 0;
3841 } else {
3842 /* Fetch socket values. */
3843 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
3844 params.assoc_value = sp->sackdelay;
3845 else
3846 params.assoc_value = 0;
3849 if (copy_to_user(optval, &params, len))
3850 return -EFAULT;
3852 if (put_user(len, optlen))
3853 return -EFAULT;
3855 return 0;
3858 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
3860 * Applications can specify protocol parameters for the default association
3861 * initialization. The option name argument to setsockopt() and getsockopt()
3862 * is SCTP_INITMSG.
3864 * Setting initialization parameters is effective only on an unconnected
3865 * socket (for UDP-style sockets only future associations are effected
3866 * by the change). With TCP-style sockets, this option is inherited by
3867 * sockets derived from a listener socket.
3869 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
3871 if (len != sizeof(struct sctp_initmsg))
3872 return -EINVAL;
3873 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
3874 return -EFAULT;
3875 return 0;
3878 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
3879 char __user *optval,
3880 int __user *optlen)
3882 sctp_assoc_t id;
3883 struct sctp_association *asoc;
3884 struct list_head *pos;
3885 int cnt = 0;
3887 if (len != sizeof(sctp_assoc_t))
3888 return -EINVAL;
3890 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
3891 return -EFAULT;
3893 /* For UDP-style sockets, id specifies the association to query. */
3894 asoc = sctp_id2assoc(sk, id);
3895 if (!asoc)
3896 return -EINVAL;
3898 list_for_each(pos, &asoc->peer.transport_addr_list) {
3899 cnt ++;
3902 return cnt;
3906 * Old API for getting list of peer addresses. Does not work for 32-bit
3907 * programs running on a 64-bit kernel
3909 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
3910 char __user *optval,
3911 int __user *optlen)
3913 struct sctp_association *asoc;
3914 struct list_head *pos;
3915 int cnt = 0;
3916 struct sctp_getaddrs_old getaddrs;
3917 struct sctp_transport *from;
3918 void __user *to;
3919 union sctp_addr temp;
3920 struct sctp_sock *sp = sctp_sk(sk);
3921 int addrlen;
3923 if (len != sizeof(struct sctp_getaddrs_old))
3924 return -EINVAL;
3926 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
3927 return -EFAULT;
3929 if (getaddrs.addr_num <= 0) return -EINVAL;
3931 /* For UDP-style sockets, id specifies the association to query. */
3932 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3933 if (!asoc)
3934 return -EINVAL;
3936 to = (void __user *)getaddrs.addrs;
3937 list_for_each(pos, &asoc->peer.transport_addr_list) {
3938 from = list_entry(pos, struct sctp_transport, transports);
3939 memcpy(&temp, &from->ipaddr, sizeof(temp));
3940 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3941 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3942 if (copy_to_user(to, &temp, addrlen))
3943 return -EFAULT;
3944 to += addrlen ;
3945 cnt ++;
3946 if (cnt >= getaddrs.addr_num) break;
3948 getaddrs.addr_num = cnt;
3949 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
3950 return -EFAULT;
3952 return 0;
3955 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
3956 char __user *optval, int __user *optlen)
3958 struct sctp_association *asoc;
3959 struct list_head *pos;
3960 int cnt = 0;
3961 struct sctp_getaddrs getaddrs;
3962 struct sctp_transport *from;
3963 void __user *to;
3964 union sctp_addr temp;
3965 struct sctp_sock *sp = sctp_sk(sk);
3966 int addrlen;
3967 size_t space_left;
3968 int bytes_copied;
3970 if (len < sizeof(struct sctp_getaddrs))
3971 return -EINVAL;
3973 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
3974 return -EFAULT;
3976 /* For UDP-style sockets, id specifies the association to query. */
3977 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
3978 if (!asoc)
3979 return -EINVAL;
3981 to = optval + offsetof(struct sctp_getaddrs,addrs);
3982 space_left = len - sizeof(struct sctp_getaddrs) -
3983 offsetof(struct sctp_getaddrs,addrs);
3985 list_for_each(pos, &asoc->peer.transport_addr_list) {
3986 from = list_entry(pos, struct sctp_transport, transports);
3987 memcpy(&temp, &from->ipaddr, sizeof(temp));
3988 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
3989 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
3990 if(space_left < addrlen)
3991 return -ENOMEM;
3992 if (copy_to_user(to, &temp, addrlen))
3993 return -EFAULT;
3994 to += addrlen;
3995 cnt++;
3996 space_left -= addrlen;
3999 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4000 return -EFAULT;
4001 bytes_copied = ((char __user *)to) - optval;
4002 if (put_user(bytes_copied, optlen))
4003 return -EFAULT;
4005 return 0;
4008 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4009 char __user *optval,
4010 int __user *optlen)
4012 sctp_assoc_t id;
4013 struct sctp_bind_addr *bp;
4014 struct sctp_association *asoc;
4015 struct list_head *pos, *temp;
4016 struct sctp_sockaddr_entry *addr;
4017 rwlock_t *addr_lock;
4018 int cnt = 0;
4020 if (len != sizeof(sctp_assoc_t))
4021 return -EINVAL;
4023 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4024 return -EFAULT;
4027 * For UDP-style sockets, id specifies the association to query.
4028 * If the id field is set to the value '0' then the locally bound
4029 * addresses are returned without regard to any particular
4030 * association.
4032 if (0 == id) {
4033 bp = &sctp_sk(sk)->ep->base.bind_addr;
4034 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4035 } else {
4036 asoc = sctp_id2assoc(sk, id);
4037 if (!asoc)
4038 return -EINVAL;
4039 bp = &asoc->base.bind_addr;
4040 addr_lock = &asoc->base.addr_lock;
4043 sctp_read_lock(addr_lock);
4045 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4046 * addresses from the global local address list.
4048 if (sctp_list_single_entry(&bp->address_list)) {
4049 addr = list_entry(bp->address_list.next,
4050 struct sctp_sockaddr_entry, list);
4051 if (sctp_is_any(&addr->a)) {
4052 list_for_each_safe(pos, temp, &sctp_local_addr_list) {
4053 addr = list_entry(pos,
4054 struct sctp_sockaddr_entry,
4055 list);
4056 if ((PF_INET == sk->sk_family) &&
4057 (AF_INET6 == addr->a.sa.sa_family))
4058 continue;
4059 cnt++;
4061 } else {
4062 cnt = 1;
4064 goto done;
4067 list_for_each(pos, &bp->address_list) {
4068 cnt ++;
4071 done:
4072 sctp_read_unlock(addr_lock);
4073 return cnt;
4076 /* Helper function that copies local addresses to user and returns the number
4077 * of addresses copied.
4079 static int sctp_copy_laddrs_to_user_old(struct sock *sk, __u16 port, int max_addrs,
4080 void __user *to)
4082 struct list_head *pos, *next;
4083 struct sctp_sockaddr_entry *addr;
4084 union sctp_addr temp;
4085 int cnt = 0;
4086 int addrlen;
4088 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4089 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4090 if ((PF_INET == sk->sk_family) &&
4091 (AF_INET6 == addr->a.sa.sa_family))
4092 continue;
4093 memcpy(&temp, &addr->a, sizeof(temp));
4094 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4095 &temp);
4096 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4097 if (copy_to_user(to, &temp, addrlen))
4098 return -EFAULT;
4100 to += addrlen;
4101 cnt ++;
4102 if (cnt >= max_addrs) break;
4105 return cnt;
4108 static int sctp_copy_laddrs_to_user(struct sock *sk, __u16 port,
4109 void __user **to, size_t space_left)
4111 struct list_head *pos, *next;
4112 struct sctp_sockaddr_entry *addr;
4113 union sctp_addr temp;
4114 int cnt = 0;
4115 int addrlen;
4117 list_for_each_safe(pos, next, &sctp_local_addr_list) {
4118 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4119 if ((PF_INET == sk->sk_family) &&
4120 (AF_INET6 == addr->a.sa.sa_family))
4121 continue;
4122 memcpy(&temp, &addr->a, sizeof(temp));
4123 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4124 &temp);
4125 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4126 if(space_left<addrlen)
4127 return -ENOMEM;
4128 if (copy_to_user(*to, &temp, addrlen))
4129 return -EFAULT;
4131 *to += addrlen;
4132 cnt ++;
4133 space_left -= addrlen;
4136 return cnt;
4139 /* Old API for getting list of local addresses. Does not work for 32-bit
4140 * programs running on a 64-bit kernel
4142 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4143 char __user *optval, int __user *optlen)
4145 struct sctp_bind_addr *bp;
4146 struct sctp_association *asoc;
4147 struct list_head *pos;
4148 int cnt = 0;
4149 struct sctp_getaddrs_old getaddrs;
4150 struct sctp_sockaddr_entry *addr;
4151 void __user *to;
4152 union sctp_addr temp;
4153 struct sctp_sock *sp = sctp_sk(sk);
4154 int addrlen;
4155 rwlock_t *addr_lock;
4156 int err = 0;
4158 if (len != sizeof(struct sctp_getaddrs_old))
4159 return -EINVAL;
4161 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs_old)))
4162 return -EFAULT;
4164 if (getaddrs.addr_num <= 0) return -EINVAL;
4166 * For UDP-style sockets, id specifies the association to query.
4167 * If the id field is set to the value '0' then the locally bound
4168 * addresses are returned without regard to any particular
4169 * association.
4171 if (0 == getaddrs.assoc_id) {
4172 bp = &sctp_sk(sk)->ep->base.bind_addr;
4173 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4174 } else {
4175 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4176 if (!asoc)
4177 return -EINVAL;
4178 bp = &asoc->base.bind_addr;
4179 addr_lock = &asoc->base.addr_lock;
4182 to = getaddrs.addrs;
4184 sctp_read_lock(addr_lock);
4186 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4187 * addresses from the global local address list.
4189 if (sctp_list_single_entry(&bp->address_list)) {
4190 addr = list_entry(bp->address_list.next,
4191 struct sctp_sockaddr_entry, list);
4192 if (sctp_is_any(&addr->a)) {
4193 cnt = sctp_copy_laddrs_to_user_old(sk, bp->port,
4194 getaddrs.addr_num,
4195 to);
4196 if (cnt < 0) {
4197 err = cnt;
4198 goto unlock;
4200 goto copy_getaddrs;
4204 list_for_each(pos, &bp->address_list) {
4205 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4206 memcpy(&temp, &addr->a, sizeof(temp));
4207 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4208 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4209 if (copy_to_user(to, &temp, addrlen)) {
4210 err = -EFAULT;
4211 goto unlock;
4213 to += addrlen;
4214 cnt ++;
4215 if (cnt >= getaddrs.addr_num) break;
4218 copy_getaddrs:
4219 getaddrs.addr_num = cnt;
4220 if (copy_to_user(optval, &getaddrs, sizeof(struct sctp_getaddrs_old)))
4221 err = -EFAULT;
4223 unlock:
4224 sctp_read_unlock(addr_lock);
4225 return err;
4228 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4229 char __user *optval, int __user *optlen)
4231 struct sctp_bind_addr *bp;
4232 struct sctp_association *asoc;
4233 struct list_head *pos;
4234 int cnt = 0;
4235 struct sctp_getaddrs getaddrs;
4236 struct sctp_sockaddr_entry *addr;
4237 void __user *to;
4238 union sctp_addr temp;
4239 struct sctp_sock *sp = sctp_sk(sk);
4240 int addrlen;
4241 rwlock_t *addr_lock;
4242 int err = 0;
4243 size_t space_left;
4244 int bytes_copied;
4246 if (len <= sizeof(struct sctp_getaddrs))
4247 return -EINVAL;
4249 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4250 return -EFAULT;
4253 * For UDP-style sockets, id specifies the association to query.
4254 * If the id field is set to the value '0' then the locally bound
4255 * addresses are returned without regard to any particular
4256 * association.
4258 if (0 == getaddrs.assoc_id) {
4259 bp = &sctp_sk(sk)->ep->base.bind_addr;
4260 addr_lock = &sctp_sk(sk)->ep->base.addr_lock;
4261 } else {
4262 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4263 if (!asoc)
4264 return -EINVAL;
4265 bp = &asoc->base.bind_addr;
4266 addr_lock = &asoc->base.addr_lock;
4269 to = optval + offsetof(struct sctp_getaddrs,addrs);
4270 space_left = len - sizeof(struct sctp_getaddrs) -
4271 offsetof(struct sctp_getaddrs,addrs);
4273 sctp_read_lock(addr_lock);
4275 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4276 * addresses from the global local address list.
4278 if (sctp_list_single_entry(&bp->address_list)) {
4279 addr = list_entry(bp->address_list.next,
4280 struct sctp_sockaddr_entry, list);
4281 if (sctp_is_any(&addr->a)) {
4282 cnt = sctp_copy_laddrs_to_user(sk, bp->port,
4283 &to, space_left);
4284 if (cnt < 0) {
4285 err = cnt;
4286 goto unlock;
4288 goto copy_getaddrs;
4292 list_for_each(pos, &bp->address_list) {
4293 addr = list_entry(pos, struct sctp_sockaddr_entry, list);
4294 memcpy(&temp, &addr->a, sizeof(temp));
4295 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4296 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4297 if(space_left < addrlen)
4298 return -ENOMEM; /*fixme: right error?*/
4299 if (copy_to_user(to, &temp, addrlen)) {
4300 err = -EFAULT;
4301 goto unlock;
4303 to += addrlen;
4304 cnt ++;
4305 space_left -= addrlen;
4308 copy_getaddrs:
4309 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4310 return -EFAULT;
4311 bytes_copied = ((char __user *)to) - optval;
4312 if (put_user(bytes_copied, optlen))
4313 return -EFAULT;
4315 unlock:
4316 sctp_read_unlock(addr_lock);
4317 return err;
4320 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4322 * Requests that the local SCTP stack use the enclosed peer address as
4323 * the association primary. The enclosed address must be one of the
4324 * association peer's addresses.
4326 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4327 char __user *optval, int __user *optlen)
4329 struct sctp_prim prim;
4330 struct sctp_association *asoc;
4331 struct sctp_sock *sp = sctp_sk(sk);
4333 if (len != sizeof(struct sctp_prim))
4334 return -EINVAL;
4336 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
4337 return -EFAULT;
4339 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4340 if (!asoc)
4341 return -EINVAL;
4343 if (!asoc->peer.primary_path)
4344 return -ENOTCONN;
4346 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4347 asoc->peer.primary_path->af_specific->sockaddr_len);
4349 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4350 (union sctp_addr *)&prim.ssp_addr);
4352 if (copy_to_user(optval, &prim, sizeof(struct sctp_prim)))
4353 return -EFAULT;
4355 return 0;
4359 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4361 * Requests that the local endpoint set the specified Adaptation Layer
4362 * Indication parameter for all future INIT and INIT-ACK exchanges.
4364 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4365 char __user *optval, int __user *optlen)
4367 struct sctp_setadaptation adaptation;
4369 if (len != sizeof(struct sctp_setadaptation))
4370 return -EINVAL;
4372 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4373 if (copy_to_user(optval, &adaptation, len))
4374 return -EFAULT;
4376 return 0;
4381 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4383 * Applications that wish to use the sendto() system call may wish to
4384 * specify a default set of parameters that would normally be supplied
4385 * through the inclusion of ancillary data. This socket option allows
4386 * such an application to set the default sctp_sndrcvinfo structure.
4389 * The application that wishes to use this socket option simply passes
4390 * in to this call the sctp_sndrcvinfo structure defined in Section
4391 * 5.2.2) The input parameters accepted by this call include
4392 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4393 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4394 * to this call if the caller is using the UDP model.
4396 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4398 static int sctp_getsockopt_default_send_param(struct sock *sk,
4399 int len, char __user *optval,
4400 int __user *optlen)
4402 struct sctp_sndrcvinfo info;
4403 struct sctp_association *asoc;
4404 struct sctp_sock *sp = sctp_sk(sk);
4406 if (len != sizeof(struct sctp_sndrcvinfo))
4407 return -EINVAL;
4408 if (copy_from_user(&info, optval, sizeof(struct sctp_sndrcvinfo)))
4409 return -EFAULT;
4411 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4412 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4413 return -EINVAL;
4415 if (asoc) {
4416 info.sinfo_stream = asoc->default_stream;
4417 info.sinfo_flags = asoc->default_flags;
4418 info.sinfo_ppid = asoc->default_ppid;
4419 info.sinfo_context = asoc->default_context;
4420 info.sinfo_timetolive = asoc->default_timetolive;
4421 } else {
4422 info.sinfo_stream = sp->default_stream;
4423 info.sinfo_flags = sp->default_flags;
4424 info.sinfo_ppid = sp->default_ppid;
4425 info.sinfo_context = sp->default_context;
4426 info.sinfo_timetolive = sp->default_timetolive;
4429 if (copy_to_user(optval, &info, sizeof(struct sctp_sndrcvinfo)))
4430 return -EFAULT;
4432 return 0;
4437 * 7.1.5 SCTP_NODELAY
4439 * Turn on/off any Nagle-like algorithm. This means that packets are
4440 * generally sent as soon as possible and no unnecessary delays are
4441 * introduced, at the cost of more packets in the network. Expects an
4442 * integer boolean flag.
4445 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4446 char __user *optval, int __user *optlen)
4448 int val;
4450 if (len < sizeof(int))
4451 return -EINVAL;
4453 len = sizeof(int);
4454 val = (sctp_sk(sk)->nodelay == 1);
4455 if (put_user(len, optlen))
4456 return -EFAULT;
4457 if (copy_to_user(optval, &val, len))
4458 return -EFAULT;
4459 return 0;
4464 * 7.1.1 SCTP_RTOINFO
4466 * The protocol parameters used to initialize and bound retransmission
4467 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4468 * and modify these parameters.
4469 * All parameters are time values, in milliseconds. A value of 0, when
4470 * modifying the parameters, indicates that the current value should not
4471 * be changed.
4474 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4475 char __user *optval,
4476 int __user *optlen) {
4477 struct sctp_rtoinfo rtoinfo;
4478 struct sctp_association *asoc;
4480 if (len != sizeof (struct sctp_rtoinfo))
4481 return -EINVAL;
4483 if (copy_from_user(&rtoinfo, optval, sizeof (struct sctp_rtoinfo)))
4484 return -EFAULT;
4486 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4488 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4489 return -EINVAL;
4491 /* Values corresponding to the specific association. */
4492 if (asoc) {
4493 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4494 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4495 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4496 } else {
4497 /* Values corresponding to the endpoint. */
4498 struct sctp_sock *sp = sctp_sk(sk);
4500 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4501 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4502 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4505 if (put_user(len, optlen))
4506 return -EFAULT;
4508 if (copy_to_user(optval, &rtoinfo, len))
4509 return -EFAULT;
4511 return 0;
4516 * 7.1.2 SCTP_ASSOCINFO
4518 * This option is used to tune the the maximum retransmission attempts
4519 * of the association.
4520 * Returns an error if the new association retransmission value is
4521 * greater than the sum of the retransmission value of the peer.
4522 * See [SCTP] for more information.
4525 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4526 char __user *optval,
4527 int __user *optlen)
4530 struct sctp_assocparams assocparams;
4531 struct sctp_association *asoc;
4532 struct list_head *pos;
4533 int cnt = 0;
4535 if (len != sizeof (struct sctp_assocparams))
4536 return -EINVAL;
4538 if (copy_from_user(&assocparams, optval,
4539 sizeof (struct sctp_assocparams)))
4540 return -EFAULT;
4542 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4544 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4545 return -EINVAL;
4547 /* Values correspoinding to the specific association */
4548 if (asoc) {
4549 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4550 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4551 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4552 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4553 * 1000) +
4554 (asoc->cookie_life.tv_usec
4555 / 1000);
4557 list_for_each(pos, &asoc->peer.transport_addr_list) {
4558 cnt ++;
4561 assocparams.sasoc_number_peer_destinations = cnt;
4562 } else {
4563 /* Values corresponding to the endpoint */
4564 struct sctp_sock *sp = sctp_sk(sk);
4566 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4567 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4568 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4569 assocparams.sasoc_cookie_life =
4570 sp->assocparams.sasoc_cookie_life;
4571 assocparams.sasoc_number_peer_destinations =
4572 sp->assocparams.
4573 sasoc_number_peer_destinations;
4576 if (put_user(len, optlen))
4577 return -EFAULT;
4579 if (copy_to_user(optval, &assocparams, len))
4580 return -EFAULT;
4582 return 0;
4586 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4588 * This socket option is a boolean flag which turns on or off mapped V4
4589 * addresses. If this option is turned on and the socket is type
4590 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4591 * If this option is turned off, then no mapping will be done of V4
4592 * addresses and a user will receive both PF_INET6 and PF_INET type
4593 * addresses on the socket.
4595 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4596 char __user *optval, int __user *optlen)
4598 int val;
4599 struct sctp_sock *sp = sctp_sk(sk);
4601 if (len < sizeof(int))
4602 return -EINVAL;
4604 len = sizeof(int);
4605 val = sp->v4mapped;
4606 if (put_user(len, optlen))
4607 return -EFAULT;
4608 if (copy_to_user(optval, &val, len))
4609 return -EFAULT;
4611 return 0;
4615 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4616 * (chapter and verse is quoted at sctp_setsockopt_context())
4618 static int sctp_getsockopt_context(struct sock *sk, int len,
4619 char __user *optval, int __user *optlen)
4621 struct sctp_assoc_value params;
4622 struct sctp_sock *sp;
4623 struct sctp_association *asoc;
4625 if (len != sizeof(struct sctp_assoc_value))
4626 return -EINVAL;
4628 if (copy_from_user(&params, optval, len))
4629 return -EFAULT;
4631 sp = sctp_sk(sk);
4633 if (params.assoc_id != 0) {
4634 asoc = sctp_id2assoc(sk, params.assoc_id);
4635 if (!asoc)
4636 return -EINVAL;
4637 params.assoc_value = asoc->default_rcv_context;
4638 } else {
4639 params.assoc_value = sp->default_rcv_context;
4642 if (put_user(len, optlen))
4643 return -EFAULT;
4644 if (copy_to_user(optval, &params, len))
4645 return -EFAULT;
4647 return 0;
4651 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4653 * This socket option specifies the maximum size to put in any outgoing
4654 * SCTP chunk. If a message is larger than this size it will be
4655 * fragmented by SCTP into the specified size. Note that the underlying
4656 * SCTP implementation may fragment into smaller sized chunks when the
4657 * PMTU of the underlying association is smaller than the value set by
4658 * the user.
4660 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4661 char __user *optval, int __user *optlen)
4663 int val;
4665 if (len < sizeof(int))
4666 return -EINVAL;
4668 len = sizeof(int);
4670 val = sctp_sk(sk)->user_frag;
4671 if (put_user(len, optlen))
4672 return -EFAULT;
4673 if (copy_to_user(optval, &val, len))
4674 return -EFAULT;
4676 return 0;
4680 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4681 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4683 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4684 char __user *optval, int __user *optlen)
4686 int val;
4688 if (len < sizeof(int))
4689 return -EINVAL;
4691 len = sizeof(int);
4693 val = sctp_sk(sk)->frag_interleave;
4694 if (put_user(len, optlen))
4695 return -EFAULT;
4696 if (copy_to_user(optval, &val, len))
4697 return -EFAULT;
4699 return 0;
4703 * 7.1.25. Set or Get the sctp partial delivery point
4704 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4706 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4707 char __user *optval,
4708 int __user *optlen)
4710 u32 val;
4712 if (len < sizeof(u32))
4713 return -EINVAL;
4715 len = sizeof(u32);
4717 val = sctp_sk(sk)->pd_point;
4718 if (put_user(len, optlen))
4719 return -EFAULT;
4720 if (copy_to_user(optval, &val, len))
4721 return -EFAULT;
4723 return -ENOTSUPP;
4727 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
4728 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
4730 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
4731 char __user *optval,
4732 int __user *optlen)
4734 int val;
4736 if (len < sizeof(int))
4737 return -EINVAL;
4739 len = sizeof(int);
4741 val = sctp_sk(sk)->max_burst;
4742 if (put_user(len, optlen))
4743 return -EFAULT;
4744 if (copy_to_user(optval, &val, len))
4745 return -EFAULT;
4747 return -ENOTSUPP;
4750 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
4751 char __user *optval, int __user *optlen)
4753 int retval = 0;
4754 int len;
4756 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
4757 sk, optname);
4759 /* I can hardly begin to describe how wrong this is. This is
4760 * so broken as to be worse than useless. The API draft
4761 * REALLY is NOT helpful here... I am not convinced that the
4762 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
4763 * are at all well-founded.
4765 if (level != SOL_SCTP) {
4766 struct sctp_af *af = sctp_sk(sk)->pf->af;
4768 retval = af->getsockopt(sk, level, optname, optval, optlen);
4769 return retval;
4772 if (get_user(len, optlen))
4773 return -EFAULT;
4775 sctp_lock_sock(sk);
4777 switch (optname) {
4778 case SCTP_STATUS:
4779 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
4780 break;
4781 case SCTP_DISABLE_FRAGMENTS:
4782 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
4783 optlen);
4784 break;
4785 case SCTP_EVENTS:
4786 retval = sctp_getsockopt_events(sk, len, optval, optlen);
4787 break;
4788 case SCTP_AUTOCLOSE:
4789 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
4790 break;
4791 case SCTP_SOCKOPT_PEELOFF:
4792 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
4793 break;
4794 case SCTP_PEER_ADDR_PARAMS:
4795 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
4796 optlen);
4797 break;
4798 case SCTP_DELAYED_ACK_TIME:
4799 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
4800 optlen);
4801 break;
4802 case SCTP_INITMSG:
4803 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
4804 break;
4805 case SCTP_GET_PEER_ADDRS_NUM_OLD:
4806 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
4807 optlen);
4808 break;
4809 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
4810 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
4811 optlen);
4812 break;
4813 case SCTP_GET_PEER_ADDRS_OLD:
4814 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
4815 optlen);
4816 break;
4817 case SCTP_GET_LOCAL_ADDRS_OLD:
4818 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
4819 optlen);
4820 break;
4821 case SCTP_GET_PEER_ADDRS:
4822 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
4823 optlen);
4824 break;
4825 case SCTP_GET_LOCAL_ADDRS:
4826 retval = sctp_getsockopt_local_addrs(sk, len, optval,
4827 optlen);
4828 break;
4829 case SCTP_DEFAULT_SEND_PARAM:
4830 retval = sctp_getsockopt_default_send_param(sk, len,
4831 optval, optlen);
4832 break;
4833 case SCTP_PRIMARY_ADDR:
4834 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
4835 break;
4836 case SCTP_NODELAY:
4837 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
4838 break;
4839 case SCTP_RTOINFO:
4840 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
4841 break;
4842 case SCTP_ASSOCINFO:
4843 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
4844 break;
4845 case SCTP_I_WANT_MAPPED_V4_ADDR:
4846 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
4847 break;
4848 case SCTP_MAXSEG:
4849 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
4850 break;
4851 case SCTP_GET_PEER_ADDR_INFO:
4852 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
4853 optlen);
4854 break;
4855 case SCTP_ADAPTATION_LAYER:
4856 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
4857 optlen);
4858 break;
4859 case SCTP_CONTEXT:
4860 retval = sctp_getsockopt_context(sk, len, optval, optlen);
4861 break;
4862 case SCTP_FRAGMENT_INTERLEAVE:
4863 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
4864 optlen);
4865 break;
4866 case SCTP_PARTIAL_DELIVERY_POINT:
4867 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
4868 optlen);
4869 break;
4870 case SCTP_MAX_BURST:
4871 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
4872 break;
4873 default:
4874 retval = -ENOPROTOOPT;
4875 break;
4878 sctp_release_sock(sk);
4879 return retval;
4882 static void sctp_hash(struct sock *sk)
4884 /* STUB */
4887 static void sctp_unhash(struct sock *sk)
4889 /* STUB */
4892 /* Check if port is acceptable. Possibly find first available port.
4894 * The port hash table (contained in the 'global' SCTP protocol storage
4895 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
4896 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
4897 * list (the list number is the port number hashed out, so as you
4898 * would expect from a hash function, all the ports in a given list have
4899 * such a number that hashes out to the same list number; you were
4900 * expecting that, right?); so each list has a set of ports, with a
4901 * link to the socket (struct sock) that uses it, the port number and
4902 * a fastreuse flag (FIXME: NPI ipg).
4904 static struct sctp_bind_bucket *sctp_bucket_create(
4905 struct sctp_bind_hashbucket *head, unsigned short snum);
4907 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
4909 struct sctp_bind_hashbucket *head; /* hash list */
4910 struct sctp_bind_bucket *pp; /* hash list port iterator */
4911 unsigned short snum;
4912 int ret;
4914 snum = ntohs(addr->v4.sin_port);
4916 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
4917 sctp_local_bh_disable();
4919 if (snum == 0) {
4920 /* Search for an available port.
4922 * 'sctp_port_rover' was the last port assigned, so
4923 * we start to search from 'sctp_port_rover +
4924 * 1'. What we do is first check if port 'rover' is
4925 * already in the hash table; if not, we use that; if
4926 * it is, we try next.
4928 int low = sysctl_local_port_range[0];
4929 int high = sysctl_local_port_range[1];
4930 int remaining = (high - low) + 1;
4931 int rover;
4932 int index;
4934 sctp_spin_lock(&sctp_port_alloc_lock);
4935 rover = sctp_port_rover;
4936 do {
4937 rover++;
4938 if ((rover < low) || (rover > high))
4939 rover = low;
4940 index = sctp_phashfn(rover);
4941 head = &sctp_port_hashtable[index];
4942 sctp_spin_lock(&head->lock);
4943 for (pp = head->chain; pp; pp = pp->next)
4944 if (pp->port == rover)
4945 goto next;
4946 break;
4947 next:
4948 sctp_spin_unlock(&head->lock);
4949 } while (--remaining > 0);
4950 sctp_port_rover = rover;
4951 sctp_spin_unlock(&sctp_port_alloc_lock);
4953 /* Exhausted local port range during search? */
4954 ret = 1;
4955 if (remaining <= 0)
4956 goto fail;
4958 /* OK, here is the one we will use. HEAD (the port
4959 * hash table list entry) is non-NULL and we hold it's
4960 * mutex.
4962 snum = rover;
4963 } else {
4964 /* We are given an specific port number; we verify
4965 * that it is not being used. If it is used, we will
4966 * exahust the search in the hash list corresponding
4967 * to the port number (snum) - we detect that with the
4968 * port iterator, pp being NULL.
4970 head = &sctp_port_hashtable[sctp_phashfn(snum)];
4971 sctp_spin_lock(&head->lock);
4972 for (pp = head->chain; pp; pp = pp->next) {
4973 if (pp->port == snum)
4974 goto pp_found;
4977 pp = NULL;
4978 goto pp_not_found;
4979 pp_found:
4980 if (!hlist_empty(&pp->owner)) {
4981 /* We had a port hash table hit - there is an
4982 * available port (pp != NULL) and it is being
4983 * used by other socket (pp->owner not empty); that other
4984 * socket is going to be sk2.
4986 int reuse = sk->sk_reuse;
4987 struct sock *sk2;
4988 struct hlist_node *node;
4990 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
4991 if (pp->fastreuse && sk->sk_reuse)
4992 goto success;
4994 /* Run through the list of sockets bound to the port
4995 * (pp->port) [via the pointers bind_next and
4996 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
4997 * we get the endpoint they describe and run through
4998 * the endpoint's list of IP (v4 or v6) addresses,
4999 * comparing each of the addresses with the address of
5000 * the socket sk. If we find a match, then that means
5001 * that this port/socket (sk) combination are already
5002 * in an endpoint.
5004 sk_for_each_bound(sk2, node, &pp->owner) {
5005 struct sctp_endpoint *ep2;
5006 ep2 = sctp_sk(sk2)->ep;
5008 if (reuse && sk2->sk_reuse)
5009 continue;
5011 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5012 sctp_sk(sk))) {
5013 ret = (long)sk2;
5014 goto fail_unlock;
5017 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5019 pp_not_found:
5020 /* If there was a hash table miss, create a new port. */
5021 ret = 1;
5022 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5023 goto fail_unlock;
5025 /* In either case (hit or miss), make sure fastreuse is 1 only
5026 * if sk->sk_reuse is too (that is, if the caller requested
5027 * SO_REUSEADDR on this socket -sk-).
5029 if (hlist_empty(&pp->owner))
5030 pp->fastreuse = sk->sk_reuse ? 1 : 0;
5031 else if (pp->fastreuse && !sk->sk_reuse)
5032 pp->fastreuse = 0;
5034 /* We are set, so fill up all the data in the hash table
5035 * entry, tie the socket list information with the rest of the
5036 * sockets FIXME: Blurry, NPI (ipg).
5038 success:
5039 inet_sk(sk)->num = snum;
5040 if (!sctp_sk(sk)->bind_hash) {
5041 sk_add_bind_node(sk, &pp->owner);
5042 sctp_sk(sk)->bind_hash = pp;
5044 ret = 0;
5046 fail_unlock:
5047 sctp_spin_unlock(&head->lock);
5049 fail:
5050 sctp_local_bh_enable();
5051 return ret;
5054 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5055 * port is requested.
5057 static int sctp_get_port(struct sock *sk, unsigned short snum)
5059 long ret;
5060 union sctp_addr addr;
5061 struct sctp_af *af = sctp_sk(sk)->pf->af;
5063 /* Set up a dummy address struct from the sk. */
5064 af->from_sk(&addr, sk);
5065 addr.v4.sin_port = htons(snum);
5067 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5068 ret = sctp_get_port_local(sk, &addr);
5070 return (ret ? 1 : 0);
5074 * 3.1.3 listen() - UDP Style Syntax
5076 * By default, new associations are not accepted for UDP style sockets.
5077 * An application uses listen() to mark a socket as being able to
5078 * accept new associations.
5080 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5082 struct sctp_sock *sp = sctp_sk(sk);
5083 struct sctp_endpoint *ep = sp->ep;
5085 /* Only UDP style sockets that are not peeled off are allowed to
5086 * listen().
5088 if (!sctp_style(sk, UDP))
5089 return -EINVAL;
5091 /* If backlog is zero, disable listening. */
5092 if (!backlog) {
5093 if (sctp_sstate(sk, CLOSED))
5094 return 0;
5096 sctp_unhash_endpoint(ep);
5097 sk->sk_state = SCTP_SS_CLOSED;
5100 /* Return if we are already listening. */
5101 if (sctp_sstate(sk, LISTENING))
5102 return 0;
5105 * If a bind() or sctp_bindx() is not called prior to a listen()
5106 * call that allows new associations to be accepted, the system
5107 * picks an ephemeral port and will choose an address set equivalent
5108 * to binding with a wildcard address.
5110 * This is not currently spelled out in the SCTP sockets
5111 * extensions draft, but follows the practice as seen in TCP
5112 * sockets.
5114 if (!ep->base.bind_addr.port) {
5115 if (sctp_autobind(sk))
5116 return -EAGAIN;
5118 sk->sk_state = SCTP_SS_LISTENING;
5119 sctp_hash_endpoint(ep);
5120 return 0;
5124 * 4.1.3 listen() - TCP Style Syntax
5126 * Applications uses listen() to ready the SCTP endpoint for accepting
5127 * inbound associations.
5129 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5131 struct sctp_sock *sp = sctp_sk(sk);
5132 struct sctp_endpoint *ep = sp->ep;
5134 /* If backlog is zero, disable listening. */
5135 if (!backlog) {
5136 if (sctp_sstate(sk, CLOSED))
5137 return 0;
5139 sctp_unhash_endpoint(ep);
5140 sk->sk_state = SCTP_SS_CLOSED;
5143 if (sctp_sstate(sk, LISTENING))
5144 return 0;
5147 * If a bind() or sctp_bindx() is not called prior to a listen()
5148 * call that allows new associations to be accepted, the system
5149 * picks an ephemeral port and will choose an address set equivalent
5150 * to binding with a wildcard address.
5152 * This is not currently spelled out in the SCTP sockets
5153 * extensions draft, but follows the practice as seen in TCP
5154 * sockets.
5156 if (!ep->base.bind_addr.port) {
5157 if (sctp_autobind(sk))
5158 return -EAGAIN;
5160 sk->sk_state = SCTP_SS_LISTENING;
5161 sk->sk_max_ack_backlog = backlog;
5162 sctp_hash_endpoint(ep);
5163 return 0;
5167 * Move a socket to LISTENING state.
5169 int sctp_inet_listen(struct socket *sock, int backlog)
5171 struct sock *sk = sock->sk;
5172 struct crypto_hash *tfm = NULL;
5173 int err = -EINVAL;
5175 if (unlikely(backlog < 0))
5176 goto out;
5178 sctp_lock_sock(sk);
5180 if (sock->state != SS_UNCONNECTED)
5181 goto out;
5183 /* Allocate HMAC for generating cookie. */
5184 if (sctp_hmac_alg) {
5185 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5186 if (!tfm) {
5187 err = -ENOSYS;
5188 goto out;
5192 switch (sock->type) {
5193 case SOCK_SEQPACKET:
5194 err = sctp_seqpacket_listen(sk, backlog);
5195 break;
5196 case SOCK_STREAM:
5197 err = sctp_stream_listen(sk, backlog);
5198 break;
5199 default:
5200 break;
5203 if (err)
5204 goto cleanup;
5206 /* Store away the transform reference. */
5207 sctp_sk(sk)->hmac = tfm;
5208 out:
5209 sctp_release_sock(sk);
5210 return err;
5211 cleanup:
5212 crypto_free_hash(tfm);
5213 goto out;
5217 * This function is done by modeling the current datagram_poll() and the
5218 * tcp_poll(). Note that, based on these implementations, we don't
5219 * lock the socket in this function, even though it seems that,
5220 * ideally, locking or some other mechanisms can be used to ensure
5221 * the integrity of the counters (sndbuf and wmem_alloc) used
5222 * in this place. We assume that we don't need locks either until proven
5223 * otherwise.
5225 * Another thing to note is that we include the Async I/O support
5226 * here, again, by modeling the current TCP/UDP code. We don't have
5227 * a good way to test with it yet.
5229 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5231 struct sock *sk = sock->sk;
5232 struct sctp_sock *sp = sctp_sk(sk);
5233 unsigned int mask;
5235 poll_wait(file, sk->sk_sleep, wait);
5237 /* A TCP-style listening socket becomes readable when the accept queue
5238 * is not empty.
5240 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5241 return (!list_empty(&sp->ep->asocs)) ?
5242 (POLLIN | POLLRDNORM) : 0;
5244 mask = 0;
5246 /* Is there any exceptional events? */
5247 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5248 mask |= POLLERR;
5249 if (sk->sk_shutdown & RCV_SHUTDOWN)
5250 mask |= POLLRDHUP;
5251 if (sk->sk_shutdown == SHUTDOWN_MASK)
5252 mask |= POLLHUP;
5254 /* Is it readable? Reconsider this code with TCP-style support. */
5255 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5256 (sk->sk_shutdown & RCV_SHUTDOWN))
5257 mask |= POLLIN | POLLRDNORM;
5259 /* The association is either gone or not ready. */
5260 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5261 return mask;
5263 /* Is it writable? */
5264 if (sctp_writeable(sk)) {
5265 mask |= POLLOUT | POLLWRNORM;
5266 } else {
5267 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5269 * Since the socket is not locked, the buffer
5270 * might be made available after the writeable check and
5271 * before the bit is set. This could cause a lost I/O
5272 * signal. tcp_poll() has a race breaker for this race
5273 * condition. Based on their implementation, we put
5274 * in the following code to cover it as well.
5276 if (sctp_writeable(sk))
5277 mask |= POLLOUT | POLLWRNORM;
5279 return mask;
5282 /********************************************************************
5283 * 2nd Level Abstractions
5284 ********************************************************************/
5286 static struct sctp_bind_bucket *sctp_bucket_create(
5287 struct sctp_bind_hashbucket *head, unsigned short snum)
5289 struct sctp_bind_bucket *pp;
5291 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5292 SCTP_DBG_OBJCNT_INC(bind_bucket);
5293 if (pp) {
5294 pp->port = snum;
5295 pp->fastreuse = 0;
5296 INIT_HLIST_HEAD(&pp->owner);
5297 if ((pp->next = head->chain) != NULL)
5298 pp->next->pprev = &pp->next;
5299 head->chain = pp;
5300 pp->pprev = &head->chain;
5302 return pp;
5305 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5306 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5308 if (pp && hlist_empty(&pp->owner)) {
5309 if (pp->next)
5310 pp->next->pprev = pp->pprev;
5311 *(pp->pprev) = pp->next;
5312 kmem_cache_free(sctp_bucket_cachep, pp);
5313 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5317 /* Release this socket's reference to a local port. */
5318 static inline void __sctp_put_port(struct sock *sk)
5320 struct sctp_bind_hashbucket *head =
5321 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5322 struct sctp_bind_bucket *pp;
5324 sctp_spin_lock(&head->lock);
5325 pp = sctp_sk(sk)->bind_hash;
5326 __sk_del_bind_node(sk);
5327 sctp_sk(sk)->bind_hash = NULL;
5328 inet_sk(sk)->num = 0;
5329 sctp_bucket_destroy(pp);
5330 sctp_spin_unlock(&head->lock);
5333 void sctp_put_port(struct sock *sk)
5335 sctp_local_bh_disable();
5336 __sctp_put_port(sk);
5337 sctp_local_bh_enable();
5341 * The system picks an ephemeral port and choose an address set equivalent
5342 * to binding with a wildcard address.
5343 * One of those addresses will be the primary address for the association.
5344 * This automatically enables the multihoming capability of SCTP.
5346 static int sctp_autobind(struct sock *sk)
5348 union sctp_addr autoaddr;
5349 struct sctp_af *af;
5350 __be16 port;
5352 /* Initialize a local sockaddr structure to INADDR_ANY. */
5353 af = sctp_sk(sk)->pf->af;
5355 port = htons(inet_sk(sk)->num);
5356 af->inaddr_any(&autoaddr, port);
5358 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5361 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5363 * From RFC 2292
5364 * 4.2 The cmsghdr Structure *
5366 * When ancillary data is sent or received, any number of ancillary data
5367 * objects can be specified by the msg_control and msg_controllen members of
5368 * the msghdr structure, because each object is preceded by
5369 * a cmsghdr structure defining the object's length (the cmsg_len member).
5370 * Historically Berkeley-derived implementations have passed only one object
5371 * at a time, but this API allows multiple objects to be
5372 * passed in a single call to sendmsg() or recvmsg(). The following example
5373 * shows two ancillary data objects in a control buffer.
5375 * |<--------------------------- msg_controllen -------------------------->|
5376 * | |
5378 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5380 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5381 * | | |
5383 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5385 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5386 * | | | | |
5388 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5389 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5391 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5393 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5397 * msg_control
5398 * points here
5400 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5401 sctp_cmsgs_t *cmsgs)
5403 struct cmsghdr *cmsg;
5405 for (cmsg = CMSG_FIRSTHDR(msg);
5406 cmsg != NULL;
5407 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5408 if (!CMSG_OK(msg, cmsg))
5409 return -EINVAL;
5411 /* Should we parse this header or ignore? */
5412 if (cmsg->cmsg_level != IPPROTO_SCTP)
5413 continue;
5415 /* Strictly check lengths following example in SCM code. */
5416 switch (cmsg->cmsg_type) {
5417 case SCTP_INIT:
5418 /* SCTP Socket API Extension
5419 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5421 * This cmsghdr structure provides information for
5422 * initializing new SCTP associations with sendmsg().
5423 * The SCTP_INITMSG socket option uses this same data
5424 * structure. This structure is not used for
5425 * recvmsg().
5427 * cmsg_level cmsg_type cmsg_data[]
5428 * ------------ ------------ ----------------------
5429 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5431 if (cmsg->cmsg_len !=
5432 CMSG_LEN(sizeof(struct sctp_initmsg)))
5433 return -EINVAL;
5434 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5435 break;
5437 case SCTP_SNDRCV:
5438 /* SCTP Socket API Extension
5439 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5441 * This cmsghdr structure specifies SCTP options for
5442 * sendmsg() and describes SCTP header information
5443 * about a received message through recvmsg().
5445 * cmsg_level cmsg_type cmsg_data[]
5446 * ------------ ------------ ----------------------
5447 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5449 if (cmsg->cmsg_len !=
5450 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5451 return -EINVAL;
5453 cmsgs->info =
5454 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5456 /* Minimally, validate the sinfo_flags. */
5457 if (cmsgs->info->sinfo_flags &
5458 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5459 SCTP_ABORT | SCTP_EOF))
5460 return -EINVAL;
5461 break;
5463 default:
5464 return -EINVAL;
5467 return 0;
5471 * Wait for a packet..
5472 * Note: This function is the same function as in core/datagram.c
5473 * with a few modifications to make lksctp work.
5475 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5477 int error;
5478 DEFINE_WAIT(wait);
5480 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5482 /* Socket errors? */
5483 error = sock_error(sk);
5484 if (error)
5485 goto out;
5487 if (!skb_queue_empty(&sk->sk_receive_queue))
5488 goto ready;
5490 /* Socket shut down? */
5491 if (sk->sk_shutdown & RCV_SHUTDOWN)
5492 goto out;
5494 /* Sequenced packets can come disconnected. If so we report the
5495 * problem.
5497 error = -ENOTCONN;
5499 /* Is there a good reason to think that we may receive some data? */
5500 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5501 goto out;
5503 /* Handle signals. */
5504 if (signal_pending(current))
5505 goto interrupted;
5507 /* Let another process have a go. Since we are going to sleep
5508 * anyway. Note: This may cause odd behaviors if the message
5509 * does not fit in the user's buffer, but this seems to be the
5510 * only way to honor MSG_DONTWAIT realistically.
5512 sctp_release_sock(sk);
5513 *timeo_p = schedule_timeout(*timeo_p);
5514 sctp_lock_sock(sk);
5516 ready:
5517 finish_wait(sk->sk_sleep, &wait);
5518 return 0;
5520 interrupted:
5521 error = sock_intr_errno(*timeo_p);
5523 out:
5524 finish_wait(sk->sk_sleep, &wait);
5525 *err = error;
5526 return error;
5529 /* Receive a datagram.
5530 * Note: This is pretty much the same routine as in core/datagram.c
5531 * with a few changes to make lksctp work.
5533 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5534 int noblock, int *err)
5536 int error;
5537 struct sk_buff *skb;
5538 long timeo;
5540 timeo = sock_rcvtimeo(sk, noblock);
5542 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5543 timeo, MAX_SCHEDULE_TIMEOUT);
5545 do {
5546 /* Again only user level code calls this function,
5547 * so nothing interrupt level
5548 * will suddenly eat the receive_queue.
5550 * Look at current nfs client by the way...
5551 * However, this function was corrent in any case. 8)
5553 if (flags & MSG_PEEK) {
5554 spin_lock_bh(&sk->sk_receive_queue.lock);
5555 skb = skb_peek(&sk->sk_receive_queue);
5556 if (skb)
5557 atomic_inc(&skb->users);
5558 spin_unlock_bh(&sk->sk_receive_queue.lock);
5559 } else {
5560 skb = skb_dequeue(&sk->sk_receive_queue);
5563 if (skb)
5564 return skb;
5566 /* Caller is allowed not to check sk->sk_err before calling. */
5567 error = sock_error(sk);
5568 if (error)
5569 goto no_packet;
5571 if (sk->sk_shutdown & RCV_SHUTDOWN)
5572 break;
5574 /* User doesn't want to wait. */
5575 error = -EAGAIN;
5576 if (!timeo)
5577 goto no_packet;
5578 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
5580 return NULL;
5582 no_packet:
5583 *err = error;
5584 return NULL;
5587 /* If sndbuf has changed, wake up per association sndbuf waiters. */
5588 static void __sctp_write_space(struct sctp_association *asoc)
5590 struct sock *sk = asoc->base.sk;
5591 struct socket *sock = sk->sk_socket;
5593 if ((sctp_wspace(asoc) > 0) && sock) {
5594 if (waitqueue_active(&asoc->wait))
5595 wake_up_interruptible(&asoc->wait);
5597 if (sctp_writeable(sk)) {
5598 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
5599 wake_up_interruptible(sk->sk_sleep);
5601 /* Note that we try to include the Async I/O support
5602 * here by modeling from the current TCP/UDP code.
5603 * We have not tested with it yet.
5605 if (sock->fasync_list &&
5606 !(sk->sk_shutdown & SEND_SHUTDOWN))
5607 sock_wake_async(sock, 2, POLL_OUT);
5612 /* Do accounting for the sndbuf space.
5613 * Decrement the used sndbuf space of the corresponding association by the
5614 * data size which was just transmitted(freed).
5616 static void sctp_wfree(struct sk_buff *skb)
5618 struct sctp_association *asoc;
5619 struct sctp_chunk *chunk;
5620 struct sock *sk;
5622 /* Get the saved chunk pointer. */
5623 chunk = *((struct sctp_chunk **)(skb->cb));
5624 asoc = chunk->asoc;
5625 sk = asoc->base.sk;
5626 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
5627 sizeof(struct sk_buff) +
5628 sizeof(struct sctp_chunk);
5630 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
5632 sock_wfree(skb);
5633 __sctp_write_space(asoc);
5635 sctp_association_put(asoc);
5638 /* Do accounting for the receive space on the socket.
5639 * Accounting for the association is done in ulpevent.c
5640 * We set this as a destructor for the cloned data skbs so that
5641 * accounting is done at the correct time.
5643 void sctp_sock_rfree(struct sk_buff *skb)
5645 struct sock *sk = skb->sk;
5646 struct sctp_ulpevent *event = sctp_skb2event(skb);
5648 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
5652 /* Helper function to wait for space in the sndbuf. */
5653 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
5654 size_t msg_len)
5656 struct sock *sk = asoc->base.sk;
5657 int err = 0;
5658 long current_timeo = *timeo_p;
5659 DEFINE_WAIT(wait);
5661 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
5662 asoc, (long)(*timeo_p), msg_len);
5664 /* Increment the association's refcnt. */
5665 sctp_association_hold(asoc);
5667 /* Wait on the association specific sndbuf space. */
5668 for (;;) {
5669 prepare_to_wait_exclusive(&asoc->wait, &wait,
5670 TASK_INTERRUPTIBLE);
5671 if (!*timeo_p)
5672 goto do_nonblock;
5673 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5674 asoc->base.dead)
5675 goto do_error;
5676 if (signal_pending(current))
5677 goto do_interrupted;
5678 if (msg_len <= sctp_wspace(asoc))
5679 break;
5681 /* Let another process have a go. Since we are going
5682 * to sleep anyway.
5684 sctp_release_sock(sk);
5685 current_timeo = schedule_timeout(current_timeo);
5686 BUG_ON(sk != asoc->base.sk);
5687 sctp_lock_sock(sk);
5689 *timeo_p = current_timeo;
5692 out:
5693 finish_wait(&asoc->wait, &wait);
5695 /* Release the association's refcnt. */
5696 sctp_association_put(asoc);
5698 return err;
5700 do_error:
5701 err = -EPIPE;
5702 goto out;
5704 do_interrupted:
5705 err = sock_intr_errno(*timeo_p);
5706 goto out;
5708 do_nonblock:
5709 err = -EAGAIN;
5710 goto out;
5713 /* If socket sndbuf has changed, wake up all per association waiters. */
5714 void sctp_write_space(struct sock *sk)
5716 struct sctp_association *asoc;
5717 struct list_head *pos;
5719 /* Wake up the tasks in each wait queue. */
5720 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
5721 asoc = list_entry(pos, struct sctp_association, asocs);
5722 __sctp_write_space(asoc);
5726 /* Is there any sndbuf space available on the socket?
5728 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
5729 * associations on the same socket. For a UDP-style socket with
5730 * multiple associations, it is possible for it to be "unwriteable"
5731 * prematurely. I assume that this is acceptable because
5732 * a premature "unwriteable" is better than an accidental "writeable" which
5733 * would cause an unwanted block under certain circumstances. For the 1-1
5734 * UDP-style sockets or TCP-style sockets, this code should work.
5735 * - Daisy
5737 static int sctp_writeable(struct sock *sk)
5739 int amt = 0;
5741 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
5742 if (amt < 0)
5743 amt = 0;
5744 return amt;
5747 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
5748 * returns immediately with EINPROGRESS.
5750 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
5752 struct sock *sk = asoc->base.sk;
5753 int err = 0;
5754 long current_timeo = *timeo_p;
5755 DEFINE_WAIT(wait);
5757 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
5758 (long)(*timeo_p));
5760 /* Increment the association's refcnt. */
5761 sctp_association_hold(asoc);
5763 for (;;) {
5764 prepare_to_wait_exclusive(&asoc->wait, &wait,
5765 TASK_INTERRUPTIBLE);
5766 if (!*timeo_p)
5767 goto do_nonblock;
5768 if (sk->sk_shutdown & RCV_SHUTDOWN)
5769 break;
5770 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
5771 asoc->base.dead)
5772 goto do_error;
5773 if (signal_pending(current))
5774 goto do_interrupted;
5776 if (sctp_state(asoc, ESTABLISHED))
5777 break;
5779 /* Let another process have a go. Since we are going
5780 * to sleep anyway.
5782 sctp_release_sock(sk);
5783 current_timeo = schedule_timeout(current_timeo);
5784 sctp_lock_sock(sk);
5786 *timeo_p = current_timeo;
5789 out:
5790 finish_wait(&asoc->wait, &wait);
5792 /* Release the association's refcnt. */
5793 sctp_association_put(asoc);
5795 return err;
5797 do_error:
5798 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
5799 err = -ETIMEDOUT;
5800 else
5801 err = -ECONNREFUSED;
5802 goto out;
5804 do_interrupted:
5805 err = sock_intr_errno(*timeo_p);
5806 goto out;
5808 do_nonblock:
5809 err = -EINPROGRESS;
5810 goto out;
5813 static int sctp_wait_for_accept(struct sock *sk, long timeo)
5815 struct sctp_endpoint *ep;
5816 int err = 0;
5817 DEFINE_WAIT(wait);
5819 ep = sctp_sk(sk)->ep;
5822 for (;;) {
5823 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
5824 TASK_INTERRUPTIBLE);
5826 if (list_empty(&ep->asocs)) {
5827 sctp_release_sock(sk);
5828 timeo = schedule_timeout(timeo);
5829 sctp_lock_sock(sk);
5832 err = -EINVAL;
5833 if (!sctp_sstate(sk, LISTENING))
5834 break;
5836 err = 0;
5837 if (!list_empty(&ep->asocs))
5838 break;
5840 err = sock_intr_errno(timeo);
5841 if (signal_pending(current))
5842 break;
5844 err = -EAGAIN;
5845 if (!timeo)
5846 break;
5849 finish_wait(sk->sk_sleep, &wait);
5851 return err;
5854 void sctp_wait_for_close(struct sock *sk, long timeout)
5856 DEFINE_WAIT(wait);
5858 do {
5859 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5860 if (list_empty(&sctp_sk(sk)->ep->asocs))
5861 break;
5862 sctp_release_sock(sk);
5863 timeout = schedule_timeout(timeout);
5864 sctp_lock_sock(sk);
5865 } while (!signal_pending(current) && timeout);
5867 finish_wait(sk->sk_sleep, &wait);
5870 static void sctp_sock_rfree_frag(struct sk_buff *skb)
5872 struct sk_buff *frag;
5874 if (!skb->data_len)
5875 goto done;
5877 /* Don't forget the fragments. */
5878 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5879 sctp_sock_rfree_frag(frag);
5881 done:
5882 sctp_sock_rfree(skb);
5885 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
5887 struct sk_buff *frag;
5889 if (!skb->data_len)
5890 goto done;
5892 /* Don't forget the fragments. */
5893 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
5894 sctp_skb_set_owner_r_frag(frag, sk);
5896 done:
5897 sctp_skb_set_owner_r(skb, sk);
5900 /* Populate the fields of the newsk from the oldsk and migrate the assoc
5901 * and its messages to the newsk.
5903 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
5904 struct sctp_association *assoc,
5905 sctp_socket_type_t type)
5907 struct sctp_sock *oldsp = sctp_sk(oldsk);
5908 struct sctp_sock *newsp = sctp_sk(newsk);
5909 struct sctp_bind_bucket *pp; /* hash list port iterator */
5910 struct sctp_endpoint *newep = newsp->ep;
5911 struct sk_buff *skb, *tmp;
5912 struct sctp_ulpevent *event;
5913 int flags = 0;
5915 /* Migrate socket buffer sizes and all the socket level options to the
5916 * new socket.
5918 newsk->sk_sndbuf = oldsk->sk_sndbuf;
5919 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
5920 /* Brute force copy old sctp opt. */
5921 inet_sk_copy_descendant(newsk, oldsk);
5923 /* Restore the ep value that was overwritten with the above structure
5924 * copy.
5926 newsp->ep = newep;
5927 newsp->hmac = NULL;
5929 /* Hook this new socket in to the bind_hash list. */
5930 pp = sctp_sk(oldsk)->bind_hash;
5931 sk_add_bind_node(newsk, &pp->owner);
5932 sctp_sk(newsk)->bind_hash = pp;
5933 inet_sk(newsk)->num = inet_sk(oldsk)->num;
5935 /* Copy the bind_addr list from the original endpoint to the new
5936 * endpoint so that we can handle restarts properly
5938 if (PF_INET6 == assoc->base.sk->sk_family)
5939 flags = SCTP_ADDR6_ALLOWED;
5940 if (assoc->peer.ipv4_address)
5941 flags |= SCTP_ADDR4_PEERSUPP;
5942 if (assoc->peer.ipv6_address)
5943 flags |= SCTP_ADDR6_PEERSUPP;
5944 sctp_bind_addr_copy(&newsp->ep->base.bind_addr,
5945 &oldsp->ep->base.bind_addr,
5946 SCTP_SCOPE_GLOBAL, GFP_KERNEL, flags);
5948 /* Move any messages in the old socket's receive queue that are for the
5949 * peeled off association to the new socket's receive queue.
5951 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
5952 event = sctp_skb2event(skb);
5953 if (event->asoc == assoc) {
5954 sctp_sock_rfree_frag(skb);
5955 __skb_unlink(skb, &oldsk->sk_receive_queue);
5956 __skb_queue_tail(&newsk->sk_receive_queue, skb);
5957 sctp_skb_set_owner_r_frag(skb, newsk);
5961 /* Clean up any messages pending delivery due to partial
5962 * delivery. Three cases:
5963 * 1) No partial deliver; no work.
5964 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
5965 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
5967 skb_queue_head_init(&newsp->pd_lobby);
5968 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
5970 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
5971 struct sk_buff_head *queue;
5973 /* Decide which queue to move pd_lobby skbs to. */
5974 if (assoc->ulpq.pd_mode) {
5975 queue = &newsp->pd_lobby;
5976 } else
5977 queue = &newsk->sk_receive_queue;
5979 /* Walk through the pd_lobby, looking for skbs that
5980 * need moved to the new socket.
5982 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
5983 event = sctp_skb2event(skb);
5984 if (event->asoc == assoc) {
5985 sctp_sock_rfree_frag(skb);
5986 __skb_unlink(skb, &oldsp->pd_lobby);
5987 __skb_queue_tail(queue, skb);
5988 sctp_skb_set_owner_r_frag(skb, newsk);
5992 /* Clear up any skbs waiting for the partial
5993 * delivery to finish.
5995 if (assoc->ulpq.pd_mode)
5996 sctp_clear_pd(oldsk, NULL);
6000 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6001 sctp_sock_rfree_frag(skb);
6002 sctp_skb_set_owner_r_frag(skb, newsk);
6005 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6006 sctp_sock_rfree_frag(skb);
6007 sctp_skb_set_owner_r_frag(skb, newsk);
6010 /* Set the type of socket to indicate that it is peeled off from the
6011 * original UDP-style socket or created with the accept() call on a
6012 * TCP-style socket..
6014 newsp->type = type;
6016 /* Mark the new socket "in-use" by the user so that any packets
6017 * that may arrive on the association after we've moved it are
6018 * queued to the backlog. This prevents a potential race between
6019 * backlog processing on the old socket and new-packet processing
6020 * on the new socket.
6022 sctp_lock_sock(newsk);
6023 sctp_assoc_migrate(assoc, newsk);
6025 /* If the association on the newsk is already closed before accept()
6026 * is called, set RCV_SHUTDOWN flag.
6028 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6029 newsk->sk_shutdown |= RCV_SHUTDOWN;
6031 newsk->sk_state = SCTP_SS_ESTABLISHED;
6032 sctp_release_sock(newsk);
6035 /* This proto struct describes the ULP interface for SCTP. */
6036 struct proto sctp_prot = {
6037 .name = "SCTP",
6038 .owner = THIS_MODULE,
6039 .close = sctp_close,
6040 .connect = sctp_connect,
6041 .disconnect = sctp_disconnect,
6042 .accept = sctp_accept,
6043 .ioctl = sctp_ioctl,
6044 .init = sctp_init_sock,
6045 .destroy = sctp_destroy_sock,
6046 .shutdown = sctp_shutdown,
6047 .setsockopt = sctp_setsockopt,
6048 .getsockopt = sctp_getsockopt,
6049 .sendmsg = sctp_sendmsg,
6050 .recvmsg = sctp_recvmsg,
6051 .bind = sctp_bind,
6052 .backlog_rcv = sctp_backlog_rcv,
6053 .hash = sctp_hash,
6054 .unhash = sctp_unhash,
6055 .get_port = sctp_get_port,
6056 .obj_size = sizeof(struct sctp_sock),
6059 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6060 struct proto sctpv6_prot = {
6061 .name = "SCTPv6",
6062 .owner = THIS_MODULE,
6063 .close = sctp_close,
6064 .connect = sctp_connect,
6065 .disconnect = sctp_disconnect,
6066 .accept = sctp_accept,
6067 .ioctl = sctp_ioctl,
6068 .init = sctp_init_sock,
6069 .destroy = sctp_destroy_sock,
6070 .shutdown = sctp_shutdown,
6071 .setsockopt = sctp_setsockopt,
6072 .getsockopt = sctp_getsockopt,
6073 .sendmsg = sctp_sendmsg,
6074 .recvmsg = sctp_recvmsg,
6075 .bind = sctp_bind,
6076 .backlog_rcv = sctp_backlog_rcv,
6077 .hash = sctp_hash,
6078 .unhash = sctp_unhash,
6079 .get_port = sctp_get_port,
6080 .obj_size = sizeof(struct sctp6_sock),
6082 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */