vt: add an activate and lock
[linux-2.6/mini2440.git] / net / sctp / socket.c
blob89af37a6c87168a4b68f60e5211b57882fb9c25c
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;
1364 sk->sk_state = SCTP_SS_CLOSING;
1366 ep = sctp_sk(sk)->ep;
1368 /* Walk all associations on an endpoint. */
1369 list_for_each_safe(pos, temp, &ep->asocs) {
1370 asoc = list_entry(pos, struct sctp_association, asocs);
1372 if (sctp_style(sk, TCP)) {
1373 /* A closed association can still be in the list if
1374 * it belongs to a TCP-style listening socket that is
1375 * not yet accepted. If so, free it. If not, send an
1376 * ABORT or SHUTDOWN based on the linger options.
1378 if (sctp_state(asoc, CLOSED)) {
1379 sctp_unhash_established(asoc);
1380 sctp_association_free(asoc);
1381 continue;
1385 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1386 struct sctp_chunk *chunk;
1388 chunk = sctp_make_abort_user(asoc, NULL, 0);
1389 if (chunk)
1390 sctp_primitive_ABORT(asoc, chunk);
1391 } else
1392 sctp_primitive_SHUTDOWN(asoc, NULL);
1395 /* Clean up any skbs sitting on the receive queue. */
1396 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1397 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1399 /* On a TCP-style socket, block for at most linger_time if set. */
1400 if (sctp_style(sk, TCP) && timeout)
1401 sctp_wait_for_close(sk, timeout);
1403 /* This will run the backlog queue. */
1404 sctp_release_sock(sk);
1406 /* Supposedly, no process has access to the socket, but
1407 * the net layers still may.
1409 sctp_local_bh_disable();
1410 sctp_bh_lock_sock(sk);
1412 /* Hold the sock, since sk_common_release() will put sock_put()
1413 * and we have just a little more cleanup.
1415 sock_hold(sk);
1416 sk_common_release(sk);
1418 sctp_bh_unlock_sock(sk);
1419 sctp_local_bh_enable();
1421 sock_put(sk);
1423 SCTP_DBG_OBJCNT_DEC(sock);
1426 /* Handle EPIPE error. */
1427 static int sctp_error(struct sock *sk, int flags, int err)
1429 if (err == -EPIPE)
1430 err = sock_error(sk) ? : -EPIPE;
1431 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1432 send_sig(SIGPIPE, current, 0);
1433 return err;
1436 /* API 3.1.3 sendmsg() - UDP Style Syntax
1438 * An application uses sendmsg() and recvmsg() calls to transmit data to
1439 * and receive data from its peer.
1441 * ssize_t sendmsg(int socket, const struct msghdr *message,
1442 * int flags);
1444 * socket - the socket descriptor of the endpoint.
1445 * message - pointer to the msghdr structure which contains a single
1446 * user message and possibly some ancillary data.
1448 * See Section 5 for complete description of the data
1449 * structures.
1451 * flags - flags sent or received with the user message, see Section
1452 * 5 for complete description of the flags.
1454 * Note: This function could use a rewrite especially when explicit
1455 * connect support comes in.
1457 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1459 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1461 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1462 struct msghdr *msg, size_t msg_len)
1464 struct sctp_sock *sp;
1465 struct sctp_endpoint *ep;
1466 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1467 struct sctp_transport *transport, *chunk_tp;
1468 struct sctp_chunk *chunk;
1469 union sctp_addr to;
1470 struct sockaddr *msg_name = NULL;
1471 struct sctp_sndrcvinfo default_sinfo = { 0 };
1472 struct sctp_sndrcvinfo *sinfo;
1473 struct sctp_initmsg *sinit;
1474 sctp_assoc_t associd = 0;
1475 sctp_cmsgs_t cmsgs = { NULL };
1476 int err;
1477 sctp_scope_t scope;
1478 long timeo;
1479 __u16 sinfo_flags = 0;
1480 struct sctp_datamsg *datamsg;
1481 int msg_flags = msg->msg_flags;
1483 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1484 sk, msg, msg_len);
1486 err = 0;
1487 sp = sctp_sk(sk);
1488 ep = sp->ep;
1490 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1492 /* We cannot send a message over a TCP-style listening socket. */
1493 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1494 err = -EPIPE;
1495 goto out_nounlock;
1498 /* Parse out the SCTP CMSGs. */
1499 err = sctp_msghdr_parse(msg, &cmsgs);
1501 if (err) {
1502 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1503 goto out_nounlock;
1506 /* Fetch the destination address for this packet. This
1507 * address only selects the association--it is not necessarily
1508 * the address we will send to.
1509 * For a peeled-off socket, msg_name is ignored.
1511 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1512 int msg_namelen = msg->msg_namelen;
1514 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1515 msg_namelen);
1516 if (err)
1517 return err;
1519 if (msg_namelen > sizeof(to))
1520 msg_namelen = sizeof(to);
1521 memcpy(&to, msg->msg_name, msg_namelen);
1522 msg_name = msg->msg_name;
1525 sinfo = cmsgs.info;
1526 sinit = cmsgs.init;
1528 /* Did the user specify SNDRCVINFO? */
1529 if (sinfo) {
1530 sinfo_flags = sinfo->sinfo_flags;
1531 associd = sinfo->sinfo_assoc_id;
1534 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1535 msg_len, sinfo_flags);
1537 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1538 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1539 err = -EINVAL;
1540 goto out_nounlock;
1543 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1544 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1545 * If SCTP_ABORT is set, the message length could be non zero with
1546 * the msg_iov set to the user abort reason.
1548 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1549 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1550 err = -EINVAL;
1551 goto out_nounlock;
1554 /* If SCTP_ADDR_OVER is set, there must be an address
1555 * specified in msg_name.
1557 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1558 err = -EINVAL;
1559 goto out_nounlock;
1562 transport = NULL;
1564 SCTP_DEBUG_PRINTK("About to look up association.\n");
1566 sctp_lock_sock(sk);
1568 /* If a msg_name has been specified, assume this is to be used. */
1569 if (msg_name) {
1570 /* Look for a matching association on the endpoint. */
1571 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1572 if (!asoc) {
1573 /* If we could not find a matching association on the
1574 * endpoint, make sure that it is not a TCP-style
1575 * socket that already has an association or there is
1576 * no peeled-off association on another socket.
1578 if ((sctp_style(sk, TCP) &&
1579 sctp_sstate(sk, ESTABLISHED)) ||
1580 sctp_endpoint_is_peeled_off(ep, &to)) {
1581 err = -EADDRNOTAVAIL;
1582 goto out_unlock;
1585 } else {
1586 asoc = sctp_id2assoc(sk, associd);
1587 if (!asoc) {
1588 err = -EPIPE;
1589 goto out_unlock;
1593 if (asoc) {
1594 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1596 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1597 * socket that has an association in CLOSED state. This can
1598 * happen when an accepted socket has an association that is
1599 * already CLOSED.
1601 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1602 err = -EPIPE;
1603 goto out_unlock;
1606 if (sinfo_flags & SCTP_EOF) {
1607 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1608 asoc);
1609 sctp_primitive_SHUTDOWN(asoc, NULL);
1610 err = 0;
1611 goto out_unlock;
1613 if (sinfo_flags & SCTP_ABORT) {
1615 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1616 if (!chunk) {
1617 err = -ENOMEM;
1618 goto out_unlock;
1621 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1622 sctp_primitive_ABORT(asoc, chunk);
1623 err = 0;
1624 goto out_unlock;
1628 /* Do we need to create the association? */
1629 if (!asoc) {
1630 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1632 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1633 err = -EINVAL;
1634 goto out_unlock;
1637 /* Check for invalid stream against the stream counts,
1638 * either the default or the user specified stream counts.
1640 if (sinfo) {
1641 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1642 /* Check against the defaults. */
1643 if (sinfo->sinfo_stream >=
1644 sp->initmsg.sinit_num_ostreams) {
1645 err = -EINVAL;
1646 goto out_unlock;
1648 } else {
1649 /* Check against the requested. */
1650 if (sinfo->sinfo_stream >=
1651 sinit->sinit_num_ostreams) {
1652 err = -EINVAL;
1653 goto out_unlock;
1659 * API 3.1.2 bind() - UDP Style Syntax
1660 * If a bind() or sctp_bindx() is not called prior to a
1661 * sendmsg() call that initiates a new association, the
1662 * system picks an ephemeral port and will choose an address
1663 * set equivalent to binding with a wildcard address.
1665 if (!ep->base.bind_addr.port) {
1666 if (sctp_autobind(sk)) {
1667 err = -EAGAIN;
1668 goto out_unlock;
1670 } else {
1672 * If an unprivileged user inherits a one-to-many
1673 * style socket with open associations on a privileged
1674 * port, it MAY be permitted to accept new associations,
1675 * but it SHOULD NOT be permitted to open new
1676 * associations.
1678 if (ep->base.bind_addr.port < PROT_SOCK &&
1679 !capable(CAP_NET_BIND_SERVICE)) {
1680 err = -EACCES;
1681 goto out_unlock;
1685 scope = sctp_scope(&to);
1686 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1687 if (!new_asoc) {
1688 err = -ENOMEM;
1689 goto out_unlock;
1691 asoc = new_asoc;
1693 /* If the SCTP_INIT ancillary data is specified, set all
1694 * the association init values accordingly.
1696 if (sinit) {
1697 if (sinit->sinit_num_ostreams) {
1698 asoc->c.sinit_num_ostreams =
1699 sinit->sinit_num_ostreams;
1701 if (sinit->sinit_max_instreams) {
1702 asoc->c.sinit_max_instreams =
1703 sinit->sinit_max_instreams;
1705 if (sinit->sinit_max_attempts) {
1706 asoc->max_init_attempts
1707 = sinit->sinit_max_attempts;
1709 if (sinit->sinit_max_init_timeo) {
1710 asoc->max_init_timeo =
1711 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1715 /* Prime the peer's transport structures. */
1716 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1717 if (!transport) {
1718 err = -ENOMEM;
1719 goto out_free;
1721 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1722 if (err < 0) {
1723 err = -ENOMEM;
1724 goto out_free;
1728 /* ASSERT: we have a valid association at this point. */
1729 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1731 if (!sinfo) {
1732 /* If the user didn't specify SNDRCVINFO, make up one with
1733 * some defaults.
1735 default_sinfo.sinfo_stream = asoc->default_stream;
1736 default_sinfo.sinfo_flags = asoc->default_flags;
1737 default_sinfo.sinfo_ppid = asoc->default_ppid;
1738 default_sinfo.sinfo_context = asoc->default_context;
1739 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1740 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1741 sinfo = &default_sinfo;
1744 /* API 7.1.7, the sndbuf size per association bounds the
1745 * maximum size of data that can be sent in a single send call.
1747 if (msg_len > sk->sk_sndbuf) {
1748 err = -EMSGSIZE;
1749 goto out_free;
1752 if (asoc->pmtu_pending)
1753 sctp_assoc_pending_pmtu(asoc);
1755 /* If fragmentation is disabled and the message length exceeds the
1756 * association fragmentation point, return EMSGSIZE. The I-D
1757 * does not specify what this error is, but this looks like
1758 * a great fit.
1760 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1761 err = -EMSGSIZE;
1762 goto out_free;
1765 if (sinfo) {
1766 /* Check for invalid stream. */
1767 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1768 err = -EINVAL;
1769 goto out_free;
1773 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1774 if (!sctp_wspace(asoc)) {
1775 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1776 if (err)
1777 goto out_free;
1780 /* If an address is passed with the sendto/sendmsg call, it is used
1781 * to override the primary destination address in the TCP model, or
1782 * when SCTP_ADDR_OVER flag is set in the UDP model.
1784 if ((sctp_style(sk, TCP) && msg_name) ||
1785 (sinfo_flags & SCTP_ADDR_OVER)) {
1786 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1787 if (!chunk_tp) {
1788 err = -EINVAL;
1789 goto out_free;
1791 } else
1792 chunk_tp = NULL;
1794 /* Auto-connect, if we aren't connected already. */
1795 if (sctp_state(asoc, CLOSED)) {
1796 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1797 if (err < 0)
1798 goto out_free;
1799 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1802 /* Break the message into multiple chunks of maximum size. */
1803 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1804 if (!datamsg) {
1805 err = -ENOMEM;
1806 goto out_free;
1809 /* Now send the (possibly) fragmented message. */
1810 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1811 sctp_chunk_hold(chunk);
1813 /* Do accounting for the write space. */
1814 sctp_set_owner_w(chunk);
1816 chunk->transport = chunk_tp;
1819 /* Send it to the lower layers. Note: all chunks
1820 * must either fail or succeed. The lower layer
1821 * works that way today. Keep it that way or this
1822 * breaks.
1824 err = sctp_primitive_SEND(asoc, datamsg);
1825 /* Did the lower layer accept the chunk? */
1826 if (err)
1827 sctp_datamsg_free(datamsg);
1828 else
1829 sctp_datamsg_put(datamsg);
1831 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1833 if (err)
1834 goto out_free;
1835 else
1836 err = msg_len;
1838 /* If we are already past ASSOCIATE, the lower
1839 * layers are responsible for association cleanup.
1841 goto out_unlock;
1843 out_free:
1844 if (new_asoc)
1845 sctp_association_free(asoc);
1846 out_unlock:
1847 sctp_release_sock(sk);
1849 out_nounlock:
1850 return sctp_error(sk, msg_flags, err);
1852 #if 0
1853 do_sock_err:
1854 if (msg_len)
1855 err = msg_len;
1856 else
1857 err = sock_error(sk);
1858 goto out;
1860 do_interrupted:
1861 if (msg_len)
1862 err = msg_len;
1863 goto out;
1864 #endif /* 0 */
1867 /* This is an extended version of skb_pull() that removes the data from the
1868 * start of a skb even when data is spread across the list of skb's in the
1869 * frag_list. len specifies the total amount of data that needs to be removed.
1870 * when 'len' bytes could be removed from the skb, it returns 0.
1871 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1872 * could not be removed.
1874 static int sctp_skb_pull(struct sk_buff *skb, int len)
1876 struct sk_buff *list;
1877 int skb_len = skb_headlen(skb);
1878 int rlen;
1880 if (len <= skb_len) {
1881 __skb_pull(skb, len);
1882 return 0;
1884 len -= skb_len;
1885 __skb_pull(skb, skb_len);
1887 skb_walk_frags(skb, list) {
1888 rlen = sctp_skb_pull(list, len);
1889 skb->len -= (len-rlen);
1890 skb->data_len -= (len-rlen);
1892 if (!rlen)
1893 return 0;
1895 len = rlen;
1898 return len;
1901 /* API 3.1.3 recvmsg() - UDP Style Syntax
1903 * ssize_t recvmsg(int socket, struct msghdr *message,
1904 * int flags);
1906 * socket - the socket descriptor of the endpoint.
1907 * message - pointer to the msghdr structure which contains a single
1908 * user message and possibly some ancillary data.
1910 * See Section 5 for complete description of the data
1911 * structures.
1913 * flags - flags sent or received with the user message, see Section
1914 * 5 for complete description of the flags.
1916 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1918 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1919 struct msghdr *msg, size_t len, int noblock,
1920 int flags, int *addr_len)
1922 struct sctp_ulpevent *event = NULL;
1923 struct sctp_sock *sp = sctp_sk(sk);
1924 struct sk_buff *skb;
1925 int copied;
1926 int err = 0;
1927 int skb_len;
1929 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1930 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1931 "len", len, "knoblauch", noblock,
1932 "flags", flags, "addr_len", addr_len);
1934 sctp_lock_sock(sk);
1936 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1937 err = -ENOTCONN;
1938 goto out;
1941 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1942 if (!skb)
1943 goto out;
1945 /* Get the total length of the skb including any skb's in the
1946 * frag_list.
1948 skb_len = skb->len;
1950 copied = skb_len;
1951 if (copied > len)
1952 copied = len;
1954 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1956 event = sctp_skb2event(skb);
1958 if (err)
1959 goto out_free;
1961 sock_recv_timestamp(msg, sk, skb);
1962 if (sctp_ulpevent_is_notification(event)) {
1963 msg->msg_flags |= MSG_NOTIFICATION;
1964 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1965 } else {
1966 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1969 /* Check if we allow SCTP_SNDRCVINFO. */
1970 if (sp->subscribe.sctp_data_io_event)
1971 sctp_ulpevent_read_sndrcvinfo(event, msg);
1972 #if 0
1973 /* FIXME: we should be calling IP/IPv6 layers. */
1974 if (sk->sk_protinfo.af_inet.cmsg_flags)
1975 ip_cmsg_recv(msg, skb);
1976 #endif
1978 err = copied;
1980 /* If skb's length exceeds the user's buffer, update the skb and
1981 * push it back to the receive_queue so that the next call to
1982 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1984 if (skb_len > copied) {
1985 msg->msg_flags &= ~MSG_EOR;
1986 if (flags & MSG_PEEK)
1987 goto out_free;
1988 sctp_skb_pull(skb, copied);
1989 skb_queue_head(&sk->sk_receive_queue, skb);
1991 /* When only partial message is copied to the user, increase
1992 * rwnd by that amount. If all the data in the skb is read,
1993 * rwnd is updated when the event is freed.
1995 if (!sctp_ulpevent_is_notification(event))
1996 sctp_assoc_rwnd_increase(event->asoc, copied);
1997 goto out;
1998 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1999 (event->msg_flags & MSG_EOR))
2000 msg->msg_flags |= MSG_EOR;
2001 else
2002 msg->msg_flags &= ~MSG_EOR;
2004 out_free:
2005 if (flags & MSG_PEEK) {
2006 /* Release the skb reference acquired after peeking the skb in
2007 * sctp_skb_recv_datagram().
2009 kfree_skb(skb);
2010 } else {
2011 /* Free the event which includes releasing the reference to
2012 * the owner of the skb, freeing the skb and updating the
2013 * rwnd.
2015 sctp_ulpevent_free(event);
2017 out:
2018 sctp_release_sock(sk);
2019 return err;
2022 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2024 * This option is a on/off flag. If enabled no SCTP message
2025 * fragmentation will be performed. Instead if a message being sent
2026 * exceeds the current PMTU size, the message will NOT be sent and
2027 * instead a error will be indicated to the user.
2029 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2030 char __user *optval, int optlen)
2032 int val;
2034 if (optlen < sizeof(int))
2035 return -EINVAL;
2037 if (get_user(val, (int __user *)optval))
2038 return -EFAULT;
2040 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2042 return 0;
2045 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2046 int optlen)
2048 if (optlen > sizeof(struct sctp_event_subscribe))
2049 return -EINVAL;
2050 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2051 return -EFAULT;
2052 return 0;
2055 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2057 * This socket option is applicable to the UDP-style socket only. When
2058 * set it will cause associations that are idle for more than the
2059 * specified number of seconds to automatically close. An association
2060 * being idle is defined an association that has NOT sent or received
2061 * user data. The special value of '0' indicates that no automatic
2062 * close of any associations should be performed. The option expects an
2063 * integer defining the number of seconds of idle time before an
2064 * association is closed.
2066 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2067 int optlen)
2069 struct sctp_sock *sp = sctp_sk(sk);
2071 /* Applicable to UDP-style socket only */
2072 if (sctp_style(sk, TCP))
2073 return -EOPNOTSUPP;
2074 if (optlen != sizeof(int))
2075 return -EINVAL;
2076 if (copy_from_user(&sp->autoclose, optval, optlen))
2077 return -EFAULT;
2079 return 0;
2082 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2084 * Applications can enable or disable heartbeats for any peer address of
2085 * an association, modify an address's heartbeat interval, force a
2086 * heartbeat to be sent immediately, and adjust the address's maximum
2087 * number of retransmissions sent before an address is considered
2088 * unreachable. The following structure is used to access and modify an
2089 * address's parameters:
2091 * struct sctp_paddrparams {
2092 * sctp_assoc_t spp_assoc_id;
2093 * struct sockaddr_storage spp_address;
2094 * uint32_t spp_hbinterval;
2095 * uint16_t spp_pathmaxrxt;
2096 * uint32_t spp_pathmtu;
2097 * uint32_t spp_sackdelay;
2098 * uint32_t spp_flags;
2099 * };
2101 * spp_assoc_id - (one-to-many style socket) This is filled in the
2102 * application, and identifies the association for
2103 * this query.
2104 * spp_address - This specifies which address is of interest.
2105 * spp_hbinterval - This contains the value of the heartbeat interval,
2106 * in milliseconds. If a value of zero
2107 * is present in this field then no changes are to
2108 * be made to this parameter.
2109 * spp_pathmaxrxt - This contains the maximum number of
2110 * retransmissions before this address shall be
2111 * considered unreachable. If a value of zero
2112 * is present in this field then no changes are to
2113 * be made to this parameter.
2114 * spp_pathmtu - When Path MTU discovery is disabled the value
2115 * specified here will be the "fixed" path mtu.
2116 * Note that if the spp_address field is empty
2117 * then all associations on this address will
2118 * have this fixed path mtu set upon them.
2120 * spp_sackdelay - When delayed sack is enabled, this value specifies
2121 * the number of milliseconds that sacks will be delayed
2122 * for. This value will apply to all addresses of an
2123 * association if the spp_address field is empty. Note
2124 * also, that if delayed sack is enabled and this
2125 * value is set to 0, no change is made to the last
2126 * recorded delayed sack timer value.
2128 * spp_flags - These flags are used to control various features
2129 * on an association. The flag field may contain
2130 * zero or more of the following options.
2132 * SPP_HB_ENABLE - Enable heartbeats on the
2133 * specified address. Note that if the address
2134 * field is empty all addresses for the association
2135 * have heartbeats enabled upon them.
2137 * SPP_HB_DISABLE - Disable heartbeats on the
2138 * speicifed address. Note that if the address
2139 * field is empty all addresses for the association
2140 * will have their heartbeats disabled. Note also
2141 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2142 * mutually exclusive, only one of these two should
2143 * be specified. Enabling both fields will have
2144 * undetermined results.
2146 * SPP_HB_DEMAND - Request a user initiated heartbeat
2147 * to be made immediately.
2149 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2150 * heartbeat delayis to be set to the value of 0
2151 * milliseconds.
2153 * SPP_PMTUD_ENABLE - This field will enable PMTU
2154 * discovery upon the specified address. Note that
2155 * if the address feild is empty then all addresses
2156 * on the association are effected.
2158 * SPP_PMTUD_DISABLE - This field will disable PMTU
2159 * discovery upon the specified address. Note that
2160 * if the address feild is empty then all addresses
2161 * on the association are effected. Not also that
2162 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2163 * exclusive. Enabling both will have undetermined
2164 * results.
2166 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2167 * on delayed sack. The time specified in spp_sackdelay
2168 * is used to specify the sack delay for this address. Note
2169 * that if spp_address is empty then all addresses will
2170 * enable delayed sack and take on the sack delay
2171 * value specified in spp_sackdelay.
2172 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2173 * off delayed sack. If the spp_address field is blank then
2174 * delayed sack is disabled for the entire association. Note
2175 * also that this field is mutually exclusive to
2176 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2177 * results.
2179 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2180 struct sctp_transport *trans,
2181 struct sctp_association *asoc,
2182 struct sctp_sock *sp,
2183 int hb_change,
2184 int pmtud_change,
2185 int sackdelay_change)
2187 int error;
2189 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2190 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2191 if (error)
2192 return error;
2195 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2196 * this field is ignored. Note also that a value of zero indicates
2197 * the current setting should be left unchanged.
2199 if (params->spp_flags & SPP_HB_ENABLE) {
2201 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2202 * set. This lets us use 0 value when this flag
2203 * is set.
2205 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2206 params->spp_hbinterval = 0;
2208 if (params->spp_hbinterval ||
2209 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2210 if (trans) {
2211 trans->hbinterval =
2212 msecs_to_jiffies(params->spp_hbinterval);
2213 } else if (asoc) {
2214 asoc->hbinterval =
2215 msecs_to_jiffies(params->spp_hbinterval);
2216 } else {
2217 sp->hbinterval = params->spp_hbinterval;
2222 if (hb_change) {
2223 if (trans) {
2224 trans->param_flags =
2225 (trans->param_flags & ~SPP_HB) | hb_change;
2226 } else if (asoc) {
2227 asoc->param_flags =
2228 (asoc->param_flags & ~SPP_HB) | hb_change;
2229 } else {
2230 sp->param_flags =
2231 (sp->param_flags & ~SPP_HB) | hb_change;
2235 /* When Path MTU discovery is disabled the value specified here will
2236 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2237 * include the flag SPP_PMTUD_DISABLE for this field to have any
2238 * effect).
2240 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2241 if (trans) {
2242 trans->pathmtu = params->spp_pathmtu;
2243 sctp_assoc_sync_pmtu(asoc);
2244 } else if (asoc) {
2245 asoc->pathmtu = params->spp_pathmtu;
2246 sctp_frag_point(asoc, params->spp_pathmtu);
2247 } else {
2248 sp->pathmtu = params->spp_pathmtu;
2252 if (pmtud_change) {
2253 if (trans) {
2254 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2255 (params->spp_flags & SPP_PMTUD_ENABLE);
2256 trans->param_flags =
2257 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2258 if (update) {
2259 sctp_transport_pmtu(trans);
2260 sctp_assoc_sync_pmtu(asoc);
2262 } else if (asoc) {
2263 asoc->param_flags =
2264 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2265 } else {
2266 sp->param_flags =
2267 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2271 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2272 * value of this field is ignored. Note also that a value of zero
2273 * indicates the current setting should be left unchanged.
2275 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2276 if (trans) {
2277 trans->sackdelay =
2278 msecs_to_jiffies(params->spp_sackdelay);
2279 } else if (asoc) {
2280 asoc->sackdelay =
2281 msecs_to_jiffies(params->spp_sackdelay);
2282 } else {
2283 sp->sackdelay = params->spp_sackdelay;
2287 if (sackdelay_change) {
2288 if (trans) {
2289 trans->param_flags =
2290 (trans->param_flags & ~SPP_SACKDELAY) |
2291 sackdelay_change;
2292 } else if (asoc) {
2293 asoc->param_flags =
2294 (asoc->param_flags & ~SPP_SACKDELAY) |
2295 sackdelay_change;
2296 } else {
2297 sp->param_flags =
2298 (sp->param_flags & ~SPP_SACKDELAY) |
2299 sackdelay_change;
2303 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2304 * of this field is ignored. Note also that a value of zero
2305 * indicates the current setting should be left unchanged.
2307 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2308 if (trans) {
2309 trans->pathmaxrxt = params->spp_pathmaxrxt;
2310 } else if (asoc) {
2311 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2312 } else {
2313 sp->pathmaxrxt = params->spp_pathmaxrxt;
2317 return 0;
2320 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2321 char __user *optval, int optlen)
2323 struct sctp_paddrparams params;
2324 struct sctp_transport *trans = NULL;
2325 struct sctp_association *asoc = NULL;
2326 struct sctp_sock *sp = sctp_sk(sk);
2327 int error;
2328 int hb_change, pmtud_change, sackdelay_change;
2330 if (optlen != sizeof(struct sctp_paddrparams))
2331 return - EINVAL;
2333 if (copy_from_user(&params, optval, optlen))
2334 return -EFAULT;
2336 /* Validate flags and value parameters. */
2337 hb_change = params.spp_flags & SPP_HB;
2338 pmtud_change = params.spp_flags & SPP_PMTUD;
2339 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2341 if (hb_change == SPP_HB ||
2342 pmtud_change == SPP_PMTUD ||
2343 sackdelay_change == SPP_SACKDELAY ||
2344 params.spp_sackdelay > 500 ||
2345 (params.spp_pathmtu
2346 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2347 return -EINVAL;
2349 /* If an address other than INADDR_ANY is specified, and
2350 * no transport is found, then the request is invalid.
2352 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2353 trans = sctp_addr_id2transport(sk, &params.spp_address,
2354 params.spp_assoc_id);
2355 if (!trans)
2356 return -EINVAL;
2359 /* Get association, if assoc_id != 0 and the socket is a one
2360 * to many style socket, and an association was not found, then
2361 * the id was invalid.
2363 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2364 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2365 return -EINVAL;
2367 /* Heartbeat demand can only be sent on a transport or
2368 * association, but not a socket.
2370 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2371 return -EINVAL;
2373 /* Process parameters. */
2374 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2375 hb_change, pmtud_change,
2376 sackdelay_change);
2378 if (error)
2379 return error;
2381 /* If changes are for association, also apply parameters to each
2382 * transport.
2384 if (!trans && asoc) {
2385 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2386 transports) {
2387 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2388 hb_change, pmtud_change,
2389 sackdelay_change);
2393 return 0;
2397 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2399 * This option will effect the way delayed acks are performed. This
2400 * option allows you to get or set the delayed ack time, in
2401 * milliseconds. It also allows changing the delayed ack frequency.
2402 * Changing the frequency to 1 disables the delayed sack algorithm. If
2403 * the assoc_id is 0, then this sets or gets the endpoints default
2404 * values. If the assoc_id field is non-zero, then the set or get
2405 * effects the specified association for the one to many model (the
2406 * assoc_id field is ignored by the one to one model). Note that if
2407 * sack_delay or sack_freq are 0 when setting this option, then the
2408 * current values will remain unchanged.
2410 * struct sctp_sack_info {
2411 * sctp_assoc_t sack_assoc_id;
2412 * uint32_t sack_delay;
2413 * uint32_t sack_freq;
2414 * };
2416 * sack_assoc_id - This parameter, indicates which association the user
2417 * is performing an action upon. Note that if this field's value is
2418 * zero then the endpoints default value is changed (effecting future
2419 * associations only).
2421 * sack_delay - This parameter contains the number of milliseconds that
2422 * the user is requesting the delayed ACK timer be set to. Note that
2423 * this value is defined in the standard to be between 200 and 500
2424 * milliseconds.
2426 * sack_freq - This parameter contains the number of packets that must
2427 * be received before a sack is sent without waiting for the delay
2428 * timer to expire. The default value for this is 2, setting this
2429 * value to 1 will disable the delayed sack algorithm.
2432 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2433 char __user *optval, int optlen)
2435 struct sctp_sack_info params;
2436 struct sctp_transport *trans = NULL;
2437 struct sctp_association *asoc = NULL;
2438 struct sctp_sock *sp = sctp_sk(sk);
2440 if (optlen == sizeof(struct sctp_sack_info)) {
2441 if (copy_from_user(&params, optval, optlen))
2442 return -EFAULT;
2444 if (params.sack_delay == 0 && params.sack_freq == 0)
2445 return 0;
2446 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2447 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2448 "in delayed_ack socket option deprecated\n");
2449 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2450 if (copy_from_user(&params, optval, optlen))
2451 return -EFAULT;
2453 if (params.sack_delay == 0)
2454 params.sack_freq = 1;
2455 else
2456 params.sack_freq = 0;
2457 } else
2458 return - EINVAL;
2460 /* Validate value parameter. */
2461 if (params.sack_delay > 500)
2462 return -EINVAL;
2464 /* Get association, if sack_assoc_id != 0 and the socket is a one
2465 * to many style socket, and an association was not found, then
2466 * the id was invalid.
2468 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2469 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2470 return -EINVAL;
2472 if (params.sack_delay) {
2473 if (asoc) {
2474 asoc->sackdelay =
2475 msecs_to_jiffies(params.sack_delay);
2476 asoc->param_flags =
2477 (asoc->param_flags & ~SPP_SACKDELAY) |
2478 SPP_SACKDELAY_ENABLE;
2479 } else {
2480 sp->sackdelay = params.sack_delay;
2481 sp->param_flags =
2482 (sp->param_flags & ~SPP_SACKDELAY) |
2483 SPP_SACKDELAY_ENABLE;
2487 if (params.sack_freq == 1) {
2488 if (asoc) {
2489 asoc->param_flags =
2490 (asoc->param_flags & ~SPP_SACKDELAY) |
2491 SPP_SACKDELAY_DISABLE;
2492 } else {
2493 sp->param_flags =
2494 (sp->param_flags & ~SPP_SACKDELAY) |
2495 SPP_SACKDELAY_DISABLE;
2497 } else if (params.sack_freq > 1) {
2498 if (asoc) {
2499 asoc->sackfreq = params.sack_freq;
2500 asoc->param_flags =
2501 (asoc->param_flags & ~SPP_SACKDELAY) |
2502 SPP_SACKDELAY_ENABLE;
2503 } else {
2504 sp->sackfreq = params.sack_freq;
2505 sp->param_flags =
2506 (sp->param_flags & ~SPP_SACKDELAY) |
2507 SPP_SACKDELAY_ENABLE;
2511 /* If change is for association, also apply to each transport. */
2512 if (asoc) {
2513 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2514 transports) {
2515 if (params.sack_delay) {
2516 trans->sackdelay =
2517 msecs_to_jiffies(params.sack_delay);
2518 trans->param_flags =
2519 (trans->param_flags & ~SPP_SACKDELAY) |
2520 SPP_SACKDELAY_ENABLE;
2522 if (params.sack_freq == 1) {
2523 trans->param_flags =
2524 (trans->param_flags & ~SPP_SACKDELAY) |
2525 SPP_SACKDELAY_DISABLE;
2526 } else if (params.sack_freq > 1) {
2527 trans->sackfreq = params.sack_freq;
2528 trans->param_flags =
2529 (trans->param_flags & ~SPP_SACKDELAY) |
2530 SPP_SACKDELAY_ENABLE;
2535 return 0;
2538 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2540 * Applications can specify protocol parameters for the default association
2541 * initialization. The option name argument to setsockopt() and getsockopt()
2542 * is SCTP_INITMSG.
2544 * Setting initialization parameters is effective only on an unconnected
2545 * socket (for UDP-style sockets only future associations are effected
2546 * by the change). With TCP-style sockets, this option is inherited by
2547 * sockets derived from a listener socket.
2549 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2551 struct sctp_initmsg sinit;
2552 struct sctp_sock *sp = sctp_sk(sk);
2554 if (optlen != sizeof(struct sctp_initmsg))
2555 return -EINVAL;
2556 if (copy_from_user(&sinit, optval, optlen))
2557 return -EFAULT;
2559 if (sinit.sinit_num_ostreams)
2560 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2561 if (sinit.sinit_max_instreams)
2562 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2563 if (sinit.sinit_max_attempts)
2564 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2565 if (sinit.sinit_max_init_timeo)
2566 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2568 return 0;
2572 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2574 * Applications that wish to use the sendto() system call may wish to
2575 * specify a default set of parameters that would normally be supplied
2576 * through the inclusion of ancillary data. This socket option allows
2577 * such an application to set the default sctp_sndrcvinfo structure.
2578 * The application that wishes to use this socket option simply passes
2579 * in to this call the sctp_sndrcvinfo structure defined in Section
2580 * 5.2.2) The input parameters accepted by this call include
2581 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2582 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2583 * to this call if the caller is using the UDP model.
2585 static int sctp_setsockopt_default_send_param(struct sock *sk,
2586 char __user *optval, int optlen)
2588 struct sctp_sndrcvinfo info;
2589 struct sctp_association *asoc;
2590 struct sctp_sock *sp = sctp_sk(sk);
2592 if (optlen != sizeof(struct sctp_sndrcvinfo))
2593 return -EINVAL;
2594 if (copy_from_user(&info, optval, optlen))
2595 return -EFAULT;
2597 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2598 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2599 return -EINVAL;
2601 if (asoc) {
2602 asoc->default_stream = info.sinfo_stream;
2603 asoc->default_flags = info.sinfo_flags;
2604 asoc->default_ppid = info.sinfo_ppid;
2605 asoc->default_context = info.sinfo_context;
2606 asoc->default_timetolive = info.sinfo_timetolive;
2607 } else {
2608 sp->default_stream = info.sinfo_stream;
2609 sp->default_flags = info.sinfo_flags;
2610 sp->default_ppid = info.sinfo_ppid;
2611 sp->default_context = info.sinfo_context;
2612 sp->default_timetolive = info.sinfo_timetolive;
2615 return 0;
2618 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2620 * Requests that the local SCTP stack use the enclosed peer address as
2621 * the association primary. The enclosed address must be one of the
2622 * association peer's addresses.
2624 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2625 int optlen)
2627 struct sctp_prim prim;
2628 struct sctp_transport *trans;
2630 if (optlen != sizeof(struct sctp_prim))
2631 return -EINVAL;
2633 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2634 return -EFAULT;
2636 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2637 if (!trans)
2638 return -EINVAL;
2640 sctp_assoc_set_primary(trans->asoc, trans);
2642 return 0;
2646 * 7.1.5 SCTP_NODELAY
2648 * Turn on/off any Nagle-like algorithm. This means that packets are
2649 * generally sent as soon as possible and no unnecessary delays are
2650 * introduced, at the cost of more packets in the network. Expects an
2651 * integer boolean flag.
2653 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2654 int optlen)
2656 int val;
2658 if (optlen < sizeof(int))
2659 return -EINVAL;
2660 if (get_user(val, (int __user *)optval))
2661 return -EFAULT;
2663 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2664 return 0;
2669 * 7.1.1 SCTP_RTOINFO
2671 * The protocol parameters used to initialize and bound retransmission
2672 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2673 * and modify these parameters.
2674 * All parameters are time values, in milliseconds. A value of 0, when
2675 * modifying the parameters, indicates that the current value should not
2676 * be changed.
2679 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2680 struct sctp_rtoinfo rtoinfo;
2681 struct sctp_association *asoc;
2683 if (optlen != sizeof (struct sctp_rtoinfo))
2684 return -EINVAL;
2686 if (copy_from_user(&rtoinfo, optval, optlen))
2687 return -EFAULT;
2689 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2691 /* Set the values to the specific association */
2692 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2693 return -EINVAL;
2695 if (asoc) {
2696 if (rtoinfo.srto_initial != 0)
2697 asoc->rto_initial =
2698 msecs_to_jiffies(rtoinfo.srto_initial);
2699 if (rtoinfo.srto_max != 0)
2700 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2701 if (rtoinfo.srto_min != 0)
2702 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2703 } else {
2704 /* If there is no association or the association-id = 0
2705 * set the values to the endpoint.
2707 struct sctp_sock *sp = sctp_sk(sk);
2709 if (rtoinfo.srto_initial != 0)
2710 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2711 if (rtoinfo.srto_max != 0)
2712 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2713 if (rtoinfo.srto_min != 0)
2714 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2717 return 0;
2722 * 7.1.2 SCTP_ASSOCINFO
2724 * This option is used to tune the maximum retransmission attempts
2725 * of the association.
2726 * Returns an error if the new association retransmission value is
2727 * greater than the sum of the retransmission value of the peer.
2728 * See [SCTP] for more information.
2731 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2734 struct sctp_assocparams assocparams;
2735 struct sctp_association *asoc;
2737 if (optlen != sizeof(struct sctp_assocparams))
2738 return -EINVAL;
2739 if (copy_from_user(&assocparams, optval, optlen))
2740 return -EFAULT;
2742 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2744 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2745 return -EINVAL;
2747 /* Set the values to the specific association */
2748 if (asoc) {
2749 if (assocparams.sasoc_asocmaxrxt != 0) {
2750 __u32 path_sum = 0;
2751 int paths = 0;
2752 struct sctp_transport *peer_addr;
2754 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2755 transports) {
2756 path_sum += peer_addr->pathmaxrxt;
2757 paths++;
2760 /* Only validate asocmaxrxt if we have more than
2761 * one path/transport. We do this because path
2762 * retransmissions are only counted when we have more
2763 * then one path.
2765 if (paths > 1 &&
2766 assocparams.sasoc_asocmaxrxt > path_sum)
2767 return -EINVAL;
2769 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2772 if (assocparams.sasoc_cookie_life != 0) {
2773 asoc->cookie_life.tv_sec =
2774 assocparams.sasoc_cookie_life / 1000;
2775 asoc->cookie_life.tv_usec =
2776 (assocparams.sasoc_cookie_life % 1000)
2777 * 1000;
2779 } else {
2780 /* Set the values to the endpoint */
2781 struct sctp_sock *sp = sctp_sk(sk);
2783 if (assocparams.sasoc_asocmaxrxt != 0)
2784 sp->assocparams.sasoc_asocmaxrxt =
2785 assocparams.sasoc_asocmaxrxt;
2786 if (assocparams.sasoc_cookie_life != 0)
2787 sp->assocparams.sasoc_cookie_life =
2788 assocparams.sasoc_cookie_life;
2790 return 0;
2794 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2796 * This socket option is a boolean flag which turns on or off mapped V4
2797 * addresses. If this option is turned on and the socket is type
2798 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2799 * If this option is turned off, then no mapping will be done of V4
2800 * addresses and a user will receive both PF_INET6 and PF_INET type
2801 * addresses on the socket.
2803 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2805 int val;
2806 struct sctp_sock *sp = sctp_sk(sk);
2808 if (optlen < sizeof(int))
2809 return -EINVAL;
2810 if (get_user(val, (int __user *)optval))
2811 return -EFAULT;
2812 if (val)
2813 sp->v4mapped = 1;
2814 else
2815 sp->v4mapped = 0;
2817 return 0;
2821 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2822 * This option will get or set the maximum size to put in any outgoing
2823 * SCTP DATA chunk. If a message is larger than this size it will be
2824 * fragmented by SCTP into the specified size. Note that the underlying
2825 * SCTP implementation may fragment into smaller sized chunks when the
2826 * PMTU of the underlying association is smaller than the value set by
2827 * the user. The default value for this option is '0' which indicates
2828 * the user is NOT limiting fragmentation and only the PMTU will effect
2829 * SCTP's choice of DATA chunk size. Note also that values set larger
2830 * than the maximum size of an IP datagram will effectively let SCTP
2831 * control fragmentation (i.e. the same as setting this option to 0).
2833 * The following structure is used to access and modify this parameter:
2835 * struct sctp_assoc_value {
2836 * sctp_assoc_t assoc_id;
2837 * uint32_t assoc_value;
2838 * };
2840 * assoc_id: This parameter is ignored for one-to-one style sockets.
2841 * For one-to-many style sockets this parameter indicates which
2842 * association the user is performing an action upon. Note that if
2843 * this field's value is zero then the endpoints default value is
2844 * changed (effecting future associations only).
2845 * assoc_value: This parameter specifies the maximum size in bytes.
2847 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2849 struct sctp_assoc_value params;
2850 struct sctp_association *asoc;
2851 struct sctp_sock *sp = sctp_sk(sk);
2852 int val;
2854 if (optlen == sizeof(int)) {
2855 printk(KERN_WARNING
2856 "SCTP: Use of int in maxseg socket option deprecated\n");
2857 printk(KERN_WARNING
2858 "SCTP: Use struct sctp_assoc_value instead\n");
2859 if (copy_from_user(&val, optval, optlen))
2860 return -EFAULT;
2861 params.assoc_id = 0;
2862 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2863 if (copy_from_user(&params, optval, optlen))
2864 return -EFAULT;
2865 val = params.assoc_value;
2866 } else
2867 return -EINVAL;
2869 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2870 return -EINVAL;
2872 asoc = sctp_id2assoc(sk, params.assoc_id);
2873 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2874 return -EINVAL;
2876 if (asoc) {
2877 if (val == 0) {
2878 val = asoc->pathmtu;
2879 val -= sp->pf->af->net_header_len;
2880 val -= sizeof(struct sctphdr) +
2881 sizeof(struct sctp_data_chunk);
2883 asoc->user_frag = val;
2884 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2885 } else {
2886 sp->user_frag = val;
2889 return 0;
2894 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2896 * Requests that the peer mark the enclosed address as the association
2897 * primary. The enclosed address must be one of the association's
2898 * locally bound addresses. The following structure is used to make a
2899 * set primary request:
2901 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2902 int optlen)
2904 struct sctp_sock *sp;
2905 struct sctp_endpoint *ep;
2906 struct sctp_association *asoc = NULL;
2907 struct sctp_setpeerprim prim;
2908 struct sctp_chunk *chunk;
2909 int err;
2911 sp = sctp_sk(sk);
2912 ep = sp->ep;
2914 if (!sctp_addip_enable)
2915 return -EPERM;
2917 if (optlen != sizeof(struct sctp_setpeerprim))
2918 return -EINVAL;
2920 if (copy_from_user(&prim, optval, optlen))
2921 return -EFAULT;
2923 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2924 if (!asoc)
2925 return -EINVAL;
2927 if (!asoc->peer.asconf_capable)
2928 return -EPERM;
2930 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2931 return -EPERM;
2933 if (!sctp_state(asoc, ESTABLISHED))
2934 return -ENOTCONN;
2936 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2937 return -EADDRNOTAVAIL;
2939 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2940 chunk = sctp_make_asconf_set_prim(asoc,
2941 (union sctp_addr *)&prim.sspp_addr);
2942 if (!chunk)
2943 return -ENOMEM;
2945 err = sctp_send_asconf(asoc, chunk);
2947 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2949 return err;
2952 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2953 int optlen)
2955 struct sctp_setadaptation adaptation;
2957 if (optlen != sizeof(struct sctp_setadaptation))
2958 return -EINVAL;
2959 if (copy_from_user(&adaptation, optval, optlen))
2960 return -EFAULT;
2962 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2964 return 0;
2968 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2970 * The context field in the sctp_sndrcvinfo structure is normally only
2971 * used when a failed message is retrieved holding the value that was
2972 * sent down on the actual send call. This option allows the setting of
2973 * a default context on an association basis that will be received on
2974 * reading messages from the peer. This is especially helpful in the
2975 * one-2-many model for an application to keep some reference to an
2976 * internal state machine that is processing messages on the
2977 * association. Note that the setting of this value only effects
2978 * received messages from the peer and does not effect the value that is
2979 * saved with outbound messages.
2981 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2982 int optlen)
2984 struct sctp_assoc_value params;
2985 struct sctp_sock *sp;
2986 struct sctp_association *asoc;
2988 if (optlen != sizeof(struct sctp_assoc_value))
2989 return -EINVAL;
2990 if (copy_from_user(&params, optval, optlen))
2991 return -EFAULT;
2993 sp = sctp_sk(sk);
2995 if (params.assoc_id != 0) {
2996 asoc = sctp_id2assoc(sk, params.assoc_id);
2997 if (!asoc)
2998 return -EINVAL;
2999 asoc->default_rcv_context = params.assoc_value;
3000 } else {
3001 sp->default_rcv_context = params.assoc_value;
3004 return 0;
3008 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3010 * This options will at a minimum specify if the implementation is doing
3011 * fragmented interleave. Fragmented interleave, for a one to many
3012 * socket, is when subsequent calls to receive a message may return
3013 * parts of messages from different associations. Some implementations
3014 * may allow you to turn this value on or off. If so, when turned off,
3015 * no fragment interleave will occur (which will cause a head of line
3016 * blocking amongst multiple associations sharing the same one to many
3017 * socket). When this option is turned on, then each receive call may
3018 * come from a different association (thus the user must receive data
3019 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3020 * association each receive belongs to.
3022 * This option takes a boolean value. A non-zero value indicates that
3023 * fragmented interleave is on. A value of zero indicates that
3024 * fragmented interleave is off.
3026 * Note that it is important that an implementation that allows this
3027 * option to be turned on, have it off by default. Otherwise an unaware
3028 * application using the one to many model may become confused and act
3029 * incorrectly.
3031 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3032 char __user *optval,
3033 int optlen)
3035 int val;
3037 if (optlen != sizeof(int))
3038 return -EINVAL;
3039 if (get_user(val, (int __user *)optval))
3040 return -EFAULT;
3042 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3044 return 0;
3048 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3049 * (SCTP_PARTIAL_DELIVERY_POINT)
3051 * This option will set or get the SCTP partial delivery point. This
3052 * point is the size of a message where the partial delivery API will be
3053 * invoked to help free up rwnd space for the peer. Setting this to a
3054 * lower value will cause partial deliveries to happen more often. The
3055 * calls argument is an integer that sets or gets the partial delivery
3056 * point. Note also that the call will fail if the user attempts to set
3057 * this value larger than the socket receive buffer size.
3059 * Note that any single message having a length smaller than or equal to
3060 * the SCTP partial delivery point will be delivered in one single read
3061 * call as long as the user provided buffer is large enough to hold the
3062 * message.
3064 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3065 char __user *optval,
3066 int optlen)
3068 u32 val;
3070 if (optlen != sizeof(u32))
3071 return -EINVAL;
3072 if (get_user(val, (int __user *)optval))
3073 return -EFAULT;
3075 /* Note: We double the receive buffer from what the user sets
3076 * it to be, also initial rwnd is based on rcvbuf/2.
3078 if (val > (sk->sk_rcvbuf >> 1))
3079 return -EINVAL;
3081 sctp_sk(sk)->pd_point = val;
3083 return 0; /* is this the right error code? */
3087 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3089 * This option will allow a user to change the maximum burst of packets
3090 * that can be emitted by this association. Note that the default value
3091 * is 4, and some implementations may restrict this setting so that it
3092 * can only be lowered.
3094 * NOTE: This text doesn't seem right. Do this on a socket basis with
3095 * future associations inheriting the socket value.
3097 static int sctp_setsockopt_maxburst(struct sock *sk,
3098 char __user *optval,
3099 int optlen)
3101 struct sctp_assoc_value params;
3102 struct sctp_sock *sp;
3103 struct sctp_association *asoc;
3104 int val;
3105 int assoc_id = 0;
3107 if (optlen == sizeof(int)) {
3108 printk(KERN_WARNING
3109 "SCTP: Use of int in max_burst socket option deprecated\n");
3110 printk(KERN_WARNING
3111 "SCTP: Use struct sctp_assoc_value instead\n");
3112 if (copy_from_user(&val, optval, optlen))
3113 return -EFAULT;
3114 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3115 if (copy_from_user(&params, optval, optlen))
3116 return -EFAULT;
3117 val = params.assoc_value;
3118 assoc_id = params.assoc_id;
3119 } else
3120 return -EINVAL;
3122 sp = sctp_sk(sk);
3124 if (assoc_id != 0) {
3125 asoc = sctp_id2assoc(sk, assoc_id);
3126 if (!asoc)
3127 return -EINVAL;
3128 asoc->max_burst = val;
3129 } else
3130 sp->max_burst = val;
3132 return 0;
3136 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3138 * This set option adds a chunk type that the user is requesting to be
3139 * received only in an authenticated way. Changes to the list of chunks
3140 * will only effect future associations on the socket.
3142 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3143 char __user *optval,
3144 int optlen)
3146 struct sctp_authchunk val;
3148 if (!sctp_auth_enable)
3149 return -EACCES;
3151 if (optlen != sizeof(struct sctp_authchunk))
3152 return -EINVAL;
3153 if (copy_from_user(&val, optval, optlen))
3154 return -EFAULT;
3156 switch (val.sauth_chunk) {
3157 case SCTP_CID_INIT:
3158 case SCTP_CID_INIT_ACK:
3159 case SCTP_CID_SHUTDOWN_COMPLETE:
3160 case SCTP_CID_AUTH:
3161 return -EINVAL;
3164 /* add this chunk id to the endpoint */
3165 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3169 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3171 * This option gets or sets the list of HMAC algorithms that the local
3172 * endpoint requires the peer to use.
3174 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3175 char __user *optval,
3176 int optlen)
3178 struct sctp_hmacalgo *hmacs;
3179 u32 idents;
3180 int err;
3182 if (!sctp_auth_enable)
3183 return -EACCES;
3185 if (optlen < sizeof(struct sctp_hmacalgo))
3186 return -EINVAL;
3188 hmacs = kmalloc(optlen, GFP_KERNEL);
3189 if (!hmacs)
3190 return -ENOMEM;
3192 if (copy_from_user(hmacs, optval, optlen)) {
3193 err = -EFAULT;
3194 goto out;
3197 idents = hmacs->shmac_num_idents;
3198 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3199 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3200 err = -EINVAL;
3201 goto out;
3204 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3205 out:
3206 kfree(hmacs);
3207 return err;
3211 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3213 * This option will set a shared secret key which is used to build an
3214 * association shared key.
3216 static int sctp_setsockopt_auth_key(struct sock *sk,
3217 char __user *optval,
3218 int optlen)
3220 struct sctp_authkey *authkey;
3221 struct sctp_association *asoc;
3222 int ret;
3224 if (!sctp_auth_enable)
3225 return -EACCES;
3227 if (optlen <= sizeof(struct sctp_authkey))
3228 return -EINVAL;
3230 authkey = kmalloc(optlen, GFP_KERNEL);
3231 if (!authkey)
3232 return -ENOMEM;
3234 if (copy_from_user(authkey, optval, optlen)) {
3235 ret = -EFAULT;
3236 goto out;
3239 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3240 ret = -EINVAL;
3241 goto out;
3244 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3245 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3246 ret = -EINVAL;
3247 goto out;
3250 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3251 out:
3252 kfree(authkey);
3253 return ret;
3257 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3259 * This option will get or set the active shared key to be used to build
3260 * the association shared key.
3262 static int sctp_setsockopt_active_key(struct sock *sk,
3263 char __user *optval,
3264 int optlen)
3266 struct sctp_authkeyid val;
3267 struct sctp_association *asoc;
3269 if (!sctp_auth_enable)
3270 return -EACCES;
3272 if (optlen != sizeof(struct sctp_authkeyid))
3273 return -EINVAL;
3274 if (copy_from_user(&val, optval, optlen))
3275 return -EFAULT;
3277 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3278 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3279 return -EINVAL;
3281 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3282 val.scact_keynumber);
3286 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3288 * This set option will delete a shared secret key from use.
3290 static int sctp_setsockopt_del_key(struct sock *sk,
3291 char __user *optval,
3292 int optlen)
3294 struct sctp_authkeyid val;
3295 struct sctp_association *asoc;
3297 if (!sctp_auth_enable)
3298 return -EACCES;
3300 if (optlen != sizeof(struct sctp_authkeyid))
3301 return -EINVAL;
3302 if (copy_from_user(&val, optval, optlen))
3303 return -EFAULT;
3305 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3306 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3307 return -EINVAL;
3309 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3310 val.scact_keynumber);
3315 /* API 6.2 setsockopt(), getsockopt()
3317 * Applications use setsockopt() and getsockopt() to set or retrieve
3318 * socket options. Socket options are used to change the default
3319 * behavior of sockets calls. They are described in Section 7.
3321 * The syntax is:
3323 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3324 * int __user *optlen);
3325 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3326 * int optlen);
3328 * sd - the socket descript.
3329 * level - set to IPPROTO_SCTP for all SCTP options.
3330 * optname - the option name.
3331 * optval - the buffer to store the value of the option.
3332 * optlen - the size of the buffer.
3334 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3335 char __user *optval, int optlen)
3337 int retval = 0;
3339 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3340 sk, optname);
3342 /* I can hardly begin to describe how wrong this is. This is
3343 * so broken as to be worse than useless. The API draft
3344 * REALLY is NOT helpful here... I am not convinced that the
3345 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3346 * are at all well-founded.
3348 if (level != SOL_SCTP) {
3349 struct sctp_af *af = sctp_sk(sk)->pf->af;
3350 retval = af->setsockopt(sk, level, optname, optval, optlen);
3351 goto out_nounlock;
3354 sctp_lock_sock(sk);
3356 switch (optname) {
3357 case SCTP_SOCKOPT_BINDX_ADD:
3358 /* 'optlen' is the size of the addresses buffer. */
3359 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3360 optlen, SCTP_BINDX_ADD_ADDR);
3361 break;
3363 case SCTP_SOCKOPT_BINDX_REM:
3364 /* 'optlen' is the size of the addresses buffer. */
3365 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3366 optlen, SCTP_BINDX_REM_ADDR);
3367 break;
3369 case SCTP_SOCKOPT_CONNECTX_OLD:
3370 /* 'optlen' is the size of the addresses buffer. */
3371 retval = sctp_setsockopt_connectx_old(sk,
3372 (struct sockaddr __user *)optval,
3373 optlen);
3374 break;
3376 case SCTP_SOCKOPT_CONNECTX:
3377 /* 'optlen' is the size of the addresses buffer. */
3378 retval = sctp_setsockopt_connectx(sk,
3379 (struct sockaddr __user *)optval,
3380 optlen);
3381 break;
3383 case SCTP_DISABLE_FRAGMENTS:
3384 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3385 break;
3387 case SCTP_EVENTS:
3388 retval = sctp_setsockopt_events(sk, optval, optlen);
3389 break;
3391 case SCTP_AUTOCLOSE:
3392 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3393 break;
3395 case SCTP_PEER_ADDR_PARAMS:
3396 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3397 break;
3399 case SCTP_DELAYED_ACK:
3400 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3401 break;
3402 case SCTP_PARTIAL_DELIVERY_POINT:
3403 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3404 break;
3406 case SCTP_INITMSG:
3407 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3408 break;
3409 case SCTP_DEFAULT_SEND_PARAM:
3410 retval = sctp_setsockopt_default_send_param(sk, optval,
3411 optlen);
3412 break;
3413 case SCTP_PRIMARY_ADDR:
3414 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3415 break;
3416 case SCTP_SET_PEER_PRIMARY_ADDR:
3417 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3418 break;
3419 case SCTP_NODELAY:
3420 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3421 break;
3422 case SCTP_RTOINFO:
3423 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3424 break;
3425 case SCTP_ASSOCINFO:
3426 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3427 break;
3428 case SCTP_I_WANT_MAPPED_V4_ADDR:
3429 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3430 break;
3431 case SCTP_MAXSEG:
3432 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3433 break;
3434 case SCTP_ADAPTATION_LAYER:
3435 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3436 break;
3437 case SCTP_CONTEXT:
3438 retval = sctp_setsockopt_context(sk, optval, optlen);
3439 break;
3440 case SCTP_FRAGMENT_INTERLEAVE:
3441 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3442 break;
3443 case SCTP_MAX_BURST:
3444 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3445 break;
3446 case SCTP_AUTH_CHUNK:
3447 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3448 break;
3449 case SCTP_HMAC_IDENT:
3450 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3451 break;
3452 case SCTP_AUTH_KEY:
3453 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3454 break;
3455 case SCTP_AUTH_ACTIVE_KEY:
3456 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3457 break;
3458 case SCTP_AUTH_DELETE_KEY:
3459 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3460 break;
3461 default:
3462 retval = -ENOPROTOOPT;
3463 break;
3466 sctp_release_sock(sk);
3468 out_nounlock:
3469 return retval;
3472 /* API 3.1.6 connect() - UDP Style Syntax
3474 * An application may use the connect() call in the UDP model to initiate an
3475 * association without sending data.
3477 * The syntax is:
3479 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3481 * sd: the socket descriptor to have a new association added to.
3483 * nam: the address structure (either struct sockaddr_in or struct
3484 * sockaddr_in6 defined in RFC2553 [7]).
3486 * len: the size of the address.
3488 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3489 int addr_len)
3491 int err = 0;
3492 struct sctp_af *af;
3494 sctp_lock_sock(sk);
3496 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3497 __func__, sk, addr, addr_len);
3499 /* Validate addr_len before calling common connect/connectx routine. */
3500 af = sctp_get_af_specific(addr->sa_family);
3501 if (!af || addr_len < af->sockaddr_len) {
3502 err = -EINVAL;
3503 } else {
3504 /* Pass correct addr len to common routine (so it knows there
3505 * is only one address being passed.
3507 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3510 sctp_release_sock(sk);
3511 return err;
3514 /* FIXME: Write comments. */
3515 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3517 return -EOPNOTSUPP; /* STUB */
3520 /* 4.1.4 accept() - TCP Style Syntax
3522 * Applications use accept() call to remove an established SCTP
3523 * association from the accept queue of the endpoint. A new socket
3524 * descriptor will be returned from accept() to represent the newly
3525 * formed association.
3527 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3529 struct sctp_sock *sp;
3530 struct sctp_endpoint *ep;
3531 struct sock *newsk = NULL;
3532 struct sctp_association *asoc;
3533 long timeo;
3534 int error = 0;
3536 sctp_lock_sock(sk);
3538 sp = sctp_sk(sk);
3539 ep = sp->ep;
3541 if (!sctp_style(sk, TCP)) {
3542 error = -EOPNOTSUPP;
3543 goto out;
3546 if (!sctp_sstate(sk, LISTENING)) {
3547 error = -EINVAL;
3548 goto out;
3551 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3553 error = sctp_wait_for_accept(sk, timeo);
3554 if (error)
3555 goto out;
3557 /* We treat the list of associations on the endpoint as the accept
3558 * queue and pick the first association on the list.
3560 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3562 newsk = sp->pf->create_accept_sk(sk, asoc);
3563 if (!newsk) {
3564 error = -ENOMEM;
3565 goto out;
3568 /* Populate the fields of the newsk from the oldsk and migrate the
3569 * asoc to the newsk.
3571 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3573 out:
3574 sctp_release_sock(sk);
3575 *err = error;
3576 return newsk;
3579 /* The SCTP ioctl handler. */
3580 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3582 return -ENOIOCTLCMD;
3585 /* This is the function which gets called during socket creation to
3586 * initialized the SCTP-specific portion of the sock.
3587 * The sock structure should already be zero-filled memory.
3589 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3591 struct sctp_endpoint *ep;
3592 struct sctp_sock *sp;
3594 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3596 sp = sctp_sk(sk);
3598 /* Initialize the SCTP per socket area. */
3599 switch (sk->sk_type) {
3600 case SOCK_SEQPACKET:
3601 sp->type = SCTP_SOCKET_UDP;
3602 break;
3603 case SOCK_STREAM:
3604 sp->type = SCTP_SOCKET_TCP;
3605 break;
3606 default:
3607 return -ESOCKTNOSUPPORT;
3610 /* Initialize default send parameters. These parameters can be
3611 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3613 sp->default_stream = 0;
3614 sp->default_ppid = 0;
3615 sp->default_flags = 0;
3616 sp->default_context = 0;
3617 sp->default_timetolive = 0;
3619 sp->default_rcv_context = 0;
3620 sp->max_burst = sctp_max_burst;
3622 /* Initialize default setup parameters. These parameters
3623 * can be modified with the SCTP_INITMSG socket option or
3624 * overridden by the SCTP_INIT CMSG.
3626 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3627 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3628 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3629 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3631 /* Initialize default RTO related parameters. These parameters can
3632 * be modified for with the SCTP_RTOINFO socket option.
3634 sp->rtoinfo.srto_initial = sctp_rto_initial;
3635 sp->rtoinfo.srto_max = sctp_rto_max;
3636 sp->rtoinfo.srto_min = sctp_rto_min;
3638 /* Initialize default association related parameters. These parameters
3639 * can be modified with the SCTP_ASSOCINFO socket option.
3641 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3642 sp->assocparams.sasoc_number_peer_destinations = 0;
3643 sp->assocparams.sasoc_peer_rwnd = 0;
3644 sp->assocparams.sasoc_local_rwnd = 0;
3645 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3647 /* Initialize default event subscriptions. By default, all the
3648 * options are off.
3650 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3652 /* Default Peer Address Parameters. These defaults can
3653 * be modified via SCTP_PEER_ADDR_PARAMS
3655 sp->hbinterval = sctp_hb_interval;
3656 sp->pathmaxrxt = sctp_max_retrans_path;
3657 sp->pathmtu = 0; // allow default discovery
3658 sp->sackdelay = sctp_sack_timeout;
3659 sp->sackfreq = 2;
3660 sp->param_flags = SPP_HB_ENABLE |
3661 SPP_PMTUD_ENABLE |
3662 SPP_SACKDELAY_ENABLE;
3664 /* If enabled no SCTP message fragmentation will be performed.
3665 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3667 sp->disable_fragments = 0;
3669 /* Enable Nagle algorithm by default. */
3670 sp->nodelay = 0;
3672 /* Enable by default. */
3673 sp->v4mapped = 1;
3675 /* Auto-close idle associations after the configured
3676 * number of seconds. A value of 0 disables this
3677 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3678 * for UDP-style sockets only.
3680 sp->autoclose = 0;
3682 /* User specified fragmentation limit. */
3683 sp->user_frag = 0;
3685 sp->adaptation_ind = 0;
3687 sp->pf = sctp_get_pf_specific(sk->sk_family);
3689 /* Control variables for partial data delivery. */
3690 atomic_set(&sp->pd_mode, 0);
3691 skb_queue_head_init(&sp->pd_lobby);
3692 sp->frag_interleave = 0;
3694 /* Create a per socket endpoint structure. Even if we
3695 * change the data structure relationships, this may still
3696 * be useful for storing pre-connect address information.
3698 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3699 if (!ep)
3700 return -ENOMEM;
3702 sp->ep = ep;
3703 sp->hmac = NULL;
3705 SCTP_DBG_OBJCNT_INC(sock);
3706 percpu_counter_inc(&sctp_sockets_allocated);
3708 local_bh_disable();
3709 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3710 local_bh_enable();
3712 return 0;
3715 /* Cleanup any SCTP per socket resources. */
3716 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3718 struct sctp_endpoint *ep;
3720 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3722 /* Release our hold on the endpoint. */
3723 ep = sctp_sk(sk)->ep;
3724 sctp_endpoint_free(ep);
3725 percpu_counter_dec(&sctp_sockets_allocated);
3726 local_bh_disable();
3727 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3728 local_bh_enable();
3731 /* API 4.1.7 shutdown() - TCP Style Syntax
3732 * int shutdown(int socket, int how);
3734 * sd - the socket descriptor of the association to be closed.
3735 * how - Specifies the type of shutdown. The values are
3736 * as follows:
3737 * SHUT_RD
3738 * Disables further receive operations. No SCTP
3739 * protocol action is taken.
3740 * SHUT_WR
3741 * Disables further send operations, and initiates
3742 * the SCTP shutdown sequence.
3743 * SHUT_RDWR
3744 * Disables further send and receive operations
3745 * and initiates the SCTP shutdown sequence.
3747 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3749 struct sctp_endpoint *ep;
3750 struct sctp_association *asoc;
3752 if (!sctp_style(sk, TCP))
3753 return;
3755 if (how & SEND_SHUTDOWN) {
3756 ep = sctp_sk(sk)->ep;
3757 if (!list_empty(&ep->asocs)) {
3758 asoc = list_entry(ep->asocs.next,
3759 struct sctp_association, asocs);
3760 sctp_primitive_SHUTDOWN(asoc, NULL);
3765 /* 7.2.1 Association Status (SCTP_STATUS)
3767 * Applications can retrieve current status information about an
3768 * association, including association state, peer receiver window size,
3769 * number of unacked data chunks, and number of data chunks pending
3770 * receipt. This information is read-only.
3772 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3773 char __user *optval,
3774 int __user *optlen)
3776 struct sctp_status status;
3777 struct sctp_association *asoc = NULL;
3778 struct sctp_transport *transport;
3779 sctp_assoc_t associd;
3780 int retval = 0;
3782 if (len < sizeof(status)) {
3783 retval = -EINVAL;
3784 goto out;
3787 len = sizeof(status);
3788 if (copy_from_user(&status, optval, len)) {
3789 retval = -EFAULT;
3790 goto out;
3793 associd = status.sstat_assoc_id;
3794 asoc = sctp_id2assoc(sk, associd);
3795 if (!asoc) {
3796 retval = -EINVAL;
3797 goto out;
3800 transport = asoc->peer.primary_path;
3802 status.sstat_assoc_id = sctp_assoc2id(asoc);
3803 status.sstat_state = asoc->state;
3804 status.sstat_rwnd = asoc->peer.rwnd;
3805 status.sstat_unackdata = asoc->unack_data;
3807 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3808 status.sstat_instrms = asoc->c.sinit_max_instreams;
3809 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3810 status.sstat_fragmentation_point = asoc->frag_point;
3811 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3812 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3813 transport->af_specific->sockaddr_len);
3814 /* Map ipv4 address into v4-mapped-on-v6 address. */
3815 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3816 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3817 status.sstat_primary.spinfo_state = transport->state;
3818 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3819 status.sstat_primary.spinfo_srtt = transport->srtt;
3820 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3821 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3823 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3824 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3826 if (put_user(len, optlen)) {
3827 retval = -EFAULT;
3828 goto out;
3831 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3832 len, status.sstat_state, status.sstat_rwnd,
3833 status.sstat_assoc_id);
3835 if (copy_to_user(optval, &status, len)) {
3836 retval = -EFAULT;
3837 goto out;
3840 out:
3841 return (retval);
3845 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3847 * Applications can retrieve information about a specific peer address
3848 * of an association, including its reachability state, congestion
3849 * window, and retransmission timer values. This information is
3850 * read-only.
3852 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3853 char __user *optval,
3854 int __user *optlen)
3856 struct sctp_paddrinfo pinfo;
3857 struct sctp_transport *transport;
3858 int retval = 0;
3860 if (len < sizeof(pinfo)) {
3861 retval = -EINVAL;
3862 goto out;
3865 len = sizeof(pinfo);
3866 if (copy_from_user(&pinfo, optval, len)) {
3867 retval = -EFAULT;
3868 goto out;
3871 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3872 pinfo.spinfo_assoc_id);
3873 if (!transport)
3874 return -EINVAL;
3876 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3877 pinfo.spinfo_state = transport->state;
3878 pinfo.spinfo_cwnd = transport->cwnd;
3879 pinfo.spinfo_srtt = transport->srtt;
3880 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3881 pinfo.spinfo_mtu = transport->pathmtu;
3883 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3884 pinfo.spinfo_state = SCTP_ACTIVE;
3886 if (put_user(len, optlen)) {
3887 retval = -EFAULT;
3888 goto out;
3891 if (copy_to_user(optval, &pinfo, len)) {
3892 retval = -EFAULT;
3893 goto out;
3896 out:
3897 return (retval);
3900 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3902 * This option is a on/off flag. If enabled no SCTP message
3903 * fragmentation will be performed. Instead if a message being sent
3904 * exceeds the current PMTU size, the message will NOT be sent and
3905 * instead a error will be indicated to the user.
3907 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3908 char __user *optval, int __user *optlen)
3910 int val;
3912 if (len < sizeof(int))
3913 return -EINVAL;
3915 len = sizeof(int);
3916 val = (sctp_sk(sk)->disable_fragments == 1);
3917 if (put_user(len, optlen))
3918 return -EFAULT;
3919 if (copy_to_user(optval, &val, len))
3920 return -EFAULT;
3921 return 0;
3924 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3926 * This socket option is used to specify various notifications and
3927 * ancillary data the user wishes to receive.
3929 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3930 int __user *optlen)
3932 if (len < sizeof(struct sctp_event_subscribe))
3933 return -EINVAL;
3934 len = sizeof(struct sctp_event_subscribe);
3935 if (put_user(len, optlen))
3936 return -EFAULT;
3937 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3938 return -EFAULT;
3939 return 0;
3942 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3944 * This socket option is applicable to the UDP-style socket only. When
3945 * set it will cause associations that are idle for more than the
3946 * specified number of seconds to automatically close. An association
3947 * being idle is defined an association that has NOT sent or received
3948 * user data. The special value of '0' indicates that no automatic
3949 * close of any associations should be performed. The option expects an
3950 * integer defining the number of seconds of idle time before an
3951 * association is closed.
3953 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3955 /* Applicable to UDP-style socket only */
3956 if (sctp_style(sk, TCP))
3957 return -EOPNOTSUPP;
3958 if (len < sizeof(int))
3959 return -EINVAL;
3960 len = sizeof(int);
3961 if (put_user(len, optlen))
3962 return -EFAULT;
3963 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3964 return -EFAULT;
3965 return 0;
3968 /* Helper routine to branch off an association to a new socket. */
3969 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3970 struct socket **sockp)
3972 struct sock *sk = asoc->base.sk;
3973 struct socket *sock;
3974 struct sctp_af *af;
3975 int err = 0;
3977 /* An association cannot be branched off from an already peeled-off
3978 * socket, nor is this supported for tcp style sockets.
3980 if (!sctp_style(sk, UDP))
3981 return -EINVAL;
3983 /* Create a new socket. */
3984 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3985 if (err < 0)
3986 return err;
3988 sctp_copy_sock(sock->sk, sk, asoc);
3990 /* Make peeled-off sockets more like 1-1 accepted sockets.
3991 * Set the daddr and initialize id to something more random
3993 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3994 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3996 /* Populate the fields of the newsk from the oldsk and migrate the
3997 * asoc to the newsk.
3999 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4001 *sockp = sock;
4003 return err;
4006 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4008 sctp_peeloff_arg_t peeloff;
4009 struct socket *newsock;
4010 int retval = 0;
4011 struct sctp_association *asoc;
4013 if (len < sizeof(sctp_peeloff_arg_t))
4014 return -EINVAL;
4015 len = sizeof(sctp_peeloff_arg_t);
4016 if (copy_from_user(&peeloff, optval, len))
4017 return -EFAULT;
4019 asoc = sctp_id2assoc(sk, peeloff.associd);
4020 if (!asoc) {
4021 retval = -EINVAL;
4022 goto out;
4025 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4027 retval = sctp_do_peeloff(asoc, &newsock);
4028 if (retval < 0)
4029 goto out;
4031 /* Map the socket to an unused fd that can be returned to the user. */
4032 retval = sock_map_fd(newsock, 0);
4033 if (retval < 0) {
4034 sock_release(newsock);
4035 goto out;
4038 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4039 __func__, sk, asoc, newsock->sk, retval);
4041 /* Return the fd mapped to the new socket. */
4042 peeloff.sd = retval;
4043 if (put_user(len, optlen))
4044 return -EFAULT;
4045 if (copy_to_user(optval, &peeloff, len))
4046 retval = -EFAULT;
4048 out:
4049 return retval;
4052 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4054 * Applications can enable or disable heartbeats for any peer address of
4055 * an association, modify an address's heartbeat interval, force a
4056 * heartbeat to be sent immediately, and adjust the address's maximum
4057 * number of retransmissions sent before an address is considered
4058 * unreachable. The following structure is used to access and modify an
4059 * address's parameters:
4061 * struct sctp_paddrparams {
4062 * sctp_assoc_t spp_assoc_id;
4063 * struct sockaddr_storage spp_address;
4064 * uint32_t spp_hbinterval;
4065 * uint16_t spp_pathmaxrxt;
4066 * uint32_t spp_pathmtu;
4067 * uint32_t spp_sackdelay;
4068 * uint32_t spp_flags;
4069 * };
4071 * spp_assoc_id - (one-to-many style socket) This is filled in the
4072 * application, and identifies the association for
4073 * this query.
4074 * spp_address - This specifies which address is of interest.
4075 * spp_hbinterval - This contains the value of the heartbeat interval,
4076 * in milliseconds. If a value of zero
4077 * is present in this field then no changes are to
4078 * be made to this parameter.
4079 * spp_pathmaxrxt - This contains the maximum number of
4080 * retransmissions before this address shall be
4081 * considered unreachable. If a value of zero
4082 * is present in this field then no changes are to
4083 * be made to this parameter.
4084 * spp_pathmtu - When Path MTU discovery is disabled the value
4085 * specified here will be the "fixed" path mtu.
4086 * Note that if the spp_address field is empty
4087 * then all associations on this address will
4088 * have this fixed path mtu set upon them.
4090 * spp_sackdelay - When delayed sack is enabled, this value specifies
4091 * the number of milliseconds that sacks will be delayed
4092 * for. This value will apply to all addresses of an
4093 * association if the spp_address field is empty. Note
4094 * also, that if delayed sack is enabled and this
4095 * value is set to 0, no change is made to the last
4096 * recorded delayed sack timer value.
4098 * spp_flags - These flags are used to control various features
4099 * on an association. The flag field may contain
4100 * zero or more of the following options.
4102 * SPP_HB_ENABLE - Enable heartbeats on the
4103 * specified address. Note that if the address
4104 * field is empty all addresses for the association
4105 * have heartbeats enabled upon them.
4107 * SPP_HB_DISABLE - Disable heartbeats on the
4108 * speicifed address. Note that if the address
4109 * field is empty all addresses for the association
4110 * will have their heartbeats disabled. Note also
4111 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4112 * mutually exclusive, only one of these two should
4113 * be specified. Enabling both fields will have
4114 * undetermined results.
4116 * SPP_HB_DEMAND - Request a user initiated heartbeat
4117 * to be made immediately.
4119 * SPP_PMTUD_ENABLE - This field will enable PMTU
4120 * discovery upon the specified address. Note that
4121 * if the address feild is empty then all addresses
4122 * on the association are effected.
4124 * SPP_PMTUD_DISABLE - This field will disable PMTU
4125 * discovery upon the specified address. Note that
4126 * if the address feild is empty then all addresses
4127 * on the association are effected. Not also that
4128 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4129 * exclusive. Enabling both will have undetermined
4130 * results.
4132 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4133 * on delayed sack. The time specified in spp_sackdelay
4134 * is used to specify the sack delay for this address. Note
4135 * that if spp_address is empty then all addresses will
4136 * enable delayed sack and take on the sack delay
4137 * value specified in spp_sackdelay.
4138 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4139 * off delayed sack. If the spp_address field is blank then
4140 * delayed sack is disabled for the entire association. Note
4141 * also that this field is mutually exclusive to
4142 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4143 * results.
4145 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4146 char __user *optval, int __user *optlen)
4148 struct sctp_paddrparams params;
4149 struct sctp_transport *trans = NULL;
4150 struct sctp_association *asoc = NULL;
4151 struct sctp_sock *sp = sctp_sk(sk);
4153 if (len < sizeof(struct sctp_paddrparams))
4154 return -EINVAL;
4155 len = sizeof(struct sctp_paddrparams);
4156 if (copy_from_user(&params, optval, len))
4157 return -EFAULT;
4159 /* If an address other than INADDR_ANY is specified, and
4160 * no transport is found, then the request is invalid.
4162 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4163 trans = sctp_addr_id2transport(sk, &params.spp_address,
4164 params.spp_assoc_id);
4165 if (!trans) {
4166 SCTP_DEBUG_PRINTK("Failed no transport\n");
4167 return -EINVAL;
4171 /* Get association, if assoc_id != 0 and the socket is a one
4172 * to many style socket, and an association was not found, then
4173 * the id was invalid.
4175 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4176 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4177 SCTP_DEBUG_PRINTK("Failed no association\n");
4178 return -EINVAL;
4181 if (trans) {
4182 /* Fetch transport values. */
4183 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4184 params.spp_pathmtu = trans->pathmtu;
4185 params.spp_pathmaxrxt = trans->pathmaxrxt;
4186 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4188 /*draft-11 doesn't say what to return in spp_flags*/
4189 params.spp_flags = trans->param_flags;
4190 } else if (asoc) {
4191 /* Fetch association values. */
4192 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4193 params.spp_pathmtu = asoc->pathmtu;
4194 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4195 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4197 /*draft-11 doesn't say what to return in spp_flags*/
4198 params.spp_flags = asoc->param_flags;
4199 } else {
4200 /* Fetch socket values. */
4201 params.spp_hbinterval = sp->hbinterval;
4202 params.spp_pathmtu = sp->pathmtu;
4203 params.spp_sackdelay = sp->sackdelay;
4204 params.spp_pathmaxrxt = sp->pathmaxrxt;
4206 /*draft-11 doesn't say what to return in spp_flags*/
4207 params.spp_flags = sp->param_flags;
4210 if (copy_to_user(optval, &params, len))
4211 return -EFAULT;
4213 if (put_user(len, optlen))
4214 return -EFAULT;
4216 return 0;
4220 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4222 * This option will effect the way delayed acks are performed. This
4223 * option allows you to get or set the delayed ack time, in
4224 * milliseconds. It also allows changing the delayed ack frequency.
4225 * Changing the frequency to 1 disables the delayed sack algorithm. If
4226 * the assoc_id is 0, then this sets or gets the endpoints default
4227 * values. If the assoc_id field is non-zero, then the set or get
4228 * effects the specified association for the one to many model (the
4229 * assoc_id field is ignored by the one to one model). Note that if
4230 * sack_delay or sack_freq are 0 when setting this option, then the
4231 * current values will remain unchanged.
4233 * struct sctp_sack_info {
4234 * sctp_assoc_t sack_assoc_id;
4235 * uint32_t sack_delay;
4236 * uint32_t sack_freq;
4237 * };
4239 * sack_assoc_id - This parameter, indicates which association the user
4240 * is performing an action upon. Note that if this field's value is
4241 * zero then the endpoints default value is changed (effecting future
4242 * associations only).
4244 * sack_delay - This parameter contains the number of milliseconds that
4245 * the user is requesting the delayed ACK timer be set to. Note that
4246 * this value is defined in the standard to be between 200 and 500
4247 * milliseconds.
4249 * sack_freq - This parameter contains the number of packets that must
4250 * be received before a sack is sent without waiting for the delay
4251 * timer to expire. The default value for this is 2, setting this
4252 * value to 1 will disable the delayed sack algorithm.
4254 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4255 char __user *optval,
4256 int __user *optlen)
4258 struct sctp_sack_info params;
4259 struct sctp_association *asoc = NULL;
4260 struct sctp_sock *sp = sctp_sk(sk);
4262 if (len >= sizeof(struct sctp_sack_info)) {
4263 len = sizeof(struct sctp_sack_info);
4265 if (copy_from_user(&params, optval, len))
4266 return -EFAULT;
4267 } else if (len == sizeof(struct sctp_assoc_value)) {
4268 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4269 "in delayed_ack socket option deprecated\n");
4270 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4271 if (copy_from_user(&params, optval, len))
4272 return -EFAULT;
4273 } else
4274 return - EINVAL;
4276 /* Get association, if sack_assoc_id != 0 and the socket is a one
4277 * to many style socket, and an association was not found, then
4278 * the id was invalid.
4280 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4281 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4282 return -EINVAL;
4284 if (asoc) {
4285 /* Fetch association values. */
4286 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4287 params.sack_delay = jiffies_to_msecs(
4288 asoc->sackdelay);
4289 params.sack_freq = asoc->sackfreq;
4291 } else {
4292 params.sack_delay = 0;
4293 params.sack_freq = 1;
4295 } else {
4296 /* Fetch socket values. */
4297 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4298 params.sack_delay = sp->sackdelay;
4299 params.sack_freq = sp->sackfreq;
4300 } else {
4301 params.sack_delay = 0;
4302 params.sack_freq = 1;
4306 if (copy_to_user(optval, &params, len))
4307 return -EFAULT;
4309 if (put_user(len, optlen))
4310 return -EFAULT;
4312 return 0;
4315 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4317 * Applications can specify protocol parameters for the default association
4318 * initialization. The option name argument to setsockopt() and getsockopt()
4319 * is SCTP_INITMSG.
4321 * Setting initialization parameters is effective only on an unconnected
4322 * socket (for UDP-style sockets only future associations are effected
4323 * by the change). With TCP-style sockets, this option is inherited by
4324 * sockets derived from a listener socket.
4326 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4328 if (len < sizeof(struct sctp_initmsg))
4329 return -EINVAL;
4330 len = sizeof(struct sctp_initmsg);
4331 if (put_user(len, optlen))
4332 return -EFAULT;
4333 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4334 return -EFAULT;
4335 return 0;
4338 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4339 char __user *optval,
4340 int __user *optlen)
4342 sctp_assoc_t id;
4343 struct sctp_association *asoc;
4344 struct list_head *pos;
4345 int cnt = 0;
4347 if (len < sizeof(sctp_assoc_t))
4348 return -EINVAL;
4350 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4351 return -EFAULT;
4353 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4354 "socket option deprecated\n");
4355 /* For UDP-style sockets, id specifies the association to query. */
4356 asoc = sctp_id2assoc(sk, id);
4357 if (!asoc)
4358 return -EINVAL;
4360 list_for_each(pos, &asoc->peer.transport_addr_list) {
4361 cnt ++;
4364 return cnt;
4368 * Old API for getting list of peer addresses. Does not work for 32-bit
4369 * programs running on a 64-bit kernel
4371 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4372 char __user *optval,
4373 int __user *optlen)
4375 struct sctp_association *asoc;
4376 int cnt = 0;
4377 struct sctp_getaddrs_old getaddrs;
4378 struct sctp_transport *from;
4379 void __user *to;
4380 union sctp_addr temp;
4381 struct sctp_sock *sp = sctp_sk(sk);
4382 int addrlen;
4384 if (len < sizeof(struct sctp_getaddrs_old))
4385 return -EINVAL;
4387 len = sizeof(struct sctp_getaddrs_old);
4389 if (copy_from_user(&getaddrs, optval, len))
4390 return -EFAULT;
4392 if (getaddrs.addr_num <= 0) return -EINVAL;
4394 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4395 "socket option deprecated\n");
4397 /* For UDP-style sockets, id specifies the association to query. */
4398 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4399 if (!asoc)
4400 return -EINVAL;
4402 to = (void __user *)getaddrs.addrs;
4403 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4404 transports) {
4405 memcpy(&temp, &from->ipaddr, sizeof(temp));
4406 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4407 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4408 if (copy_to_user(to, &temp, addrlen))
4409 return -EFAULT;
4410 to += addrlen ;
4411 cnt ++;
4412 if (cnt >= getaddrs.addr_num) break;
4414 getaddrs.addr_num = cnt;
4415 if (put_user(len, optlen))
4416 return -EFAULT;
4417 if (copy_to_user(optval, &getaddrs, len))
4418 return -EFAULT;
4420 return 0;
4423 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4424 char __user *optval, int __user *optlen)
4426 struct sctp_association *asoc;
4427 int cnt = 0;
4428 struct sctp_getaddrs getaddrs;
4429 struct sctp_transport *from;
4430 void __user *to;
4431 union sctp_addr temp;
4432 struct sctp_sock *sp = sctp_sk(sk);
4433 int addrlen;
4434 size_t space_left;
4435 int bytes_copied;
4437 if (len < sizeof(struct sctp_getaddrs))
4438 return -EINVAL;
4440 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4441 return -EFAULT;
4443 /* For UDP-style sockets, id specifies the association to query. */
4444 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4445 if (!asoc)
4446 return -EINVAL;
4448 to = optval + offsetof(struct sctp_getaddrs,addrs);
4449 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4451 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4452 transports) {
4453 memcpy(&temp, &from->ipaddr, sizeof(temp));
4454 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4455 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4456 if (space_left < addrlen)
4457 return -ENOMEM;
4458 if (copy_to_user(to, &temp, addrlen))
4459 return -EFAULT;
4460 to += addrlen;
4461 cnt++;
4462 space_left -= addrlen;
4465 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4466 return -EFAULT;
4467 bytes_copied = ((char __user *)to) - optval;
4468 if (put_user(bytes_copied, optlen))
4469 return -EFAULT;
4471 return 0;
4474 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4475 char __user *optval,
4476 int __user *optlen)
4478 sctp_assoc_t id;
4479 struct sctp_bind_addr *bp;
4480 struct sctp_association *asoc;
4481 struct sctp_sockaddr_entry *addr;
4482 int cnt = 0;
4484 if (len < sizeof(sctp_assoc_t))
4485 return -EINVAL;
4487 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4488 return -EFAULT;
4490 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4491 "socket option deprecated\n");
4494 * For UDP-style sockets, id specifies the association to query.
4495 * If the id field is set to the value '0' then the locally bound
4496 * addresses are returned without regard to any particular
4497 * association.
4499 if (0 == id) {
4500 bp = &sctp_sk(sk)->ep->base.bind_addr;
4501 } else {
4502 asoc = sctp_id2assoc(sk, id);
4503 if (!asoc)
4504 return -EINVAL;
4505 bp = &asoc->base.bind_addr;
4508 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4509 * addresses from the global local address list.
4511 if (sctp_list_single_entry(&bp->address_list)) {
4512 addr = list_entry(bp->address_list.next,
4513 struct sctp_sockaddr_entry, list);
4514 if (sctp_is_any(sk, &addr->a)) {
4515 rcu_read_lock();
4516 list_for_each_entry_rcu(addr,
4517 &sctp_local_addr_list, list) {
4518 if (!addr->valid)
4519 continue;
4521 if ((PF_INET == sk->sk_family) &&
4522 (AF_INET6 == addr->a.sa.sa_family))
4523 continue;
4525 if ((PF_INET6 == sk->sk_family) &&
4526 inet_v6_ipv6only(sk) &&
4527 (AF_INET == addr->a.sa.sa_family))
4528 continue;
4530 cnt++;
4532 rcu_read_unlock();
4533 } else {
4534 cnt = 1;
4536 goto done;
4539 /* Protection on the bound address list is not needed,
4540 * since in the socket option context we hold the socket lock,
4541 * so there is no way that the bound address list can change.
4543 list_for_each_entry(addr, &bp->address_list, list) {
4544 cnt ++;
4546 done:
4547 return cnt;
4550 /* Helper function that copies local addresses to user and returns the number
4551 * of addresses copied.
4553 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4554 int max_addrs, void *to,
4555 int *bytes_copied)
4557 struct sctp_sockaddr_entry *addr;
4558 union sctp_addr temp;
4559 int cnt = 0;
4560 int addrlen;
4562 rcu_read_lock();
4563 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4564 if (!addr->valid)
4565 continue;
4567 if ((PF_INET == sk->sk_family) &&
4568 (AF_INET6 == addr->a.sa.sa_family))
4569 continue;
4570 if ((PF_INET6 == sk->sk_family) &&
4571 inet_v6_ipv6only(sk) &&
4572 (AF_INET == addr->a.sa.sa_family))
4573 continue;
4574 memcpy(&temp, &addr->a, sizeof(temp));
4575 if (!temp.v4.sin_port)
4576 temp.v4.sin_port = htons(port);
4578 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4579 &temp);
4580 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4581 memcpy(to, &temp, addrlen);
4583 to += addrlen;
4584 *bytes_copied += addrlen;
4585 cnt ++;
4586 if (cnt >= max_addrs) break;
4588 rcu_read_unlock();
4590 return cnt;
4593 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4594 size_t space_left, int *bytes_copied)
4596 struct sctp_sockaddr_entry *addr;
4597 union sctp_addr temp;
4598 int cnt = 0;
4599 int addrlen;
4601 rcu_read_lock();
4602 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4603 if (!addr->valid)
4604 continue;
4606 if ((PF_INET == sk->sk_family) &&
4607 (AF_INET6 == addr->a.sa.sa_family))
4608 continue;
4609 if ((PF_INET6 == sk->sk_family) &&
4610 inet_v6_ipv6only(sk) &&
4611 (AF_INET == addr->a.sa.sa_family))
4612 continue;
4613 memcpy(&temp, &addr->a, sizeof(temp));
4614 if (!temp.v4.sin_port)
4615 temp.v4.sin_port = htons(port);
4617 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4618 &temp);
4619 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4620 if (space_left < addrlen) {
4621 cnt = -ENOMEM;
4622 break;
4624 memcpy(to, &temp, addrlen);
4626 to += addrlen;
4627 cnt ++;
4628 space_left -= addrlen;
4629 *bytes_copied += addrlen;
4631 rcu_read_unlock();
4633 return cnt;
4636 /* Old API for getting list of local addresses. Does not work for 32-bit
4637 * programs running on a 64-bit kernel
4639 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4640 char __user *optval, int __user *optlen)
4642 struct sctp_bind_addr *bp;
4643 struct sctp_association *asoc;
4644 int cnt = 0;
4645 struct sctp_getaddrs_old getaddrs;
4646 struct sctp_sockaddr_entry *addr;
4647 void __user *to;
4648 union sctp_addr temp;
4649 struct sctp_sock *sp = sctp_sk(sk);
4650 int addrlen;
4651 int err = 0;
4652 void *addrs;
4653 void *buf;
4654 int bytes_copied = 0;
4656 if (len < sizeof(struct sctp_getaddrs_old))
4657 return -EINVAL;
4659 len = sizeof(struct sctp_getaddrs_old);
4660 if (copy_from_user(&getaddrs, optval, len))
4661 return -EFAULT;
4663 if (getaddrs.addr_num <= 0 ||
4664 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4665 return -EINVAL;
4667 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4668 "socket option deprecated\n");
4671 * For UDP-style sockets, id specifies the association to query.
4672 * If the id field is set to the value '0' then the locally bound
4673 * addresses are returned without regard to any particular
4674 * association.
4676 if (0 == getaddrs.assoc_id) {
4677 bp = &sctp_sk(sk)->ep->base.bind_addr;
4678 } else {
4679 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4680 if (!asoc)
4681 return -EINVAL;
4682 bp = &asoc->base.bind_addr;
4685 to = getaddrs.addrs;
4687 /* Allocate space for a local instance of packed array to hold all
4688 * the data. We store addresses here first and then put write them
4689 * to the user in one shot.
4691 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4692 GFP_KERNEL);
4693 if (!addrs)
4694 return -ENOMEM;
4696 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4697 * addresses from the global local address list.
4699 if (sctp_list_single_entry(&bp->address_list)) {
4700 addr = list_entry(bp->address_list.next,
4701 struct sctp_sockaddr_entry, list);
4702 if (sctp_is_any(sk, &addr->a)) {
4703 cnt = sctp_copy_laddrs_old(sk, bp->port,
4704 getaddrs.addr_num,
4705 addrs, &bytes_copied);
4706 goto copy_getaddrs;
4710 buf = addrs;
4711 /* Protection on the bound address list is not needed since
4712 * in the socket option context we hold a socket lock and
4713 * thus the bound address list can't change.
4715 list_for_each_entry(addr, &bp->address_list, list) {
4716 memcpy(&temp, &addr->a, sizeof(temp));
4717 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4718 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4719 memcpy(buf, &temp, addrlen);
4720 buf += addrlen;
4721 bytes_copied += addrlen;
4722 cnt ++;
4723 if (cnt >= getaddrs.addr_num) break;
4726 copy_getaddrs:
4727 /* copy the entire address list into the user provided space */
4728 if (copy_to_user(to, addrs, bytes_copied)) {
4729 err = -EFAULT;
4730 goto error;
4733 /* copy the leading structure back to user */
4734 getaddrs.addr_num = cnt;
4735 if (copy_to_user(optval, &getaddrs, len))
4736 err = -EFAULT;
4738 error:
4739 kfree(addrs);
4740 return err;
4743 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4744 char __user *optval, int __user *optlen)
4746 struct sctp_bind_addr *bp;
4747 struct sctp_association *asoc;
4748 int cnt = 0;
4749 struct sctp_getaddrs getaddrs;
4750 struct sctp_sockaddr_entry *addr;
4751 void __user *to;
4752 union sctp_addr temp;
4753 struct sctp_sock *sp = sctp_sk(sk);
4754 int addrlen;
4755 int err = 0;
4756 size_t space_left;
4757 int bytes_copied = 0;
4758 void *addrs;
4759 void *buf;
4761 if (len < sizeof(struct sctp_getaddrs))
4762 return -EINVAL;
4764 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4765 return -EFAULT;
4768 * For UDP-style sockets, id specifies the association to query.
4769 * If the id field is set to the value '0' then the locally bound
4770 * addresses are returned without regard to any particular
4771 * association.
4773 if (0 == getaddrs.assoc_id) {
4774 bp = &sctp_sk(sk)->ep->base.bind_addr;
4775 } else {
4776 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4777 if (!asoc)
4778 return -EINVAL;
4779 bp = &asoc->base.bind_addr;
4782 to = optval + offsetof(struct sctp_getaddrs,addrs);
4783 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4785 addrs = kmalloc(space_left, GFP_KERNEL);
4786 if (!addrs)
4787 return -ENOMEM;
4789 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4790 * addresses from the global local address list.
4792 if (sctp_list_single_entry(&bp->address_list)) {
4793 addr = list_entry(bp->address_list.next,
4794 struct sctp_sockaddr_entry, list);
4795 if (sctp_is_any(sk, &addr->a)) {
4796 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4797 space_left, &bytes_copied);
4798 if (cnt < 0) {
4799 err = cnt;
4800 goto out;
4802 goto copy_getaddrs;
4806 buf = addrs;
4807 /* Protection on the bound address list is not needed since
4808 * in the socket option context we hold a socket lock and
4809 * thus the bound address list can't change.
4811 list_for_each_entry(addr, &bp->address_list, list) {
4812 memcpy(&temp, &addr->a, sizeof(temp));
4813 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4814 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4815 if (space_left < addrlen) {
4816 err = -ENOMEM; /*fixme: right error?*/
4817 goto out;
4819 memcpy(buf, &temp, addrlen);
4820 buf += addrlen;
4821 bytes_copied += addrlen;
4822 cnt ++;
4823 space_left -= addrlen;
4826 copy_getaddrs:
4827 if (copy_to_user(to, addrs, bytes_copied)) {
4828 err = -EFAULT;
4829 goto out;
4831 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4832 err = -EFAULT;
4833 goto out;
4835 if (put_user(bytes_copied, optlen))
4836 err = -EFAULT;
4837 out:
4838 kfree(addrs);
4839 return err;
4842 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4844 * Requests that the local SCTP stack use the enclosed peer address as
4845 * the association primary. The enclosed address must be one of the
4846 * association peer's addresses.
4848 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4849 char __user *optval, int __user *optlen)
4851 struct sctp_prim prim;
4852 struct sctp_association *asoc;
4853 struct sctp_sock *sp = sctp_sk(sk);
4855 if (len < sizeof(struct sctp_prim))
4856 return -EINVAL;
4858 len = sizeof(struct sctp_prim);
4860 if (copy_from_user(&prim, optval, len))
4861 return -EFAULT;
4863 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4864 if (!asoc)
4865 return -EINVAL;
4867 if (!asoc->peer.primary_path)
4868 return -ENOTCONN;
4870 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4871 asoc->peer.primary_path->af_specific->sockaddr_len);
4873 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4874 (union sctp_addr *)&prim.ssp_addr);
4876 if (put_user(len, optlen))
4877 return -EFAULT;
4878 if (copy_to_user(optval, &prim, len))
4879 return -EFAULT;
4881 return 0;
4885 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4887 * Requests that the local endpoint set the specified Adaptation Layer
4888 * Indication parameter for all future INIT and INIT-ACK exchanges.
4890 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4891 char __user *optval, int __user *optlen)
4893 struct sctp_setadaptation adaptation;
4895 if (len < sizeof(struct sctp_setadaptation))
4896 return -EINVAL;
4898 len = sizeof(struct sctp_setadaptation);
4900 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4902 if (put_user(len, optlen))
4903 return -EFAULT;
4904 if (copy_to_user(optval, &adaptation, len))
4905 return -EFAULT;
4907 return 0;
4912 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4914 * Applications that wish to use the sendto() system call may wish to
4915 * specify a default set of parameters that would normally be supplied
4916 * through the inclusion of ancillary data. This socket option allows
4917 * such an application to set the default sctp_sndrcvinfo structure.
4920 * The application that wishes to use this socket option simply passes
4921 * in to this call the sctp_sndrcvinfo structure defined in Section
4922 * 5.2.2) The input parameters accepted by this call include
4923 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4924 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4925 * to this call if the caller is using the UDP model.
4927 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4929 static int sctp_getsockopt_default_send_param(struct sock *sk,
4930 int len, char __user *optval,
4931 int __user *optlen)
4933 struct sctp_sndrcvinfo info;
4934 struct sctp_association *asoc;
4935 struct sctp_sock *sp = sctp_sk(sk);
4937 if (len < sizeof(struct sctp_sndrcvinfo))
4938 return -EINVAL;
4940 len = sizeof(struct sctp_sndrcvinfo);
4942 if (copy_from_user(&info, optval, len))
4943 return -EFAULT;
4945 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4946 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4947 return -EINVAL;
4949 if (asoc) {
4950 info.sinfo_stream = asoc->default_stream;
4951 info.sinfo_flags = asoc->default_flags;
4952 info.sinfo_ppid = asoc->default_ppid;
4953 info.sinfo_context = asoc->default_context;
4954 info.sinfo_timetolive = asoc->default_timetolive;
4955 } else {
4956 info.sinfo_stream = sp->default_stream;
4957 info.sinfo_flags = sp->default_flags;
4958 info.sinfo_ppid = sp->default_ppid;
4959 info.sinfo_context = sp->default_context;
4960 info.sinfo_timetolive = sp->default_timetolive;
4963 if (put_user(len, optlen))
4964 return -EFAULT;
4965 if (copy_to_user(optval, &info, len))
4966 return -EFAULT;
4968 return 0;
4973 * 7.1.5 SCTP_NODELAY
4975 * Turn on/off any Nagle-like algorithm. This means that packets are
4976 * generally sent as soon as possible and no unnecessary delays are
4977 * introduced, at the cost of more packets in the network. Expects an
4978 * integer boolean flag.
4981 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4982 char __user *optval, int __user *optlen)
4984 int val;
4986 if (len < sizeof(int))
4987 return -EINVAL;
4989 len = sizeof(int);
4990 val = (sctp_sk(sk)->nodelay == 1);
4991 if (put_user(len, optlen))
4992 return -EFAULT;
4993 if (copy_to_user(optval, &val, len))
4994 return -EFAULT;
4995 return 0;
5000 * 7.1.1 SCTP_RTOINFO
5002 * The protocol parameters used to initialize and bound retransmission
5003 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5004 * and modify these parameters.
5005 * All parameters are time values, in milliseconds. A value of 0, when
5006 * modifying the parameters, indicates that the current value should not
5007 * be changed.
5010 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5011 char __user *optval,
5012 int __user *optlen) {
5013 struct sctp_rtoinfo rtoinfo;
5014 struct sctp_association *asoc;
5016 if (len < sizeof (struct sctp_rtoinfo))
5017 return -EINVAL;
5019 len = sizeof(struct sctp_rtoinfo);
5021 if (copy_from_user(&rtoinfo, optval, len))
5022 return -EFAULT;
5024 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5026 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5027 return -EINVAL;
5029 /* Values corresponding to the specific association. */
5030 if (asoc) {
5031 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5032 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5033 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5034 } else {
5035 /* Values corresponding to the endpoint. */
5036 struct sctp_sock *sp = sctp_sk(sk);
5038 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5039 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5040 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5043 if (put_user(len, optlen))
5044 return -EFAULT;
5046 if (copy_to_user(optval, &rtoinfo, len))
5047 return -EFAULT;
5049 return 0;
5054 * 7.1.2 SCTP_ASSOCINFO
5056 * This option is used to tune the maximum retransmission attempts
5057 * of the association.
5058 * Returns an error if the new association retransmission value is
5059 * greater than the sum of the retransmission value of the peer.
5060 * See [SCTP] for more information.
5063 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5064 char __user *optval,
5065 int __user *optlen)
5068 struct sctp_assocparams assocparams;
5069 struct sctp_association *asoc;
5070 struct list_head *pos;
5071 int cnt = 0;
5073 if (len < sizeof (struct sctp_assocparams))
5074 return -EINVAL;
5076 len = sizeof(struct sctp_assocparams);
5078 if (copy_from_user(&assocparams, optval, len))
5079 return -EFAULT;
5081 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5083 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5084 return -EINVAL;
5086 /* Values correspoinding to the specific association */
5087 if (asoc) {
5088 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5089 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5090 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5091 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5092 * 1000) +
5093 (asoc->cookie_life.tv_usec
5094 / 1000);
5096 list_for_each(pos, &asoc->peer.transport_addr_list) {
5097 cnt ++;
5100 assocparams.sasoc_number_peer_destinations = cnt;
5101 } else {
5102 /* Values corresponding to the endpoint */
5103 struct sctp_sock *sp = sctp_sk(sk);
5105 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5106 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5107 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5108 assocparams.sasoc_cookie_life =
5109 sp->assocparams.sasoc_cookie_life;
5110 assocparams.sasoc_number_peer_destinations =
5111 sp->assocparams.
5112 sasoc_number_peer_destinations;
5115 if (put_user(len, optlen))
5116 return -EFAULT;
5118 if (copy_to_user(optval, &assocparams, len))
5119 return -EFAULT;
5121 return 0;
5125 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5127 * This socket option is a boolean flag which turns on or off mapped V4
5128 * addresses. If this option is turned on and the socket is type
5129 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5130 * If this option is turned off, then no mapping will be done of V4
5131 * addresses and a user will receive both PF_INET6 and PF_INET type
5132 * addresses on the socket.
5134 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5135 char __user *optval, int __user *optlen)
5137 int val;
5138 struct sctp_sock *sp = sctp_sk(sk);
5140 if (len < sizeof(int))
5141 return -EINVAL;
5143 len = sizeof(int);
5144 val = sp->v4mapped;
5145 if (put_user(len, optlen))
5146 return -EFAULT;
5147 if (copy_to_user(optval, &val, len))
5148 return -EFAULT;
5150 return 0;
5154 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5155 * (chapter and verse is quoted at sctp_setsockopt_context())
5157 static int sctp_getsockopt_context(struct sock *sk, int len,
5158 char __user *optval, int __user *optlen)
5160 struct sctp_assoc_value params;
5161 struct sctp_sock *sp;
5162 struct sctp_association *asoc;
5164 if (len < sizeof(struct sctp_assoc_value))
5165 return -EINVAL;
5167 len = sizeof(struct sctp_assoc_value);
5169 if (copy_from_user(&params, optval, len))
5170 return -EFAULT;
5172 sp = sctp_sk(sk);
5174 if (params.assoc_id != 0) {
5175 asoc = sctp_id2assoc(sk, params.assoc_id);
5176 if (!asoc)
5177 return -EINVAL;
5178 params.assoc_value = asoc->default_rcv_context;
5179 } else {
5180 params.assoc_value = sp->default_rcv_context;
5183 if (put_user(len, optlen))
5184 return -EFAULT;
5185 if (copy_to_user(optval, &params, len))
5186 return -EFAULT;
5188 return 0;
5192 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5193 * This option will get or set the maximum size to put in any outgoing
5194 * SCTP DATA chunk. If a message is larger than this size it will be
5195 * fragmented by SCTP into the specified size. Note that the underlying
5196 * SCTP implementation may fragment into smaller sized chunks when the
5197 * PMTU of the underlying association is smaller than the value set by
5198 * the user. The default value for this option is '0' which indicates
5199 * the user is NOT limiting fragmentation and only the PMTU will effect
5200 * SCTP's choice of DATA chunk size. Note also that values set larger
5201 * than the maximum size of an IP datagram will effectively let SCTP
5202 * control fragmentation (i.e. the same as setting this option to 0).
5204 * The following structure is used to access and modify this parameter:
5206 * struct sctp_assoc_value {
5207 * sctp_assoc_t assoc_id;
5208 * uint32_t assoc_value;
5209 * };
5211 * assoc_id: This parameter is ignored for one-to-one style sockets.
5212 * For one-to-many style sockets this parameter indicates which
5213 * association the user is performing an action upon. Note that if
5214 * this field's value is zero then the endpoints default value is
5215 * changed (effecting future associations only).
5216 * assoc_value: This parameter specifies the maximum size in bytes.
5218 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5219 char __user *optval, int __user *optlen)
5221 struct sctp_assoc_value params;
5222 struct sctp_association *asoc;
5224 if (len == sizeof(int)) {
5225 printk(KERN_WARNING
5226 "SCTP: Use of int in maxseg socket option deprecated\n");
5227 printk(KERN_WARNING
5228 "SCTP: Use struct sctp_assoc_value instead\n");
5229 params.assoc_id = 0;
5230 } else if (len >= sizeof(struct sctp_assoc_value)) {
5231 len = sizeof(struct sctp_assoc_value);
5232 if (copy_from_user(&params, optval, sizeof(params)))
5233 return -EFAULT;
5234 } else
5235 return -EINVAL;
5237 asoc = sctp_id2assoc(sk, params.assoc_id);
5238 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5239 return -EINVAL;
5241 if (asoc)
5242 params.assoc_value = asoc->frag_point;
5243 else
5244 params.assoc_value = sctp_sk(sk)->user_frag;
5246 if (put_user(len, optlen))
5247 return -EFAULT;
5248 if (len == sizeof(int)) {
5249 if (copy_to_user(optval, &params.assoc_value, len))
5250 return -EFAULT;
5251 } else {
5252 if (copy_to_user(optval, &params, len))
5253 return -EFAULT;
5256 return 0;
5260 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5261 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5263 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5264 char __user *optval, int __user *optlen)
5266 int val;
5268 if (len < sizeof(int))
5269 return -EINVAL;
5271 len = sizeof(int);
5273 val = sctp_sk(sk)->frag_interleave;
5274 if (put_user(len, optlen))
5275 return -EFAULT;
5276 if (copy_to_user(optval, &val, len))
5277 return -EFAULT;
5279 return 0;
5283 * 7.1.25. Set or Get the sctp partial delivery point
5284 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5286 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5287 char __user *optval,
5288 int __user *optlen)
5290 u32 val;
5292 if (len < sizeof(u32))
5293 return -EINVAL;
5295 len = sizeof(u32);
5297 val = sctp_sk(sk)->pd_point;
5298 if (put_user(len, optlen))
5299 return -EFAULT;
5300 if (copy_to_user(optval, &val, len))
5301 return -EFAULT;
5303 return -ENOTSUPP;
5307 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5308 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5310 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5311 char __user *optval,
5312 int __user *optlen)
5314 struct sctp_assoc_value params;
5315 struct sctp_sock *sp;
5316 struct sctp_association *asoc;
5318 if (len == sizeof(int)) {
5319 printk(KERN_WARNING
5320 "SCTP: Use of int in max_burst socket option deprecated\n");
5321 printk(KERN_WARNING
5322 "SCTP: Use struct sctp_assoc_value instead\n");
5323 params.assoc_id = 0;
5324 } else if (len >= sizeof(struct sctp_assoc_value)) {
5325 len = sizeof(struct sctp_assoc_value);
5326 if (copy_from_user(&params, optval, len))
5327 return -EFAULT;
5328 } else
5329 return -EINVAL;
5331 sp = sctp_sk(sk);
5333 if (params.assoc_id != 0) {
5334 asoc = sctp_id2assoc(sk, params.assoc_id);
5335 if (!asoc)
5336 return -EINVAL;
5337 params.assoc_value = asoc->max_burst;
5338 } else
5339 params.assoc_value = sp->max_burst;
5341 if (len == sizeof(int)) {
5342 if (copy_to_user(optval, &params.assoc_value, len))
5343 return -EFAULT;
5344 } else {
5345 if (copy_to_user(optval, &params, len))
5346 return -EFAULT;
5349 return 0;
5353 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5354 char __user *optval, int __user *optlen)
5356 struct sctp_hmacalgo __user *p = (void __user *)optval;
5357 struct sctp_hmac_algo_param *hmacs;
5358 __u16 data_len = 0;
5359 u32 num_idents;
5361 if (!sctp_auth_enable)
5362 return -EACCES;
5364 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5365 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5367 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5368 return -EINVAL;
5370 len = sizeof(struct sctp_hmacalgo) + data_len;
5371 num_idents = data_len / sizeof(u16);
5373 if (put_user(len, optlen))
5374 return -EFAULT;
5375 if (put_user(num_idents, &p->shmac_num_idents))
5376 return -EFAULT;
5377 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5378 return -EFAULT;
5379 return 0;
5382 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5383 char __user *optval, int __user *optlen)
5385 struct sctp_authkeyid val;
5386 struct sctp_association *asoc;
5388 if (!sctp_auth_enable)
5389 return -EACCES;
5391 if (len < sizeof(struct sctp_authkeyid))
5392 return -EINVAL;
5393 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5394 return -EFAULT;
5396 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5397 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5398 return -EINVAL;
5400 if (asoc)
5401 val.scact_keynumber = asoc->active_key_id;
5402 else
5403 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5405 len = sizeof(struct sctp_authkeyid);
5406 if (put_user(len, optlen))
5407 return -EFAULT;
5408 if (copy_to_user(optval, &val, len))
5409 return -EFAULT;
5411 return 0;
5414 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5415 char __user *optval, int __user *optlen)
5417 struct sctp_authchunks __user *p = (void __user *)optval;
5418 struct sctp_authchunks val;
5419 struct sctp_association *asoc;
5420 struct sctp_chunks_param *ch;
5421 u32 num_chunks = 0;
5422 char __user *to;
5424 if (!sctp_auth_enable)
5425 return -EACCES;
5427 if (len < sizeof(struct sctp_authchunks))
5428 return -EINVAL;
5430 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5431 return -EFAULT;
5433 to = p->gauth_chunks;
5434 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5435 if (!asoc)
5436 return -EINVAL;
5438 ch = asoc->peer.peer_chunks;
5439 if (!ch)
5440 goto num;
5442 /* See if the user provided enough room for all the data */
5443 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5444 if (len < num_chunks)
5445 return -EINVAL;
5447 if (copy_to_user(to, ch->chunks, num_chunks))
5448 return -EFAULT;
5449 num:
5450 len = sizeof(struct sctp_authchunks) + num_chunks;
5451 if (put_user(len, optlen)) return -EFAULT;
5452 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5453 return -EFAULT;
5454 return 0;
5457 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5458 char __user *optval, int __user *optlen)
5460 struct sctp_authchunks __user *p = (void __user *)optval;
5461 struct sctp_authchunks val;
5462 struct sctp_association *asoc;
5463 struct sctp_chunks_param *ch;
5464 u32 num_chunks = 0;
5465 char __user *to;
5467 if (!sctp_auth_enable)
5468 return -EACCES;
5470 if (len < sizeof(struct sctp_authchunks))
5471 return -EINVAL;
5473 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5474 return -EFAULT;
5476 to = p->gauth_chunks;
5477 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5478 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5479 return -EINVAL;
5481 if (asoc)
5482 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5483 else
5484 ch = sctp_sk(sk)->ep->auth_chunk_list;
5486 if (!ch)
5487 goto num;
5489 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5490 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5491 return -EINVAL;
5493 if (copy_to_user(to, ch->chunks, num_chunks))
5494 return -EFAULT;
5495 num:
5496 len = sizeof(struct sctp_authchunks) + num_chunks;
5497 if (put_user(len, optlen))
5498 return -EFAULT;
5499 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5500 return -EFAULT;
5502 return 0;
5506 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5507 * This option gets the current number of associations that are attached
5508 * to a one-to-many style socket. The option value is an uint32_t.
5510 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5511 char __user *optval, int __user *optlen)
5513 struct sctp_sock *sp = sctp_sk(sk);
5514 struct sctp_association *asoc;
5515 u32 val = 0;
5517 if (sctp_style(sk, TCP))
5518 return -EOPNOTSUPP;
5520 if (len < sizeof(u32))
5521 return -EINVAL;
5523 len = sizeof(u32);
5525 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5526 val++;
5529 if (put_user(len, optlen))
5530 return -EFAULT;
5531 if (copy_to_user(optval, &val, len))
5532 return -EFAULT;
5534 return 0;
5537 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5538 char __user *optval, int __user *optlen)
5540 int retval = 0;
5541 int len;
5543 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5544 sk, optname);
5546 /* I can hardly begin to describe how wrong this is. This is
5547 * so broken as to be worse than useless. The API draft
5548 * REALLY is NOT helpful here... I am not convinced that the
5549 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5550 * are at all well-founded.
5552 if (level != SOL_SCTP) {
5553 struct sctp_af *af = sctp_sk(sk)->pf->af;
5555 retval = af->getsockopt(sk, level, optname, optval, optlen);
5556 return retval;
5559 if (get_user(len, optlen))
5560 return -EFAULT;
5562 sctp_lock_sock(sk);
5564 switch (optname) {
5565 case SCTP_STATUS:
5566 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5567 break;
5568 case SCTP_DISABLE_FRAGMENTS:
5569 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5570 optlen);
5571 break;
5572 case SCTP_EVENTS:
5573 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5574 break;
5575 case SCTP_AUTOCLOSE:
5576 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5577 break;
5578 case SCTP_SOCKOPT_PEELOFF:
5579 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5580 break;
5581 case SCTP_PEER_ADDR_PARAMS:
5582 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5583 optlen);
5584 break;
5585 case SCTP_DELAYED_ACK:
5586 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5587 optlen);
5588 break;
5589 case SCTP_INITMSG:
5590 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5591 break;
5592 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5593 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5594 optlen);
5595 break;
5596 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5597 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5598 optlen);
5599 break;
5600 case SCTP_GET_PEER_ADDRS_OLD:
5601 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5602 optlen);
5603 break;
5604 case SCTP_GET_LOCAL_ADDRS_OLD:
5605 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5606 optlen);
5607 break;
5608 case SCTP_GET_PEER_ADDRS:
5609 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5610 optlen);
5611 break;
5612 case SCTP_GET_LOCAL_ADDRS:
5613 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5614 optlen);
5615 break;
5616 case SCTP_SOCKOPT_CONNECTX3:
5617 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5618 break;
5619 case SCTP_DEFAULT_SEND_PARAM:
5620 retval = sctp_getsockopt_default_send_param(sk, len,
5621 optval, optlen);
5622 break;
5623 case SCTP_PRIMARY_ADDR:
5624 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5625 break;
5626 case SCTP_NODELAY:
5627 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5628 break;
5629 case SCTP_RTOINFO:
5630 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5631 break;
5632 case SCTP_ASSOCINFO:
5633 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5634 break;
5635 case SCTP_I_WANT_MAPPED_V4_ADDR:
5636 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5637 break;
5638 case SCTP_MAXSEG:
5639 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5640 break;
5641 case SCTP_GET_PEER_ADDR_INFO:
5642 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5643 optlen);
5644 break;
5645 case SCTP_ADAPTATION_LAYER:
5646 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5647 optlen);
5648 break;
5649 case SCTP_CONTEXT:
5650 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5651 break;
5652 case SCTP_FRAGMENT_INTERLEAVE:
5653 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5654 optlen);
5655 break;
5656 case SCTP_PARTIAL_DELIVERY_POINT:
5657 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5658 optlen);
5659 break;
5660 case SCTP_MAX_BURST:
5661 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5662 break;
5663 case SCTP_AUTH_KEY:
5664 case SCTP_AUTH_CHUNK:
5665 case SCTP_AUTH_DELETE_KEY:
5666 retval = -EOPNOTSUPP;
5667 break;
5668 case SCTP_HMAC_IDENT:
5669 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5670 break;
5671 case SCTP_AUTH_ACTIVE_KEY:
5672 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5673 break;
5674 case SCTP_PEER_AUTH_CHUNKS:
5675 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5676 optlen);
5677 break;
5678 case SCTP_LOCAL_AUTH_CHUNKS:
5679 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5680 optlen);
5681 break;
5682 case SCTP_GET_ASSOC_NUMBER:
5683 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5684 break;
5685 default:
5686 retval = -ENOPROTOOPT;
5687 break;
5690 sctp_release_sock(sk);
5691 return retval;
5694 static void sctp_hash(struct sock *sk)
5696 /* STUB */
5699 static void sctp_unhash(struct sock *sk)
5701 /* STUB */
5704 /* Check if port is acceptable. Possibly find first available port.
5706 * The port hash table (contained in the 'global' SCTP protocol storage
5707 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5708 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5709 * list (the list number is the port number hashed out, so as you
5710 * would expect from a hash function, all the ports in a given list have
5711 * such a number that hashes out to the same list number; you were
5712 * expecting that, right?); so each list has a set of ports, with a
5713 * link to the socket (struct sock) that uses it, the port number and
5714 * a fastreuse flag (FIXME: NPI ipg).
5716 static struct sctp_bind_bucket *sctp_bucket_create(
5717 struct sctp_bind_hashbucket *head, unsigned short snum);
5719 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5721 struct sctp_bind_hashbucket *head; /* hash list */
5722 struct sctp_bind_bucket *pp; /* hash list port iterator */
5723 struct hlist_node *node;
5724 unsigned short snum;
5725 int ret;
5727 snum = ntohs(addr->v4.sin_port);
5729 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5730 sctp_local_bh_disable();
5732 if (snum == 0) {
5733 /* Search for an available port. */
5734 int low, high, remaining, index;
5735 unsigned int rover;
5737 inet_get_local_port_range(&low, &high);
5738 remaining = (high - low) + 1;
5739 rover = net_random() % remaining + low;
5741 do {
5742 rover++;
5743 if ((rover < low) || (rover > high))
5744 rover = low;
5745 index = sctp_phashfn(rover);
5746 head = &sctp_port_hashtable[index];
5747 sctp_spin_lock(&head->lock);
5748 sctp_for_each_hentry(pp, node, &head->chain)
5749 if (pp->port == rover)
5750 goto next;
5751 break;
5752 next:
5753 sctp_spin_unlock(&head->lock);
5754 } while (--remaining > 0);
5756 /* Exhausted local port range during search? */
5757 ret = 1;
5758 if (remaining <= 0)
5759 goto fail;
5761 /* OK, here is the one we will use. HEAD (the port
5762 * hash table list entry) is non-NULL and we hold it's
5763 * mutex.
5765 snum = rover;
5766 } else {
5767 /* We are given an specific port number; we verify
5768 * that it is not being used. If it is used, we will
5769 * exahust the search in the hash list corresponding
5770 * to the port number (snum) - we detect that with the
5771 * port iterator, pp being NULL.
5773 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5774 sctp_spin_lock(&head->lock);
5775 sctp_for_each_hentry(pp, node, &head->chain) {
5776 if (pp->port == snum)
5777 goto pp_found;
5780 pp = NULL;
5781 goto pp_not_found;
5782 pp_found:
5783 if (!hlist_empty(&pp->owner)) {
5784 /* We had a port hash table hit - there is an
5785 * available port (pp != NULL) and it is being
5786 * used by other socket (pp->owner not empty); that other
5787 * socket is going to be sk2.
5789 int reuse = sk->sk_reuse;
5790 struct sock *sk2;
5791 struct hlist_node *node;
5793 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5794 if (pp->fastreuse && sk->sk_reuse &&
5795 sk->sk_state != SCTP_SS_LISTENING)
5796 goto success;
5798 /* Run through the list of sockets bound to the port
5799 * (pp->port) [via the pointers bind_next and
5800 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5801 * we get the endpoint they describe and run through
5802 * the endpoint's list of IP (v4 or v6) addresses,
5803 * comparing each of the addresses with the address of
5804 * the socket sk. If we find a match, then that means
5805 * that this port/socket (sk) combination are already
5806 * in an endpoint.
5808 sk_for_each_bound(sk2, node, &pp->owner) {
5809 struct sctp_endpoint *ep2;
5810 ep2 = sctp_sk(sk2)->ep;
5812 if (sk == sk2 ||
5813 (reuse && sk2->sk_reuse &&
5814 sk2->sk_state != SCTP_SS_LISTENING))
5815 continue;
5817 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5818 sctp_sk(sk2), sctp_sk(sk))) {
5819 ret = (long)sk2;
5820 goto fail_unlock;
5823 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5825 pp_not_found:
5826 /* If there was a hash table miss, create a new port. */
5827 ret = 1;
5828 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5829 goto fail_unlock;
5831 /* In either case (hit or miss), make sure fastreuse is 1 only
5832 * if sk->sk_reuse is too (that is, if the caller requested
5833 * SO_REUSEADDR on this socket -sk-).
5835 if (hlist_empty(&pp->owner)) {
5836 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5837 pp->fastreuse = 1;
5838 else
5839 pp->fastreuse = 0;
5840 } else if (pp->fastreuse &&
5841 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5842 pp->fastreuse = 0;
5844 /* We are set, so fill up all the data in the hash table
5845 * entry, tie the socket list information with the rest of the
5846 * sockets FIXME: Blurry, NPI (ipg).
5848 success:
5849 if (!sctp_sk(sk)->bind_hash) {
5850 inet_sk(sk)->num = snum;
5851 sk_add_bind_node(sk, &pp->owner);
5852 sctp_sk(sk)->bind_hash = pp;
5854 ret = 0;
5856 fail_unlock:
5857 sctp_spin_unlock(&head->lock);
5859 fail:
5860 sctp_local_bh_enable();
5861 return ret;
5864 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5865 * port is requested.
5867 static int sctp_get_port(struct sock *sk, unsigned short snum)
5869 long ret;
5870 union sctp_addr addr;
5871 struct sctp_af *af = sctp_sk(sk)->pf->af;
5873 /* Set up a dummy address struct from the sk. */
5874 af->from_sk(&addr, sk);
5875 addr.v4.sin_port = htons(snum);
5877 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5878 ret = sctp_get_port_local(sk, &addr);
5880 return (ret ? 1 : 0);
5884 * Move a socket to LISTENING state.
5886 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5888 struct sctp_sock *sp = sctp_sk(sk);
5889 struct sctp_endpoint *ep = sp->ep;
5890 struct crypto_hash *tfm = NULL;
5892 /* Allocate HMAC for generating cookie. */
5893 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5894 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5895 if (IS_ERR(tfm)) {
5896 if (net_ratelimit()) {
5897 printk(KERN_INFO
5898 "SCTP: failed to load transform for %s: %ld\n",
5899 sctp_hmac_alg, PTR_ERR(tfm));
5901 return -ENOSYS;
5903 sctp_sk(sk)->hmac = tfm;
5907 * If a bind() or sctp_bindx() is not called prior to a listen()
5908 * call that allows new associations to be accepted, the system
5909 * picks an ephemeral port and will choose an address set equivalent
5910 * to binding with a wildcard address.
5912 * This is not currently spelled out in the SCTP sockets
5913 * extensions draft, but follows the practice as seen in TCP
5914 * sockets.
5917 sk->sk_state = SCTP_SS_LISTENING;
5918 if (!ep->base.bind_addr.port) {
5919 if (sctp_autobind(sk))
5920 return -EAGAIN;
5921 } else {
5922 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5923 sk->sk_state = SCTP_SS_CLOSED;
5924 return -EADDRINUSE;
5928 sk->sk_max_ack_backlog = backlog;
5929 sctp_hash_endpoint(ep);
5930 return 0;
5934 * 4.1.3 / 5.1.3 listen()
5936 * By default, new associations are not accepted for UDP style sockets.
5937 * An application uses listen() to mark a socket as being able to
5938 * accept new associations.
5940 * On TCP style sockets, applications use listen() to ready the SCTP
5941 * endpoint for accepting inbound associations.
5943 * On both types of endpoints a backlog of '0' disables listening.
5945 * Move a socket to LISTENING state.
5947 int sctp_inet_listen(struct socket *sock, int backlog)
5949 struct sock *sk = sock->sk;
5950 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5951 int err = -EINVAL;
5953 if (unlikely(backlog < 0))
5954 return err;
5956 sctp_lock_sock(sk);
5958 /* Peeled-off sockets are not allowed to listen(). */
5959 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5960 goto out;
5962 if (sock->state != SS_UNCONNECTED)
5963 goto out;
5965 /* If backlog is zero, disable listening. */
5966 if (!backlog) {
5967 if (sctp_sstate(sk, CLOSED))
5968 goto out;
5970 err = 0;
5971 sctp_unhash_endpoint(ep);
5972 sk->sk_state = SCTP_SS_CLOSED;
5973 if (sk->sk_reuse)
5974 sctp_sk(sk)->bind_hash->fastreuse = 1;
5975 goto out;
5978 /* If we are already listening, just update the backlog */
5979 if (sctp_sstate(sk, LISTENING))
5980 sk->sk_max_ack_backlog = backlog;
5981 else {
5982 err = sctp_listen_start(sk, backlog);
5983 if (err)
5984 goto out;
5987 err = 0;
5988 out:
5989 sctp_release_sock(sk);
5990 return err;
5994 * This function is done by modeling the current datagram_poll() and the
5995 * tcp_poll(). Note that, based on these implementations, we don't
5996 * lock the socket in this function, even though it seems that,
5997 * ideally, locking or some other mechanisms can be used to ensure
5998 * the integrity of the counters (sndbuf and wmem_alloc) used
5999 * in this place. We assume that we don't need locks either until proven
6000 * otherwise.
6002 * Another thing to note is that we include the Async I/O support
6003 * here, again, by modeling the current TCP/UDP code. We don't have
6004 * a good way to test with it yet.
6006 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6008 struct sock *sk = sock->sk;
6009 struct sctp_sock *sp = sctp_sk(sk);
6010 unsigned int mask;
6012 poll_wait(file, sk->sk_sleep, wait);
6014 /* A TCP-style listening socket becomes readable when the accept queue
6015 * is not empty.
6017 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6018 return (!list_empty(&sp->ep->asocs)) ?
6019 (POLLIN | POLLRDNORM) : 0;
6021 mask = 0;
6023 /* Is there any exceptional events? */
6024 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6025 mask |= POLLERR;
6026 if (sk->sk_shutdown & RCV_SHUTDOWN)
6027 mask |= POLLRDHUP;
6028 if (sk->sk_shutdown == SHUTDOWN_MASK)
6029 mask |= POLLHUP;
6031 /* Is it readable? Reconsider this code with TCP-style support. */
6032 if (!skb_queue_empty(&sk->sk_receive_queue) ||
6033 (sk->sk_shutdown & RCV_SHUTDOWN))
6034 mask |= POLLIN | POLLRDNORM;
6036 /* The association is either gone or not ready. */
6037 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6038 return mask;
6040 /* Is it writable? */
6041 if (sctp_writeable(sk)) {
6042 mask |= POLLOUT | POLLWRNORM;
6043 } else {
6044 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6046 * Since the socket is not locked, the buffer
6047 * might be made available after the writeable check and
6048 * before the bit is set. This could cause a lost I/O
6049 * signal. tcp_poll() has a race breaker for this race
6050 * condition. Based on their implementation, we put
6051 * in the following code to cover it as well.
6053 if (sctp_writeable(sk))
6054 mask |= POLLOUT | POLLWRNORM;
6056 return mask;
6059 /********************************************************************
6060 * 2nd Level Abstractions
6061 ********************************************************************/
6063 static struct sctp_bind_bucket *sctp_bucket_create(
6064 struct sctp_bind_hashbucket *head, unsigned short snum)
6066 struct sctp_bind_bucket *pp;
6068 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6069 if (pp) {
6070 SCTP_DBG_OBJCNT_INC(bind_bucket);
6071 pp->port = snum;
6072 pp->fastreuse = 0;
6073 INIT_HLIST_HEAD(&pp->owner);
6074 hlist_add_head(&pp->node, &head->chain);
6076 return pp;
6079 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6080 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6082 if (pp && hlist_empty(&pp->owner)) {
6083 __hlist_del(&pp->node);
6084 kmem_cache_free(sctp_bucket_cachep, pp);
6085 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6089 /* Release this socket's reference to a local port. */
6090 static inline void __sctp_put_port(struct sock *sk)
6092 struct sctp_bind_hashbucket *head =
6093 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
6094 struct sctp_bind_bucket *pp;
6096 sctp_spin_lock(&head->lock);
6097 pp = sctp_sk(sk)->bind_hash;
6098 __sk_del_bind_node(sk);
6099 sctp_sk(sk)->bind_hash = NULL;
6100 inet_sk(sk)->num = 0;
6101 sctp_bucket_destroy(pp);
6102 sctp_spin_unlock(&head->lock);
6105 void sctp_put_port(struct sock *sk)
6107 sctp_local_bh_disable();
6108 __sctp_put_port(sk);
6109 sctp_local_bh_enable();
6113 * The system picks an ephemeral port and choose an address set equivalent
6114 * to binding with a wildcard address.
6115 * One of those addresses will be the primary address for the association.
6116 * This automatically enables the multihoming capability of SCTP.
6118 static int sctp_autobind(struct sock *sk)
6120 union sctp_addr autoaddr;
6121 struct sctp_af *af;
6122 __be16 port;
6124 /* Initialize a local sockaddr structure to INADDR_ANY. */
6125 af = sctp_sk(sk)->pf->af;
6127 port = htons(inet_sk(sk)->num);
6128 af->inaddr_any(&autoaddr, port);
6130 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6133 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6135 * From RFC 2292
6136 * 4.2 The cmsghdr Structure *
6138 * When ancillary data is sent or received, any number of ancillary data
6139 * objects can be specified by the msg_control and msg_controllen members of
6140 * the msghdr structure, because each object is preceded by
6141 * a cmsghdr structure defining the object's length (the cmsg_len member).
6142 * Historically Berkeley-derived implementations have passed only one object
6143 * at a time, but this API allows multiple objects to be
6144 * passed in a single call to sendmsg() or recvmsg(). The following example
6145 * shows two ancillary data objects in a control buffer.
6147 * |<--------------------------- msg_controllen -------------------------->|
6148 * | |
6150 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6152 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6153 * | | |
6155 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6157 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6158 * | | | | |
6160 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6161 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6163 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6165 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6169 * msg_control
6170 * points here
6172 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6173 sctp_cmsgs_t *cmsgs)
6175 struct cmsghdr *cmsg;
6176 struct msghdr *my_msg = (struct msghdr *)msg;
6178 for (cmsg = CMSG_FIRSTHDR(msg);
6179 cmsg != NULL;
6180 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6181 if (!CMSG_OK(my_msg, cmsg))
6182 return -EINVAL;
6184 /* Should we parse this header or ignore? */
6185 if (cmsg->cmsg_level != IPPROTO_SCTP)
6186 continue;
6188 /* Strictly check lengths following example in SCM code. */
6189 switch (cmsg->cmsg_type) {
6190 case SCTP_INIT:
6191 /* SCTP Socket API Extension
6192 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6194 * This cmsghdr structure provides information for
6195 * initializing new SCTP associations with sendmsg().
6196 * The SCTP_INITMSG socket option uses this same data
6197 * structure. This structure is not used for
6198 * recvmsg().
6200 * cmsg_level cmsg_type cmsg_data[]
6201 * ------------ ------------ ----------------------
6202 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6204 if (cmsg->cmsg_len !=
6205 CMSG_LEN(sizeof(struct sctp_initmsg)))
6206 return -EINVAL;
6207 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6208 break;
6210 case SCTP_SNDRCV:
6211 /* SCTP Socket API Extension
6212 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6214 * This cmsghdr structure specifies SCTP options for
6215 * sendmsg() and describes SCTP header information
6216 * about a received message through recvmsg().
6218 * cmsg_level cmsg_type cmsg_data[]
6219 * ------------ ------------ ----------------------
6220 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6222 if (cmsg->cmsg_len !=
6223 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6224 return -EINVAL;
6226 cmsgs->info =
6227 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6229 /* Minimally, validate the sinfo_flags. */
6230 if (cmsgs->info->sinfo_flags &
6231 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6232 SCTP_ABORT | SCTP_EOF))
6233 return -EINVAL;
6234 break;
6236 default:
6237 return -EINVAL;
6240 return 0;
6244 * Wait for a packet..
6245 * Note: This function is the same function as in core/datagram.c
6246 * with a few modifications to make lksctp work.
6248 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6250 int error;
6251 DEFINE_WAIT(wait);
6253 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6255 /* Socket errors? */
6256 error = sock_error(sk);
6257 if (error)
6258 goto out;
6260 if (!skb_queue_empty(&sk->sk_receive_queue))
6261 goto ready;
6263 /* Socket shut down? */
6264 if (sk->sk_shutdown & RCV_SHUTDOWN)
6265 goto out;
6267 /* Sequenced packets can come disconnected. If so we report the
6268 * problem.
6270 error = -ENOTCONN;
6272 /* Is there a good reason to think that we may receive some data? */
6273 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6274 goto out;
6276 /* Handle signals. */
6277 if (signal_pending(current))
6278 goto interrupted;
6280 /* Let another process have a go. Since we are going to sleep
6281 * anyway. Note: This may cause odd behaviors if the message
6282 * does not fit in the user's buffer, but this seems to be the
6283 * only way to honor MSG_DONTWAIT realistically.
6285 sctp_release_sock(sk);
6286 *timeo_p = schedule_timeout(*timeo_p);
6287 sctp_lock_sock(sk);
6289 ready:
6290 finish_wait(sk->sk_sleep, &wait);
6291 return 0;
6293 interrupted:
6294 error = sock_intr_errno(*timeo_p);
6296 out:
6297 finish_wait(sk->sk_sleep, &wait);
6298 *err = error;
6299 return error;
6302 /* Receive a datagram.
6303 * Note: This is pretty much the same routine as in core/datagram.c
6304 * with a few changes to make lksctp work.
6306 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6307 int noblock, int *err)
6309 int error;
6310 struct sk_buff *skb;
6311 long timeo;
6313 timeo = sock_rcvtimeo(sk, noblock);
6315 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6316 timeo, MAX_SCHEDULE_TIMEOUT);
6318 do {
6319 /* Again only user level code calls this function,
6320 * so nothing interrupt level
6321 * will suddenly eat the receive_queue.
6323 * Look at current nfs client by the way...
6324 * However, this function was corrent in any case. 8)
6326 if (flags & MSG_PEEK) {
6327 spin_lock_bh(&sk->sk_receive_queue.lock);
6328 skb = skb_peek(&sk->sk_receive_queue);
6329 if (skb)
6330 atomic_inc(&skb->users);
6331 spin_unlock_bh(&sk->sk_receive_queue.lock);
6332 } else {
6333 skb = skb_dequeue(&sk->sk_receive_queue);
6336 if (skb)
6337 return skb;
6339 /* Caller is allowed not to check sk->sk_err before calling. */
6340 error = sock_error(sk);
6341 if (error)
6342 goto no_packet;
6344 if (sk->sk_shutdown & RCV_SHUTDOWN)
6345 break;
6347 /* User doesn't want to wait. */
6348 error = -EAGAIN;
6349 if (!timeo)
6350 goto no_packet;
6351 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6353 return NULL;
6355 no_packet:
6356 *err = error;
6357 return NULL;
6360 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6361 static void __sctp_write_space(struct sctp_association *asoc)
6363 struct sock *sk = asoc->base.sk;
6364 struct socket *sock = sk->sk_socket;
6366 if ((sctp_wspace(asoc) > 0) && sock) {
6367 if (waitqueue_active(&asoc->wait))
6368 wake_up_interruptible(&asoc->wait);
6370 if (sctp_writeable(sk)) {
6371 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6372 wake_up_interruptible(sk->sk_sleep);
6374 /* Note that we try to include the Async I/O support
6375 * here by modeling from the current TCP/UDP code.
6376 * We have not tested with it yet.
6378 if (sock->fasync_list &&
6379 !(sk->sk_shutdown & SEND_SHUTDOWN))
6380 sock_wake_async(sock,
6381 SOCK_WAKE_SPACE, POLL_OUT);
6386 /* Do accounting for the sndbuf space.
6387 * Decrement the used sndbuf space of the corresponding association by the
6388 * data size which was just transmitted(freed).
6390 static void sctp_wfree(struct sk_buff *skb)
6392 struct sctp_association *asoc;
6393 struct sctp_chunk *chunk;
6394 struct sock *sk;
6396 /* Get the saved chunk pointer. */
6397 chunk = *((struct sctp_chunk **)(skb->cb));
6398 asoc = chunk->asoc;
6399 sk = asoc->base.sk;
6400 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6401 sizeof(struct sk_buff) +
6402 sizeof(struct sctp_chunk);
6404 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6407 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6409 sk->sk_wmem_queued -= skb->truesize;
6410 sk_mem_uncharge(sk, skb->truesize);
6412 sock_wfree(skb);
6413 __sctp_write_space(asoc);
6415 sctp_association_put(asoc);
6418 /* Do accounting for the receive space on the socket.
6419 * Accounting for the association is done in ulpevent.c
6420 * We set this as a destructor for the cloned data skbs so that
6421 * accounting is done at the correct time.
6423 void sctp_sock_rfree(struct sk_buff *skb)
6425 struct sock *sk = skb->sk;
6426 struct sctp_ulpevent *event = sctp_skb2event(skb);
6428 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6431 * Mimic the behavior of sock_rfree
6433 sk_mem_uncharge(sk, event->rmem_len);
6437 /* Helper function to wait for space in the sndbuf. */
6438 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6439 size_t msg_len)
6441 struct sock *sk = asoc->base.sk;
6442 int err = 0;
6443 long current_timeo = *timeo_p;
6444 DEFINE_WAIT(wait);
6446 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6447 asoc, (long)(*timeo_p), msg_len);
6449 /* Increment the association's refcnt. */
6450 sctp_association_hold(asoc);
6452 /* Wait on the association specific sndbuf space. */
6453 for (;;) {
6454 prepare_to_wait_exclusive(&asoc->wait, &wait,
6455 TASK_INTERRUPTIBLE);
6456 if (!*timeo_p)
6457 goto do_nonblock;
6458 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6459 asoc->base.dead)
6460 goto do_error;
6461 if (signal_pending(current))
6462 goto do_interrupted;
6463 if (msg_len <= sctp_wspace(asoc))
6464 break;
6466 /* Let another process have a go. Since we are going
6467 * to sleep anyway.
6469 sctp_release_sock(sk);
6470 current_timeo = schedule_timeout(current_timeo);
6471 BUG_ON(sk != asoc->base.sk);
6472 sctp_lock_sock(sk);
6474 *timeo_p = current_timeo;
6477 out:
6478 finish_wait(&asoc->wait, &wait);
6480 /* Release the association's refcnt. */
6481 sctp_association_put(asoc);
6483 return err;
6485 do_error:
6486 err = -EPIPE;
6487 goto out;
6489 do_interrupted:
6490 err = sock_intr_errno(*timeo_p);
6491 goto out;
6493 do_nonblock:
6494 err = -EAGAIN;
6495 goto out;
6498 /* If socket sndbuf has changed, wake up all per association waiters. */
6499 void sctp_write_space(struct sock *sk)
6501 struct sctp_association *asoc;
6503 /* Wake up the tasks in each wait queue. */
6504 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6505 __sctp_write_space(asoc);
6509 /* Is there any sndbuf space available on the socket?
6511 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6512 * associations on the same socket. For a UDP-style socket with
6513 * multiple associations, it is possible for it to be "unwriteable"
6514 * prematurely. I assume that this is acceptable because
6515 * a premature "unwriteable" is better than an accidental "writeable" which
6516 * would cause an unwanted block under certain circumstances. For the 1-1
6517 * UDP-style sockets or TCP-style sockets, this code should work.
6518 * - Daisy
6520 static int sctp_writeable(struct sock *sk)
6522 int amt = 0;
6524 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6525 if (amt < 0)
6526 amt = 0;
6527 return amt;
6530 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6531 * returns immediately with EINPROGRESS.
6533 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6535 struct sock *sk = asoc->base.sk;
6536 int err = 0;
6537 long current_timeo = *timeo_p;
6538 DEFINE_WAIT(wait);
6540 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6541 (long)(*timeo_p));
6543 /* Increment the association's refcnt. */
6544 sctp_association_hold(asoc);
6546 for (;;) {
6547 prepare_to_wait_exclusive(&asoc->wait, &wait,
6548 TASK_INTERRUPTIBLE);
6549 if (!*timeo_p)
6550 goto do_nonblock;
6551 if (sk->sk_shutdown & RCV_SHUTDOWN)
6552 break;
6553 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6554 asoc->base.dead)
6555 goto do_error;
6556 if (signal_pending(current))
6557 goto do_interrupted;
6559 if (sctp_state(asoc, ESTABLISHED))
6560 break;
6562 /* Let another process have a go. Since we are going
6563 * to sleep anyway.
6565 sctp_release_sock(sk);
6566 current_timeo = schedule_timeout(current_timeo);
6567 sctp_lock_sock(sk);
6569 *timeo_p = current_timeo;
6572 out:
6573 finish_wait(&asoc->wait, &wait);
6575 /* Release the association's refcnt. */
6576 sctp_association_put(asoc);
6578 return err;
6580 do_error:
6581 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6582 err = -ETIMEDOUT;
6583 else
6584 err = -ECONNREFUSED;
6585 goto out;
6587 do_interrupted:
6588 err = sock_intr_errno(*timeo_p);
6589 goto out;
6591 do_nonblock:
6592 err = -EINPROGRESS;
6593 goto out;
6596 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6598 struct sctp_endpoint *ep;
6599 int err = 0;
6600 DEFINE_WAIT(wait);
6602 ep = sctp_sk(sk)->ep;
6605 for (;;) {
6606 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6607 TASK_INTERRUPTIBLE);
6609 if (list_empty(&ep->asocs)) {
6610 sctp_release_sock(sk);
6611 timeo = schedule_timeout(timeo);
6612 sctp_lock_sock(sk);
6615 err = -EINVAL;
6616 if (!sctp_sstate(sk, LISTENING))
6617 break;
6619 err = 0;
6620 if (!list_empty(&ep->asocs))
6621 break;
6623 err = sock_intr_errno(timeo);
6624 if (signal_pending(current))
6625 break;
6627 err = -EAGAIN;
6628 if (!timeo)
6629 break;
6632 finish_wait(sk->sk_sleep, &wait);
6634 return err;
6637 static void sctp_wait_for_close(struct sock *sk, long timeout)
6639 DEFINE_WAIT(wait);
6641 do {
6642 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6643 if (list_empty(&sctp_sk(sk)->ep->asocs))
6644 break;
6645 sctp_release_sock(sk);
6646 timeout = schedule_timeout(timeout);
6647 sctp_lock_sock(sk);
6648 } while (!signal_pending(current) && timeout);
6650 finish_wait(sk->sk_sleep, &wait);
6653 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6655 struct sk_buff *frag;
6657 if (!skb->data_len)
6658 goto done;
6660 /* Don't forget the fragments. */
6661 skb_walk_frags(skb, frag)
6662 sctp_skb_set_owner_r_frag(frag, sk);
6664 done:
6665 sctp_skb_set_owner_r(skb, sk);
6668 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6669 struct sctp_association *asoc)
6671 struct inet_sock *inet = inet_sk(sk);
6672 struct inet_sock *newinet = inet_sk(newsk);
6674 newsk->sk_type = sk->sk_type;
6675 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6676 newsk->sk_flags = sk->sk_flags;
6677 newsk->sk_no_check = sk->sk_no_check;
6678 newsk->sk_reuse = sk->sk_reuse;
6680 newsk->sk_shutdown = sk->sk_shutdown;
6681 newsk->sk_destruct = inet_sock_destruct;
6682 newsk->sk_family = sk->sk_family;
6683 newsk->sk_protocol = IPPROTO_SCTP;
6684 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6685 newsk->sk_sndbuf = sk->sk_sndbuf;
6686 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6687 newsk->sk_lingertime = sk->sk_lingertime;
6688 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6689 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6691 newinet = inet_sk(newsk);
6693 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6694 * getsockname() and getpeername()
6696 newinet->sport = inet->sport;
6697 newinet->saddr = inet->saddr;
6698 newinet->rcv_saddr = inet->rcv_saddr;
6699 newinet->dport = htons(asoc->peer.port);
6700 newinet->pmtudisc = inet->pmtudisc;
6701 newinet->id = asoc->next_tsn ^ jiffies;
6703 newinet->uc_ttl = inet->uc_ttl;
6704 newinet->mc_loop = 1;
6705 newinet->mc_ttl = 1;
6706 newinet->mc_index = 0;
6707 newinet->mc_list = NULL;
6710 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6711 * and its messages to the newsk.
6713 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6714 struct sctp_association *assoc,
6715 sctp_socket_type_t type)
6717 struct sctp_sock *oldsp = sctp_sk(oldsk);
6718 struct sctp_sock *newsp = sctp_sk(newsk);
6719 struct sctp_bind_bucket *pp; /* hash list port iterator */
6720 struct sctp_endpoint *newep = newsp->ep;
6721 struct sk_buff *skb, *tmp;
6722 struct sctp_ulpevent *event;
6723 struct sctp_bind_hashbucket *head;
6725 /* Migrate socket buffer sizes and all the socket level options to the
6726 * new socket.
6728 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6729 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6730 /* Brute force copy old sctp opt. */
6731 inet_sk_copy_descendant(newsk, oldsk);
6733 /* Restore the ep value that was overwritten with the above structure
6734 * copy.
6736 newsp->ep = newep;
6737 newsp->hmac = NULL;
6739 /* Hook this new socket in to the bind_hash list. */
6740 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6741 sctp_local_bh_disable();
6742 sctp_spin_lock(&head->lock);
6743 pp = sctp_sk(oldsk)->bind_hash;
6744 sk_add_bind_node(newsk, &pp->owner);
6745 sctp_sk(newsk)->bind_hash = pp;
6746 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6747 sctp_spin_unlock(&head->lock);
6748 sctp_local_bh_enable();
6750 /* Copy the bind_addr list from the original endpoint to the new
6751 * endpoint so that we can handle restarts properly
6753 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6754 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6756 /* Move any messages in the old socket's receive queue that are for the
6757 * peeled off association to the new socket's receive queue.
6759 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6760 event = sctp_skb2event(skb);
6761 if (event->asoc == assoc) {
6762 __skb_unlink(skb, &oldsk->sk_receive_queue);
6763 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6764 sctp_skb_set_owner_r_frag(skb, newsk);
6768 /* Clean up any messages pending delivery due to partial
6769 * delivery. Three cases:
6770 * 1) No partial deliver; no work.
6771 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6772 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6774 skb_queue_head_init(&newsp->pd_lobby);
6775 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6777 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6778 struct sk_buff_head *queue;
6780 /* Decide which queue to move pd_lobby skbs to. */
6781 if (assoc->ulpq.pd_mode) {
6782 queue = &newsp->pd_lobby;
6783 } else
6784 queue = &newsk->sk_receive_queue;
6786 /* Walk through the pd_lobby, looking for skbs that
6787 * need moved to the new socket.
6789 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6790 event = sctp_skb2event(skb);
6791 if (event->asoc == assoc) {
6792 __skb_unlink(skb, &oldsp->pd_lobby);
6793 __skb_queue_tail(queue, skb);
6794 sctp_skb_set_owner_r_frag(skb, newsk);
6798 /* Clear up any skbs waiting for the partial
6799 * delivery to finish.
6801 if (assoc->ulpq.pd_mode)
6802 sctp_clear_pd(oldsk, NULL);
6806 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6807 sctp_skb_set_owner_r_frag(skb, newsk);
6809 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6810 sctp_skb_set_owner_r_frag(skb, newsk);
6812 /* Set the type of socket to indicate that it is peeled off from the
6813 * original UDP-style socket or created with the accept() call on a
6814 * TCP-style socket..
6816 newsp->type = type;
6818 /* Mark the new socket "in-use" by the user so that any packets
6819 * that may arrive on the association after we've moved it are
6820 * queued to the backlog. This prevents a potential race between
6821 * backlog processing on the old socket and new-packet processing
6822 * on the new socket.
6824 * The caller has just allocated newsk so we can guarantee that other
6825 * paths won't try to lock it and then oldsk.
6827 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6828 sctp_assoc_migrate(assoc, newsk);
6830 /* If the association on the newsk is already closed before accept()
6831 * is called, set RCV_SHUTDOWN flag.
6833 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6834 newsk->sk_shutdown |= RCV_SHUTDOWN;
6836 newsk->sk_state = SCTP_SS_ESTABLISHED;
6837 sctp_release_sock(newsk);
6841 /* This proto struct describes the ULP interface for SCTP. */
6842 struct proto sctp_prot = {
6843 .name = "SCTP",
6844 .owner = THIS_MODULE,
6845 .close = sctp_close,
6846 .connect = sctp_connect,
6847 .disconnect = sctp_disconnect,
6848 .accept = sctp_accept,
6849 .ioctl = sctp_ioctl,
6850 .init = sctp_init_sock,
6851 .destroy = sctp_destroy_sock,
6852 .shutdown = sctp_shutdown,
6853 .setsockopt = sctp_setsockopt,
6854 .getsockopt = sctp_getsockopt,
6855 .sendmsg = sctp_sendmsg,
6856 .recvmsg = sctp_recvmsg,
6857 .bind = sctp_bind,
6858 .backlog_rcv = sctp_backlog_rcv,
6859 .hash = sctp_hash,
6860 .unhash = sctp_unhash,
6861 .get_port = sctp_get_port,
6862 .obj_size = sizeof(struct sctp_sock),
6863 .sysctl_mem = sysctl_sctp_mem,
6864 .sysctl_rmem = sysctl_sctp_rmem,
6865 .sysctl_wmem = sysctl_sctp_wmem,
6866 .memory_pressure = &sctp_memory_pressure,
6867 .enter_memory_pressure = sctp_enter_memory_pressure,
6868 .memory_allocated = &sctp_memory_allocated,
6869 .sockets_allocated = &sctp_sockets_allocated,
6872 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6874 struct proto sctpv6_prot = {
6875 .name = "SCTPv6",
6876 .owner = THIS_MODULE,
6877 .close = sctp_close,
6878 .connect = sctp_connect,
6879 .disconnect = sctp_disconnect,
6880 .accept = sctp_accept,
6881 .ioctl = sctp_ioctl,
6882 .init = sctp_init_sock,
6883 .destroy = sctp_destroy_sock,
6884 .shutdown = sctp_shutdown,
6885 .setsockopt = sctp_setsockopt,
6886 .getsockopt = sctp_getsockopt,
6887 .sendmsg = sctp_sendmsg,
6888 .recvmsg = sctp_recvmsg,
6889 .bind = sctp_bind,
6890 .backlog_rcv = sctp_backlog_rcv,
6891 .hash = sctp_hash,
6892 .unhash = sctp_unhash,
6893 .get_port = sctp_get_port,
6894 .obj_size = sizeof(struct sctp6_sock),
6895 .sysctl_mem = sysctl_sctp_mem,
6896 .sysctl_rmem = sysctl_sctp_rmem,
6897 .sysctl_wmem = sysctl_sctp_wmem,
6898 .memory_pressure = &sctp_memory_pressure,
6899 .enter_memory_pressure = sctp_enter_memory_pressure,
6900 .memory_allocated = &sctp_memory_allocated,
6901 .sockets_allocated = &sctp_sockets_allocated,
6903 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */