neighbor free after disconnect
[cor_2_6_31.git] / net / sctp / socket.c
blob971890dbfea020be8f75b72269bcf34456abc102
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
60 #include <linux/types.h>
61 #include <linux/kernel.h>
62 #include <linux/wait.h>
63 #include <linux/time.h>
64 #include <linux/ip.h>
65 #include <linux/capability.h>
66 #include <linux/fcntl.h>
67 #include <linux/poll.h>
68 #include <linux/init.h>
69 #include <linux/crypto.h>
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;
111 extern int sysctl_sctp_mem[3];
112 extern int sysctl_sctp_rmem[3];
113 extern int sysctl_sctp_wmem[3];
115 static int sctp_memory_pressure;
116 static atomic_t sctp_memory_allocated;
117 struct percpu_counter sctp_sockets_allocated;
119 static void sctp_enter_memory_pressure(struct sock *sk)
121 sctp_memory_pressure = 1;
125 /* Get the sndbuf space available at the time on the association. */
126 static inline int sctp_wspace(struct sctp_association *asoc)
128 int amt;
130 if (asoc->ep->sndbuf_policy)
131 amt = asoc->sndbuf_used;
132 else
133 amt = sk_wmem_alloc_get(asoc->base.sk);
135 if (amt >= asoc->base.sk->sk_sndbuf) {
136 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
137 amt = 0;
138 else {
139 amt = sk_stream_wspace(asoc->base.sk);
140 if (amt < 0)
141 amt = 0;
143 } else {
144 amt = asoc->base.sk->sk_sndbuf - amt;
146 return amt;
149 /* Increment the used sndbuf space count of the corresponding association by
150 * the size of the outgoing data chunk.
151 * Also, set the skb destructor for sndbuf accounting later.
153 * Since it is always 1-1 between chunk and skb, and also a new skb is always
154 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
155 * destructor in the data chunk skb for the purpose of the sndbuf space
156 * tracking.
158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
160 struct sctp_association *asoc = chunk->asoc;
161 struct sock *sk = asoc->base.sk;
163 /* The sndbuf space is tracked per association. */
164 sctp_association_hold(asoc);
166 skb_set_owner_w(chunk->skb, sk);
168 chunk->skb->destructor = sctp_wfree;
169 /* Save the chunk pointer in skb for sctp_wfree to use later. */
170 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
172 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
173 sizeof(struct sk_buff) +
174 sizeof(struct sctp_chunk);
176 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
177 sk->sk_wmem_queued += chunk->skb->truesize;
178 sk_mem_charge(sk, chunk->skb->truesize);
181 /* Verify that this is a valid address. */
182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
183 int len)
185 struct sctp_af *af;
187 /* Verify basic sockaddr. */
188 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
189 if (!af)
190 return -EINVAL;
192 /* Is this a valid SCTP address? */
193 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
194 return -EINVAL;
196 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
197 return -EINVAL;
199 return 0;
202 /* Look up the association by its id. If this is not a UDP-style
203 * socket, the ID field is always ignored.
205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
207 struct sctp_association *asoc = NULL;
209 /* If this is not a UDP-style socket, assoc id should be ignored. */
210 if (!sctp_style(sk, UDP)) {
211 /* Return NULL if the socket state is not ESTABLISHED. It
212 * could be a TCP-style listening socket or a socket which
213 * hasn't yet called connect() to establish an association.
215 if (!sctp_sstate(sk, ESTABLISHED))
216 return NULL;
218 /* Get the first and the only association from the list. */
219 if (!list_empty(&sctp_sk(sk)->ep->asocs))
220 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
221 struct sctp_association, asocs);
222 return asoc;
225 /* Otherwise this is a UDP-style socket. */
226 if (!id || (id == (sctp_assoc_t)-1))
227 return NULL;
229 spin_lock_bh(&sctp_assocs_id_lock);
230 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
231 spin_unlock_bh(&sctp_assocs_id_lock);
233 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
234 return NULL;
236 return asoc;
239 /* Look up the transport from an address and an assoc id. If both address and
240 * id are specified, the associations matching the address and the id should be
241 * the same.
243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
244 struct sockaddr_storage *addr,
245 sctp_assoc_t id)
247 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
248 struct sctp_transport *transport;
249 union sctp_addr *laddr = (union sctp_addr *)addr;
251 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
252 laddr,
253 &transport);
255 if (!addr_asoc)
256 return NULL;
258 id_asoc = sctp_id2assoc(sk, id);
259 if (id_asoc && (id_asoc != addr_asoc))
260 return NULL;
262 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
263 (union sctp_addr *)addr);
265 return transport;
268 /* API 3.1.2 bind() - UDP Style Syntax
269 * The syntax of bind() is,
271 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
273 * sd - the socket descriptor returned by socket().
274 * addr - the address structure (struct sockaddr_in or struct
275 * sockaddr_in6 [RFC 2553]),
276 * addr_len - the size of the address structure.
278 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
280 int retval = 0;
282 sctp_lock_sock(sk);
284 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
285 sk, addr, addr_len);
287 /* Disallow binding twice. */
288 if (!sctp_sk(sk)->ep->base.bind_addr.port)
289 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
290 addr_len);
291 else
292 retval = -EINVAL;
294 sctp_release_sock(sk);
296 return retval;
299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
301 /* Verify this is a valid sockaddr. */
302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
303 union sctp_addr *addr, int len)
305 struct sctp_af *af;
307 /* Check minimum size. */
308 if (len < sizeof (struct sockaddr))
309 return NULL;
311 /* V4 mapped address are really of AF_INET family */
312 if (addr->sa.sa_family == AF_INET6 &&
313 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
314 if (!opt->pf->af_supported(AF_INET, opt))
315 return NULL;
316 } else {
317 /* Does this PF support this AF? */
318 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
319 return NULL;
322 /* If we get this far, af is valid. */
323 af = sctp_get_af_specific(addr->sa.sa_family);
325 if (len < af->sockaddr_len)
326 return NULL;
328 return af;
331 /* Bind a local address either to an endpoint or to an association. */
332 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
334 struct sctp_sock *sp = sctp_sk(sk);
335 struct sctp_endpoint *ep = sp->ep;
336 struct sctp_bind_addr *bp = &ep->base.bind_addr;
337 struct sctp_af *af;
338 unsigned short snum;
339 int ret = 0;
341 /* Common sockaddr verification. */
342 af = sctp_sockaddr_af(sp, addr, len);
343 if (!af) {
344 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
345 sk, addr, len);
346 return -EINVAL;
349 snum = ntohs(addr->v4.sin_port);
351 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
352 ", port: %d, new port: %d, len: %d)\n",
354 addr,
355 bp->port, snum,
356 len);
358 /* PF specific bind() address verification. */
359 if (!sp->pf->bind_verify(sp, addr))
360 return -EADDRNOTAVAIL;
362 /* We must either be unbound, or bind to the same port.
363 * It's OK to allow 0 ports if we are already bound.
364 * We'll just inhert an already bound port in this case
366 if (bp->port) {
367 if (!snum)
368 snum = bp->port;
369 else if (snum != bp->port) {
370 SCTP_DEBUG_PRINTK("sctp_do_bind:"
371 " New port %d does not match existing port "
372 "%d.\n", snum, bp->port);
373 return -EINVAL;
377 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
378 return -EACCES;
380 /* See if the address matches any of the addresses we may have
381 * already bound before checking against other endpoints.
383 if (sctp_bind_addr_match(bp, addr, sp))
384 return -EINVAL;
386 /* Make sure we are allowed to bind here.
387 * The function sctp_get_port_local() does duplicate address
388 * detection.
390 addr->v4.sin_port = htons(snum);
391 if ((ret = sctp_get_port_local(sk, addr))) {
392 return -EADDRINUSE;
395 /* Refresh ephemeral port. */
396 if (!bp->port)
397 bp->port = inet_sk(sk)->num;
399 /* Add the address to the bind address list.
400 * Use GFP_ATOMIC since BHs will be disabled.
402 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
404 /* Copy back into socket for getsockname() use. */
405 if (!ret) {
406 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
407 af->to_sk_saddr(addr, sk);
410 return ret;
413 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
415 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
416 * at any one time. If a sender, after sending an ASCONF chunk, decides
417 * it needs to transfer another ASCONF Chunk, it MUST wait until the
418 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
419 * subsequent ASCONF. Note this restriction binds each side, so at any
420 * time two ASCONF may be in-transit on any given association (one sent
421 * from each endpoint).
423 static int sctp_send_asconf(struct sctp_association *asoc,
424 struct sctp_chunk *chunk)
426 int retval = 0;
428 /* If there is an outstanding ASCONF chunk, queue it for later
429 * transmission.
431 if (asoc->addip_last_asconf) {
432 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
433 goto out;
436 /* Hold the chunk until an ASCONF_ACK is received. */
437 sctp_chunk_hold(chunk);
438 retval = sctp_primitive_ASCONF(asoc, chunk);
439 if (retval)
440 sctp_chunk_free(chunk);
441 else
442 asoc->addip_last_asconf = chunk;
444 out:
445 return retval;
448 /* Add a list of addresses as bind addresses to local endpoint or
449 * association.
451 * Basically run through each address specified in the addrs/addrcnt
452 * array/length pair, determine if it is IPv6 or IPv4 and call
453 * sctp_do_bind() on it.
455 * If any of them fails, then the operation will be reversed and the
456 * ones that were added will be removed.
458 * Only sctp_setsockopt_bindx() is supposed to call this function.
460 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
462 int cnt;
463 int retval = 0;
464 void *addr_buf;
465 struct sockaddr *sa_addr;
466 struct sctp_af *af;
468 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
469 sk, addrs, addrcnt);
471 addr_buf = addrs;
472 for (cnt = 0; cnt < addrcnt; cnt++) {
473 /* The list may contain either IPv4 or IPv6 address;
474 * determine the address length for walking thru the list.
476 sa_addr = (struct sockaddr *)addr_buf;
477 af = sctp_get_af_specific(sa_addr->sa_family);
478 if (!af) {
479 retval = -EINVAL;
480 goto err_bindx_add;
483 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
484 af->sockaddr_len);
486 addr_buf += af->sockaddr_len;
488 err_bindx_add:
489 if (retval < 0) {
490 /* Failed. Cleanup the ones that have been added */
491 if (cnt > 0)
492 sctp_bindx_rem(sk, addrs, cnt);
493 return retval;
497 return retval;
500 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
501 * associations that are part of the endpoint indicating that a list of local
502 * addresses are added to the endpoint.
504 * If any of the addresses is already in the bind address list of the
505 * association, we do not send the chunk for that association. But it will not
506 * affect other associations.
508 * Only sctp_setsockopt_bindx() is supposed to call this function.
510 static int sctp_send_asconf_add_ip(struct sock *sk,
511 struct sockaddr *addrs,
512 int addrcnt)
514 struct sctp_sock *sp;
515 struct sctp_endpoint *ep;
516 struct sctp_association *asoc;
517 struct sctp_bind_addr *bp;
518 struct sctp_chunk *chunk;
519 struct sctp_sockaddr_entry *laddr;
520 union sctp_addr *addr;
521 union sctp_addr saveaddr;
522 void *addr_buf;
523 struct sctp_af *af;
524 struct list_head *p;
525 int i;
526 int retval = 0;
528 if (!sctp_addip_enable)
529 return retval;
531 sp = sctp_sk(sk);
532 ep = sp->ep;
534 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
535 __func__, sk, addrs, addrcnt);
537 list_for_each_entry(asoc, &ep->asocs, asocs) {
539 if (!asoc->peer.asconf_capable)
540 continue;
542 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
543 continue;
545 if (!sctp_state(asoc, ESTABLISHED))
546 continue;
548 /* Check if any address in the packed array of addresses is
549 * in the bind address list of the association. If so,
550 * do not send the asconf chunk to its peer, but continue with
551 * other associations.
553 addr_buf = addrs;
554 for (i = 0; i < addrcnt; i++) {
555 addr = (union sctp_addr *)addr_buf;
556 af = sctp_get_af_specific(addr->v4.sin_family);
557 if (!af) {
558 retval = -EINVAL;
559 goto out;
562 if (sctp_assoc_lookup_laddr(asoc, addr))
563 break;
565 addr_buf += af->sockaddr_len;
567 if (i < addrcnt)
568 continue;
570 /* Use the first valid address in bind addr list of
571 * association as Address Parameter of ASCONF CHUNK.
573 bp = &asoc->base.bind_addr;
574 p = bp->address_list.next;
575 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
576 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
577 addrcnt, SCTP_PARAM_ADD_IP);
578 if (!chunk) {
579 retval = -ENOMEM;
580 goto out;
583 retval = sctp_send_asconf(asoc, chunk);
584 if (retval)
585 goto out;
587 /* Add the new addresses to the bind address list with
588 * use_as_src set to 0.
590 addr_buf = addrs;
591 for (i = 0; i < addrcnt; i++) {
592 addr = (union sctp_addr *)addr_buf;
593 af = sctp_get_af_specific(addr->v4.sin_family);
594 memcpy(&saveaddr, addr, af->sockaddr_len);
595 retval = sctp_add_bind_addr(bp, &saveaddr,
596 SCTP_ADDR_NEW, GFP_ATOMIC);
597 addr_buf += af->sockaddr_len;
601 out:
602 return retval;
605 /* Remove a list of addresses from bind addresses list. Do not remove the
606 * last address.
608 * Basically run through each address specified in the addrs/addrcnt
609 * array/length pair, determine if it is IPv6 or IPv4 and call
610 * sctp_del_bind() on it.
612 * If any of them fails, then the operation will be reversed and the
613 * ones that were removed will be added back.
615 * At least one address has to be left; if only one address is
616 * available, the operation will return -EBUSY.
618 * Only sctp_setsockopt_bindx() is supposed to call this function.
620 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
622 struct sctp_sock *sp = sctp_sk(sk);
623 struct sctp_endpoint *ep = sp->ep;
624 int cnt;
625 struct sctp_bind_addr *bp = &ep->base.bind_addr;
626 int retval = 0;
627 void *addr_buf;
628 union sctp_addr *sa_addr;
629 struct sctp_af *af;
631 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
632 sk, addrs, addrcnt);
634 addr_buf = addrs;
635 for (cnt = 0; cnt < addrcnt; cnt++) {
636 /* If the bind address list is empty or if there is only one
637 * bind address, there is nothing more to be removed (we need
638 * at least one address here).
640 if (list_empty(&bp->address_list) ||
641 (sctp_list_single_entry(&bp->address_list))) {
642 retval = -EBUSY;
643 goto err_bindx_rem;
646 sa_addr = (union sctp_addr *)addr_buf;
647 af = sctp_get_af_specific(sa_addr->sa.sa_family);
648 if (!af) {
649 retval = -EINVAL;
650 goto err_bindx_rem;
653 if (!af->addr_valid(sa_addr, sp, NULL)) {
654 retval = -EADDRNOTAVAIL;
655 goto err_bindx_rem;
658 if (sa_addr->v4.sin_port != htons(bp->port)) {
659 retval = -EINVAL;
660 goto err_bindx_rem;
663 /* FIXME - There is probably a need to check if sk->sk_saddr and
664 * sk->sk_rcv_addr are currently set to one of the addresses to
665 * be removed. This is something which needs to be looked into
666 * when we are fixing the outstanding issues with multi-homing
667 * socket routing and failover schemes. Refer to comments in
668 * sctp_do_bind(). -daisy
670 retval = sctp_del_bind_addr(bp, sa_addr);
672 addr_buf += af->sockaddr_len;
673 err_bindx_rem:
674 if (retval < 0) {
675 /* Failed. Add the ones that has been removed back */
676 if (cnt > 0)
677 sctp_bindx_add(sk, addrs, cnt);
678 return retval;
682 return retval;
685 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
686 * the associations that are part of the endpoint indicating that a list of
687 * local addresses are removed from the endpoint.
689 * If any of the addresses is already in the bind address list of the
690 * association, we do not send the chunk for that association. But it will not
691 * affect other associations.
693 * Only sctp_setsockopt_bindx() is supposed to call this function.
695 static int sctp_send_asconf_del_ip(struct sock *sk,
696 struct sockaddr *addrs,
697 int addrcnt)
699 struct sctp_sock *sp;
700 struct sctp_endpoint *ep;
701 struct sctp_association *asoc;
702 struct sctp_transport *transport;
703 struct sctp_bind_addr *bp;
704 struct sctp_chunk *chunk;
705 union sctp_addr *laddr;
706 void *addr_buf;
707 struct sctp_af *af;
708 struct sctp_sockaddr_entry *saddr;
709 int i;
710 int retval = 0;
712 if (!sctp_addip_enable)
713 return retval;
715 sp = sctp_sk(sk);
716 ep = sp->ep;
718 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
719 __func__, sk, addrs, addrcnt);
721 list_for_each_entry(asoc, &ep->asocs, asocs) {
723 if (!asoc->peer.asconf_capable)
724 continue;
726 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
727 continue;
729 if (!sctp_state(asoc, ESTABLISHED))
730 continue;
732 /* Check if any address in the packed array of addresses is
733 * not present in the bind address list of the association.
734 * If so, do not send the asconf chunk to its peer, but
735 * continue with other associations.
737 addr_buf = addrs;
738 for (i = 0; i < addrcnt; i++) {
739 laddr = (union sctp_addr *)addr_buf;
740 af = sctp_get_af_specific(laddr->v4.sin_family);
741 if (!af) {
742 retval = -EINVAL;
743 goto out;
746 if (!sctp_assoc_lookup_laddr(asoc, laddr))
747 break;
749 addr_buf += af->sockaddr_len;
751 if (i < addrcnt)
752 continue;
754 /* Find one address in the association's bind address list
755 * that is not in the packed array of addresses. This is to
756 * make sure that we do not delete all the addresses in the
757 * association.
759 bp = &asoc->base.bind_addr;
760 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
761 addrcnt, sp);
762 if (!laddr)
763 continue;
765 /* We do not need RCU protection throughout this loop
766 * because this is done under a socket lock from the
767 * setsockopt call.
769 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
770 SCTP_PARAM_DEL_IP);
771 if (!chunk) {
772 retval = -ENOMEM;
773 goto out;
776 /* Reset use_as_src flag for the addresses in the bind address
777 * list that are to be deleted.
779 addr_buf = addrs;
780 for (i = 0; i < addrcnt; i++) {
781 laddr = (union sctp_addr *)addr_buf;
782 af = sctp_get_af_specific(laddr->v4.sin_family);
783 list_for_each_entry(saddr, &bp->address_list, list) {
784 if (sctp_cmp_addr_exact(&saddr->a, laddr))
785 saddr->state = SCTP_ADDR_DEL;
787 addr_buf += af->sockaddr_len;
790 /* Update the route and saddr entries for all the transports
791 * as some of the addresses in the bind address list are
792 * about to be deleted and cannot be used as source addresses.
794 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
795 transports) {
796 dst_release(transport->dst);
797 sctp_transport_route(transport, NULL,
798 sctp_sk(asoc->base.sk));
801 retval = sctp_send_asconf(asoc, chunk);
803 out:
804 return retval;
807 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
809 * API 8.1
810 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
811 * int flags);
813 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
814 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
815 * or IPv6 addresses.
817 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
818 * Section 3.1.2 for this usage.
820 * addrs is a pointer to an array of one or more socket addresses. Each
821 * address is contained in its appropriate structure (i.e. struct
822 * sockaddr_in or struct sockaddr_in6) the family of the address type
823 * must be used to distinguish the address length (note that this
824 * representation is termed a "packed array" of addresses). The caller
825 * specifies the number of addresses in the array with addrcnt.
827 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
828 * -1, and sets errno to the appropriate error code.
830 * For SCTP, the port given in each socket address must be the same, or
831 * sctp_bindx() will fail, setting errno to EINVAL.
833 * The flags parameter is formed from the bitwise OR of zero or more of
834 * the following currently defined flags:
836 * SCTP_BINDX_ADD_ADDR
838 * SCTP_BINDX_REM_ADDR
840 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
841 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
842 * addresses from the association. The two flags are mutually exclusive;
843 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
844 * not remove all addresses from an association; sctp_bindx() will
845 * reject such an attempt with EINVAL.
847 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
848 * additional addresses with an endpoint after calling bind(). Or use
849 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
850 * socket is associated with so that no new association accepted will be
851 * associated with those addresses. If the endpoint supports dynamic
852 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
853 * endpoint to send the appropriate message to the peer to change the
854 * peers address lists.
856 * Adding and removing addresses from a connected association is
857 * optional functionality. Implementations that do not support this
858 * functionality should return EOPNOTSUPP.
860 * Basically do nothing but copying the addresses from user to kernel
861 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
862 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
863 * from userspace.
865 * We don't use copy_from_user() for optimization: we first do the
866 * sanity checks (buffer size -fast- and access check-healthy
867 * pointer); if all of those succeed, then we can alloc the memory
868 * (expensive operation) needed to copy the data to kernel. Then we do
869 * the copying without checking the user space area
870 * (__copy_from_user()).
872 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
873 * it.
875 * sk The sk of the socket
876 * addrs The pointer to the addresses in user land
877 * addrssize Size of the addrs buffer
878 * op Operation to perform (add or remove, see the flags of
879 * sctp_bindx)
881 * Returns 0 if ok, <0 errno code on error.
883 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
884 struct sockaddr __user *addrs,
885 int addrs_size, int op)
887 struct sockaddr *kaddrs;
888 int err;
889 int addrcnt = 0;
890 int walk_size = 0;
891 struct sockaddr *sa_addr;
892 void *addr_buf;
893 struct sctp_af *af;
895 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
896 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
898 if (unlikely(addrs_size <= 0))
899 return -EINVAL;
901 /* Check the user passed a healthy pointer. */
902 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
903 return -EFAULT;
905 /* Alloc space for the address array in kernel memory. */
906 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
907 if (unlikely(!kaddrs))
908 return -ENOMEM;
910 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
911 kfree(kaddrs);
912 return -EFAULT;
915 /* Walk through the addrs buffer and count the number of addresses. */
916 addr_buf = kaddrs;
917 while (walk_size < addrs_size) {
918 sa_addr = (struct sockaddr *)addr_buf;
919 af = sctp_get_af_specific(sa_addr->sa_family);
921 /* If the address family is not supported or if this address
922 * causes the address buffer to overflow return EINVAL.
924 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
925 kfree(kaddrs);
926 return -EINVAL;
928 addrcnt++;
929 addr_buf += af->sockaddr_len;
930 walk_size += af->sockaddr_len;
933 /* Do the work. */
934 switch (op) {
935 case SCTP_BINDX_ADD_ADDR:
936 err = sctp_bindx_add(sk, kaddrs, addrcnt);
937 if (err)
938 goto out;
939 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
940 break;
942 case SCTP_BINDX_REM_ADDR:
943 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
944 if (err)
945 goto out;
946 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
947 break;
949 default:
950 err = -EINVAL;
951 break;
954 out:
955 kfree(kaddrs);
957 return err;
960 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
962 * Common routine for handling connect() and sctp_connectx().
963 * Connect will come in with just a single address.
965 static int __sctp_connect(struct sock* sk,
966 struct sockaddr *kaddrs,
967 int addrs_size,
968 sctp_assoc_t *assoc_id)
970 struct sctp_sock *sp;
971 struct sctp_endpoint *ep;
972 struct sctp_association *asoc = NULL;
973 struct sctp_association *asoc2;
974 struct sctp_transport *transport;
975 union sctp_addr to;
976 struct sctp_af *af;
977 sctp_scope_t scope;
978 long timeo;
979 int err = 0;
980 int addrcnt = 0;
981 int walk_size = 0;
982 union sctp_addr *sa_addr = NULL;
983 void *addr_buf;
984 unsigned short port;
985 unsigned int f_flags = 0;
987 sp = sctp_sk(sk);
988 ep = sp->ep;
990 /* connect() cannot be done on a socket that is already in ESTABLISHED
991 * state - UDP-style peeled off socket or a TCP-style socket that
992 * is already connected.
993 * It cannot be done even on a TCP-style listening socket.
995 if (sctp_sstate(sk, ESTABLISHED) ||
996 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
997 err = -EISCONN;
998 goto out_free;
1001 /* Walk through the addrs buffer and count the number of addresses. */
1002 addr_buf = kaddrs;
1003 while (walk_size < addrs_size) {
1004 sa_addr = (union sctp_addr *)addr_buf;
1005 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1006 port = ntohs(sa_addr->v4.sin_port);
1008 /* If the address family is not supported or if this address
1009 * causes the address buffer to overflow return EINVAL.
1011 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1012 err = -EINVAL;
1013 goto out_free;
1016 /* Save current address so we can work with it */
1017 memcpy(&to, sa_addr, af->sockaddr_len);
1019 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1020 if (err)
1021 goto out_free;
1023 /* Make sure the destination port is correctly set
1024 * in all addresses.
1026 if (asoc && asoc->peer.port && asoc->peer.port != port)
1027 goto out_free;
1030 /* Check if there already is a matching association on the
1031 * endpoint (other than the one created here).
1033 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1034 if (asoc2 && asoc2 != asoc) {
1035 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1036 err = -EISCONN;
1037 else
1038 err = -EALREADY;
1039 goto out_free;
1042 /* If we could not find a matching association on the endpoint,
1043 * make sure that there is no peeled-off association matching
1044 * the peer address even on another socket.
1046 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1047 err = -EADDRNOTAVAIL;
1048 goto out_free;
1051 if (!asoc) {
1052 /* If a bind() or sctp_bindx() is not called prior to
1053 * an sctp_connectx() call, the system picks an
1054 * ephemeral port and will choose an address set
1055 * equivalent to binding with a wildcard address.
1057 if (!ep->base.bind_addr.port) {
1058 if (sctp_autobind(sk)) {
1059 err = -EAGAIN;
1060 goto out_free;
1062 } else {
1064 * If an unprivileged user inherits a 1-many
1065 * style socket with open associations on a
1066 * privileged port, it MAY be permitted to
1067 * accept new associations, but it SHOULD NOT
1068 * be permitted to open new associations.
1070 if (ep->base.bind_addr.port < PROT_SOCK &&
1071 !capable(CAP_NET_BIND_SERVICE)) {
1072 err = -EACCES;
1073 goto out_free;
1077 scope = sctp_scope(&to);
1078 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1079 if (!asoc) {
1080 err = -ENOMEM;
1081 goto out_free;
1085 /* Prime the peer's transport structures. */
1086 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1087 SCTP_UNKNOWN);
1088 if (!transport) {
1089 err = -ENOMEM;
1090 goto out_free;
1093 addrcnt++;
1094 addr_buf += af->sockaddr_len;
1095 walk_size += af->sockaddr_len;
1098 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1099 if (err < 0) {
1100 goto out_free;
1103 /* In case the user of sctp_connectx() wants an association
1104 * id back, assign one now.
1106 if (assoc_id) {
1107 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1108 if (err < 0)
1109 goto out_free;
1112 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1113 if (err < 0) {
1114 goto out_free;
1117 /* Initialize sk's dport and daddr for getpeername() */
1118 inet_sk(sk)->dport = htons(asoc->peer.port);
1119 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1120 af->to_sk_daddr(sa_addr, sk);
1121 sk->sk_err = 0;
1123 /* in-kernel sockets don't generally have a file allocated to them
1124 * if all they do is call sock_create_kern().
1126 if (sk->sk_socket->file)
1127 f_flags = sk->sk_socket->file->f_flags;
1129 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1131 err = sctp_wait_for_connect(asoc, &timeo);
1132 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1133 *assoc_id = asoc->assoc_id;
1135 /* Don't free association on exit. */
1136 asoc = NULL;
1138 out_free:
1140 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1141 " kaddrs: %p err: %d\n",
1142 asoc, kaddrs, err);
1143 if (asoc)
1144 sctp_association_free(asoc);
1145 return err;
1148 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1150 * API 8.9
1151 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1152 * sctp_assoc_t *asoc);
1154 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1155 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1156 * or IPv6 addresses.
1158 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1159 * Section 3.1.2 for this usage.
1161 * addrs is a pointer to an array of one or more socket addresses. Each
1162 * address is contained in its appropriate structure (i.e. struct
1163 * sockaddr_in or struct sockaddr_in6) the family of the address type
1164 * must be used to distengish the address length (note that this
1165 * representation is termed a "packed array" of addresses). The caller
1166 * specifies the number of addresses in the array with addrcnt.
1168 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1169 * the association id of the new association. On failure, sctp_connectx()
1170 * returns -1, and sets errno to the appropriate error code. The assoc_id
1171 * is not touched by the kernel.
1173 * For SCTP, the port given in each socket address must be the same, or
1174 * sctp_connectx() will fail, setting errno to EINVAL.
1176 * An application can use sctp_connectx to initiate an association with
1177 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1178 * allows a caller to specify multiple addresses at which a peer can be
1179 * reached. The way the SCTP stack uses the list of addresses to set up
1180 * the association is implementation dependant. This function only
1181 * specifies that the stack will try to make use of all the addresses in
1182 * the list when needed.
1184 * Note that the list of addresses passed in is only used for setting up
1185 * the association. It does not necessarily equal the set of addresses
1186 * the peer uses for the resulting association. If the caller wants to
1187 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1188 * retrieve them after the association has been set up.
1190 * Basically do nothing but copying the addresses from user to kernel
1191 * land and invoking either sctp_connectx(). This is used for tunneling
1192 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1194 * We don't use copy_from_user() for optimization: we first do the
1195 * sanity checks (buffer size -fast- and access check-healthy
1196 * pointer); if all of those succeed, then we can alloc the memory
1197 * (expensive operation) needed to copy the data to kernel. Then we do
1198 * the copying without checking the user space area
1199 * (__copy_from_user()).
1201 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1202 * it.
1204 * sk The sk of the socket
1205 * addrs The pointer to the addresses in user land
1206 * addrssize Size of the addrs buffer
1208 * Returns >=0 if ok, <0 errno code on error.
1210 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1211 struct sockaddr __user *addrs,
1212 int addrs_size,
1213 sctp_assoc_t *assoc_id)
1215 int err = 0;
1216 struct sockaddr *kaddrs;
1218 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1219 __func__, sk, addrs, addrs_size);
1221 if (unlikely(addrs_size <= 0))
1222 return -EINVAL;
1224 /* Check the user passed a healthy pointer. */
1225 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1226 return -EFAULT;
1228 /* Alloc space for the address array in kernel memory. */
1229 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1230 if (unlikely(!kaddrs))
1231 return -ENOMEM;
1233 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1234 err = -EFAULT;
1235 } else {
1236 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1239 kfree(kaddrs);
1241 return err;
1245 * This is an older interface. It's kept for backward compatibility
1246 * to the option that doesn't provide association id.
1248 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1249 struct sockaddr __user *addrs,
1250 int addrs_size)
1252 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1256 * New interface for the API. The since the API is done with a socket
1257 * option, to make it simple we feed back the association id is as a return
1258 * indication to the call. Error is always negative and association id is
1259 * always positive.
1261 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1262 struct sockaddr __user *addrs,
1263 int addrs_size)
1265 sctp_assoc_t assoc_id = 0;
1266 int err = 0;
1268 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1270 if (err)
1271 return err;
1272 else
1273 return assoc_id;
1277 * New (hopefully final) interface for the API. The option buffer is used
1278 * both for the returned association id and the addresses.
1280 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1281 char __user *optval,
1282 int __user *optlen)
1284 sctp_assoc_t assoc_id = 0;
1285 int err = 0;
1287 if (len < sizeof(assoc_id))
1288 return -EINVAL;
1290 err = __sctp_setsockopt_connectx(sk,
1291 (struct sockaddr __user *)(optval + sizeof(assoc_id)),
1292 len - sizeof(assoc_id), &assoc_id);
1294 if (err == 0 || err == -EINPROGRESS) {
1295 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1296 return -EFAULT;
1297 if (put_user(sizeof(assoc_id), optlen))
1298 return -EFAULT;
1301 return err;
1304 /* API 3.1.4 close() - UDP Style Syntax
1305 * Applications use close() to perform graceful shutdown (as described in
1306 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1307 * by a UDP-style socket.
1309 * The syntax is
1311 * ret = close(int sd);
1313 * sd - the socket descriptor of the associations to be closed.
1315 * To gracefully shutdown a specific association represented by the
1316 * UDP-style socket, an application should use the sendmsg() call,
1317 * passing no user data, but including the appropriate flag in the
1318 * ancillary data (see Section xxxx).
1320 * If sd in the close() call is a branched-off socket representing only
1321 * one association, the shutdown is performed on that association only.
1323 * 4.1.6 close() - TCP Style Syntax
1325 * Applications use close() to gracefully close down an association.
1327 * The syntax is:
1329 * int close(int sd);
1331 * sd - the socket descriptor of the association to be closed.
1333 * After an application calls close() on a socket descriptor, no further
1334 * socket operations will succeed on that descriptor.
1336 * API 7.1.4 SO_LINGER
1338 * An application using the TCP-style socket can use this option to
1339 * perform the SCTP ABORT primitive. The linger option structure is:
1341 * struct linger {
1342 * int l_onoff; // option on/off
1343 * int l_linger; // linger time
1344 * };
1346 * To enable the option, set l_onoff to 1. If the l_linger value is set
1347 * to 0, calling close() is the same as the ABORT primitive. If the
1348 * value is set to a negative value, the setsockopt() call will return
1349 * an error. If the value is set to a positive value linger_time, the
1350 * close() can be blocked for at most linger_time ms. If the graceful
1351 * shutdown phase does not finish during this period, close() will
1352 * return but the graceful shutdown phase continues in the system.
1354 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1356 struct sctp_endpoint *ep;
1357 struct sctp_association *asoc;
1358 struct list_head *pos, *temp;
1360 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1362 sctp_lock_sock(sk);
1363 sk->sk_shutdown = SHUTDOWN_MASK;
1365 ep = sctp_sk(sk)->ep;
1367 /* Walk all associations on an endpoint. */
1368 list_for_each_safe(pos, temp, &ep->asocs) {
1369 asoc = list_entry(pos, struct sctp_association, asocs);
1371 if (sctp_style(sk, TCP)) {
1372 /* A closed association can still be in the list if
1373 * it belongs to a TCP-style listening socket that is
1374 * not yet accepted. If so, free it. If not, send an
1375 * ABORT or SHUTDOWN based on the linger options.
1377 if (sctp_state(asoc, CLOSED)) {
1378 sctp_unhash_established(asoc);
1379 sctp_association_free(asoc);
1380 continue;
1384 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1385 struct sctp_chunk *chunk;
1387 chunk = sctp_make_abort_user(asoc, NULL, 0);
1388 if (chunk)
1389 sctp_primitive_ABORT(asoc, chunk);
1390 } else
1391 sctp_primitive_SHUTDOWN(asoc, NULL);
1394 /* Clean up any skbs sitting on the receive queue. */
1395 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1396 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1398 /* On a TCP-style socket, block for at most linger_time if set. */
1399 if (sctp_style(sk, TCP) && timeout)
1400 sctp_wait_for_close(sk, timeout);
1402 /* This will run the backlog queue. */
1403 sctp_release_sock(sk);
1405 /* Supposedly, no process has access to the socket, but
1406 * the net layers still may.
1408 sctp_local_bh_disable();
1409 sctp_bh_lock_sock(sk);
1411 /* Hold the sock, since sk_common_release() will put sock_put()
1412 * and we have just a little more cleanup.
1414 sock_hold(sk);
1415 sk_common_release(sk);
1417 sctp_bh_unlock_sock(sk);
1418 sctp_local_bh_enable();
1420 sock_put(sk);
1422 SCTP_DBG_OBJCNT_DEC(sock);
1425 /* Handle EPIPE error. */
1426 static int sctp_error(struct sock *sk, int flags, int err)
1428 if (err == -EPIPE)
1429 err = sock_error(sk) ? : -EPIPE;
1430 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1431 send_sig(SIGPIPE, current, 0);
1432 return err;
1435 /* API 3.1.3 sendmsg() - UDP Style Syntax
1437 * An application uses sendmsg() and recvmsg() calls to transmit data to
1438 * and receive data from its peer.
1440 * ssize_t sendmsg(int socket, const struct msghdr *message,
1441 * int flags);
1443 * socket - the socket descriptor of the endpoint.
1444 * message - pointer to the msghdr structure which contains a single
1445 * user message and possibly some ancillary data.
1447 * See Section 5 for complete description of the data
1448 * structures.
1450 * flags - flags sent or received with the user message, see Section
1451 * 5 for complete description of the flags.
1453 * Note: This function could use a rewrite especially when explicit
1454 * connect support comes in.
1456 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1458 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1460 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1461 struct msghdr *msg, size_t msg_len)
1463 struct sctp_sock *sp;
1464 struct sctp_endpoint *ep;
1465 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1466 struct sctp_transport *transport, *chunk_tp;
1467 struct sctp_chunk *chunk;
1468 union sctp_addr to;
1469 struct sockaddr *msg_name = NULL;
1470 struct sctp_sndrcvinfo default_sinfo = { 0 };
1471 struct sctp_sndrcvinfo *sinfo;
1472 struct sctp_initmsg *sinit;
1473 sctp_assoc_t associd = 0;
1474 sctp_cmsgs_t cmsgs = { NULL };
1475 int err;
1476 sctp_scope_t scope;
1477 long timeo;
1478 __u16 sinfo_flags = 0;
1479 struct sctp_datamsg *datamsg;
1480 int msg_flags = msg->msg_flags;
1482 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1483 sk, msg, msg_len);
1485 err = 0;
1486 sp = sctp_sk(sk);
1487 ep = sp->ep;
1489 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1491 /* We cannot send a message over a TCP-style listening socket. */
1492 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1493 err = -EPIPE;
1494 goto out_nounlock;
1497 /* Parse out the SCTP CMSGs. */
1498 err = sctp_msghdr_parse(msg, &cmsgs);
1500 if (err) {
1501 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1502 goto out_nounlock;
1505 /* Fetch the destination address for this packet. This
1506 * address only selects the association--it is not necessarily
1507 * the address we will send to.
1508 * For a peeled-off socket, msg_name is ignored.
1510 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1511 int msg_namelen = msg->msg_namelen;
1513 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1514 msg_namelen);
1515 if (err)
1516 return err;
1518 if (msg_namelen > sizeof(to))
1519 msg_namelen = sizeof(to);
1520 memcpy(&to, msg->msg_name, msg_namelen);
1521 msg_name = msg->msg_name;
1524 sinfo = cmsgs.info;
1525 sinit = cmsgs.init;
1527 /* Did the user specify SNDRCVINFO? */
1528 if (sinfo) {
1529 sinfo_flags = sinfo->sinfo_flags;
1530 associd = sinfo->sinfo_assoc_id;
1533 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1534 msg_len, sinfo_flags);
1536 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1537 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1538 err = -EINVAL;
1539 goto out_nounlock;
1542 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1543 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1544 * If SCTP_ABORT is set, the message length could be non zero with
1545 * the msg_iov set to the user abort reason.
1547 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1548 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1549 err = -EINVAL;
1550 goto out_nounlock;
1553 /* If SCTP_ADDR_OVER is set, there must be an address
1554 * specified in msg_name.
1556 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1557 err = -EINVAL;
1558 goto out_nounlock;
1561 transport = NULL;
1563 SCTP_DEBUG_PRINTK("About to look up association.\n");
1565 sctp_lock_sock(sk);
1567 /* If a msg_name has been specified, assume this is to be used. */
1568 if (msg_name) {
1569 /* Look for a matching association on the endpoint. */
1570 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1571 if (!asoc) {
1572 /* If we could not find a matching association on the
1573 * endpoint, make sure that it is not a TCP-style
1574 * socket that already has an association or there is
1575 * no peeled-off association on another socket.
1577 if ((sctp_style(sk, TCP) &&
1578 sctp_sstate(sk, ESTABLISHED)) ||
1579 sctp_endpoint_is_peeled_off(ep, &to)) {
1580 err = -EADDRNOTAVAIL;
1581 goto out_unlock;
1584 } else {
1585 asoc = sctp_id2assoc(sk, associd);
1586 if (!asoc) {
1587 err = -EPIPE;
1588 goto out_unlock;
1592 if (asoc) {
1593 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1595 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1596 * socket that has an association in CLOSED state. This can
1597 * happen when an accepted socket has an association that is
1598 * already CLOSED.
1600 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1601 err = -EPIPE;
1602 goto out_unlock;
1605 if (sinfo_flags & SCTP_EOF) {
1606 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1607 asoc);
1608 sctp_primitive_SHUTDOWN(asoc, NULL);
1609 err = 0;
1610 goto out_unlock;
1612 if (sinfo_flags & SCTP_ABORT) {
1614 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1615 if (!chunk) {
1616 err = -ENOMEM;
1617 goto out_unlock;
1620 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1621 sctp_primitive_ABORT(asoc, chunk);
1622 err = 0;
1623 goto out_unlock;
1627 /* Do we need to create the association? */
1628 if (!asoc) {
1629 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1631 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1632 err = -EINVAL;
1633 goto out_unlock;
1636 /* Check for invalid stream against the stream counts,
1637 * either the default or the user specified stream counts.
1639 if (sinfo) {
1640 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1641 /* Check against the defaults. */
1642 if (sinfo->sinfo_stream >=
1643 sp->initmsg.sinit_num_ostreams) {
1644 err = -EINVAL;
1645 goto out_unlock;
1647 } else {
1648 /* Check against the requested. */
1649 if (sinfo->sinfo_stream >=
1650 sinit->sinit_num_ostreams) {
1651 err = -EINVAL;
1652 goto out_unlock;
1658 * API 3.1.2 bind() - UDP Style Syntax
1659 * If a bind() or sctp_bindx() is not called prior to a
1660 * sendmsg() call that initiates a new association, the
1661 * system picks an ephemeral port and will choose an address
1662 * set equivalent to binding with a wildcard address.
1664 if (!ep->base.bind_addr.port) {
1665 if (sctp_autobind(sk)) {
1666 err = -EAGAIN;
1667 goto out_unlock;
1669 } else {
1671 * If an unprivileged user inherits a one-to-many
1672 * style socket with open associations on a privileged
1673 * port, it MAY be permitted to accept new associations,
1674 * but it SHOULD NOT be permitted to open new
1675 * associations.
1677 if (ep->base.bind_addr.port < PROT_SOCK &&
1678 !capable(CAP_NET_BIND_SERVICE)) {
1679 err = -EACCES;
1680 goto out_unlock;
1684 scope = sctp_scope(&to);
1685 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1686 if (!new_asoc) {
1687 err = -ENOMEM;
1688 goto out_unlock;
1690 asoc = new_asoc;
1692 /* If the SCTP_INIT ancillary data is specified, set all
1693 * the association init values accordingly.
1695 if (sinit) {
1696 if (sinit->sinit_num_ostreams) {
1697 asoc->c.sinit_num_ostreams =
1698 sinit->sinit_num_ostreams;
1700 if (sinit->sinit_max_instreams) {
1701 asoc->c.sinit_max_instreams =
1702 sinit->sinit_max_instreams;
1704 if (sinit->sinit_max_attempts) {
1705 asoc->max_init_attempts
1706 = sinit->sinit_max_attempts;
1708 if (sinit->sinit_max_init_timeo) {
1709 asoc->max_init_timeo =
1710 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1714 /* Prime the peer's transport structures. */
1715 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1716 if (!transport) {
1717 err = -ENOMEM;
1718 goto out_free;
1720 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1721 if (err < 0) {
1722 err = -ENOMEM;
1723 goto out_free;
1727 /* ASSERT: we have a valid association at this point. */
1728 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1730 if (!sinfo) {
1731 /* If the user didn't specify SNDRCVINFO, make up one with
1732 * some defaults.
1734 default_sinfo.sinfo_stream = asoc->default_stream;
1735 default_sinfo.sinfo_flags = asoc->default_flags;
1736 default_sinfo.sinfo_ppid = asoc->default_ppid;
1737 default_sinfo.sinfo_context = asoc->default_context;
1738 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1739 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1740 sinfo = &default_sinfo;
1743 /* API 7.1.7, the sndbuf size per association bounds the
1744 * maximum size of data that can be sent in a single send call.
1746 if (msg_len > sk->sk_sndbuf) {
1747 err = -EMSGSIZE;
1748 goto out_free;
1751 if (asoc->pmtu_pending)
1752 sctp_assoc_pending_pmtu(asoc);
1754 /* If fragmentation is disabled and the message length exceeds the
1755 * association fragmentation point, return EMSGSIZE. The I-D
1756 * does not specify what this error is, but this looks like
1757 * a great fit.
1759 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1760 err = -EMSGSIZE;
1761 goto out_free;
1764 if (sinfo) {
1765 /* Check for invalid stream. */
1766 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1767 err = -EINVAL;
1768 goto out_free;
1772 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1773 if (!sctp_wspace(asoc)) {
1774 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1775 if (err)
1776 goto out_free;
1779 /* If an address is passed with the sendto/sendmsg call, it is used
1780 * to override the primary destination address in the TCP model, or
1781 * when SCTP_ADDR_OVER flag is set in the UDP model.
1783 if ((sctp_style(sk, TCP) && msg_name) ||
1784 (sinfo_flags & SCTP_ADDR_OVER)) {
1785 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1786 if (!chunk_tp) {
1787 err = -EINVAL;
1788 goto out_free;
1790 } else
1791 chunk_tp = NULL;
1793 /* Auto-connect, if we aren't connected already. */
1794 if (sctp_state(asoc, CLOSED)) {
1795 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1796 if (err < 0)
1797 goto out_free;
1798 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1801 /* Break the message into multiple chunks of maximum size. */
1802 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1803 if (!datamsg) {
1804 err = -ENOMEM;
1805 goto out_free;
1808 /* Now send the (possibly) fragmented message. */
1809 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1810 sctp_chunk_hold(chunk);
1812 /* Do accounting for the write space. */
1813 sctp_set_owner_w(chunk);
1815 chunk->transport = chunk_tp;
1817 /* Send it to the lower layers. Note: all chunks
1818 * must either fail or succeed. The lower layer
1819 * works that way today. Keep it that way or this
1820 * breaks.
1822 err = sctp_primitive_SEND(asoc, chunk);
1823 /* Did the lower layer accept the chunk? */
1824 if (err)
1825 sctp_chunk_free(chunk);
1826 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1829 sctp_datamsg_put(datamsg);
1830 if (err)
1831 goto out_free;
1832 else
1833 err = msg_len;
1835 /* If we are already past ASSOCIATE, the lower
1836 * layers are responsible for association cleanup.
1838 goto out_unlock;
1840 out_free:
1841 if (new_asoc)
1842 sctp_association_free(asoc);
1843 out_unlock:
1844 sctp_release_sock(sk);
1846 out_nounlock:
1847 return sctp_error(sk, msg_flags, err);
1849 #if 0
1850 do_sock_err:
1851 if (msg_len)
1852 err = msg_len;
1853 else
1854 err = sock_error(sk);
1855 goto out;
1857 do_interrupted:
1858 if (msg_len)
1859 err = msg_len;
1860 goto out;
1861 #endif /* 0 */
1864 /* This is an extended version of skb_pull() that removes the data from the
1865 * start of a skb even when data is spread across the list of skb's in the
1866 * frag_list. len specifies the total amount of data that needs to be removed.
1867 * when 'len' bytes could be removed from the skb, it returns 0.
1868 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1869 * could not be removed.
1871 static int sctp_skb_pull(struct sk_buff *skb, int len)
1873 struct sk_buff *list;
1874 int skb_len = skb_headlen(skb);
1875 int rlen;
1877 if (len <= skb_len) {
1878 __skb_pull(skb, len);
1879 return 0;
1881 len -= skb_len;
1882 __skb_pull(skb, skb_len);
1884 skb_walk_frags(skb, list) {
1885 rlen = sctp_skb_pull(list, len);
1886 skb->len -= (len-rlen);
1887 skb->data_len -= (len-rlen);
1889 if (!rlen)
1890 return 0;
1892 len = rlen;
1895 return len;
1898 /* API 3.1.3 recvmsg() - UDP Style Syntax
1900 * ssize_t recvmsg(int socket, struct msghdr *message,
1901 * int flags);
1903 * socket - the socket descriptor of the endpoint.
1904 * message - pointer to the msghdr structure which contains a single
1905 * user message and possibly some ancillary data.
1907 * See Section 5 for complete description of the data
1908 * structures.
1910 * flags - flags sent or received with the user message, see Section
1911 * 5 for complete description of the flags.
1913 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1915 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1916 struct msghdr *msg, size_t len, int noblock,
1917 int flags, int *addr_len)
1919 struct sctp_ulpevent *event = NULL;
1920 struct sctp_sock *sp = sctp_sk(sk);
1921 struct sk_buff *skb;
1922 int copied;
1923 int err = 0;
1924 int skb_len;
1926 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1927 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1928 "len", len, "knoblauch", noblock,
1929 "flags", flags, "addr_len", addr_len);
1931 sctp_lock_sock(sk);
1933 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1934 err = -ENOTCONN;
1935 goto out;
1938 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1939 if (!skb)
1940 goto out;
1942 /* Get the total length of the skb including any skb's in the
1943 * frag_list.
1945 skb_len = skb->len;
1947 copied = skb_len;
1948 if (copied > len)
1949 copied = len;
1951 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1953 event = sctp_skb2event(skb);
1955 if (err)
1956 goto out_free;
1958 sock_recv_timestamp(msg, sk, skb);
1959 if (sctp_ulpevent_is_notification(event)) {
1960 msg->msg_flags |= MSG_NOTIFICATION;
1961 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1962 } else {
1963 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1966 /* Check if we allow SCTP_SNDRCVINFO. */
1967 if (sp->subscribe.sctp_data_io_event)
1968 sctp_ulpevent_read_sndrcvinfo(event, msg);
1969 #if 0
1970 /* FIXME: we should be calling IP/IPv6 layers. */
1971 if (sk->sk_protinfo.af_inet.cmsg_flags)
1972 ip_cmsg_recv(msg, skb);
1973 #endif
1975 err = copied;
1977 /* If skb's length exceeds the user's buffer, update the skb and
1978 * push it back to the receive_queue so that the next call to
1979 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1981 if (skb_len > copied) {
1982 msg->msg_flags &= ~MSG_EOR;
1983 if (flags & MSG_PEEK)
1984 goto out_free;
1985 sctp_skb_pull(skb, copied);
1986 skb_queue_head(&sk->sk_receive_queue, skb);
1988 /* When only partial message is copied to the user, increase
1989 * rwnd by that amount. If all the data in the skb is read,
1990 * rwnd is updated when the event is freed.
1992 if (!sctp_ulpevent_is_notification(event))
1993 sctp_assoc_rwnd_increase(event->asoc, copied);
1994 goto out;
1995 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1996 (event->msg_flags & MSG_EOR))
1997 msg->msg_flags |= MSG_EOR;
1998 else
1999 msg->msg_flags &= ~MSG_EOR;
2001 out_free:
2002 if (flags & MSG_PEEK) {
2003 /* Release the skb reference acquired after peeking the skb in
2004 * sctp_skb_recv_datagram().
2006 kfree_skb(skb);
2007 } else {
2008 /* Free the event which includes releasing the reference to
2009 * the owner of the skb, freeing the skb and updating the
2010 * rwnd.
2012 sctp_ulpevent_free(event);
2014 out:
2015 sctp_release_sock(sk);
2016 return err;
2019 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2021 * This option is a on/off flag. If enabled no SCTP message
2022 * fragmentation will be performed. Instead if a message being sent
2023 * exceeds the current PMTU size, the message will NOT be sent and
2024 * instead a error will be indicated to the user.
2026 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2027 char __user *optval, int optlen)
2029 int val;
2031 if (optlen < sizeof(int))
2032 return -EINVAL;
2034 if (get_user(val, (int __user *)optval))
2035 return -EFAULT;
2037 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2039 return 0;
2042 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2043 int optlen)
2045 if (optlen > sizeof(struct sctp_event_subscribe))
2046 return -EINVAL;
2047 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2048 return -EFAULT;
2049 return 0;
2052 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2054 * This socket option is applicable to the UDP-style socket only. When
2055 * set it will cause associations that are idle for more than the
2056 * specified number of seconds to automatically close. An association
2057 * being idle is defined an association that has NOT sent or received
2058 * user data. The special value of '0' indicates that no automatic
2059 * close of any associations should be performed. The option expects an
2060 * integer defining the number of seconds of idle time before an
2061 * association is closed.
2063 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2064 int optlen)
2066 struct sctp_sock *sp = sctp_sk(sk);
2068 /* Applicable to UDP-style socket only */
2069 if (sctp_style(sk, TCP))
2070 return -EOPNOTSUPP;
2071 if (optlen != sizeof(int))
2072 return -EINVAL;
2073 if (copy_from_user(&sp->autoclose, optval, optlen))
2074 return -EFAULT;
2076 return 0;
2079 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2081 * Applications can enable or disable heartbeats for any peer address of
2082 * an association, modify an address's heartbeat interval, force a
2083 * heartbeat to be sent immediately, and adjust the address's maximum
2084 * number of retransmissions sent before an address is considered
2085 * unreachable. The following structure is used to access and modify an
2086 * address's parameters:
2088 * struct sctp_paddrparams {
2089 * sctp_assoc_t spp_assoc_id;
2090 * struct sockaddr_storage spp_address;
2091 * uint32_t spp_hbinterval;
2092 * uint16_t spp_pathmaxrxt;
2093 * uint32_t spp_pathmtu;
2094 * uint32_t spp_sackdelay;
2095 * uint32_t spp_flags;
2096 * };
2098 * spp_assoc_id - (one-to-many style socket) This is filled in the
2099 * application, and identifies the association for
2100 * this query.
2101 * spp_address - This specifies which address is of interest.
2102 * spp_hbinterval - This contains the value of the heartbeat interval,
2103 * in milliseconds. If a value of zero
2104 * is present in this field then no changes are to
2105 * be made to this parameter.
2106 * spp_pathmaxrxt - This contains the maximum number of
2107 * retransmissions before this address shall be
2108 * considered unreachable. If a value of zero
2109 * is present in this field then no changes are to
2110 * be made to this parameter.
2111 * spp_pathmtu - When Path MTU discovery is disabled the value
2112 * specified here will be the "fixed" path mtu.
2113 * Note that if the spp_address field is empty
2114 * then all associations on this address will
2115 * have this fixed path mtu set upon them.
2117 * spp_sackdelay - When delayed sack is enabled, this value specifies
2118 * the number of milliseconds that sacks will be delayed
2119 * for. This value will apply to all addresses of an
2120 * association if the spp_address field is empty. Note
2121 * also, that if delayed sack is enabled and this
2122 * value is set to 0, no change is made to the last
2123 * recorded delayed sack timer value.
2125 * spp_flags - These flags are used to control various features
2126 * on an association. The flag field may contain
2127 * zero or more of the following options.
2129 * SPP_HB_ENABLE - Enable heartbeats on the
2130 * specified address. Note that if the address
2131 * field is empty all addresses for the association
2132 * have heartbeats enabled upon them.
2134 * SPP_HB_DISABLE - Disable heartbeats on the
2135 * speicifed address. Note that if the address
2136 * field is empty all addresses for the association
2137 * will have their heartbeats disabled. Note also
2138 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2139 * mutually exclusive, only one of these two should
2140 * be specified. Enabling both fields will have
2141 * undetermined results.
2143 * SPP_HB_DEMAND - Request a user initiated heartbeat
2144 * to be made immediately.
2146 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2147 * heartbeat delayis to be set to the value of 0
2148 * milliseconds.
2150 * SPP_PMTUD_ENABLE - This field will enable PMTU
2151 * discovery upon the specified address. Note that
2152 * if the address feild is empty then all addresses
2153 * on the association are effected.
2155 * SPP_PMTUD_DISABLE - This field will disable PMTU
2156 * discovery upon the specified address. Note that
2157 * if the address feild is empty then all addresses
2158 * on the association are effected. Not also that
2159 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2160 * exclusive. Enabling both will have undetermined
2161 * results.
2163 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2164 * on delayed sack. The time specified in spp_sackdelay
2165 * is used to specify the sack delay for this address. Note
2166 * that if spp_address is empty then all addresses will
2167 * enable delayed sack and take on the sack delay
2168 * value specified in spp_sackdelay.
2169 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2170 * off delayed sack. If the spp_address field is blank then
2171 * delayed sack is disabled for the entire association. Note
2172 * also that this field is mutually exclusive to
2173 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2174 * results.
2176 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2177 struct sctp_transport *trans,
2178 struct sctp_association *asoc,
2179 struct sctp_sock *sp,
2180 int hb_change,
2181 int pmtud_change,
2182 int sackdelay_change)
2184 int error;
2186 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2187 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2188 if (error)
2189 return error;
2192 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2193 * this field is ignored. Note also that a value of zero indicates
2194 * the current setting should be left unchanged.
2196 if (params->spp_flags & SPP_HB_ENABLE) {
2198 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2199 * set. This lets us use 0 value when this flag
2200 * is set.
2202 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2203 params->spp_hbinterval = 0;
2205 if (params->spp_hbinterval ||
2206 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2207 if (trans) {
2208 trans->hbinterval =
2209 msecs_to_jiffies(params->spp_hbinterval);
2210 } else if (asoc) {
2211 asoc->hbinterval =
2212 msecs_to_jiffies(params->spp_hbinterval);
2213 } else {
2214 sp->hbinterval = params->spp_hbinterval;
2219 if (hb_change) {
2220 if (trans) {
2221 trans->param_flags =
2222 (trans->param_flags & ~SPP_HB) | hb_change;
2223 } else if (asoc) {
2224 asoc->param_flags =
2225 (asoc->param_flags & ~SPP_HB) | hb_change;
2226 } else {
2227 sp->param_flags =
2228 (sp->param_flags & ~SPP_HB) | hb_change;
2232 /* When Path MTU discovery is disabled the value specified here will
2233 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2234 * include the flag SPP_PMTUD_DISABLE for this field to have any
2235 * effect).
2237 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2238 if (trans) {
2239 trans->pathmtu = params->spp_pathmtu;
2240 sctp_assoc_sync_pmtu(asoc);
2241 } else if (asoc) {
2242 asoc->pathmtu = params->spp_pathmtu;
2243 sctp_frag_point(sp, params->spp_pathmtu);
2244 } else {
2245 sp->pathmtu = params->spp_pathmtu;
2249 if (pmtud_change) {
2250 if (trans) {
2251 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2252 (params->spp_flags & SPP_PMTUD_ENABLE);
2253 trans->param_flags =
2254 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2255 if (update) {
2256 sctp_transport_pmtu(trans);
2257 sctp_assoc_sync_pmtu(asoc);
2259 } else if (asoc) {
2260 asoc->param_flags =
2261 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2262 } else {
2263 sp->param_flags =
2264 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2268 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2269 * value of this field is ignored. Note also that a value of zero
2270 * indicates the current setting should be left unchanged.
2272 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2273 if (trans) {
2274 trans->sackdelay =
2275 msecs_to_jiffies(params->spp_sackdelay);
2276 } else if (asoc) {
2277 asoc->sackdelay =
2278 msecs_to_jiffies(params->spp_sackdelay);
2279 } else {
2280 sp->sackdelay = params->spp_sackdelay;
2284 if (sackdelay_change) {
2285 if (trans) {
2286 trans->param_flags =
2287 (trans->param_flags & ~SPP_SACKDELAY) |
2288 sackdelay_change;
2289 } else if (asoc) {
2290 asoc->param_flags =
2291 (asoc->param_flags & ~SPP_SACKDELAY) |
2292 sackdelay_change;
2293 } else {
2294 sp->param_flags =
2295 (sp->param_flags & ~SPP_SACKDELAY) |
2296 sackdelay_change;
2300 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2301 * of this field is ignored. Note also that a value of zero
2302 * indicates the current setting should be left unchanged.
2304 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2305 if (trans) {
2306 trans->pathmaxrxt = params->spp_pathmaxrxt;
2307 } else if (asoc) {
2308 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2309 } else {
2310 sp->pathmaxrxt = params->spp_pathmaxrxt;
2314 return 0;
2317 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2318 char __user *optval, int optlen)
2320 struct sctp_paddrparams params;
2321 struct sctp_transport *trans = NULL;
2322 struct sctp_association *asoc = NULL;
2323 struct sctp_sock *sp = sctp_sk(sk);
2324 int error;
2325 int hb_change, pmtud_change, sackdelay_change;
2327 if (optlen != sizeof(struct sctp_paddrparams))
2328 return - EINVAL;
2330 if (copy_from_user(&params, optval, optlen))
2331 return -EFAULT;
2333 /* Validate flags and value parameters. */
2334 hb_change = params.spp_flags & SPP_HB;
2335 pmtud_change = params.spp_flags & SPP_PMTUD;
2336 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2338 if (hb_change == SPP_HB ||
2339 pmtud_change == SPP_PMTUD ||
2340 sackdelay_change == SPP_SACKDELAY ||
2341 params.spp_sackdelay > 500 ||
2342 (params.spp_pathmtu
2343 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2344 return -EINVAL;
2346 /* If an address other than INADDR_ANY is specified, and
2347 * no transport is found, then the request is invalid.
2349 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2350 trans = sctp_addr_id2transport(sk, &params.spp_address,
2351 params.spp_assoc_id);
2352 if (!trans)
2353 return -EINVAL;
2356 /* Get association, if assoc_id != 0 and the socket is a one
2357 * to many style socket, and an association was not found, then
2358 * the id was invalid.
2360 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2361 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2362 return -EINVAL;
2364 /* Heartbeat demand can only be sent on a transport or
2365 * association, but not a socket.
2367 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2368 return -EINVAL;
2370 /* Process parameters. */
2371 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2372 hb_change, pmtud_change,
2373 sackdelay_change);
2375 if (error)
2376 return error;
2378 /* If changes are for association, also apply parameters to each
2379 * transport.
2381 if (!trans && asoc) {
2382 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2383 transports) {
2384 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2385 hb_change, pmtud_change,
2386 sackdelay_change);
2390 return 0;
2394 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2396 * This option will effect the way delayed acks are performed. This
2397 * option allows you to get or set the delayed ack time, in
2398 * milliseconds. It also allows changing the delayed ack frequency.
2399 * Changing the frequency to 1 disables the delayed sack algorithm. If
2400 * the assoc_id is 0, then this sets or gets the endpoints default
2401 * values. If the assoc_id field is non-zero, then the set or get
2402 * effects the specified association for the one to many model (the
2403 * assoc_id field is ignored by the one to one model). Note that if
2404 * sack_delay or sack_freq are 0 when setting this option, then the
2405 * current values will remain unchanged.
2407 * struct sctp_sack_info {
2408 * sctp_assoc_t sack_assoc_id;
2409 * uint32_t sack_delay;
2410 * uint32_t sack_freq;
2411 * };
2413 * sack_assoc_id - This parameter, indicates which association the user
2414 * is performing an action upon. Note that if this field's value is
2415 * zero then the endpoints default value is changed (effecting future
2416 * associations only).
2418 * sack_delay - This parameter contains the number of milliseconds that
2419 * the user is requesting the delayed ACK timer be set to. Note that
2420 * this value is defined in the standard to be between 200 and 500
2421 * milliseconds.
2423 * sack_freq - This parameter contains the number of packets that must
2424 * be received before a sack is sent without waiting for the delay
2425 * timer to expire. The default value for this is 2, setting this
2426 * value to 1 will disable the delayed sack algorithm.
2429 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2430 char __user *optval, int optlen)
2432 struct sctp_sack_info params;
2433 struct sctp_transport *trans = NULL;
2434 struct sctp_association *asoc = NULL;
2435 struct sctp_sock *sp = sctp_sk(sk);
2437 if (optlen == sizeof(struct sctp_sack_info)) {
2438 if (copy_from_user(&params, optval, optlen))
2439 return -EFAULT;
2441 if (params.sack_delay == 0 && params.sack_freq == 0)
2442 return 0;
2443 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2444 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2445 "in delayed_ack socket option deprecated\n");
2446 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2447 if (copy_from_user(&params, optval, optlen))
2448 return -EFAULT;
2450 if (params.sack_delay == 0)
2451 params.sack_freq = 1;
2452 else
2453 params.sack_freq = 0;
2454 } else
2455 return - EINVAL;
2457 /* Validate value parameter. */
2458 if (params.sack_delay > 500)
2459 return -EINVAL;
2461 /* Get association, if sack_assoc_id != 0 and the socket is a one
2462 * to many style socket, and an association was not found, then
2463 * the id was invalid.
2465 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2466 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2467 return -EINVAL;
2469 if (params.sack_delay) {
2470 if (asoc) {
2471 asoc->sackdelay =
2472 msecs_to_jiffies(params.sack_delay);
2473 asoc->param_flags =
2474 (asoc->param_flags & ~SPP_SACKDELAY) |
2475 SPP_SACKDELAY_ENABLE;
2476 } else {
2477 sp->sackdelay = params.sack_delay;
2478 sp->param_flags =
2479 (sp->param_flags & ~SPP_SACKDELAY) |
2480 SPP_SACKDELAY_ENABLE;
2484 if (params.sack_freq == 1) {
2485 if (asoc) {
2486 asoc->param_flags =
2487 (asoc->param_flags & ~SPP_SACKDELAY) |
2488 SPP_SACKDELAY_DISABLE;
2489 } else {
2490 sp->param_flags =
2491 (sp->param_flags & ~SPP_SACKDELAY) |
2492 SPP_SACKDELAY_DISABLE;
2494 } else if (params.sack_freq > 1) {
2495 if (asoc) {
2496 asoc->sackfreq = params.sack_freq;
2497 asoc->param_flags =
2498 (asoc->param_flags & ~SPP_SACKDELAY) |
2499 SPP_SACKDELAY_ENABLE;
2500 } else {
2501 sp->sackfreq = params.sack_freq;
2502 sp->param_flags =
2503 (sp->param_flags & ~SPP_SACKDELAY) |
2504 SPP_SACKDELAY_ENABLE;
2508 /* If change is for association, also apply to each transport. */
2509 if (asoc) {
2510 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2511 transports) {
2512 if (params.sack_delay) {
2513 trans->sackdelay =
2514 msecs_to_jiffies(params.sack_delay);
2515 trans->param_flags =
2516 (trans->param_flags & ~SPP_SACKDELAY) |
2517 SPP_SACKDELAY_ENABLE;
2519 if (params.sack_freq == 1) {
2520 trans->param_flags =
2521 (trans->param_flags & ~SPP_SACKDELAY) |
2522 SPP_SACKDELAY_DISABLE;
2523 } else if (params.sack_freq > 1) {
2524 trans->sackfreq = params.sack_freq;
2525 trans->param_flags =
2526 (trans->param_flags & ~SPP_SACKDELAY) |
2527 SPP_SACKDELAY_ENABLE;
2532 return 0;
2535 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2537 * Applications can specify protocol parameters for the default association
2538 * initialization. The option name argument to setsockopt() and getsockopt()
2539 * is SCTP_INITMSG.
2541 * Setting initialization parameters is effective only on an unconnected
2542 * socket (for UDP-style sockets only future associations are effected
2543 * by the change). With TCP-style sockets, this option is inherited by
2544 * sockets derived from a listener socket.
2546 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2548 struct sctp_initmsg sinit;
2549 struct sctp_sock *sp = sctp_sk(sk);
2551 if (optlen != sizeof(struct sctp_initmsg))
2552 return -EINVAL;
2553 if (copy_from_user(&sinit, optval, optlen))
2554 return -EFAULT;
2556 if (sinit.sinit_num_ostreams)
2557 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2558 if (sinit.sinit_max_instreams)
2559 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2560 if (sinit.sinit_max_attempts)
2561 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2562 if (sinit.sinit_max_init_timeo)
2563 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2565 return 0;
2569 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2571 * Applications that wish to use the sendto() system call may wish to
2572 * specify a default set of parameters that would normally be supplied
2573 * through the inclusion of ancillary data. This socket option allows
2574 * such an application to set the default sctp_sndrcvinfo structure.
2575 * The application that wishes to use this socket option simply passes
2576 * in to this call the sctp_sndrcvinfo structure defined in Section
2577 * 5.2.2) The input parameters accepted by this call include
2578 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2579 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2580 * to this call if the caller is using the UDP model.
2582 static int sctp_setsockopt_default_send_param(struct sock *sk,
2583 char __user *optval, int optlen)
2585 struct sctp_sndrcvinfo info;
2586 struct sctp_association *asoc;
2587 struct sctp_sock *sp = sctp_sk(sk);
2589 if (optlen != sizeof(struct sctp_sndrcvinfo))
2590 return -EINVAL;
2591 if (copy_from_user(&info, optval, optlen))
2592 return -EFAULT;
2594 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2595 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2596 return -EINVAL;
2598 if (asoc) {
2599 asoc->default_stream = info.sinfo_stream;
2600 asoc->default_flags = info.sinfo_flags;
2601 asoc->default_ppid = info.sinfo_ppid;
2602 asoc->default_context = info.sinfo_context;
2603 asoc->default_timetolive = info.sinfo_timetolive;
2604 } else {
2605 sp->default_stream = info.sinfo_stream;
2606 sp->default_flags = info.sinfo_flags;
2607 sp->default_ppid = info.sinfo_ppid;
2608 sp->default_context = info.sinfo_context;
2609 sp->default_timetolive = info.sinfo_timetolive;
2612 return 0;
2615 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2617 * Requests that the local SCTP stack use the enclosed peer address as
2618 * the association primary. The enclosed address must be one of the
2619 * association peer's addresses.
2621 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2622 int optlen)
2624 struct sctp_prim prim;
2625 struct sctp_transport *trans;
2627 if (optlen != sizeof(struct sctp_prim))
2628 return -EINVAL;
2630 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2631 return -EFAULT;
2633 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2634 if (!trans)
2635 return -EINVAL;
2637 sctp_assoc_set_primary(trans->asoc, trans);
2639 return 0;
2643 * 7.1.5 SCTP_NODELAY
2645 * Turn on/off any Nagle-like algorithm. This means that packets are
2646 * generally sent as soon as possible and no unnecessary delays are
2647 * introduced, at the cost of more packets in the network. Expects an
2648 * integer boolean flag.
2650 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2651 int optlen)
2653 int val;
2655 if (optlen < sizeof(int))
2656 return -EINVAL;
2657 if (get_user(val, (int __user *)optval))
2658 return -EFAULT;
2660 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2661 return 0;
2666 * 7.1.1 SCTP_RTOINFO
2668 * The protocol parameters used to initialize and bound retransmission
2669 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2670 * and modify these parameters.
2671 * All parameters are time values, in milliseconds. A value of 0, when
2672 * modifying the parameters, indicates that the current value should not
2673 * be changed.
2676 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2677 struct sctp_rtoinfo rtoinfo;
2678 struct sctp_association *asoc;
2680 if (optlen != sizeof (struct sctp_rtoinfo))
2681 return -EINVAL;
2683 if (copy_from_user(&rtoinfo, optval, optlen))
2684 return -EFAULT;
2686 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2688 /* Set the values to the specific association */
2689 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2690 return -EINVAL;
2692 if (asoc) {
2693 if (rtoinfo.srto_initial != 0)
2694 asoc->rto_initial =
2695 msecs_to_jiffies(rtoinfo.srto_initial);
2696 if (rtoinfo.srto_max != 0)
2697 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2698 if (rtoinfo.srto_min != 0)
2699 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2700 } else {
2701 /* If there is no association or the association-id = 0
2702 * set the values to the endpoint.
2704 struct sctp_sock *sp = sctp_sk(sk);
2706 if (rtoinfo.srto_initial != 0)
2707 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2708 if (rtoinfo.srto_max != 0)
2709 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2710 if (rtoinfo.srto_min != 0)
2711 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2714 return 0;
2719 * 7.1.2 SCTP_ASSOCINFO
2721 * This option is used to tune the maximum retransmission attempts
2722 * of the association.
2723 * Returns an error if the new association retransmission value is
2724 * greater than the sum of the retransmission value of the peer.
2725 * See [SCTP] for more information.
2728 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2731 struct sctp_assocparams assocparams;
2732 struct sctp_association *asoc;
2734 if (optlen != sizeof(struct sctp_assocparams))
2735 return -EINVAL;
2736 if (copy_from_user(&assocparams, optval, optlen))
2737 return -EFAULT;
2739 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2741 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2742 return -EINVAL;
2744 /* Set the values to the specific association */
2745 if (asoc) {
2746 if (assocparams.sasoc_asocmaxrxt != 0) {
2747 __u32 path_sum = 0;
2748 int paths = 0;
2749 struct sctp_transport *peer_addr;
2751 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2752 transports) {
2753 path_sum += peer_addr->pathmaxrxt;
2754 paths++;
2757 /* Only validate asocmaxrxt if we have more than
2758 * one path/transport. We do this because path
2759 * retransmissions are only counted when we have more
2760 * then one path.
2762 if (paths > 1 &&
2763 assocparams.sasoc_asocmaxrxt > path_sum)
2764 return -EINVAL;
2766 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2769 if (assocparams.sasoc_cookie_life != 0) {
2770 asoc->cookie_life.tv_sec =
2771 assocparams.sasoc_cookie_life / 1000;
2772 asoc->cookie_life.tv_usec =
2773 (assocparams.sasoc_cookie_life % 1000)
2774 * 1000;
2776 } else {
2777 /* Set the values to the endpoint */
2778 struct sctp_sock *sp = sctp_sk(sk);
2780 if (assocparams.sasoc_asocmaxrxt != 0)
2781 sp->assocparams.sasoc_asocmaxrxt =
2782 assocparams.sasoc_asocmaxrxt;
2783 if (assocparams.sasoc_cookie_life != 0)
2784 sp->assocparams.sasoc_cookie_life =
2785 assocparams.sasoc_cookie_life;
2787 return 0;
2791 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2793 * This socket option is a boolean flag which turns on or off mapped V4
2794 * addresses. If this option is turned on and the socket is type
2795 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2796 * If this option is turned off, then no mapping will be done of V4
2797 * addresses and a user will receive both PF_INET6 and PF_INET type
2798 * addresses on the socket.
2800 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2802 int val;
2803 struct sctp_sock *sp = sctp_sk(sk);
2805 if (optlen < sizeof(int))
2806 return -EINVAL;
2807 if (get_user(val, (int __user *)optval))
2808 return -EFAULT;
2809 if (val)
2810 sp->v4mapped = 1;
2811 else
2812 sp->v4mapped = 0;
2814 return 0;
2818 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2819 * This option will get or set the maximum size to put in any outgoing
2820 * SCTP DATA chunk. If a message is larger than this size it will be
2821 * fragmented by SCTP into the specified size. Note that the underlying
2822 * SCTP implementation may fragment into smaller sized chunks when the
2823 * PMTU of the underlying association is smaller than the value set by
2824 * the user. The default value for this option is '0' which indicates
2825 * the user is NOT limiting fragmentation and only the PMTU will effect
2826 * SCTP's choice of DATA chunk size. Note also that values set larger
2827 * than the maximum size of an IP datagram will effectively let SCTP
2828 * control fragmentation (i.e. the same as setting this option to 0).
2830 * The following structure is used to access and modify this parameter:
2832 * struct sctp_assoc_value {
2833 * sctp_assoc_t assoc_id;
2834 * uint32_t assoc_value;
2835 * };
2837 * assoc_id: This parameter is ignored for one-to-one style sockets.
2838 * For one-to-many style sockets this parameter indicates which
2839 * association the user is performing an action upon. Note that if
2840 * this field's value is zero then the endpoints default value is
2841 * changed (effecting future associations only).
2842 * assoc_value: This parameter specifies the maximum size in bytes.
2844 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2846 struct sctp_assoc_value params;
2847 struct sctp_association *asoc;
2848 struct sctp_sock *sp = sctp_sk(sk);
2849 int val;
2851 if (optlen == sizeof(int)) {
2852 printk(KERN_WARNING
2853 "SCTP: Use of int in maxseg socket option deprecated\n");
2854 printk(KERN_WARNING
2855 "SCTP: Use struct sctp_assoc_value instead\n");
2856 if (copy_from_user(&val, optval, optlen))
2857 return -EFAULT;
2858 params.assoc_id = 0;
2859 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2860 if (copy_from_user(&params, optval, optlen))
2861 return -EFAULT;
2862 val = params.assoc_value;
2863 } else
2864 return -EINVAL;
2866 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2867 return -EINVAL;
2869 asoc = sctp_id2assoc(sk, params.assoc_id);
2870 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2871 return -EINVAL;
2873 if (asoc) {
2874 if (val == 0) {
2875 val = asoc->pathmtu;
2876 val -= sp->pf->af->net_header_len;
2877 val -= sizeof(struct sctphdr) +
2878 sizeof(struct sctp_data_chunk);
2881 asoc->frag_point = val;
2882 } else {
2883 sp->user_frag = val;
2885 /* Update the frag_point of the existing associations. */
2886 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2887 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2891 return 0;
2896 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2898 * Requests that the peer mark the enclosed address as the association
2899 * primary. The enclosed address must be one of the association's
2900 * locally bound addresses. The following structure is used to make a
2901 * set primary request:
2903 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2904 int optlen)
2906 struct sctp_sock *sp;
2907 struct sctp_endpoint *ep;
2908 struct sctp_association *asoc = NULL;
2909 struct sctp_setpeerprim prim;
2910 struct sctp_chunk *chunk;
2911 int err;
2913 sp = sctp_sk(sk);
2914 ep = sp->ep;
2916 if (!sctp_addip_enable)
2917 return -EPERM;
2919 if (optlen != sizeof(struct sctp_setpeerprim))
2920 return -EINVAL;
2922 if (copy_from_user(&prim, optval, optlen))
2923 return -EFAULT;
2925 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2926 if (!asoc)
2927 return -EINVAL;
2929 if (!asoc->peer.asconf_capable)
2930 return -EPERM;
2932 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2933 return -EPERM;
2935 if (!sctp_state(asoc, ESTABLISHED))
2936 return -ENOTCONN;
2938 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2939 return -EADDRNOTAVAIL;
2941 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2942 chunk = sctp_make_asconf_set_prim(asoc,
2943 (union sctp_addr *)&prim.sspp_addr);
2944 if (!chunk)
2945 return -ENOMEM;
2947 err = sctp_send_asconf(asoc, chunk);
2949 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2951 return err;
2954 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2955 int optlen)
2957 struct sctp_setadaptation adaptation;
2959 if (optlen != sizeof(struct sctp_setadaptation))
2960 return -EINVAL;
2961 if (copy_from_user(&adaptation, optval, optlen))
2962 return -EFAULT;
2964 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2966 return 0;
2970 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2972 * The context field in the sctp_sndrcvinfo structure is normally only
2973 * used when a failed message is retrieved holding the value that was
2974 * sent down on the actual send call. This option allows the setting of
2975 * a default context on an association basis that will be received on
2976 * reading messages from the peer. This is especially helpful in the
2977 * one-2-many model for an application to keep some reference to an
2978 * internal state machine that is processing messages on the
2979 * association. Note that the setting of this value only effects
2980 * received messages from the peer and does not effect the value that is
2981 * saved with outbound messages.
2983 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2984 int optlen)
2986 struct sctp_assoc_value params;
2987 struct sctp_sock *sp;
2988 struct sctp_association *asoc;
2990 if (optlen != sizeof(struct sctp_assoc_value))
2991 return -EINVAL;
2992 if (copy_from_user(&params, optval, optlen))
2993 return -EFAULT;
2995 sp = sctp_sk(sk);
2997 if (params.assoc_id != 0) {
2998 asoc = sctp_id2assoc(sk, params.assoc_id);
2999 if (!asoc)
3000 return -EINVAL;
3001 asoc->default_rcv_context = params.assoc_value;
3002 } else {
3003 sp->default_rcv_context = params.assoc_value;
3006 return 0;
3010 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3012 * This options will at a minimum specify if the implementation is doing
3013 * fragmented interleave. Fragmented interleave, for a one to many
3014 * socket, is when subsequent calls to receive a message may return
3015 * parts of messages from different associations. Some implementations
3016 * may allow you to turn this value on or off. If so, when turned off,
3017 * no fragment interleave will occur (which will cause a head of line
3018 * blocking amongst multiple associations sharing the same one to many
3019 * socket). When this option is turned on, then each receive call may
3020 * come from a different association (thus the user must receive data
3021 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3022 * association each receive belongs to.
3024 * This option takes a boolean value. A non-zero value indicates that
3025 * fragmented interleave is on. A value of zero indicates that
3026 * fragmented interleave is off.
3028 * Note that it is important that an implementation that allows this
3029 * option to be turned on, have it off by default. Otherwise an unaware
3030 * application using the one to many model may become confused and act
3031 * incorrectly.
3033 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3034 char __user *optval,
3035 int optlen)
3037 int val;
3039 if (optlen != sizeof(int))
3040 return -EINVAL;
3041 if (get_user(val, (int __user *)optval))
3042 return -EFAULT;
3044 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3046 return 0;
3050 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3051 * (SCTP_PARTIAL_DELIVERY_POINT)
3053 * This option will set or get the SCTP partial delivery point. This
3054 * point is the size of a message where the partial delivery API will be
3055 * invoked to help free up rwnd space for the peer. Setting this to a
3056 * lower value will cause partial deliveries to happen more often. The
3057 * calls argument is an integer that sets or gets the partial delivery
3058 * point. Note also that the call will fail if the user attempts to set
3059 * this value larger than the socket receive buffer size.
3061 * Note that any single message having a length smaller than or equal to
3062 * the SCTP partial delivery point will be delivered in one single read
3063 * call as long as the user provided buffer is large enough to hold the
3064 * message.
3066 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3067 char __user *optval,
3068 int optlen)
3070 u32 val;
3072 if (optlen != sizeof(u32))
3073 return -EINVAL;
3074 if (get_user(val, (int __user *)optval))
3075 return -EFAULT;
3077 /* Note: We double the receive buffer from what the user sets
3078 * it to be, also initial rwnd is based on rcvbuf/2.
3080 if (val > (sk->sk_rcvbuf >> 1))
3081 return -EINVAL;
3083 sctp_sk(sk)->pd_point = val;
3085 return 0; /* is this the right error code? */
3089 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3091 * This option will allow a user to change the maximum burst of packets
3092 * that can be emitted by this association. Note that the default value
3093 * is 4, and some implementations may restrict this setting so that it
3094 * can only be lowered.
3096 * NOTE: This text doesn't seem right. Do this on a socket basis with
3097 * future associations inheriting the socket value.
3099 static int sctp_setsockopt_maxburst(struct sock *sk,
3100 char __user *optval,
3101 int optlen)
3103 struct sctp_assoc_value params;
3104 struct sctp_sock *sp;
3105 struct sctp_association *asoc;
3106 int val;
3107 int assoc_id = 0;
3109 if (optlen == sizeof(int)) {
3110 printk(KERN_WARNING
3111 "SCTP: Use of int in max_burst socket option deprecated\n");
3112 printk(KERN_WARNING
3113 "SCTP: Use struct sctp_assoc_value instead\n");
3114 if (copy_from_user(&val, optval, optlen))
3115 return -EFAULT;
3116 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3117 if (copy_from_user(&params, optval, optlen))
3118 return -EFAULT;
3119 val = params.assoc_value;
3120 assoc_id = params.assoc_id;
3121 } else
3122 return -EINVAL;
3124 sp = sctp_sk(sk);
3126 if (assoc_id != 0) {
3127 asoc = sctp_id2assoc(sk, assoc_id);
3128 if (!asoc)
3129 return -EINVAL;
3130 asoc->max_burst = val;
3131 } else
3132 sp->max_burst = val;
3134 return 0;
3138 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3140 * This set option adds a chunk type that the user is requesting to be
3141 * received only in an authenticated way. Changes to the list of chunks
3142 * will only effect future associations on the socket.
3144 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3145 char __user *optval,
3146 int optlen)
3148 struct sctp_authchunk val;
3150 if (!sctp_auth_enable)
3151 return -EACCES;
3153 if (optlen != sizeof(struct sctp_authchunk))
3154 return -EINVAL;
3155 if (copy_from_user(&val, optval, optlen))
3156 return -EFAULT;
3158 switch (val.sauth_chunk) {
3159 case SCTP_CID_INIT:
3160 case SCTP_CID_INIT_ACK:
3161 case SCTP_CID_SHUTDOWN_COMPLETE:
3162 case SCTP_CID_AUTH:
3163 return -EINVAL;
3166 /* add this chunk id to the endpoint */
3167 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3171 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3173 * This option gets or sets the list of HMAC algorithms that the local
3174 * endpoint requires the peer to use.
3176 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3177 char __user *optval,
3178 int optlen)
3180 struct sctp_hmacalgo *hmacs;
3181 u32 idents;
3182 int err;
3184 if (!sctp_auth_enable)
3185 return -EACCES;
3187 if (optlen < sizeof(struct sctp_hmacalgo))
3188 return -EINVAL;
3190 hmacs = kmalloc(optlen, GFP_KERNEL);
3191 if (!hmacs)
3192 return -ENOMEM;
3194 if (copy_from_user(hmacs, optval, optlen)) {
3195 err = -EFAULT;
3196 goto out;
3199 idents = hmacs->shmac_num_idents;
3200 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3201 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3202 err = -EINVAL;
3203 goto out;
3206 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3207 out:
3208 kfree(hmacs);
3209 return err;
3213 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3215 * This option will set a shared secret key which is used to build an
3216 * association shared key.
3218 static int sctp_setsockopt_auth_key(struct sock *sk,
3219 char __user *optval,
3220 int optlen)
3222 struct sctp_authkey *authkey;
3223 struct sctp_association *asoc;
3224 int ret;
3226 if (!sctp_auth_enable)
3227 return -EACCES;
3229 if (optlen <= sizeof(struct sctp_authkey))
3230 return -EINVAL;
3232 authkey = kmalloc(optlen, GFP_KERNEL);
3233 if (!authkey)
3234 return -ENOMEM;
3236 if (copy_from_user(authkey, optval, optlen)) {
3237 ret = -EFAULT;
3238 goto out;
3241 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3242 ret = -EINVAL;
3243 goto out;
3246 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3247 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3248 ret = -EINVAL;
3249 goto out;
3252 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3253 out:
3254 kfree(authkey);
3255 return ret;
3259 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3261 * This option will get or set the active shared key to be used to build
3262 * the association shared key.
3264 static int sctp_setsockopt_active_key(struct sock *sk,
3265 char __user *optval,
3266 int optlen)
3268 struct sctp_authkeyid val;
3269 struct sctp_association *asoc;
3271 if (!sctp_auth_enable)
3272 return -EACCES;
3274 if (optlen != sizeof(struct sctp_authkeyid))
3275 return -EINVAL;
3276 if (copy_from_user(&val, optval, optlen))
3277 return -EFAULT;
3279 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3280 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3281 return -EINVAL;
3283 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3284 val.scact_keynumber);
3288 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3290 * This set option will delete a shared secret key from use.
3292 static int sctp_setsockopt_del_key(struct sock *sk,
3293 char __user *optval,
3294 int optlen)
3296 struct sctp_authkeyid val;
3297 struct sctp_association *asoc;
3299 if (!sctp_auth_enable)
3300 return -EACCES;
3302 if (optlen != sizeof(struct sctp_authkeyid))
3303 return -EINVAL;
3304 if (copy_from_user(&val, optval, optlen))
3305 return -EFAULT;
3307 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3308 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3309 return -EINVAL;
3311 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3312 val.scact_keynumber);
3317 /* API 6.2 setsockopt(), getsockopt()
3319 * Applications use setsockopt() and getsockopt() to set or retrieve
3320 * socket options. Socket options are used to change the default
3321 * behavior of sockets calls. They are described in Section 7.
3323 * The syntax is:
3325 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3326 * int __user *optlen);
3327 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3328 * int optlen);
3330 * sd - the socket descript.
3331 * level - set to IPPROTO_SCTP for all SCTP options.
3332 * optname - the option name.
3333 * optval - the buffer to store the value of the option.
3334 * optlen - the size of the buffer.
3336 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3337 char __user *optval, int optlen)
3339 int retval = 0;
3341 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3342 sk, optname);
3344 /* I can hardly begin to describe how wrong this is. This is
3345 * so broken as to be worse than useless. The API draft
3346 * REALLY is NOT helpful here... I am not convinced that the
3347 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3348 * are at all well-founded.
3350 if (level != SOL_SCTP) {
3351 struct sctp_af *af = sctp_sk(sk)->pf->af;
3352 retval = af->setsockopt(sk, level, optname, optval, optlen);
3353 goto out_nounlock;
3356 sctp_lock_sock(sk);
3358 switch (optname) {
3359 case SCTP_SOCKOPT_BINDX_ADD:
3360 /* 'optlen' is the size of the addresses buffer. */
3361 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3362 optlen, SCTP_BINDX_ADD_ADDR);
3363 break;
3365 case SCTP_SOCKOPT_BINDX_REM:
3366 /* 'optlen' is the size of the addresses buffer. */
3367 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3368 optlen, SCTP_BINDX_REM_ADDR);
3369 break;
3371 case SCTP_SOCKOPT_CONNECTX_OLD:
3372 /* 'optlen' is the size of the addresses buffer. */
3373 retval = sctp_setsockopt_connectx_old(sk,
3374 (struct sockaddr __user *)optval,
3375 optlen);
3376 break;
3378 case SCTP_SOCKOPT_CONNECTX:
3379 /* 'optlen' is the size of the addresses buffer. */
3380 retval = sctp_setsockopt_connectx(sk,
3381 (struct sockaddr __user *)optval,
3382 optlen);
3383 break;
3385 case SCTP_DISABLE_FRAGMENTS:
3386 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3387 break;
3389 case SCTP_EVENTS:
3390 retval = sctp_setsockopt_events(sk, optval, optlen);
3391 break;
3393 case SCTP_AUTOCLOSE:
3394 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3395 break;
3397 case SCTP_PEER_ADDR_PARAMS:
3398 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3399 break;
3401 case SCTP_DELAYED_ACK:
3402 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3403 break;
3404 case SCTP_PARTIAL_DELIVERY_POINT:
3405 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3406 break;
3408 case SCTP_INITMSG:
3409 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3410 break;
3411 case SCTP_DEFAULT_SEND_PARAM:
3412 retval = sctp_setsockopt_default_send_param(sk, optval,
3413 optlen);
3414 break;
3415 case SCTP_PRIMARY_ADDR:
3416 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3417 break;
3418 case SCTP_SET_PEER_PRIMARY_ADDR:
3419 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3420 break;
3421 case SCTP_NODELAY:
3422 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3423 break;
3424 case SCTP_RTOINFO:
3425 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3426 break;
3427 case SCTP_ASSOCINFO:
3428 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3429 break;
3430 case SCTP_I_WANT_MAPPED_V4_ADDR:
3431 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3432 break;
3433 case SCTP_MAXSEG:
3434 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3435 break;
3436 case SCTP_ADAPTATION_LAYER:
3437 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3438 break;
3439 case SCTP_CONTEXT:
3440 retval = sctp_setsockopt_context(sk, optval, optlen);
3441 break;
3442 case SCTP_FRAGMENT_INTERLEAVE:
3443 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3444 break;
3445 case SCTP_MAX_BURST:
3446 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3447 break;
3448 case SCTP_AUTH_CHUNK:
3449 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3450 break;
3451 case SCTP_HMAC_IDENT:
3452 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3453 break;
3454 case SCTP_AUTH_KEY:
3455 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3456 break;
3457 case SCTP_AUTH_ACTIVE_KEY:
3458 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3459 break;
3460 case SCTP_AUTH_DELETE_KEY:
3461 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3462 break;
3463 default:
3464 retval = -ENOPROTOOPT;
3465 break;
3468 sctp_release_sock(sk);
3470 out_nounlock:
3471 return retval;
3474 /* API 3.1.6 connect() - UDP Style Syntax
3476 * An application may use the connect() call in the UDP model to initiate an
3477 * association without sending data.
3479 * The syntax is:
3481 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3483 * sd: the socket descriptor to have a new association added to.
3485 * nam: the address structure (either struct sockaddr_in or struct
3486 * sockaddr_in6 defined in RFC2553 [7]).
3488 * len: the size of the address.
3490 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3491 int addr_len)
3493 int err = 0;
3494 struct sctp_af *af;
3496 sctp_lock_sock(sk);
3498 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3499 __func__, sk, addr, addr_len);
3501 /* Validate addr_len before calling common connect/connectx routine. */
3502 af = sctp_get_af_specific(addr->sa_family);
3503 if (!af || addr_len < af->sockaddr_len) {
3504 err = -EINVAL;
3505 } else {
3506 /* Pass correct addr len to common routine (so it knows there
3507 * is only one address being passed.
3509 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3512 sctp_release_sock(sk);
3513 return err;
3516 /* FIXME: Write comments. */
3517 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3519 return -EOPNOTSUPP; /* STUB */
3522 /* 4.1.4 accept() - TCP Style Syntax
3524 * Applications use accept() call to remove an established SCTP
3525 * association from the accept queue of the endpoint. A new socket
3526 * descriptor will be returned from accept() to represent the newly
3527 * formed association.
3529 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3531 struct sctp_sock *sp;
3532 struct sctp_endpoint *ep;
3533 struct sock *newsk = NULL;
3534 struct sctp_association *asoc;
3535 long timeo;
3536 int error = 0;
3538 sctp_lock_sock(sk);
3540 sp = sctp_sk(sk);
3541 ep = sp->ep;
3543 if (!sctp_style(sk, TCP)) {
3544 error = -EOPNOTSUPP;
3545 goto out;
3548 if (!sctp_sstate(sk, LISTENING)) {
3549 error = -EINVAL;
3550 goto out;
3553 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3555 error = sctp_wait_for_accept(sk, timeo);
3556 if (error)
3557 goto out;
3559 /* We treat the list of associations on the endpoint as the accept
3560 * queue and pick the first association on the list.
3562 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3564 newsk = sp->pf->create_accept_sk(sk, asoc);
3565 if (!newsk) {
3566 error = -ENOMEM;
3567 goto out;
3570 /* Populate the fields of the newsk from the oldsk and migrate the
3571 * asoc to the newsk.
3573 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3575 out:
3576 sctp_release_sock(sk);
3577 *err = error;
3578 return newsk;
3581 /* The SCTP ioctl handler. */
3582 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3584 return -ENOIOCTLCMD;
3587 /* This is the function which gets called during socket creation to
3588 * initialized the SCTP-specific portion of the sock.
3589 * The sock structure should already be zero-filled memory.
3591 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3593 struct sctp_endpoint *ep;
3594 struct sctp_sock *sp;
3596 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3598 sp = sctp_sk(sk);
3600 /* Initialize the SCTP per socket area. */
3601 switch (sk->sk_type) {
3602 case SOCK_SEQPACKET:
3603 sp->type = SCTP_SOCKET_UDP;
3604 break;
3605 case SOCK_STREAM:
3606 sp->type = SCTP_SOCKET_TCP;
3607 break;
3608 default:
3609 return -ESOCKTNOSUPPORT;
3612 /* Initialize default send parameters. These parameters can be
3613 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3615 sp->default_stream = 0;
3616 sp->default_ppid = 0;
3617 sp->default_flags = 0;
3618 sp->default_context = 0;
3619 sp->default_timetolive = 0;
3621 sp->default_rcv_context = 0;
3622 sp->max_burst = sctp_max_burst;
3624 /* Initialize default setup parameters. These parameters
3625 * can be modified with the SCTP_INITMSG socket option or
3626 * overridden by the SCTP_INIT CMSG.
3628 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3629 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3630 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3631 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3633 /* Initialize default RTO related parameters. These parameters can
3634 * be modified for with the SCTP_RTOINFO socket option.
3636 sp->rtoinfo.srto_initial = sctp_rto_initial;
3637 sp->rtoinfo.srto_max = sctp_rto_max;
3638 sp->rtoinfo.srto_min = sctp_rto_min;
3640 /* Initialize default association related parameters. These parameters
3641 * can be modified with the SCTP_ASSOCINFO socket option.
3643 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3644 sp->assocparams.sasoc_number_peer_destinations = 0;
3645 sp->assocparams.sasoc_peer_rwnd = 0;
3646 sp->assocparams.sasoc_local_rwnd = 0;
3647 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3649 /* Initialize default event subscriptions. By default, all the
3650 * options are off.
3652 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3654 /* Default Peer Address Parameters. These defaults can
3655 * be modified via SCTP_PEER_ADDR_PARAMS
3657 sp->hbinterval = sctp_hb_interval;
3658 sp->pathmaxrxt = sctp_max_retrans_path;
3659 sp->pathmtu = 0; // allow default discovery
3660 sp->sackdelay = sctp_sack_timeout;
3661 sp->sackfreq = 2;
3662 sp->param_flags = SPP_HB_ENABLE |
3663 SPP_PMTUD_ENABLE |
3664 SPP_SACKDELAY_ENABLE;
3666 /* If enabled no SCTP message fragmentation will be performed.
3667 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3669 sp->disable_fragments = 0;
3671 /* Enable Nagle algorithm by default. */
3672 sp->nodelay = 0;
3674 /* Enable by default. */
3675 sp->v4mapped = 1;
3677 /* Auto-close idle associations after the configured
3678 * number of seconds. A value of 0 disables this
3679 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3680 * for UDP-style sockets only.
3682 sp->autoclose = 0;
3684 /* User specified fragmentation limit. */
3685 sp->user_frag = 0;
3687 sp->adaptation_ind = 0;
3689 sp->pf = sctp_get_pf_specific(sk->sk_family);
3691 /* Control variables for partial data delivery. */
3692 atomic_set(&sp->pd_mode, 0);
3693 skb_queue_head_init(&sp->pd_lobby);
3694 sp->frag_interleave = 0;
3696 /* Create a per socket endpoint structure. Even if we
3697 * change the data structure relationships, this may still
3698 * be useful for storing pre-connect address information.
3700 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3701 if (!ep)
3702 return -ENOMEM;
3704 sp->ep = ep;
3705 sp->hmac = NULL;
3707 SCTP_DBG_OBJCNT_INC(sock);
3708 percpu_counter_inc(&sctp_sockets_allocated);
3710 local_bh_disable();
3711 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3712 local_bh_enable();
3714 return 0;
3717 /* Cleanup any SCTP per socket resources. */
3718 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3720 struct sctp_endpoint *ep;
3722 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3724 /* Release our hold on the endpoint. */
3725 ep = sctp_sk(sk)->ep;
3726 sctp_endpoint_free(ep);
3727 percpu_counter_dec(&sctp_sockets_allocated);
3728 local_bh_disable();
3729 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3730 local_bh_enable();
3733 /* API 4.1.7 shutdown() - TCP Style Syntax
3734 * int shutdown(int socket, int how);
3736 * sd - the socket descriptor of the association to be closed.
3737 * how - Specifies the type of shutdown. The values are
3738 * as follows:
3739 * SHUT_RD
3740 * Disables further receive operations. No SCTP
3741 * protocol action is taken.
3742 * SHUT_WR
3743 * Disables further send operations, and initiates
3744 * the SCTP shutdown sequence.
3745 * SHUT_RDWR
3746 * Disables further send and receive operations
3747 * and initiates the SCTP shutdown sequence.
3749 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3751 struct sctp_endpoint *ep;
3752 struct sctp_association *asoc;
3754 if (!sctp_style(sk, TCP))
3755 return;
3757 if (how & SEND_SHUTDOWN) {
3758 ep = sctp_sk(sk)->ep;
3759 if (!list_empty(&ep->asocs)) {
3760 asoc = list_entry(ep->asocs.next,
3761 struct sctp_association, asocs);
3762 sctp_primitive_SHUTDOWN(asoc, NULL);
3767 /* 7.2.1 Association Status (SCTP_STATUS)
3769 * Applications can retrieve current status information about an
3770 * association, including association state, peer receiver window size,
3771 * number of unacked data chunks, and number of data chunks pending
3772 * receipt. This information is read-only.
3774 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3775 char __user *optval,
3776 int __user *optlen)
3778 struct sctp_status status;
3779 struct sctp_association *asoc = NULL;
3780 struct sctp_transport *transport;
3781 sctp_assoc_t associd;
3782 int retval = 0;
3784 if (len < sizeof(status)) {
3785 retval = -EINVAL;
3786 goto out;
3789 len = sizeof(status);
3790 if (copy_from_user(&status, optval, len)) {
3791 retval = -EFAULT;
3792 goto out;
3795 associd = status.sstat_assoc_id;
3796 asoc = sctp_id2assoc(sk, associd);
3797 if (!asoc) {
3798 retval = -EINVAL;
3799 goto out;
3802 transport = asoc->peer.primary_path;
3804 status.sstat_assoc_id = sctp_assoc2id(asoc);
3805 status.sstat_state = asoc->state;
3806 status.sstat_rwnd = asoc->peer.rwnd;
3807 status.sstat_unackdata = asoc->unack_data;
3809 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3810 status.sstat_instrms = asoc->c.sinit_max_instreams;
3811 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3812 status.sstat_fragmentation_point = asoc->frag_point;
3813 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3814 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3815 transport->af_specific->sockaddr_len);
3816 /* Map ipv4 address into v4-mapped-on-v6 address. */
3817 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3818 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3819 status.sstat_primary.spinfo_state = transport->state;
3820 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3821 status.sstat_primary.spinfo_srtt = transport->srtt;
3822 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3823 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3825 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3826 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3828 if (put_user(len, optlen)) {
3829 retval = -EFAULT;
3830 goto out;
3833 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3834 len, status.sstat_state, status.sstat_rwnd,
3835 status.sstat_assoc_id);
3837 if (copy_to_user(optval, &status, len)) {
3838 retval = -EFAULT;
3839 goto out;
3842 out:
3843 return (retval);
3847 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3849 * Applications can retrieve information about a specific peer address
3850 * of an association, including its reachability state, congestion
3851 * window, and retransmission timer values. This information is
3852 * read-only.
3854 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3855 char __user *optval,
3856 int __user *optlen)
3858 struct sctp_paddrinfo pinfo;
3859 struct sctp_transport *transport;
3860 int retval = 0;
3862 if (len < sizeof(pinfo)) {
3863 retval = -EINVAL;
3864 goto out;
3867 len = sizeof(pinfo);
3868 if (copy_from_user(&pinfo, optval, len)) {
3869 retval = -EFAULT;
3870 goto out;
3873 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3874 pinfo.spinfo_assoc_id);
3875 if (!transport)
3876 return -EINVAL;
3878 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3879 pinfo.spinfo_state = transport->state;
3880 pinfo.spinfo_cwnd = transport->cwnd;
3881 pinfo.spinfo_srtt = transport->srtt;
3882 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3883 pinfo.spinfo_mtu = transport->pathmtu;
3885 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3886 pinfo.spinfo_state = SCTP_ACTIVE;
3888 if (put_user(len, optlen)) {
3889 retval = -EFAULT;
3890 goto out;
3893 if (copy_to_user(optval, &pinfo, len)) {
3894 retval = -EFAULT;
3895 goto out;
3898 out:
3899 return (retval);
3902 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3904 * This option is a on/off flag. If enabled no SCTP message
3905 * fragmentation will be performed. Instead if a message being sent
3906 * exceeds the current PMTU size, the message will NOT be sent and
3907 * instead a error will be indicated to the user.
3909 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3910 char __user *optval, int __user *optlen)
3912 int val;
3914 if (len < sizeof(int))
3915 return -EINVAL;
3917 len = sizeof(int);
3918 val = (sctp_sk(sk)->disable_fragments == 1);
3919 if (put_user(len, optlen))
3920 return -EFAULT;
3921 if (copy_to_user(optval, &val, len))
3922 return -EFAULT;
3923 return 0;
3926 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3928 * This socket option is used to specify various notifications and
3929 * ancillary data the user wishes to receive.
3931 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3932 int __user *optlen)
3934 if (len < sizeof(struct sctp_event_subscribe))
3935 return -EINVAL;
3936 len = sizeof(struct sctp_event_subscribe);
3937 if (put_user(len, optlen))
3938 return -EFAULT;
3939 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3940 return -EFAULT;
3941 return 0;
3944 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3946 * This socket option is applicable to the UDP-style socket only. When
3947 * set it will cause associations that are idle for more than the
3948 * specified number of seconds to automatically close. An association
3949 * being idle is defined an association that has NOT sent or received
3950 * user data. The special value of '0' indicates that no automatic
3951 * close of any associations should be performed. The option expects an
3952 * integer defining the number of seconds of idle time before an
3953 * association is closed.
3955 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3957 /* Applicable to UDP-style socket only */
3958 if (sctp_style(sk, TCP))
3959 return -EOPNOTSUPP;
3960 if (len < sizeof(int))
3961 return -EINVAL;
3962 len = sizeof(int);
3963 if (put_user(len, optlen))
3964 return -EFAULT;
3965 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3966 return -EFAULT;
3967 return 0;
3970 /* Helper routine to branch off an association to a new socket. */
3971 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3972 struct socket **sockp)
3974 struct sock *sk = asoc->base.sk;
3975 struct socket *sock;
3976 struct sctp_af *af;
3977 int err = 0;
3979 /* An association cannot be branched off from an already peeled-off
3980 * socket, nor is this supported for tcp style sockets.
3982 if (!sctp_style(sk, UDP))
3983 return -EINVAL;
3985 /* Create a new socket. */
3986 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3987 if (err < 0)
3988 return err;
3990 sctp_copy_sock(sock->sk, sk, asoc);
3992 /* Make peeled-off sockets more like 1-1 accepted sockets.
3993 * Set the daddr and initialize id to something more random
3995 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3996 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3998 /* Populate the fields of the newsk from the oldsk and migrate the
3999 * asoc to the newsk.
4001 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4003 *sockp = sock;
4005 return err;
4008 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4010 sctp_peeloff_arg_t peeloff;
4011 struct socket *newsock;
4012 int retval = 0;
4013 struct sctp_association *asoc;
4015 if (len < sizeof(sctp_peeloff_arg_t))
4016 return -EINVAL;
4017 len = sizeof(sctp_peeloff_arg_t);
4018 if (copy_from_user(&peeloff, optval, len))
4019 return -EFAULT;
4021 asoc = sctp_id2assoc(sk, peeloff.associd);
4022 if (!asoc) {
4023 retval = -EINVAL;
4024 goto out;
4027 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4029 retval = sctp_do_peeloff(asoc, &newsock);
4030 if (retval < 0)
4031 goto out;
4033 /* Map the socket to an unused fd that can be returned to the user. */
4034 retval = sock_map_fd(newsock, 0);
4035 if (retval < 0) {
4036 sock_release(newsock);
4037 goto out;
4040 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4041 __func__, sk, asoc, newsock->sk, retval);
4043 /* Return the fd mapped to the new socket. */
4044 peeloff.sd = retval;
4045 if (put_user(len, optlen))
4046 return -EFAULT;
4047 if (copy_to_user(optval, &peeloff, len))
4048 retval = -EFAULT;
4050 out:
4051 return retval;
4054 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4056 * Applications can enable or disable heartbeats for any peer address of
4057 * an association, modify an address's heartbeat interval, force a
4058 * heartbeat to be sent immediately, and adjust the address's maximum
4059 * number of retransmissions sent before an address is considered
4060 * unreachable. The following structure is used to access and modify an
4061 * address's parameters:
4063 * struct sctp_paddrparams {
4064 * sctp_assoc_t spp_assoc_id;
4065 * struct sockaddr_storage spp_address;
4066 * uint32_t spp_hbinterval;
4067 * uint16_t spp_pathmaxrxt;
4068 * uint32_t spp_pathmtu;
4069 * uint32_t spp_sackdelay;
4070 * uint32_t spp_flags;
4071 * };
4073 * spp_assoc_id - (one-to-many style socket) This is filled in the
4074 * application, and identifies the association for
4075 * this query.
4076 * spp_address - This specifies which address is of interest.
4077 * spp_hbinterval - This contains the value of the heartbeat interval,
4078 * in milliseconds. If a value of zero
4079 * is present in this field then no changes are to
4080 * be made to this parameter.
4081 * spp_pathmaxrxt - This contains the maximum number of
4082 * retransmissions before this address shall be
4083 * considered unreachable. If a value of zero
4084 * is present in this field then no changes are to
4085 * be made to this parameter.
4086 * spp_pathmtu - When Path MTU discovery is disabled the value
4087 * specified here will be the "fixed" path mtu.
4088 * Note that if the spp_address field is empty
4089 * then all associations on this address will
4090 * have this fixed path mtu set upon them.
4092 * spp_sackdelay - When delayed sack is enabled, this value specifies
4093 * the number of milliseconds that sacks will be delayed
4094 * for. This value will apply to all addresses of an
4095 * association if the spp_address field is empty. Note
4096 * also, that if delayed sack is enabled and this
4097 * value is set to 0, no change is made to the last
4098 * recorded delayed sack timer value.
4100 * spp_flags - These flags are used to control various features
4101 * on an association. The flag field may contain
4102 * zero or more of the following options.
4104 * SPP_HB_ENABLE - Enable heartbeats on the
4105 * specified address. Note that if the address
4106 * field is empty all addresses for the association
4107 * have heartbeats enabled upon them.
4109 * SPP_HB_DISABLE - Disable heartbeats on the
4110 * speicifed address. Note that if the address
4111 * field is empty all addresses for the association
4112 * will have their heartbeats disabled. Note also
4113 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4114 * mutually exclusive, only one of these two should
4115 * be specified. Enabling both fields will have
4116 * undetermined results.
4118 * SPP_HB_DEMAND - Request a user initiated heartbeat
4119 * to be made immediately.
4121 * SPP_PMTUD_ENABLE - This field will enable PMTU
4122 * discovery upon the specified address. Note that
4123 * if the address feild is empty then all addresses
4124 * on the association are effected.
4126 * SPP_PMTUD_DISABLE - This field will disable PMTU
4127 * discovery upon the specified address. Note that
4128 * if the address feild is empty then all addresses
4129 * on the association are effected. Not also that
4130 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4131 * exclusive. Enabling both will have undetermined
4132 * results.
4134 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4135 * on delayed sack. The time specified in spp_sackdelay
4136 * is used to specify the sack delay for this address. Note
4137 * that if spp_address is empty then all addresses will
4138 * enable delayed sack and take on the sack delay
4139 * value specified in spp_sackdelay.
4140 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4141 * off delayed sack. If the spp_address field is blank then
4142 * delayed sack is disabled for the entire association. Note
4143 * also that this field is mutually exclusive to
4144 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4145 * results.
4147 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4148 char __user *optval, int __user *optlen)
4150 struct sctp_paddrparams params;
4151 struct sctp_transport *trans = NULL;
4152 struct sctp_association *asoc = NULL;
4153 struct sctp_sock *sp = sctp_sk(sk);
4155 if (len < sizeof(struct sctp_paddrparams))
4156 return -EINVAL;
4157 len = sizeof(struct sctp_paddrparams);
4158 if (copy_from_user(&params, optval, len))
4159 return -EFAULT;
4161 /* If an address other than INADDR_ANY is specified, and
4162 * no transport is found, then the request is invalid.
4164 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4165 trans = sctp_addr_id2transport(sk, &params.spp_address,
4166 params.spp_assoc_id);
4167 if (!trans) {
4168 SCTP_DEBUG_PRINTK("Failed no transport\n");
4169 return -EINVAL;
4173 /* Get association, if assoc_id != 0 and the socket is a one
4174 * to many style socket, and an association was not found, then
4175 * the id was invalid.
4177 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4178 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4179 SCTP_DEBUG_PRINTK("Failed no association\n");
4180 return -EINVAL;
4183 if (trans) {
4184 /* Fetch transport values. */
4185 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4186 params.spp_pathmtu = trans->pathmtu;
4187 params.spp_pathmaxrxt = trans->pathmaxrxt;
4188 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4190 /*draft-11 doesn't say what to return in spp_flags*/
4191 params.spp_flags = trans->param_flags;
4192 } else if (asoc) {
4193 /* Fetch association values. */
4194 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4195 params.spp_pathmtu = asoc->pathmtu;
4196 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4197 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4199 /*draft-11 doesn't say what to return in spp_flags*/
4200 params.spp_flags = asoc->param_flags;
4201 } else {
4202 /* Fetch socket values. */
4203 params.spp_hbinterval = sp->hbinterval;
4204 params.spp_pathmtu = sp->pathmtu;
4205 params.spp_sackdelay = sp->sackdelay;
4206 params.spp_pathmaxrxt = sp->pathmaxrxt;
4208 /*draft-11 doesn't say what to return in spp_flags*/
4209 params.spp_flags = sp->param_flags;
4212 if (copy_to_user(optval, &params, len))
4213 return -EFAULT;
4215 if (put_user(len, optlen))
4216 return -EFAULT;
4218 return 0;
4222 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4224 * This option will effect the way delayed acks are performed. This
4225 * option allows you to get or set the delayed ack time, in
4226 * milliseconds. It also allows changing the delayed ack frequency.
4227 * Changing the frequency to 1 disables the delayed sack algorithm. If
4228 * the assoc_id is 0, then this sets or gets the endpoints default
4229 * values. If the assoc_id field is non-zero, then the set or get
4230 * effects the specified association for the one to many model (the
4231 * assoc_id field is ignored by the one to one model). Note that if
4232 * sack_delay or sack_freq are 0 when setting this option, then the
4233 * current values will remain unchanged.
4235 * struct sctp_sack_info {
4236 * sctp_assoc_t sack_assoc_id;
4237 * uint32_t sack_delay;
4238 * uint32_t sack_freq;
4239 * };
4241 * sack_assoc_id - This parameter, indicates which association the user
4242 * is performing an action upon. Note that if this field's value is
4243 * zero then the endpoints default value is changed (effecting future
4244 * associations only).
4246 * sack_delay - This parameter contains the number of milliseconds that
4247 * the user is requesting the delayed ACK timer be set to. Note that
4248 * this value is defined in the standard to be between 200 and 500
4249 * milliseconds.
4251 * sack_freq - This parameter contains the number of packets that must
4252 * be received before a sack is sent without waiting for the delay
4253 * timer to expire. The default value for this is 2, setting this
4254 * value to 1 will disable the delayed sack algorithm.
4256 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4257 char __user *optval,
4258 int __user *optlen)
4260 struct sctp_sack_info params;
4261 struct sctp_association *asoc = NULL;
4262 struct sctp_sock *sp = sctp_sk(sk);
4264 if (len >= sizeof(struct sctp_sack_info)) {
4265 len = sizeof(struct sctp_sack_info);
4267 if (copy_from_user(&params, optval, len))
4268 return -EFAULT;
4269 } else if (len == sizeof(struct sctp_assoc_value)) {
4270 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4271 "in delayed_ack socket option deprecated\n");
4272 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4273 if (copy_from_user(&params, optval, len))
4274 return -EFAULT;
4275 } else
4276 return - EINVAL;
4278 /* Get association, if sack_assoc_id != 0 and the socket is a one
4279 * to many style socket, and an association was not found, then
4280 * the id was invalid.
4282 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4283 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4284 return -EINVAL;
4286 if (asoc) {
4287 /* Fetch association values. */
4288 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4289 params.sack_delay = jiffies_to_msecs(
4290 asoc->sackdelay);
4291 params.sack_freq = asoc->sackfreq;
4293 } else {
4294 params.sack_delay = 0;
4295 params.sack_freq = 1;
4297 } else {
4298 /* Fetch socket values. */
4299 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4300 params.sack_delay = sp->sackdelay;
4301 params.sack_freq = sp->sackfreq;
4302 } else {
4303 params.sack_delay = 0;
4304 params.sack_freq = 1;
4308 if (copy_to_user(optval, &params, len))
4309 return -EFAULT;
4311 if (put_user(len, optlen))
4312 return -EFAULT;
4314 return 0;
4317 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4319 * Applications can specify protocol parameters for the default association
4320 * initialization. The option name argument to setsockopt() and getsockopt()
4321 * is SCTP_INITMSG.
4323 * Setting initialization parameters is effective only on an unconnected
4324 * socket (for UDP-style sockets only future associations are effected
4325 * by the change). With TCP-style sockets, this option is inherited by
4326 * sockets derived from a listener socket.
4328 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4330 if (len < sizeof(struct sctp_initmsg))
4331 return -EINVAL;
4332 len = sizeof(struct sctp_initmsg);
4333 if (put_user(len, optlen))
4334 return -EFAULT;
4335 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4336 return -EFAULT;
4337 return 0;
4340 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4341 char __user *optval,
4342 int __user *optlen)
4344 sctp_assoc_t id;
4345 struct sctp_association *asoc;
4346 struct list_head *pos;
4347 int cnt = 0;
4349 if (len < sizeof(sctp_assoc_t))
4350 return -EINVAL;
4352 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4353 return -EFAULT;
4355 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4356 "socket option deprecated\n");
4357 /* For UDP-style sockets, id specifies the association to query. */
4358 asoc = sctp_id2assoc(sk, id);
4359 if (!asoc)
4360 return -EINVAL;
4362 list_for_each(pos, &asoc->peer.transport_addr_list) {
4363 cnt ++;
4366 return cnt;
4370 * Old API for getting list of peer addresses. Does not work for 32-bit
4371 * programs running on a 64-bit kernel
4373 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4374 char __user *optval,
4375 int __user *optlen)
4377 struct sctp_association *asoc;
4378 int cnt = 0;
4379 struct sctp_getaddrs_old getaddrs;
4380 struct sctp_transport *from;
4381 void __user *to;
4382 union sctp_addr temp;
4383 struct sctp_sock *sp = sctp_sk(sk);
4384 int addrlen;
4386 if (len < sizeof(struct sctp_getaddrs_old))
4387 return -EINVAL;
4389 len = sizeof(struct sctp_getaddrs_old);
4391 if (copy_from_user(&getaddrs, optval, len))
4392 return -EFAULT;
4394 if (getaddrs.addr_num <= 0) return -EINVAL;
4396 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4397 "socket option deprecated\n");
4399 /* For UDP-style sockets, id specifies the association to query. */
4400 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4401 if (!asoc)
4402 return -EINVAL;
4404 to = (void __user *)getaddrs.addrs;
4405 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4406 transports) {
4407 memcpy(&temp, &from->ipaddr, sizeof(temp));
4408 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4409 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4410 if (copy_to_user(to, &temp, addrlen))
4411 return -EFAULT;
4412 to += addrlen ;
4413 cnt ++;
4414 if (cnt >= getaddrs.addr_num) break;
4416 getaddrs.addr_num = cnt;
4417 if (put_user(len, optlen))
4418 return -EFAULT;
4419 if (copy_to_user(optval, &getaddrs, len))
4420 return -EFAULT;
4422 return 0;
4425 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4426 char __user *optval, int __user *optlen)
4428 struct sctp_association *asoc;
4429 int cnt = 0;
4430 struct sctp_getaddrs getaddrs;
4431 struct sctp_transport *from;
4432 void __user *to;
4433 union sctp_addr temp;
4434 struct sctp_sock *sp = sctp_sk(sk);
4435 int addrlen;
4436 size_t space_left;
4437 int bytes_copied;
4439 if (len < sizeof(struct sctp_getaddrs))
4440 return -EINVAL;
4442 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4443 return -EFAULT;
4445 /* For UDP-style sockets, id specifies the association to query. */
4446 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4447 if (!asoc)
4448 return -EINVAL;
4450 to = optval + offsetof(struct sctp_getaddrs,addrs);
4451 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4453 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4454 transports) {
4455 memcpy(&temp, &from->ipaddr, sizeof(temp));
4456 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4457 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4458 if (space_left < addrlen)
4459 return -ENOMEM;
4460 if (copy_to_user(to, &temp, addrlen))
4461 return -EFAULT;
4462 to += addrlen;
4463 cnt++;
4464 space_left -= addrlen;
4467 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4468 return -EFAULT;
4469 bytes_copied = ((char __user *)to) - optval;
4470 if (put_user(bytes_copied, optlen))
4471 return -EFAULT;
4473 return 0;
4476 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4477 char __user *optval,
4478 int __user *optlen)
4480 sctp_assoc_t id;
4481 struct sctp_bind_addr *bp;
4482 struct sctp_association *asoc;
4483 struct sctp_sockaddr_entry *addr;
4484 int cnt = 0;
4486 if (len < sizeof(sctp_assoc_t))
4487 return -EINVAL;
4489 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4490 return -EFAULT;
4492 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4493 "socket option deprecated\n");
4496 * For UDP-style sockets, id specifies the association to query.
4497 * If the id field is set to the value '0' then the locally bound
4498 * addresses are returned without regard to any particular
4499 * association.
4501 if (0 == id) {
4502 bp = &sctp_sk(sk)->ep->base.bind_addr;
4503 } else {
4504 asoc = sctp_id2assoc(sk, id);
4505 if (!asoc)
4506 return -EINVAL;
4507 bp = &asoc->base.bind_addr;
4510 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4511 * addresses from the global local address list.
4513 if (sctp_list_single_entry(&bp->address_list)) {
4514 addr = list_entry(bp->address_list.next,
4515 struct sctp_sockaddr_entry, list);
4516 if (sctp_is_any(sk, &addr->a)) {
4517 rcu_read_lock();
4518 list_for_each_entry_rcu(addr,
4519 &sctp_local_addr_list, list) {
4520 if (!addr->valid)
4521 continue;
4523 if ((PF_INET == sk->sk_family) &&
4524 (AF_INET6 == addr->a.sa.sa_family))
4525 continue;
4527 if ((PF_INET6 == sk->sk_family) &&
4528 inet_v6_ipv6only(sk) &&
4529 (AF_INET == addr->a.sa.sa_family))
4530 continue;
4532 cnt++;
4534 rcu_read_unlock();
4535 } else {
4536 cnt = 1;
4538 goto done;
4541 /* Protection on the bound address list is not needed,
4542 * since in the socket option context we hold the socket lock,
4543 * so there is no way that the bound address list can change.
4545 list_for_each_entry(addr, &bp->address_list, list) {
4546 cnt ++;
4548 done:
4549 return cnt;
4552 /* Helper function that copies local addresses to user and returns the number
4553 * of addresses copied.
4555 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4556 int max_addrs, void *to,
4557 int *bytes_copied)
4559 struct sctp_sockaddr_entry *addr;
4560 union sctp_addr temp;
4561 int cnt = 0;
4562 int addrlen;
4564 rcu_read_lock();
4565 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4566 if (!addr->valid)
4567 continue;
4569 if ((PF_INET == sk->sk_family) &&
4570 (AF_INET6 == addr->a.sa.sa_family))
4571 continue;
4572 if ((PF_INET6 == sk->sk_family) &&
4573 inet_v6_ipv6only(sk) &&
4574 (AF_INET == addr->a.sa.sa_family))
4575 continue;
4576 memcpy(&temp, &addr->a, sizeof(temp));
4577 if (!temp.v4.sin_port)
4578 temp.v4.sin_port = htons(port);
4580 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4581 &temp);
4582 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4583 memcpy(to, &temp, addrlen);
4585 to += addrlen;
4586 *bytes_copied += addrlen;
4587 cnt ++;
4588 if (cnt >= max_addrs) break;
4590 rcu_read_unlock();
4592 return cnt;
4595 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4596 size_t space_left, int *bytes_copied)
4598 struct sctp_sockaddr_entry *addr;
4599 union sctp_addr temp;
4600 int cnt = 0;
4601 int addrlen;
4603 rcu_read_lock();
4604 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4605 if (!addr->valid)
4606 continue;
4608 if ((PF_INET == sk->sk_family) &&
4609 (AF_INET6 == addr->a.sa.sa_family))
4610 continue;
4611 if ((PF_INET6 == sk->sk_family) &&
4612 inet_v6_ipv6only(sk) &&
4613 (AF_INET == addr->a.sa.sa_family))
4614 continue;
4615 memcpy(&temp, &addr->a, sizeof(temp));
4616 if (!temp.v4.sin_port)
4617 temp.v4.sin_port = htons(port);
4619 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4620 &temp);
4621 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4622 if (space_left < addrlen) {
4623 cnt = -ENOMEM;
4624 break;
4626 memcpy(to, &temp, addrlen);
4628 to += addrlen;
4629 cnt ++;
4630 space_left -= addrlen;
4631 *bytes_copied += addrlen;
4633 rcu_read_unlock();
4635 return cnt;
4638 /* Old API for getting list of local addresses. Does not work for 32-bit
4639 * programs running on a 64-bit kernel
4641 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4642 char __user *optval, int __user *optlen)
4644 struct sctp_bind_addr *bp;
4645 struct sctp_association *asoc;
4646 int cnt = 0;
4647 struct sctp_getaddrs_old getaddrs;
4648 struct sctp_sockaddr_entry *addr;
4649 void __user *to;
4650 union sctp_addr temp;
4651 struct sctp_sock *sp = sctp_sk(sk);
4652 int addrlen;
4653 int err = 0;
4654 void *addrs;
4655 void *buf;
4656 int bytes_copied = 0;
4658 if (len < sizeof(struct sctp_getaddrs_old))
4659 return -EINVAL;
4661 len = sizeof(struct sctp_getaddrs_old);
4662 if (copy_from_user(&getaddrs, optval, len))
4663 return -EFAULT;
4665 if (getaddrs.addr_num <= 0 ||
4666 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4667 return -EINVAL;
4669 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4670 "socket option deprecated\n");
4673 * For UDP-style sockets, id specifies the association to query.
4674 * If the id field is set to the value '0' then the locally bound
4675 * addresses are returned without regard to any particular
4676 * association.
4678 if (0 == getaddrs.assoc_id) {
4679 bp = &sctp_sk(sk)->ep->base.bind_addr;
4680 } else {
4681 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4682 if (!asoc)
4683 return -EINVAL;
4684 bp = &asoc->base.bind_addr;
4687 to = getaddrs.addrs;
4689 /* Allocate space for a local instance of packed array to hold all
4690 * the data. We store addresses here first and then put write them
4691 * to the user in one shot.
4693 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4694 GFP_KERNEL);
4695 if (!addrs)
4696 return -ENOMEM;
4698 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4699 * addresses from the global local address list.
4701 if (sctp_list_single_entry(&bp->address_list)) {
4702 addr = list_entry(bp->address_list.next,
4703 struct sctp_sockaddr_entry, list);
4704 if (sctp_is_any(sk, &addr->a)) {
4705 cnt = sctp_copy_laddrs_old(sk, bp->port,
4706 getaddrs.addr_num,
4707 addrs, &bytes_copied);
4708 goto copy_getaddrs;
4712 buf = addrs;
4713 /* Protection on the bound address list is not needed since
4714 * in the socket option context we hold a socket lock and
4715 * thus the bound address list can't change.
4717 list_for_each_entry(addr, &bp->address_list, list) {
4718 memcpy(&temp, &addr->a, sizeof(temp));
4719 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4720 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4721 memcpy(buf, &temp, addrlen);
4722 buf += addrlen;
4723 bytes_copied += addrlen;
4724 cnt ++;
4725 if (cnt >= getaddrs.addr_num) break;
4728 copy_getaddrs:
4729 /* copy the entire address list into the user provided space */
4730 if (copy_to_user(to, addrs, bytes_copied)) {
4731 err = -EFAULT;
4732 goto error;
4735 /* copy the leading structure back to user */
4736 getaddrs.addr_num = cnt;
4737 if (copy_to_user(optval, &getaddrs, len))
4738 err = -EFAULT;
4740 error:
4741 kfree(addrs);
4742 return err;
4745 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4746 char __user *optval, int __user *optlen)
4748 struct sctp_bind_addr *bp;
4749 struct sctp_association *asoc;
4750 int cnt = 0;
4751 struct sctp_getaddrs getaddrs;
4752 struct sctp_sockaddr_entry *addr;
4753 void __user *to;
4754 union sctp_addr temp;
4755 struct sctp_sock *sp = sctp_sk(sk);
4756 int addrlen;
4757 int err = 0;
4758 size_t space_left;
4759 int bytes_copied = 0;
4760 void *addrs;
4761 void *buf;
4763 if (len < sizeof(struct sctp_getaddrs))
4764 return -EINVAL;
4766 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4767 return -EFAULT;
4770 * For UDP-style sockets, id specifies the association to query.
4771 * If the id field is set to the value '0' then the locally bound
4772 * addresses are returned without regard to any particular
4773 * association.
4775 if (0 == getaddrs.assoc_id) {
4776 bp = &sctp_sk(sk)->ep->base.bind_addr;
4777 } else {
4778 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4779 if (!asoc)
4780 return -EINVAL;
4781 bp = &asoc->base.bind_addr;
4784 to = optval + offsetof(struct sctp_getaddrs,addrs);
4785 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4787 addrs = kmalloc(space_left, GFP_KERNEL);
4788 if (!addrs)
4789 return -ENOMEM;
4791 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4792 * addresses from the global local address list.
4794 if (sctp_list_single_entry(&bp->address_list)) {
4795 addr = list_entry(bp->address_list.next,
4796 struct sctp_sockaddr_entry, list);
4797 if (sctp_is_any(sk, &addr->a)) {
4798 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4799 space_left, &bytes_copied);
4800 if (cnt < 0) {
4801 err = cnt;
4802 goto out;
4804 goto copy_getaddrs;
4808 buf = addrs;
4809 /* Protection on the bound address list is not needed since
4810 * in the socket option context we hold a socket lock and
4811 * thus the bound address list can't change.
4813 list_for_each_entry(addr, &bp->address_list, list) {
4814 memcpy(&temp, &addr->a, sizeof(temp));
4815 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4816 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4817 if (space_left < addrlen) {
4818 err = -ENOMEM; /*fixme: right error?*/
4819 goto out;
4821 memcpy(buf, &temp, addrlen);
4822 buf += addrlen;
4823 bytes_copied += addrlen;
4824 cnt ++;
4825 space_left -= addrlen;
4828 copy_getaddrs:
4829 if (copy_to_user(to, addrs, bytes_copied)) {
4830 err = -EFAULT;
4831 goto out;
4833 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4834 err = -EFAULT;
4835 goto out;
4837 if (put_user(bytes_copied, optlen))
4838 err = -EFAULT;
4839 out:
4840 kfree(addrs);
4841 return err;
4844 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4846 * Requests that the local SCTP stack use the enclosed peer address as
4847 * the association primary. The enclosed address must be one of the
4848 * association peer's addresses.
4850 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4851 char __user *optval, int __user *optlen)
4853 struct sctp_prim prim;
4854 struct sctp_association *asoc;
4855 struct sctp_sock *sp = sctp_sk(sk);
4857 if (len < sizeof(struct sctp_prim))
4858 return -EINVAL;
4860 len = sizeof(struct sctp_prim);
4862 if (copy_from_user(&prim, optval, len))
4863 return -EFAULT;
4865 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4866 if (!asoc)
4867 return -EINVAL;
4869 if (!asoc->peer.primary_path)
4870 return -ENOTCONN;
4872 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4873 asoc->peer.primary_path->af_specific->sockaddr_len);
4875 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4876 (union sctp_addr *)&prim.ssp_addr);
4878 if (put_user(len, optlen))
4879 return -EFAULT;
4880 if (copy_to_user(optval, &prim, len))
4881 return -EFAULT;
4883 return 0;
4887 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4889 * Requests that the local endpoint set the specified Adaptation Layer
4890 * Indication parameter for all future INIT and INIT-ACK exchanges.
4892 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4893 char __user *optval, int __user *optlen)
4895 struct sctp_setadaptation adaptation;
4897 if (len < sizeof(struct sctp_setadaptation))
4898 return -EINVAL;
4900 len = sizeof(struct sctp_setadaptation);
4902 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4904 if (put_user(len, optlen))
4905 return -EFAULT;
4906 if (copy_to_user(optval, &adaptation, len))
4907 return -EFAULT;
4909 return 0;
4914 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4916 * Applications that wish to use the sendto() system call may wish to
4917 * specify a default set of parameters that would normally be supplied
4918 * through the inclusion of ancillary data. This socket option allows
4919 * such an application to set the default sctp_sndrcvinfo structure.
4922 * The application that wishes to use this socket option simply passes
4923 * in to this call the sctp_sndrcvinfo structure defined in Section
4924 * 5.2.2) The input parameters accepted by this call include
4925 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4926 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4927 * to this call if the caller is using the UDP model.
4929 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4931 static int sctp_getsockopt_default_send_param(struct sock *sk,
4932 int len, char __user *optval,
4933 int __user *optlen)
4935 struct sctp_sndrcvinfo info;
4936 struct sctp_association *asoc;
4937 struct sctp_sock *sp = sctp_sk(sk);
4939 if (len < sizeof(struct sctp_sndrcvinfo))
4940 return -EINVAL;
4942 len = sizeof(struct sctp_sndrcvinfo);
4944 if (copy_from_user(&info, optval, len))
4945 return -EFAULT;
4947 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4948 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4949 return -EINVAL;
4951 if (asoc) {
4952 info.sinfo_stream = asoc->default_stream;
4953 info.sinfo_flags = asoc->default_flags;
4954 info.sinfo_ppid = asoc->default_ppid;
4955 info.sinfo_context = asoc->default_context;
4956 info.sinfo_timetolive = asoc->default_timetolive;
4957 } else {
4958 info.sinfo_stream = sp->default_stream;
4959 info.sinfo_flags = sp->default_flags;
4960 info.sinfo_ppid = sp->default_ppid;
4961 info.sinfo_context = sp->default_context;
4962 info.sinfo_timetolive = sp->default_timetolive;
4965 if (put_user(len, optlen))
4966 return -EFAULT;
4967 if (copy_to_user(optval, &info, len))
4968 return -EFAULT;
4970 return 0;
4975 * 7.1.5 SCTP_NODELAY
4977 * Turn on/off any Nagle-like algorithm. This means that packets are
4978 * generally sent as soon as possible and no unnecessary delays are
4979 * introduced, at the cost of more packets in the network. Expects an
4980 * integer boolean flag.
4983 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4984 char __user *optval, int __user *optlen)
4986 int val;
4988 if (len < sizeof(int))
4989 return -EINVAL;
4991 len = sizeof(int);
4992 val = (sctp_sk(sk)->nodelay == 1);
4993 if (put_user(len, optlen))
4994 return -EFAULT;
4995 if (copy_to_user(optval, &val, len))
4996 return -EFAULT;
4997 return 0;
5002 * 7.1.1 SCTP_RTOINFO
5004 * The protocol parameters used to initialize and bound retransmission
5005 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5006 * and modify these parameters.
5007 * All parameters are time values, in milliseconds. A value of 0, when
5008 * modifying the parameters, indicates that the current value should not
5009 * be changed.
5012 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5013 char __user *optval,
5014 int __user *optlen) {
5015 struct sctp_rtoinfo rtoinfo;
5016 struct sctp_association *asoc;
5018 if (len < sizeof (struct sctp_rtoinfo))
5019 return -EINVAL;
5021 len = sizeof(struct sctp_rtoinfo);
5023 if (copy_from_user(&rtoinfo, optval, len))
5024 return -EFAULT;
5026 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5028 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5029 return -EINVAL;
5031 /* Values corresponding to the specific association. */
5032 if (asoc) {
5033 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5034 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5035 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5036 } else {
5037 /* Values corresponding to the endpoint. */
5038 struct sctp_sock *sp = sctp_sk(sk);
5040 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5041 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5042 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5045 if (put_user(len, optlen))
5046 return -EFAULT;
5048 if (copy_to_user(optval, &rtoinfo, len))
5049 return -EFAULT;
5051 return 0;
5056 * 7.1.2 SCTP_ASSOCINFO
5058 * This option is used to tune the maximum retransmission attempts
5059 * of the association.
5060 * Returns an error if the new association retransmission value is
5061 * greater than the sum of the retransmission value of the peer.
5062 * See [SCTP] for more information.
5065 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5066 char __user *optval,
5067 int __user *optlen)
5070 struct sctp_assocparams assocparams;
5071 struct sctp_association *asoc;
5072 struct list_head *pos;
5073 int cnt = 0;
5075 if (len < sizeof (struct sctp_assocparams))
5076 return -EINVAL;
5078 len = sizeof(struct sctp_assocparams);
5080 if (copy_from_user(&assocparams, optval, len))
5081 return -EFAULT;
5083 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5085 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5086 return -EINVAL;
5088 /* Values correspoinding to the specific association */
5089 if (asoc) {
5090 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5091 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5092 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5093 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5094 * 1000) +
5095 (asoc->cookie_life.tv_usec
5096 / 1000);
5098 list_for_each(pos, &asoc->peer.transport_addr_list) {
5099 cnt ++;
5102 assocparams.sasoc_number_peer_destinations = cnt;
5103 } else {
5104 /* Values corresponding to the endpoint */
5105 struct sctp_sock *sp = sctp_sk(sk);
5107 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5108 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5109 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5110 assocparams.sasoc_cookie_life =
5111 sp->assocparams.sasoc_cookie_life;
5112 assocparams.sasoc_number_peer_destinations =
5113 sp->assocparams.
5114 sasoc_number_peer_destinations;
5117 if (put_user(len, optlen))
5118 return -EFAULT;
5120 if (copy_to_user(optval, &assocparams, len))
5121 return -EFAULT;
5123 return 0;
5127 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5129 * This socket option is a boolean flag which turns on or off mapped V4
5130 * addresses. If this option is turned on and the socket is type
5131 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5132 * If this option is turned off, then no mapping will be done of V4
5133 * addresses and a user will receive both PF_INET6 and PF_INET type
5134 * addresses on the socket.
5136 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5137 char __user *optval, int __user *optlen)
5139 int val;
5140 struct sctp_sock *sp = sctp_sk(sk);
5142 if (len < sizeof(int))
5143 return -EINVAL;
5145 len = sizeof(int);
5146 val = sp->v4mapped;
5147 if (put_user(len, optlen))
5148 return -EFAULT;
5149 if (copy_to_user(optval, &val, len))
5150 return -EFAULT;
5152 return 0;
5156 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5157 * (chapter and verse is quoted at sctp_setsockopt_context())
5159 static int sctp_getsockopt_context(struct sock *sk, int len,
5160 char __user *optval, int __user *optlen)
5162 struct sctp_assoc_value params;
5163 struct sctp_sock *sp;
5164 struct sctp_association *asoc;
5166 if (len < sizeof(struct sctp_assoc_value))
5167 return -EINVAL;
5169 len = sizeof(struct sctp_assoc_value);
5171 if (copy_from_user(&params, optval, len))
5172 return -EFAULT;
5174 sp = sctp_sk(sk);
5176 if (params.assoc_id != 0) {
5177 asoc = sctp_id2assoc(sk, params.assoc_id);
5178 if (!asoc)
5179 return -EINVAL;
5180 params.assoc_value = asoc->default_rcv_context;
5181 } else {
5182 params.assoc_value = sp->default_rcv_context;
5185 if (put_user(len, optlen))
5186 return -EFAULT;
5187 if (copy_to_user(optval, &params, len))
5188 return -EFAULT;
5190 return 0;
5194 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5195 * This option will get or set the maximum size to put in any outgoing
5196 * SCTP DATA chunk. If a message is larger than this size it will be
5197 * fragmented by SCTP into the specified size. Note that the underlying
5198 * SCTP implementation may fragment into smaller sized chunks when the
5199 * PMTU of the underlying association is smaller than the value set by
5200 * the user. The default value for this option is '0' which indicates
5201 * the user is NOT limiting fragmentation and only the PMTU will effect
5202 * SCTP's choice of DATA chunk size. Note also that values set larger
5203 * than the maximum size of an IP datagram will effectively let SCTP
5204 * control fragmentation (i.e. the same as setting this option to 0).
5206 * The following structure is used to access and modify this parameter:
5208 * struct sctp_assoc_value {
5209 * sctp_assoc_t assoc_id;
5210 * uint32_t assoc_value;
5211 * };
5213 * assoc_id: This parameter is ignored for one-to-one style sockets.
5214 * For one-to-many style sockets this parameter indicates which
5215 * association the user is performing an action upon. Note that if
5216 * this field's value is zero then the endpoints default value is
5217 * changed (effecting future associations only).
5218 * assoc_value: This parameter specifies the maximum size in bytes.
5220 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5221 char __user *optval, int __user *optlen)
5223 struct sctp_assoc_value params;
5224 struct sctp_association *asoc;
5226 if (len == sizeof(int)) {
5227 printk(KERN_WARNING
5228 "SCTP: Use of int in maxseg socket option deprecated\n");
5229 printk(KERN_WARNING
5230 "SCTP: Use struct sctp_assoc_value instead\n");
5231 params.assoc_id = 0;
5232 } else if (len >= sizeof(struct sctp_assoc_value)) {
5233 len = sizeof(struct sctp_assoc_value);
5234 if (copy_from_user(&params, optval, sizeof(params)))
5235 return -EFAULT;
5236 } else
5237 return -EINVAL;
5239 asoc = sctp_id2assoc(sk, params.assoc_id);
5240 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5241 return -EINVAL;
5243 if (asoc)
5244 params.assoc_value = asoc->frag_point;
5245 else
5246 params.assoc_value = sctp_sk(sk)->user_frag;
5248 if (put_user(len, optlen))
5249 return -EFAULT;
5250 if (len == sizeof(int)) {
5251 if (copy_to_user(optval, &params.assoc_value, len))
5252 return -EFAULT;
5253 } else {
5254 if (copy_to_user(optval, &params, len))
5255 return -EFAULT;
5258 return 0;
5262 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5263 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5265 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5266 char __user *optval, int __user *optlen)
5268 int val;
5270 if (len < sizeof(int))
5271 return -EINVAL;
5273 len = sizeof(int);
5275 val = sctp_sk(sk)->frag_interleave;
5276 if (put_user(len, optlen))
5277 return -EFAULT;
5278 if (copy_to_user(optval, &val, len))
5279 return -EFAULT;
5281 return 0;
5285 * 7.1.25. Set or Get the sctp partial delivery point
5286 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5288 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5289 char __user *optval,
5290 int __user *optlen)
5292 u32 val;
5294 if (len < sizeof(u32))
5295 return -EINVAL;
5297 len = sizeof(u32);
5299 val = sctp_sk(sk)->pd_point;
5300 if (put_user(len, optlen))
5301 return -EFAULT;
5302 if (copy_to_user(optval, &val, len))
5303 return -EFAULT;
5305 return -ENOTSUPP;
5309 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5310 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5312 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5313 char __user *optval,
5314 int __user *optlen)
5316 struct sctp_assoc_value params;
5317 struct sctp_sock *sp;
5318 struct sctp_association *asoc;
5320 if (len == sizeof(int)) {
5321 printk(KERN_WARNING
5322 "SCTP: Use of int in max_burst socket option deprecated\n");
5323 printk(KERN_WARNING
5324 "SCTP: Use struct sctp_assoc_value instead\n");
5325 params.assoc_id = 0;
5326 } else if (len >= sizeof(struct sctp_assoc_value)) {
5327 len = sizeof(struct sctp_assoc_value);
5328 if (copy_from_user(&params, optval, len))
5329 return -EFAULT;
5330 } else
5331 return -EINVAL;
5333 sp = sctp_sk(sk);
5335 if (params.assoc_id != 0) {
5336 asoc = sctp_id2assoc(sk, params.assoc_id);
5337 if (!asoc)
5338 return -EINVAL;
5339 params.assoc_value = asoc->max_burst;
5340 } else
5341 params.assoc_value = sp->max_burst;
5343 if (len == sizeof(int)) {
5344 if (copy_to_user(optval, &params.assoc_value, len))
5345 return -EFAULT;
5346 } else {
5347 if (copy_to_user(optval, &params, len))
5348 return -EFAULT;
5351 return 0;
5355 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5356 char __user *optval, int __user *optlen)
5358 struct sctp_hmacalgo __user *p = (void __user *)optval;
5359 struct sctp_hmac_algo_param *hmacs;
5360 __u16 data_len = 0;
5361 u32 num_idents;
5363 if (!sctp_auth_enable)
5364 return -EACCES;
5366 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5367 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5369 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5370 return -EINVAL;
5372 len = sizeof(struct sctp_hmacalgo) + data_len;
5373 num_idents = data_len / sizeof(u16);
5375 if (put_user(len, optlen))
5376 return -EFAULT;
5377 if (put_user(num_idents, &p->shmac_num_idents))
5378 return -EFAULT;
5379 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5380 return -EFAULT;
5381 return 0;
5384 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5385 char __user *optval, int __user *optlen)
5387 struct sctp_authkeyid val;
5388 struct sctp_association *asoc;
5390 if (!sctp_auth_enable)
5391 return -EACCES;
5393 if (len < sizeof(struct sctp_authkeyid))
5394 return -EINVAL;
5395 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5396 return -EFAULT;
5398 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5399 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5400 return -EINVAL;
5402 if (asoc)
5403 val.scact_keynumber = asoc->active_key_id;
5404 else
5405 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5407 len = sizeof(struct sctp_authkeyid);
5408 if (put_user(len, optlen))
5409 return -EFAULT;
5410 if (copy_to_user(optval, &val, len))
5411 return -EFAULT;
5413 return 0;
5416 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5417 char __user *optval, int __user *optlen)
5419 struct sctp_authchunks __user *p = (void __user *)optval;
5420 struct sctp_authchunks val;
5421 struct sctp_association *asoc;
5422 struct sctp_chunks_param *ch;
5423 u32 num_chunks = 0;
5424 char __user *to;
5426 if (!sctp_auth_enable)
5427 return -EACCES;
5429 if (len < sizeof(struct sctp_authchunks))
5430 return -EINVAL;
5432 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5433 return -EFAULT;
5435 to = p->gauth_chunks;
5436 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5437 if (!asoc)
5438 return -EINVAL;
5440 ch = asoc->peer.peer_chunks;
5441 if (!ch)
5442 goto num;
5444 /* See if the user provided enough room for all the data */
5445 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5446 if (len < num_chunks)
5447 return -EINVAL;
5449 if (copy_to_user(to, ch->chunks, num_chunks))
5450 return -EFAULT;
5451 num:
5452 len = sizeof(struct sctp_authchunks) + num_chunks;
5453 if (put_user(len, optlen)) return -EFAULT;
5454 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5455 return -EFAULT;
5456 return 0;
5459 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5460 char __user *optval, int __user *optlen)
5462 struct sctp_authchunks __user *p = (void __user *)optval;
5463 struct sctp_authchunks val;
5464 struct sctp_association *asoc;
5465 struct sctp_chunks_param *ch;
5466 u32 num_chunks = 0;
5467 char __user *to;
5469 if (!sctp_auth_enable)
5470 return -EACCES;
5472 if (len < sizeof(struct sctp_authchunks))
5473 return -EINVAL;
5475 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5476 return -EFAULT;
5478 to = p->gauth_chunks;
5479 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5480 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5481 return -EINVAL;
5483 if (asoc)
5484 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5485 else
5486 ch = sctp_sk(sk)->ep->auth_chunk_list;
5488 if (!ch)
5489 goto num;
5491 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5492 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5493 return -EINVAL;
5495 if (copy_to_user(to, ch->chunks, num_chunks))
5496 return -EFAULT;
5497 num:
5498 len = sizeof(struct sctp_authchunks) + num_chunks;
5499 if (put_user(len, optlen))
5500 return -EFAULT;
5501 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5502 return -EFAULT;
5504 return 0;
5508 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5509 * This option gets the current number of associations that are attached
5510 * to a one-to-many style socket. The option value is an uint32_t.
5512 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5513 char __user *optval, int __user *optlen)
5515 struct sctp_sock *sp = sctp_sk(sk);
5516 struct sctp_association *asoc;
5517 u32 val = 0;
5519 if (sctp_style(sk, TCP))
5520 return -EOPNOTSUPP;
5522 if (len < sizeof(u32))
5523 return -EINVAL;
5525 len = sizeof(u32);
5527 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5528 val++;
5531 if (put_user(len, optlen))
5532 return -EFAULT;
5533 if (copy_to_user(optval, &val, len))
5534 return -EFAULT;
5536 return 0;
5539 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5540 char __user *optval, int __user *optlen)
5542 int retval = 0;
5543 int len;
5545 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5546 sk, optname);
5548 /* I can hardly begin to describe how wrong this is. This is
5549 * so broken as to be worse than useless. The API draft
5550 * REALLY is NOT helpful here... I am not convinced that the
5551 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5552 * are at all well-founded.
5554 if (level != SOL_SCTP) {
5555 struct sctp_af *af = sctp_sk(sk)->pf->af;
5557 retval = af->getsockopt(sk, level, optname, optval, optlen);
5558 return retval;
5561 if (get_user(len, optlen))
5562 return -EFAULT;
5564 sctp_lock_sock(sk);
5566 switch (optname) {
5567 case SCTP_STATUS:
5568 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5569 break;
5570 case SCTP_DISABLE_FRAGMENTS:
5571 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5572 optlen);
5573 break;
5574 case SCTP_EVENTS:
5575 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5576 break;
5577 case SCTP_AUTOCLOSE:
5578 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5579 break;
5580 case SCTP_SOCKOPT_PEELOFF:
5581 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5582 break;
5583 case SCTP_PEER_ADDR_PARAMS:
5584 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5585 optlen);
5586 break;
5587 case SCTP_DELAYED_ACK:
5588 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5589 optlen);
5590 break;
5591 case SCTP_INITMSG:
5592 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5593 break;
5594 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5595 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5596 optlen);
5597 break;
5598 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5599 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5600 optlen);
5601 break;
5602 case SCTP_GET_PEER_ADDRS_OLD:
5603 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5604 optlen);
5605 break;
5606 case SCTP_GET_LOCAL_ADDRS_OLD:
5607 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5608 optlen);
5609 break;
5610 case SCTP_GET_PEER_ADDRS:
5611 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5612 optlen);
5613 break;
5614 case SCTP_GET_LOCAL_ADDRS:
5615 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5616 optlen);
5617 break;
5618 case SCTP_SOCKOPT_CONNECTX3:
5619 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5620 break;
5621 case SCTP_DEFAULT_SEND_PARAM:
5622 retval = sctp_getsockopt_default_send_param(sk, len,
5623 optval, optlen);
5624 break;
5625 case SCTP_PRIMARY_ADDR:
5626 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5627 break;
5628 case SCTP_NODELAY:
5629 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5630 break;
5631 case SCTP_RTOINFO:
5632 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5633 break;
5634 case SCTP_ASSOCINFO:
5635 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5636 break;
5637 case SCTP_I_WANT_MAPPED_V4_ADDR:
5638 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5639 break;
5640 case SCTP_MAXSEG:
5641 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5642 break;
5643 case SCTP_GET_PEER_ADDR_INFO:
5644 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5645 optlen);
5646 break;
5647 case SCTP_ADAPTATION_LAYER:
5648 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5649 optlen);
5650 break;
5651 case SCTP_CONTEXT:
5652 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5653 break;
5654 case SCTP_FRAGMENT_INTERLEAVE:
5655 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5656 optlen);
5657 break;
5658 case SCTP_PARTIAL_DELIVERY_POINT:
5659 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5660 optlen);
5661 break;
5662 case SCTP_MAX_BURST:
5663 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5664 break;
5665 case SCTP_AUTH_KEY:
5666 case SCTP_AUTH_CHUNK:
5667 case SCTP_AUTH_DELETE_KEY:
5668 retval = -EOPNOTSUPP;
5669 break;
5670 case SCTP_HMAC_IDENT:
5671 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5672 break;
5673 case SCTP_AUTH_ACTIVE_KEY:
5674 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5675 break;
5676 case SCTP_PEER_AUTH_CHUNKS:
5677 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5678 optlen);
5679 break;
5680 case SCTP_LOCAL_AUTH_CHUNKS:
5681 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5682 optlen);
5683 break;
5684 case SCTP_GET_ASSOC_NUMBER:
5685 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5686 break;
5687 default:
5688 retval = -ENOPROTOOPT;
5689 break;
5692 sctp_release_sock(sk);
5693 return retval;
5696 static void sctp_hash(struct sock *sk)
5698 /* STUB */
5701 static void sctp_unhash(struct sock *sk)
5703 /* STUB */
5706 /* Check if port is acceptable. Possibly find first available port.
5708 * The port hash table (contained in the 'global' SCTP protocol storage
5709 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5710 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5711 * list (the list number is the port number hashed out, so as you
5712 * would expect from a hash function, all the ports in a given list have
5713 * such a number that hashes out to the same list number; you were
5714 * expecting that, right?); so each list has a set of ports, with a
5715 * link to the socket (struct sock) that uses it, the port number and
5716 * a fastreuse flag (FIXME: NPI ipg).
5718 static struct sctp_bind_bucket *sctp_bucket_create(
5719 struct sctp_bind_hashbucket *head, unsigned short snum);
5721 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5723 struct sctp_bind_hashbucket *head; /* hash list */
5724 struct sctp_bind_bucket *pp; /* hash list port iterator */
5725 struct hlist_node *node;
5726 unsigned short snum;
5727 int ret;
5729 snum = ntohs(addr->v4.sin_port);
5731 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5732 sctp_local_bh_disable();
5734 if (snum == 0) {
5735 /* Search for an available port. */
5736 int low, high, remaining, index;
5737 unsigned int rover;
5739 inet_get_local_port_range(&low, &high);
5740 remaining = (high - low) + 1;
5741 rover = net_random() % remaining + low;
5743 do {
5744 rover++;
5745 if ((rover < low) || (rover > high))
5746 rover = low;
5747 index = sctp_phashfn(rover);
5748 head = &sctp_port_hashtable[index];
5749 sctp_spin_lock(&head->lock);
5750 sctp_for_each_hentry(pp, node, &head->chain)
5751 if (pp->port == rover)
5752 goto next;
5753 break;
5754 next:
5755 sctp_spin_unlock(&head->lock);
5756 } while (--remaining > 0);
5758 /* Exhausted local port range during search? */
5759 ret = 1;
5760 if (remaining <= 0)
5761 goto fail;
5763 /* OK, here is the one we will use. HEAD (the port
5764 * hash table list entry) is non-NULL and we hold it's
5765 * mutex.
5767 snum = rover;
5768 } else {
5769 /* We are given an specific port number; we verify
5770 * that it is not being used. If it is used, we will
5771 * exahust the search in the hash list corresponding
5772 * to the port number (snum) - we detect that with the
5773 * port iterator, pp being NULL.
5775 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5776 sctp_spin_lock(&head->lock);
5777 sctp_for_each_hentry(pp, node, &head->chain) {
5778 if (pp->port == snum)
5779 goto pp_found;
5782 pp = NULL;
5783 goto pp_not_found;
5784 pp_found:
5785 if (!hlist_empty(&pp->owner)) {
5786 /* We had a port hash table hit - there is an
5787 * available port (pp != NULL) and it is being
5788 * used by other socket (pp->owner not empty); that other
5789 * socket is going to be sk2.
5791 int reuse = sk->sk_reuse;
5792 struct sock *sk2;
5793 struct hlist_node *node;
5795 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5796 if (pp->fastreuse && sk->sk_reuse &&
5797 sk->sk_state != SCTP_SS_LISTENING)
5798 goto success;
5800 /* Run through the list of sockets bound to the port
5801 * (pp->port) [via the pointers bind_next and
5802 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5803 * we get the endpoint they describe and run through
5804 * the endpoint's list of IP (v4 or v6) addresses,
5805 * comparing each of the addresses with the address of
5806 * the socket sk. If we find a match, then that means
5807 * that this port/socket (sk) combination are already
5808 * in an endpoint.
5810 sk_for_each_bound(sk2, node, &pp->owner) {
5811 struct sctp_endpoint *ep2;
5812 ep2 = sctp_sk(sk2)->ep;
5814 if (sk == sk2 ||
5815 (reuse && sk2->sk_reuse &&
5816 sk2->sk_state != SCTP_SS_LISTENING))
5817 continue;
5819 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5820 sctp_sk(sk2), sctp_sk(sk))) {
5821 ret = (long)sk2;
5822 goto fail_unlock;
5825 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5827 pp_not_found:
5828 /* If there was a hash table miss, create a new port. */
5829 ret = 1;
5830 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5831 goto fail_unlock;
5833 /* In either case (hit or miss), make sure fastreuse is 1 only
5834 * if sk->sk_reuse is too (that is, if the caller requested
5835 * SO_REUSEADDR on this socket -sk-).
5837 if (hlist_empty(&pp->owner)) {
5838 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5839 pp->fastreuse = 1;
5840 else
5841 pp->fastreuse = 0;
5842 } else if (pp->fastreuse &&
5843 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5844 pp->fastreuse = 0;
5846 /* We are set, so fill up all the data in the hash table
5847 * entry, tie the socket list information with the rest of the
5848 * sockets FIXME: Blurry, NPI (ipg).
5850 success:
5851 if (!sctp_sk(sk)->bind_hash) {
5852 inet_sk(sk)->num = snum;
5853 sk_add_bind_node(sk, &pp->owner);
5854 sctp_sk(sk)->bind_hash = pp;
5856 ret = 0;
5858 fail_unlock:
5859 sctp_spin_unlock(&head->lock);
5861 fail:
5862 sctp_local_bh_enable();
5863 return ret;
5866 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5867 * port is requested.
5869 static int sctp_get_port(struct sock *sk, unsigned short snum)
5871 long ret;
5872 union sctp_addr addr;
5873 struct sctp_af *af = sctp_sk(sk)->pf->af;
5875 /* Set up a dummy address struct from the sk. */
5876 af->from_sk(&addr, sk);
5877 addr.v4.sin_port = htons(snum);
5879 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5880 ret = sctp_get_port_local(sk, &addr);
5882 return (ret ? 1 : 0);
5886 * Move a socket to LISTENING state.
5888 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5890 struct sctp_sock *sp = sctp_sk(sk);
5891 struct sctp_endpoint *ep = sp->ep;
5892 struct crypto_hash *tfm = NULL;
5894 /* Allocate HMAC for generating cookie. */
5895 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5896 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5897 if (IS_ERR(tfm)) {
5898 if (net_ratelimit()) {
5899 printk(KERN_INFO
5900 "SCTP: failed to load transform for %s: %ld\n",
5901 sctp_hmac_alg, PTR_ERR(tfm));
5903 return -ENOSYS;
5905 sctp_sk(sk)->hmac = tfm;
5909 * If a bind() or sctp_bindx() is not called prior to a listen()
5910 * call that allows new associations to be accepted, the system
5911 * picks an ephemeral port and will choose an address set equivalent
5912 * to binding with a wildcard address.
5914 * This is not currently spelled out in the SCTP sockets
5915 * extensions draft, but follows the practice as seen in TCP
5916 * sockets.
5919 sk->sk_state = SCTP_SS_LISTENING;
5920 if (!ep->base.bind_addr.port) {
5921 if (sctp_autobind(sk))
5922 return -EAGAIN;
5923 } else {
5924 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5925 sk->sk_state = SCTP_SS_CLOSED;
5926 return -EADDRINUSE;
5930 sk->sk_max_ack_backlog = backlog;
5931 sctp_hash_endpoint(ep);
5932 return 0;
5936 * 4.1.3 / 5.1.3 listen()
5938 * By default, new associations are not accepted for UDP style sockets.
5939 * An application uses listen() to mark a socket as being able to
5940 * accept new associations.
5942 * On TCP style sockets, applications use listen() to ready the SCTP
5943 * endpoint for accepting inbound associations.
5945 * On both types of endpoints a backlog of '0' disables listening.
5947 * Move a socket to LISTENING state.
5949 int sctp_inet_listen(struct socket *sock, int backlog)
5951 struct sock *sk = sock->sk;
5952 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5953 int err = -EINVAL;
5955 if (unlikely(backlog < 0))
5956 return err;
5958 sctp_lock_sock(sk);
5960 /* Peeled-off sockets are not allowed to listen(). */
5961 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5962 goto out;
5964 if (sock->state != SS_UNCONNECTED)
5965 goto out;
5967 /* If backlog is zero, disable listening. */
5968 if (!backlog) {
5969 if (sctp_sstate(sk, CLOSED))
5970 goto out;
5972 err = 0;
5973 sctp_unhash_endpoint(ep);
5974 sk->sk_state = SCTP_SS_CLOSED;
5975 if (sk->sk_reuse)
5976 sctp_sk(sk)->bind_hash->fastreuse = 1;
5977 goto out;
5980 /* If we are already listening, just update the backlog */
5981 if (sctp_sstate(sk, LISTENING))
5982 sk->sk_max_ack_backlog = backlog;
5983 else {
5984 err = sctp_listen_start(sk, backlog);
5985 if (err)
5986 goto out;
5989 err = 0;
5990 out:
5991 sctp_release_sock(sk);
5992 return err;
5996 * This function is done by modeling the current datagram_poll() and the
5997 * tcp_poll(). Note that, based on these implementations, we don't
5998 * lock the socket in this function, even though it seems that,
5999 * ideally, locking or some other mechanisms can be used to ensure
6000 * the integrity of the counters (sndbuf and wmem_alloc) used
6001 * in this place. We assume that we don't need locks either until proven
6002 * otherwise.
6004 * Another thing to note is that we include the Async I/O support
6005 * here, again, by modeling the current TCP/UDP code. We don't have
6006 * a good way to test with it yet.
6008 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6010 struct sock *sk = sock->sk;
6011 struct sctp_sock *sp = sctp_sk(sk);
6012 unsigned int mask;
6014 poll_wait(file, sk->sk_sleep, wait);
6016 /* A TCP-style listening socket becomes readable when the accept queue
6017 * is not empty.
6019 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6020 return (!list_empty(&sp->ep->asocs)) ?
6021 (POLLIN | POLLRDNORM) : 0;
6023 mask = 0;
6025 /* Is there any exceptional events? */
6026 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6027 mask |= POLLERR;
6028 if (sk->sk_shutdown & RCV_SHUTDOWN)
6029 mask |= POLLRDHUP;
6030 if (sk->sk_shutdown == SHUTDOWN_MASK)
6031 mask |= POLLHUP;
6033 /* Is it readable? Reconsider this code with TCP-style support. */
6034 if (!skb_queue_empty(&sk->sk_receive_queue) ||
6035 (sk->sk_shutdown & RCV_SHUTDOWN))
6036 mask |= POLLIN | POLLRDNORM;
6038 /* The association is either gone or not ready. */
6039 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6040 return mask;
6042 /* Is it writable? */
6043 if (sctp_writeable(sk)) {
6044 mask |= POLLOUT | POLLWRNORM;
6045 } else {
6046 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6048 * Since the socket is not locked, the buffer
6049 * might be made available after the writeable check and
6050 * before the bit is set. This could cause a lost I/O
6051 * signal. tcp_poll() has a race breaker for this race
6052 * condition. Based on their implementation, we put
6053 * in the following code to cover it as well.
6055 if (sctp_writeable(sk))
6056 mask |= POLLOUT | POLLWRNORM;
6058 return mask;
6061 /********************************************************************
6062 * 2nd Level Abstractions
6063 ********************************************************************/
6065 static struct sctp_bind_bucket *sctp_bucket_create(
6066 struct sctp_bind_hashbucket *head, unsigned short snum)
6068 struct sctp_bind_bucket *pp;
6070 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6071 if (pp) {
6072 SCTP_DBG_OBJCNT_INC(bind_bucket);
6073 pp->port = snum;
6074 pp->fastreuse = 0;
6075 INIT_HLIST_HEAD(&pp->owner);
6076 hlist_add_head(&pp->node, &head->chain);
6078 return pp;
6081 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6082 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6084 if (pp && hlist_empty(&pp->owner)) {
6085 __hlist_del(&pp->node);
6086 kmem_cache_free(sctp_bucket_cachep, pp);
6087 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6091 /* Release this socket's reference to a local port. */
6092 static inline void __sctp_put_port(struct sock *sk)
6094 struct sctp_bind_hashbucket *head =
6095 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
6096 struct sctp_bind_bucket *pp;
6098 sctp_spin_lock(&head->lock);
6099 pp = sctp_sk(sk)->bind_hash;
6100 __sk_del_bind_node(sk);
6101 sctp_sk(sk)->bind_hash = NULL;
6102 inet_sk(sk)->num = 0;
6103 sctp_bucket_destroy(pp);
6104 sctp_spin_unlock(&head->lock);
6107 void sctp_put_port(struct sock *sk)
6109 sctp_local_bh_disable();
6110 __sctp_put_port(sk);
6111 sctp_local_bh_enable();
6115 * The system picks an ephemeral port and choose an address set equivalent
6116 * to binding with a wildcard address.
6117 * One of those addresses will be the primary address for the association.
6118 * This automatically enables the multihoming capability of SCTP.
6120 static int sctp_autobind(struct sock *sk)
6122 union sctp_addr autoaddr;
6123 struct sctp_af *af;
6124 __be16 port;
6126 /* Initialize a local sockaddr structure to INADDR_ANY. */
6127 af = sctp_sk(sk)->pf->af;
6129 port = htons(inet_sk(sk)->num);
6130 af->inaddr_any(&autoaddr, port);
6132 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6135 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6137 * From RFC 2292
6138 * 4.2 The cmsghdr Structure *
6140 * When ancillary data is sent or received, any number of ancillary data
6141 * objects can be specified by the msg_control and msg_controllen members of
6142 * the msghdr structure, because each object is preceded by
6143 * a cmsghdr structure defining the object's length (the cmsg_len member).
6144 * Historically Berkeley-derived implementations have passed only one object
6145 * at a time, but this API allows multiple objects to be
6146 * passed in a single call to sendmsg() or recvmsg(). The following example
6147 * shows two ancillary data objects in a control buffer.
6149 * |<--------------------------- msg_controllen -------------------------->|
6150 * | |
6152 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6154 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6155 * | | |
6157 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6159 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6160 * | | | | |
6162 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6163 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6165 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6167 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6171 * msg_control
6172 * points here
6174 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6175 sctp_cmsgs_t *cmsgs)
6177 struct cmsghdr *cmsg;
6178 struct msghdr *my_msg = (struct msghdr *)msg;
6180 for (cmsg = CMSG_FIRSTHDR(msg);
6181 cmsg != NULL;
6182 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6183 if (!CMSG_OK(my_msg, cmsg))
6184 return -EINVAL;
6186 /* Should we parse this header or ignore? */
6187 if (cmsg->cmsg_level != IPPROTO_SCTP)
6188 continue;
6190 /* Strictly check lengths following example in SCM code. */
6191 switch (cmsg->cmsg_type) {
6192 case SCTP_INIT:
6193 /* SCTP Socket API Extension
6194 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6196 * This cmsghdr structure provides information for
6197 * initializing new SCTP associations with sendmsg().
6198 * The SCTP_INITMSG socket option uses this same data
6199 * structure. This structure is not used for
6200 * recvmsg().
6202 * cmsg_level cmsg_type cmsg_data[]
6203 * ------------ ------------ ----------------------
6204 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6206 if (cmsg->cmsg_len !=
6207 CMSG_LEN(sizeof(struct sctp_initmsg)))
6208 return -EINVAL;
6209 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6210 break;
6212 case SCTP_SNDRCV:
6213 /* SCTP Socket API Extension
6214 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6216 * This cmsghdr structure specifies SCTP options for
6217 * sendmsg() and describes SCTP header information
6218 * about a received message through recvmsg().
6220 * cmsg_level cmsg_type cmsg_data[]
6221 * ------------ ------------ ----------------------
6222 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6224 if (cmsg->cmsg_len !=
6225 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6226 return -EINVAL;
6228 cmsgs->info =
6229 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6231 /* Minimally, validate the sinfo_flags. */
6232 if (cmsgs->info->sinfo_flags &
6233 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6234 SCTP_ABORT | SCTP_EOF))
6235 return -EINVAL;
6236 break;
6238 default:
6239 return -EINVAL;
6242 return 0;
6246 * Wait for a packet..
6247 * Note: This function is the same function as in core/datagram.c
6248 * with a few modifications to make lksctp work.
6250 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6252 int error;
6253 DEFINE_WAIT(wait);
6255 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6257 /* Socket errors? */
6258 error = sock_error(sk);
6259 if (error)
6260 goto out;
6262 if (!skb_queue_empty(&sk->sk_receive_queue))
6263 goto ready;
6265 /* Socket shut down? */
6266 if (sk->sk_shutdown & RCV_SHUTDOWN)
6267 goto out;
6269 /* Sequenced packets can come disconnected. If so we report the
6270 * problem.
6272 error = -ENOTCONN;
6274 /* Is there a good reason to think that we may receive some data? */
6275 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6276 goto out;
6278 /* Handle signals. */
6279 if (signal_pending(current))
6280 goto interrupted;
6282 /* Let another process have a go. Since we are going to sleep
6283 * anyway. Note: This may cause odd behaviors if the message
6284 * does not fit in the user's buffer, but this seems to be the
6285 * only way to honor MSG_DONTWAIT realistically.
6287 sctp_release_sock(sk);
6288 *timeo_p = schedule_timeout(*timeo_p);
6289 sctp_lock_sock(sk);
6291 ready:
6292 finish_wait(sk->sk_sleep, &wait);
6293 return 0;
6295 interrupted:
6296 error = sock_intr_errno(*timeo_p);
6298 out:
6299 finish_wait(sk->sk_sleep, &wait);
6300 *err = error;
6301 return error;
6304 /* Receive a datagram.
6305 * Note: This is pretty much the same routine as in core/datagram.c
6306 * with a few changes to make lksctp work.
6308 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6309 int noblock, int *err)
6311 int error;
6312 struct sk_buff *skb;
6313 long timeo;
6315 timeo = sock_rcvtimeo(sk, noblock);
6317 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6318 timeo, MAX_SCHEDULE_TIMEOUT);
6320 do {
6321 /* Again only user level code calls this function,
6322 * so nothing interrupt level
6323 * will suddenly eat the receive_queue.
6325 * Look at current nfs client by the way...
6326 * However, this function was corrent in any case. 8)
6328 if (flags & MSG_PEEK) {
6329 spin_lock_bh(&sk->sk_receive_queue.lock);
6330 skb = skb_peek(&sk->sk_receive_queue);
6331 if (skb)
6332 atomic_inc(&skb->users);
6333 spin_unlock_bh(&sk->sk_receive_queue.lock);
6334 } else {
6335 skb = skb_dequeue(&sk->sk_receive_queue);
6338 if (skb)
6339 return skb;
6341 /* Caller is allowed not to check sk->sk_err before calling. */
6342 error = sock_error(sk);
6343 if (error)
6344 goto no_packet;
6346 if (sk->sk_shutdown & RCV_SHUTDOWN)
6347 break;
6349 /* User doesn't want to wait. */
6350 error = -EAGAIN;
6351 if (!timeo)
6352 goto no_packet;
6353 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6355 return NULL;
6357 no_packet:
6358 *err = error;
6359 return NULL;
6362 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6363 static void __sctp_write_space(struct sctp_association *asoc)
6365 struct sock *sk = asoc->base.sk;
6366 struct socket *sock = sk->sk_socket;
6368 if ((sctp_wspace(asoc) > 0) && sock) {
6369 if (waitqueue_active(&asoc->wait))
6370 wake_up_interruptible(&asoc->wait);
6372 if (sctp_writeable(sk)) {
6373 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6374 wake_up_interruptible(sk->sk_sleep);
6376 /* Note that we try to include the Async I/O support
6377 * here by modeling from the current TCP/UDP code.
6378 * We have not tested with it yet.
6380 if (sock->fasync_list &&
6381 !(sk->sk_shutdown & SEND_SHUTDOWN))
6382 sock_wake_async(sock,
6383 SOCK_WAKE_SPACE, POLL_OUT);
6388 /* Do accounting for the sndbuf space.
6389 * Decrement the used sndbuf space of the corresponding association by the
6390 * data size which was just transmitted(freed).
6392 static void sctp_wfree(struct sk_buff *skb)
6394 struct sctp_association *asoc;
6395 struct sctp_chunk *chunk;
6396 struct sock *sk;
6398 /* Get the saved chunk pointer. */
6399 chunk = *((struct sctp_chunk **)(skb->cb));
6400 asoc = chunk->asoc;
6401 sk = asoc->base.sk;
6402 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6403 sizeof(struct sk_buff) +
6404 sizeof(struct sctp_chunk);
6406 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6409 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6411 sk->sk_wmem_queued -= skb->truesize;
6412 sk_mem_uncharge(sk, skb->truesize);
6414 sock_wfree(skb);
6415 __sctp_write_space(asoc);
6417 sctp_association_put(asoc);
6420 /* Do accounting for the receive space on the socket.
6421 * Accounting for the association is done in ulpevent.c
6422 * We set this as a destructor for the cloned data skbs so that
6423 * accounting is done at the correct time.
6425 void sctp_sock_rfree(struct sk_buff *skb)
6427 struct sock *sk = skb->sk;
6428 struct sctp_ulpevent *event = sctp_skb2event(skb);
6430 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6433 * Mimic the behavior of sock_rfree
6435 sk_mem_uncharge(sk, event->rmem_len);
6439 /* Helper function to wait for space in the sndbuf. */
6440 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6441 size_t msg_len)
6443 struct sock *sk = asoc->base.sk;
6444 int err = 0;
6445 long current_timeo = *timeo_p;
6446 DEFINE_WAIT(wait);
6448 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6449 asoc, (long)(*timeo_p), msg_len);
6451 /* Increment the association's refcnt. */
6452 sctp_association_hold(asoc);
6454 /* Wait on the association specific sndbuf space. */
6455 for (;;) {
6456 prepare_to_wait_exclusive(&asoc->wait, &wait,
6457 TASK_INTERRUPTIBLE);
6458 if (!*timeo_p)
6459 goto do_nonblock;
6460 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6461 asoc->base.dead)
6462 goto do_error;
6463 if (signal_pending(current))
6464 goto do_interrupted;
6465 if (msg_len <= sctp_wspace(asoc))
6466 break;
6468 /* Let another process have a go. Since we are going
6469 * to sleep anyway.
6471 sctp_release_sock(sk);
6472 current_timeo = schedule_timeout(current_timeo);
6473 BUG_ON(sk != asoc->base.sk);
6474 sctp_lock_sock(sk);
6476 *timeo_p = current_timeo;
6479 out:
6480 finish_wait(&asoc->wait, &wait);
6482 /* Release the association's refcnt. */
6483 sctp_association_put(asoc);
6485 return err;
6487 do_error:
6488 err = -EPIPE;
6489 goto out;
6491 do_interrupted:
6492 err = sock_intr_errno(*timeo_p);
6493 goto out;
6495 do_nonblock:
6496 err = -EAGAIN;
6497 goto out;
6500 /* If socket sndbuf has changed, wake up all per association waiters. */
6501 void sctp_write_space(struct sock *sk)
6503 struct sctp_association *asoc;
6505 /* Wake up the tasks in each wait queue. */
6506 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6507 __sctp_write_space(asoc);
6511 /* Is there any sndbuf space available on the socket?
6513 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6514 * associations on the same socket. For a UDP-style socket with
6515 * multiple associations, it is possible for it to be "unwriteable"
6516 * prematurely. I assume that this is acceptable because
6517 * a premature "unwriteable" is better than an accidental "writeable" which
6518 * would cause an unwanted block under certain circumstances. For the 1-1
6519 * UDP-style sockets or TCP-style sockets, this code should work.
6520 * - Daisy
6522 static int sctp_writeable(struct sock *sk)
6524 int amt = 0;
6526 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6527 if (amt < 0)
6528 amt = 0;
6529 return amt;
6532 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6533 * returns immediately with EINPROGRESS.
6535 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6537 struct sock *sk = asoc->base.sk;
6538 int err = 0;
6539 long current_timeo = *timeo_p;
6540 DEFINE_WAIT(wait);
6542 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6543 (long)(*timeo_p));
6545 /* Increment the association's refcnt. */
6546 sctp_association_hold(asoc);
6548 for (;;) {
6549 prepare_to_wait_exclusive(&asoc->wait, &wait,
6550 TASK_INTERRUPTIBLE);
6551 if (!*timeo_p)
6552 goto do_nonblock;
6553 if (sk->sk_shutdown & RCV_SHUTDOWN)
6554 break;
6555 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6556 asoc->base.dead)
6557 goto do_error;
6558 if (signal_pending(current))
6559 goto do_interrupted;
6561 if (sctp_state(asoc, ESTABLISHED))
6562 break;
6564 /* Let another process have a go. Since we are going
6565 * to sleep anyway.
6567 sctp_release_sock(sk);
6568 current_timeo = schedule_timeout(current_timeo);
6569 sctp_lock_sock(sk);
6571 *timeo_p = current_timeo;
6574 out:
6575 finish_wait(&asoc->wait, &wait);
6577 /* Release the association's refcnt. */
6578 sctp_association_put(asoc);
6580 return err;
6582 do_error:
6583 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6584 err = -ETIMEDOUT;
6585 else
6586 err = -ECONNREFUSED;
6587 goto out;
6589 do_interrupted:
6590 err = sock_intr_errno(*timeo_p);
6591 goto out;
6593 do_nonblock:
6594 err = -EINPROGRESS;
6595 goto out;
6598 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6600 struct sctp_endpoint *ep;
6601 int err = 0;
6602 DEFINE_WAIT(wait);
6604 ep = sctp_sk(sk)->ep;
6607 for (;;) {
6608 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6609 TASK_INTERRUPTIBLE);
6611 if (list_empty(&ep->asocs)) {
6612 sctp_release_sock(sk);
6613 timeo = schedule_timeout(timeo);
6614 sctp_lock_sock(sk);
6617 err = -EINVAL;
6618 if (!sctp_sstate(sk, LISTENING))
6619 break;
6621 err = 0;
6622 if (!list_empty(&ep->asocs))
6623 break;
6625 err = sock_intr_errno(timeo);
6626 if (signal_pending(current))
6627 break;
6629 err = -EAGAIN;
6630 if (!timeo)
6631 break;
6634 finish_wait(sk->sk_sleep, &wait);
6636 return err;
6639 static void sctp_wait_for_close(struct sock *sk, long timeout)
6641 DEFINE_WAIT(wait);
6643 do {
6644 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6645 if (list_empty(&sctp_sk(sk)->ep->asocs))
6646 break;
6647 sctp_release_sock(sk);
6648 timeout = schedule_timeout(timeout);
6649 sctp_lock_sock(sk);
6650 } while (!signal_pending(current) && timeout);
6652 finish_wait(sk->sk_sleep, &wait);
6655 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6657 struct sk_buff *frag;
6659 if (!skb->data_len)
6660 goto done;
6662 /* Don't forget the fragments. */
6663 skb_walk_frags(skb, frag)
6664 sctp_skb_set_owner_r_frag(frag, sk);
6666 done:
6667 sctp_skb_set_owner_r(skb, sk);
6670 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6671 struct sctp_association *asoc)
6673 struct inet_sock *inet = inet_sk(sk);
6674 struct inet_sock *newinet = inet_sk(newsk);
6676 newsk->sk_type = sk->sk_type;
6677 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6678 newsk->sk_flags = sk->sk_flags;
6679 newsk->sk_no_check = sk->sk_no_check;
6680 newsk->sk_reuse = sk->sk_reuse;
6682 newsk->sk_shutdown = sk->sk_shutdown;
6683 newsk->sk_destruct = inet_sock_destruct;
6684 newsk->sk_family = sk->sk_family;
6685 newsk->sk_protocol = IPPROTO_SCTP;
6686 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6687 newsk->sk_sndbuf = sk->sk_sndbuf;
6688 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6689 newsk->sk_lingertime = sk->sk_lingertime;
6690 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6691 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6693 newinet = inet_sk(newsk);
6695 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6696 * getsockname() and getpeername()
6698 newinet->sport = inet->sport;
6699 newinet->saddr = inet->saddr;
6700 newinet->rcv_saddr = inet->rcv_saddr;
6701 newinet->dport = htons(asoc->peer.port);
6702 newinet->pmtudisc = inet->pmtudisc;
6703 newinet->id = asoc->next_tsn ^ jiffies;
6705 newinet->uc_ttl = inet->uc_ttl;
6706 newinet->mc_loop = 1;
6707 newinet->mc_ttl = 1;
6708 newinet->mc_index = 0;
6709 newinet->mc_list = NULL;
6712 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6713 * and its messages to the newsk.
6715 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6716 struct sctp_association *assoc,
6717 sctp_socket_type_t type)
6719 struct sctp_sock *oldsp = sctp_sk(oldsk);
6720 struct sctp_sock *newsp = sctp_sk(newsk);
6721 struct sctp_bind_bucket *pp; /* hash list port iterator */
6722 struct sctp_endpoint *newep = newsp->ep;
6723 struct sk_buff *skb, *tmp;
6724 struct sctp_ulpevent *event;
6725 struct sctp_bind_hashbucket *head;
6727 /* Migrate socket buffer sizes and all the socket level options to the
6728 * new socket.
6730 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6731 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6732 /* Brute force copy old sctp opt. */
6733 inet_sk_copy_descendant(newsk, oldsk);
6735 /* Restore the ep value that was overwritten with the above structure
6736 * copy.
6738 newsp->ep = newep;
6739 newsp->hmac = NULL;
6741 /* Hook this new socket in to the bind_hash list. */
6742 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6743 sctp_local_bh_disable();
6744 sctp_spin_lock(&head->lock);
6745 pp = sctp_sk(oldsk)->bind_hash;
6746 sk_add_bind_node(newsk, &pp->owner);
6747 sctp_sk(newsk)->bind_hash = pp;
6748 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6749 sctp_spin_unlock(&head->lock);
6750 sctp_local_bh_enable();
6752 /* Copy the bind_addr list from the original endpoint to the new
6753 * endpoint so that we can handle restarts properly
6755 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6756 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6758 /* Move any messages in the old socket's receive queue that are for the
6759 * peeled off association to the new socket's receive queue.
6761 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6762 event = sctp_skb2event(skb);
6763 if (event->asoc == assoc) {
6764 __skb_unlink(skb, &oldsk->sk_receive_queue);
6765 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6766 sctp_skb_set_owner_r_frag(skb, newsk);
6770 /* Clean up any messages pending delivery due to partial
6771 * delivery. Three cases:
6772 * 1) No partial deliver; no work.
6773 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6774 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6776 skb_queue_head_init(&newsp->pd_lobby);
6777 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6779 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6780 struct sk_buff_head *queue;
6782 /* Decide which queue to move pd_lobby skbs to. */
6783 if (assoc->ulpq.pd_mode) {
6784 queue = &newsp->pd_lobby;
6785 } else
6786 queue = &newsk->sk_receive_queue;
6788 /* Walk through the pd_lobby, looking for skbs that
6789 * need moved to the new socket.
6791 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6792 event = sctp_skb2event(skb);
6793 if (event->asoc == assoc) {
6794 __skb_unlink(skb, &oldsp->pd_lobby);
6795 __skb_queue_tail(queue, skb);
6796 sctp_skb_set_owner_r_frag(skb, newsk);
6800 /* Clear up any skbs waiting for the partial
6801 * delivery to finish.
6803 if (assoc->ulpq.pd_mode)
6804 sctp_clear_pd(oldsk, NULL);
6808 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6809 sctp_skb_set_owner_r_frag(skb, newsk);
6811 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6812 sctp_skb_set_owner_r_frag(skb, newsk);
6814 /* Set the type of socket to indicate that it is peeled off from the
6815 * original UDP-style socket or created with the accept() call on a
6816 * TCP-style socket..
6818 newsp->type = type;
6820 /* Mark the new socket "in-use" by the user so that any packets
6821 * that may arrive on the association after we've moved it are
6822 * queued to the backlog. This prevents a potential race between
6823 * backlog processing on the old socket and new-packet processing
6824 * on the new socket.
6826 * The caller has just allocated newsk so we can guarantee that other
6827 * paths won't try to lock it and then oldsk.
6829 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6830 sctp_assoc_migrate(assoc, newsk);
6832 /* If the association on the newsk is already closed before accept()
6833 * is called, set RCV_SHUTDOWN flag.
6835 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6836 newsk->sk_shutdown |= RCV_SHUTDOWN;
6838 newsk->sk_state = SCTP_SS_ESTABLISHED;
6839 sctp_release_sock(newsk);
6843 /* This proto struct describes the ULP interface for SCTP. */
6844 struct proto sctp_prot = {
6845 .name = "SCTP",
6846 .owner = THIS_MODULE,
6847 .close = sctp_close,
6848 .connect = sctp_connect,
6849 .disconnect = sctp_disconnect,
6850 .accept = sctp_accept,
6851 .ioctl = sctp_ioctl,
6852 .init = sctp_init_sock,
6853 .destroy = sctp_destroy_sock,
6854 .shutdown = sctp_shutdown,
6855 .setsockopt = sctp_setsockopt,
6856 .getsockopt = sctp_getsockopt,
6857 .sendmsg = sctp_sendmsg,
6858 .recvmsg = sctp_recvmsg,
6859 .bind = sctp_bind,
6860 .backlog_rcv = sctp_backlog_rcv,
6861 .hash = sctp_hash,
6862 .unhash = sctp_unhash,
6863 .get_port = sctp_get_port,
6864 .obj_size = sizeof(struct sctp_sock),
6865 .sysctl_mem = sysctl_sctp_mem,
6866 .sysctl_rmem = sysctl_sctp_rmem,
6867 .sysctl_wmem = sysctl_sctp_wmem,
6868 .memory_pressure = &sctp_memory_pressure,
6869 .enter_memory_pressure = sctp_enter_memory_pressure,
6870 .memory_allocated = &sctp_memory_allocated,
6871 .sockets_allocated = &sctp_sockets_allocated,
6874 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6876 struct proto sctpv6_prot = {
6877 .name = "SCTPv6",
6878 .owner = THIS_MODULE,
6879 .close = sctp_close,
6880 .connect = sctp_connect,
6881 .disconnect = sctp_disconnect,
6882 .accept = sctp_accept,
6883 .ioctl = sctp_ioctl,
6884 .init = sctp_init_sock,
6885 .destroy = sctp_destroy_sock,
6886 .shutdown = sctp_shutdown,
6887 .setsockopt = sctp_setsockopt,
6888 .getsockopt = sctp_getsockopt,
6889 .sendmsg = sctp_sendmsg,
6890 .recvmsg = sctp_recvmsg,
6891 .bind = sctp_bind,
6892 .backlog_rcv = sctp_backlog_rcv,
6893 .hash = sctp_hash,
6894 .unhash = sctp_unhash,
6895 .get_port = sctp_get_port,
6896 .obj_size = sizeof(struct sctp6_sock),
6897 .sysctl_mem = sysctl_sctp_mem,
6898 .sysctl_rmem = sysctl_sctp_rmem,
6899 .sysctl_wmem = sysctl_sctp_wmem,
6900 .memory_pressure = &sctp_memory_pressure,
6901 .enter_memory_pressure = sctp_enter_memory_pressure,
6902 .memory_allocated = &sctp_memory_allocated,
6903 .sockets_allocated = &sctp_sockets_allocated,
6905 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */