MINI2440: Updated machine defconfig for 2.6.32
[linux-2.6/mini2440.git] / net / sctp / socket.c
blob3a95fcb17a9e35c715123316cb458efde2975157
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;
1084 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1085 GFP_KERNEL);
1086 if (err < 0) {
1087 goto out_free;
1092 /* Prime the peer's transport structures. */
1093 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1094 SCTP_UNKNOWN);
1095 if (!transport) {
1096 err = -ENOMEM;
1097 goto out_free;
1100 addrcnt++;
1101 addr_buf += af->sockaddr_len;
1102 walk_size += af->sockaddr_len;
1105 /* In case the user of sctp_connectx() wants an association
1106 * id back, assign one now.
1108 if (assoc_id) {
1109 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1110 if (err < 0)
1111 goto out_free;
1114 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1115 if (err < 0) {
1116 goto out_free;
1119 /* Initialize sk's dport and daddr for getpeername() */
1120 inet_sk(sk)->dport = htons(asoc->peer.port);
1121 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1122 af->to_sk_daddr(sa_addr, sk);
1123 sk->sk_err = 0;
1125 /* in-kernel sockets don't generally have a file allocated to them
1126 * if all they do is call sock_create_kern().
1128 if (sk->sk_socket->file)
1129 f_flags = sk->sk_socket->file->f_flags;
1131 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1133 err = sctp_wait_for_connect(asoc, &timeo);
1134 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1135 *assoc_id = asoc->assoc_id;
1137 /* Don't free association on exit. */
1138 asoc = NULL;
1140 out_free:
1142 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1143 " kaddrs: %p err: %d\n",
1144 asoc, kaddrs, err);
1145 if (asoc)
1146 sctp_association_free(asoc);
1147 return err;
1150 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1152 * API 8.9
1153 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1154 * sctp_assoc_t *asoc);
1156 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1157 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1158 * or IPv6 addresses.
1160 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1161 * Section 3.1.2 for this usage.
1163 * addrs is a pointer to an array of one or more socket addresses. Each
1164 * address is contained in its appropriate structure (i.e. struct
1165 * sockaddr_in or struct sockaddr_in6) the family of the address type
1166 * must be used to distengish the address length (note that this
1167 * representation is termed a "packed array" of addresses). The caller
1168 * specifies the number of addresses in the array with addrcnt.
1170 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1171 * the association id of the new association. On failure, sctp_connectx()
1172 * returns -1, and sets errno to the appropriate error code. The assoc_id
1173 * is not touched by the kernel.
1175 * For SCTP, the port given in each socket address must be the same, or
1176 * sctp_connectx() will fail, setting errno to EINVAL.
1178 * An application can use sctp_connectx to initiate an association with
1179 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1180 * allows a caller to specify multiple addresses at which a peer can be
1181 * reached. The way the SCTP stack uses the list of addresses to set up
1182 * the association is implementation dependant. This function only
1183 * specifies that the stack will try to make use of all the addresses in
1184 * the list when needed.
1186 * Note that the list of addresses passed in is only used for setting up
1187 * the association. It does not necessarily equal the set of addresses
1188 * the peer uses for the resulting association. If the caller wants to
1189 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1190 * retrieve them after the association has been set up.
1192 * Basically do nothing but copying the addresses from user to kernel
1193 * land and invoking either sctp_connectx(). This is used for tunneling
1194 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1196 * We don't use copy_from_user() for optimization: we first do the
1197 * sanity checks (buffer size -fast- and access check-healthy
1198 * pointer); if all of those succeed, then we can alloc the memory
1199 * (expensive operation) needed to copy the data to kernel. Then we do
1200 * the copying without checking the user space area
1201 * (__copy_from_user()).
1203 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1204 * it.
1206 * sk The sk of the socket
1207 * addrs The pointer to the addresses in user land
1208 * addrssize Size of the addrs buffer
1210 * Returns >=0 if ok, <0 errno code on error.
1212 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1213 struct sockaddr __user *addrs,
1214 int addrs_size,
1215 sctp_assoc_t *assoc_id)
1217 int err = 0;
1218 struct sockaddr *kaddrs;
1220 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1221 __func__, sk, addrs, addrs_size);
1223 if (unlikely(addrs_size <= 0))
1224 return -EINVAL;
1226 /* Check the user passed a healthy pointer. */
1227 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1228 return -EFAULT;
1230 /* Alloc space for the address array in kernel memory. */
1231 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1232 if (unlikely(!kaddrs))
1233 return -ENOMEM;
1235 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1236 err = -EFAULT;
1237 } else {
1238 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1241 kfree(kaddrs);
1243 return err;
1247 * This is an older interface. It's kept for backward compatibility
1248 * to the option that doesn't provide association id.
1250 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1251 struct sockaddr __user *addrs,
1252 int addrs_size)
1254 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1258 * New interface for the API. The since the API is done with a socket
1259 * option, to make it simple we feed back the association id is as a return
1260 * indication to the call. Error is always negative and association id is
1261 * always positive.
1263 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1264 struct sockaddr __user *addrs,
1265 int addrs_size)
1267 sctp_assoc_t assoc_id = 0;
1268 int err = 0;
1270 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1272 if (err)
1273 return err;
1274 else
1275 return assoc_id;
1279 * New (hopefully final) interface for the API.
1280 * We use the sctp_getaddrs_old structure so that use-space library
1281 * can avoid any unnecessary allocations. The only defferent part
1282 * is that we store the actual length of the address buffer into the
1283 * addrs_num structure member. That way we can re-use the existing
1284 * code.
1286 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1287 char __user *optval,
1288 int __user *optlen)
1290 struct sctp_getaddrs_old param;
1291 sctp_assoc_t assoc_id = 0;
1292 int err = 0;
1294 if (len < sizeof(param))
1295 return -EINVAL;
1297 if (copy_from_user(&param, optval, sizeof(param)))
1298 return -EFAULT;
1300 err = __sctp_setsockopt_connectx(sk,
1301 (struct sockaddr __user *)param.addrs,
1302 param.addr_num, &assoc_id);
1304 if (err == 0 || err == -EINPROGRESS) {
1305 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1306 return -EFAULT;
1307 if (put_user(sizeof(assoc_id), optlen))
1308 return -EFAULT;
1311 return err;
1314 /* API 3.1.4 close() - UDP Style Syntax
1315 * Applications use close() to perform graceful shutdown (as described in
1316 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1317 * by a UDP-style socket.
1319 * The syntax is
1321 * ret = close(int sd);
1323 * sd - the socket descriptor of the associations to be closed.
1325 * To gracefully shutdown a specific association represented by the
1326 * UDP-style socket, an application should use the sendmsg() call,
1327 * passing no user data, but including the appropriate flag in the
1328 * ancillary data (see Section xxxx).
1330 * If sd in the close() call is a branched-off socket representing only
1331 * one association, the shutdown is performed on that association only.
1333 * 4.1.6 close() - TCP Style Syntax
1335 * Applications use close() to gracefully close down an association.
1337 * The syntax is:
1339 * int close(int sd);
1341 * sd - the socket descriptor of the association to be closed.
1343 * After an application calls close() on a socket descriptor, no further
1344 * socket operations will succeed on that descriptor.
1346 * API 7.1.4 SO_LINGER
1348 * An application using the TCP-style socket can use this option to
1349 * perform the SCTP ABORT primitive. The linger option structure is:
1351 * struct linger {
1352 * int l_onoff; // option on/off
1353 * int l_linger; // linger time
1354 * };
1356 * To enable the option, set l_onoff to 1. If the l_linger value is set
1357 * to 0, calling close() is the same as the ABORT primitive. If the
1358 * value is set to a negative value, the setsockopt() call will return
1359 * an error. If the value is set to a positive value linger_time, the
1360 * close() can be blocked for at most linger_time ms. If the graceful
1361 * shutdown phase does not finish during this period, close() will
1362 * return but the graceful shutdown phase continues in the system.
1364 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1366 struct sctp_endpoint *ep;
1367 struct sctp_association *asoc;
1368 struct list_head *pos, *temp;
1370 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1372 sctp_lock_sock(sk);
1373 sk->sk_shutdown = SHUTDOWN_MASK;
1374 sk->sk_state = SCTP_SS_CLOSING;
1376 ep = sctp_sk(sk)->ep;
1378 /* Walk all associations on an endpoint. */
1379 list_for_each_safe(pos, temp, &ep->asocs) {
1380 asoc = list_entry(pos, struct sctp_association, asocs);
1382 if (sctp_style(sk, TCP)) {
1383 /* A closed association can still be in the list if
1384 * it belongs to a TCP-style listening socket that is
1385 * not yet accepted. If so, free it. If not, send an
1386 * ABORT or SHUTDOWN based on the linger options.
1388 if (sctp_state(asoc, CLOSED)) {
1389 sctp_unhash_established(asoc);
1390 sctp_association_free(asoc);
1391 continue;
1395 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1396 struct sctp_chunk *chunk;
1398 chunk = sctp_make_abort_user(asoc, NULL, 0);
1399 if (chunk)
1400 sctp_primitive_ABORT(asoc, chunk);
1401 } else
1402 sctp_primitive_SHUTDOWN(asoc, NULL);
1405 /* Clean up any skbs sitting on the receive queue. */
1406 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1407 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1409 /* On a TCP-style socket, block for at most linger_time if set. */
1410 if (sctp_style(sk, TCP) && timeout)
1411 sctp_wait_for_close(sk, timeout);
1413 /* This will run the backlog queue. */
1414 sctp_release_sock(sk);
1416 /* Supposedly, no process has access to the socket, but
1417 * the net layers still may.
1419 sctp_local_bh_disable();
1420 sctp_bh_lock_sock(sk);
1422 /* Hold the sock, since sk_common_release() will put sock_put()
1423 * and we have just a little more cleanup.
1425 sock_hold(sk);
1426 sk_common_release(sk);
1428 sctp_bh_unlock_sock(sk);
1429 sctp_local_bh_enable();
1431 sock_put(sk);
1433 SCTP_DBG_OBJCNT_DEC(sock);
1436 /* Handle EPIPE error. */
1437 static int sctp_error(struct sock *sk, int flags, int err)
1439 if (err == -EPIPE)
1440 err = sock_error(sk) ? : -EPIPE;
1441 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1442 send_sig(SIGPIPE, current, 0);
1443 return err;
1446 /* API 3.1.3 sendmsg() - UDP Style Syntax
1448 * An application uses sendmsg() and recvmsg() calls to transmit data to
1449 * and receive data from its peer.
1451 * ssize_t sendmsg(int socket, const struct msghdr *message,
1452 * int flags);
1454 * socket - the socket descriptor of the endpoint.
1455 * message - pointer to the msghdr structure which contains a single
1456 * user message and possibly some ancillary data.
1458 * See Section 5 for complete description of the data
1459 * structures.
1461 * flags - flags sent or received with the user message, see Section
1462 * 5 for complete description of the flags.
1464 * Note: This function could use a rewrite especially when explicit
1465 * connect support comes in.
1467 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1469 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1471 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1472 struct msghdr *msg, size_t msg_len)
1474 struct sctp_sock *sp;
1475 struct sctp_endpoint *ep;
1476 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1477 struct sctp_transport *transport, *chunk_tp;
1478 struct sctp_chunk *chunk;
1479 union sctp_addr to;
1480 struct sockaddr *msg_name = NULL;
1481 struct sctp_sndrcvinfo default_sinfo = { 0 };
1482 struct sctp_sndrcvinfo *sinfo;
1483 struct sctp_initmsg *sinit;
1484 sctp_assoc_t associd = 0;
1485 sctp_cmsgs_t cmsgs = { NULL };
1486 int err;
1487 sctp_scope_t scope;
1488 long timeo;
1489 __u16 sinfo_flags = 0;
1490 struct sctp_datamsg *datamsg;
1491 int msg_flags = msg->msg_flags;
1493 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1494 sk, msg, msg_len);
1496 err = 0;
1497 sp = sctp_sk(sk);
1498 ep = sp->ep;
1500 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1502 /* We cannot send a message over a TCP-style listening socket. */
1503 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1504 err = -EPIPE;
1505 goto out_nounlock;
1508 /* Parse out the SCTP CMSGs. */
1509 err = sctp_msghdr_parse(msg, &cmsgs);
1511 if (err) {
1512 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1513 goto out_nounlock;
1516 /* Fetch the destination address for this packet. This
1517 * address only selects the association--it is not necessarily
1518 * the address we will send to.
1519 * For a peeled-off socket, msg_name is ignored.
1521 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1522 int msg_namelen = msg->msg_namelen;
1524 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1525 msg_namelen);
1526 if (err)
1527 return err;
1529 if (msg_namelen > sizeof(to))
1530 msg_namelen = sizeof(to);
1531 memcpy(&to, msg->msg_name, msg_namelen);
1532 msg_name = msg->msg_name;
1535 sinfo = cmsgs.info;
1536 sinit = cmsgs.init;
1538 /* Did the user specify SNDRCVINFO? */
1539 if (sinfo) {
1540 sinfo_flags = sinfo->sinfo_flags;
1541 associd = sinfo->sinfo_assoc_id;
1544 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1545 msg_len, sinfo_flags);
1547 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1548 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1549 err = -EINVAL;
1550 goto out_nounlock;
1553 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1554 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1555 * If SCTP_ABORT is set, the message length could be non zero with
1556 * the msg_iov set to the user abort reason.
1558 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1559 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1560 err = -EINVAL;
1561 goto out_nounlock;
1564 /* If SCTP_ADDR_OVER is set, there must be an address
1565 * specified in msg_name.
1567 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1568 err = -EINVAL;
1569 goto out_nounlock;
1572 transport = NULL;
1574 SCTP_DEBUG_PRINTK("About to look up association.\n");
1576 sctp_lock_sock(sk);
1578 /* If a msg_name has been specified, assume this is to be used. */
1579 if (msg_name) {
1580 /* Look for a matching association on the endpoint. */
1581 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1582 if (!asoc) {
1583 /* If we could not find a matching association on the
1584 * endpoint, make sure that it is not a TCP-style
1585 * socket that already has an association or there is
1586 * no peeled-off association on another socket.
1588 if ((sctp_style(sk, TCP) &&
1589 sctp_sstate(sk, ESTABLISHED)) ||
1590 sctp_endpoint_is_peeled_off(ep, &to)) {
1591 err = -EADDRNOTAVAIL;
1592 goto out_unlock;
1595 } else {
1596 asoc = sctp_id2assoc(sk, associd);
1597 if (!asoc) {
1598 err = -EPIPE;
1599 goto out_unlock;
1603 if (asoc) {
1604 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1606 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1607 * socket that has an association in CLOSED state. This can
1608 * happen when an accepted socket has an association that is
1609 * already CLOSED.
1611 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1612 err = -EPIPE;
1613 goto out_unlock;
1616 if (sinfo_flags & SCTP_EOF) {
1617 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1618 asoc);
1619 sctp_primitive_SHUTDOWN(asoc, NULL);
1620 err = 0;
1621 goto out_unlock;
1623 if (sinfo_flags & SCTP_ABORT) {
1625 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1626 if (!chunk) {
1627 err = -ENOMEM;
1628 goto out_unlock;
1631 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1632 sctp_primitive_ABORT(asoc, chunk);
1633 err = 0;
1634 goto out_unlock;
1638 /* Do we need to create the association? */
1639 if (!asoc) {
1640 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1642 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1643 err = -EINVAL;
1644 goto out_unlock;
1647 /* Check for invalid stream against the stream counts,
1648 * either the default or the user specified stream counts.
1650 if (sinfo) {
1651 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1652 /* Check against the defaults. */
1653 if (sinfo->sinfo_stream >=
1654 sp->initmsg.sinit_num_ostreams) {
1655 err = -EINVAL;
1656 goto out_unlock;
1658 } else {
1659 /* Check against the requested. */
1660 if (sinfo->sinfo_stream >=
1661 sinit->sinit_num_ostreams) {
1662 err = -EINVAL;
1663 goto out_unlock;
1669 * API 3.1.2 bind() - UDP Style Syntax
1670 * If a bind() or sctp_bindx() is not called prior to a
1671 * sendmsg() call that initiates a new association, the
1672 * system picks an ephemeral port and will choose an address
1673 * set equivalent to binding with a wildcard address.
1675 if (!ep->base.bind_addr.port) {
1676 if (sctp_autobind(sk)) {
1677 err = -EAGAIN;
1678 goto out_unlock;
1680 } else {
1682 * If an unprivileged user inherits a one-to-many
1683 * style socket with open associations on a privileged
1684 * port, it MAY be permitted to accept new associations,
1685 * but it SHOULD NOT be permitted to open new
1686 * associations.
1688 if (ep->base.bind_addr.port < PROT_SOCK &&
1689 !capable(CAP_NET_BIND_SERVICE)) {
1690 err = -EACCES;
1691 goto out_unlock;
1695 scope = sctp_scope(&to);
1696 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1697 if (!new_asoc) {
1698 err = -ENOMEM;
1699 goto out_unlock;
1701 asoc = new_asoc;
1702 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1703 if (err < 0) {
1704 err = -ENOMEM;
1705 goto out_free;
1708 /* If the SCTP_INIT ancillary data is specified, set all
1709 * the association init values accordingly.
1711 if (sinit) {
1712 if (sinit->sinit_num_ostreams) {
1713 asoc->c.sinit_num_ostreams =
1714 sinit->sinit_num_ostreams;
1716 if (sinit->sinit_max_instreams) {
1717 asoc->c.sinit_max_instreams =
1718 sinit->sinit_max_instreams;
1720 if (sinit->sinit_max_attempts) {
1721 asoc->max_init_attempts
1722 = sinit->sinit_max_attempts;
1724 if (sinit->sinit_max_init_timeo) {
1725 asoc->max_init_timeo =
1726 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1730 /* Prime the peer's transport structures. */
1731 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1732 if (!transport) {
1733 err = -ENOMEM;
1734 goto out_free;
1738 /* ASSERT: we have a valid association at this point. */
1739 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1741 if (!sinfo) {
1742 /* If the user didn't specify SNDRCVINFO, make up one with
1743 * some defaults.
1745 default_sinfo.sinfo_stream = asoc->default_stream;
1746 default_sinfo.sinfo_flags = asoc->default_flags;
1747 default_sinfo.sinfo_ppid = asoc->default_ppid;
1748 default_sinfo.sinfo_context = asoc->default_context;
1749 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1750 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1751 sinfo = &default_sinfo;
1754 /* API 7.1.7, the sndbuf size per association bounds the
1755 * maximum size of data that can be sent in a single send call.
1757 if (msg_len > sk->sk_sndbuf) {
1758 err = -EMSGSIZE;
1759 goto out_free;
1762 if (asoc->pmtu_pending)
1763 sctp_assoc_pending_pmtu(asoc);
1765 /* If fragmentation is disabled and the message length exceeds the
1766 * association fragmentation point, return EMSGSIZE. The I-D
1767 * does not specify what this error is, but this looks like
1768 * a great fit.
1770 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1771 err = -EMSGSIZE;
1772 goto out_free;
1775 if (sinfo) {
1776 /* Check for invalid stream. */
1777 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1778 err = -EINVAL;
1779 goto out_free;
1783 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1784 if (!sctp_wspace(asoc)) {
1785 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1786 if (err)
1787 goto out_free;
1790 /* If an address is passed with the sendto/sendmsg call, it is used
1791 * to override the primary destination address in the TCP model, or
1792 * when SCTP_ADDR_OVER flag is set in the UDP model.
1794 if ((sctp_style(sk, TCP) && msg_name) ||
1795 (sinfo_flags & SCTP_ADDR_OVER)) {
1796 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1797 if (!chunk_tp) {
1798 err = -EINVAL;
1799 goto out_free;
1801 } else
1802 chunk_tp = NULL;
1804 /* Auto-connect, if we aren't connected already. */
1805 if (sctp_state(asoc, CLOSED)) {
1806 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1807 if (err < 0)
1808 goto out_free;
1809 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1812 /* Break the message into multiple chunks of maximum size. */
1813 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1814 if (!datamsg) {
1815 err = -ENOMEM;
1816 goto out_free;
1819 /* Now send the (possibly) fragmented message. */
1820 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1821 sctp_chunk_hold(chunk);
1823 /* Do accounting for the write space. */
1824 sctp_set_owner_w(chunk);
1826 chunk->transport = chunk_tp;
1829 /* Send it to the lower layers. Note: all chunks
1830 * must either fail or succeed. The lower layer
1831 * works that way today. Keep it that way or this
1832 * breaks.
1834 err = sctp_primitive_SEND(asoc, datamsg);
1835 /* Did the lower layer accept the chunk? */
1836 if (err)
1837 sctp_datamsg_free(datamsg);
1838 else
1839 sctp_datamsg_put(datamsg);
1841 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1843 if (err)
1844 goto out_free;
1845 else
1846 err = msg_len;
1848 /* If we are already past ASSOCIATE, the lower
1849 * layers are responsible for association cleanup.
1851 goto out_unlock;
1853 out_free:
1854 if (new_asoc)
1855 sctp_association_free(asoc);
1856 out_unlock:
1857 sctp_release_sock(sk);
1859 out_nounlock:
1860 return sctp_error(sk, msg_flags, err);
1862 #if 0
1863 do_sock_err:
1864 if (msg_len)
1865 err = msg_len;
1866 else
1867 err = sock_error(sk);
1868 goto out;
1870 do_interrupted:
1871 if (msg_len)
1872 err = msg_len;
1873 goto out;
1874 #endif /* 0 */
1877 /* This is an extended version of skb_pull() that removes the data from the
1878 * start of a skb even when data is spread across the list of skb's in the
1879 * frag_list. len specifies the total amount of data that needs to be removed.
1880 * when 'len' bytes could be removed from the skb, it returns 0.
1881 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1882 * could not be removed.
1884 static int sctp_skb_pull(struct sk_buff *skb, int len)
1886 struct sk_buff *list;
1887 int skb_len = skb_headlen(skb);
1888 int rlen;
1890 if (len <= skb_len) {
1891 __skb_pull(skb, len);
1892 return 0;
1894 len -= skb_len;
1895 __skb_pull(skb, skb_len);
1897 skb_walk_frags(skb, list) {
1898 rlen = sctp_skb_pull(list, len);
1899 skb->len -= (len-rlen);
1900 skb->data_len -= (len-rlen);
1902 if (!rlen)
1903 return 0;
1905 len = rlen;
1908 return len;
1911 /* API 3.1.3 recvmsg() - UDP Style Syntax
1913 * ssize_t recvmsg(int socket, struct msghdr *message,
1914 * int flags);
1916 * socket - the socket descriptor of the endpoint.
1917 * message - pointer to the msghdr structure which contains a single
1918 * user message and possibly some ancillary data.
1920 * See Section 5 for complete description of the data
1921 * structures.
1923 * flags - flags sent or received with the user message, see Section
1924 * 5 for complete description of the flags.
1926 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1928 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1929 struct msghdr *msg, size_t len, int noblock,
1930 int flags, int *addr_len)
1932 struct sctp_ulpevent *event = NULL;
1933 struct sctp_sock *sp = sctp_sk(sk);
1934 struct sk_buff *skb;
1935 int copied;
1936 int err = 0;
1937 int skb_len;
1939 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1940 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1941 "len", len, "knoblauch", noblock,
1942 "flags", flags, "addr_len", addr_len);
1944 sctp_lock_sock(sk);
1946 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1947 err = -ENOTCONN;
1948 goto out;
1951 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1952 if (!skb)
1953 goto out;
1955 /* Get the total length of the skb including any skb's in the
1956 * frag_list.
1958 skb_len = skb->len;
1960 copied = skb_len;
1961 if (copied > len)
1962 copied = len;
1964 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1966 event = sctp_skb2event(skb);
1968 if (err)
1969 goto out_free;
1971 sock_recv_timestamp(msg, sk, skb);
1972 if (sctp_ulpevent_is_notification(event)) {
1973 msg->msg_flags |= MSG_NOTIFICATION;
1974 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1975 } else {
1976 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1979 /* Check if we allow SCTP_SNDRCVINFO. */
1980 if (sp->subscribe.sctp_data_io_event)
1981 sctp_ulpevent_read_sndrcvinfo(event, msg);
1982 #if 0
1983 /* FIXME: we should be calling IP/IPv6 layers. */
1984 if (sk->sk_protinfo.af_inet.cmsg_flags)
1985 ip_cmsg_recv(msg, skb);
1986 #endif
1988 err = copied;
1990 /* If skb's length exceeds the user's buffer, update the skb and
1991 * push it back to the receive_queue so that the next call to
1992 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1994 if (skb_len > copied) {
1995 msg->msg_flags &= ~MSG_EOR;
1996 if (flags & MSG_PEEK)
1997 goto out_free;
1998 sctp_skb_pull(skb, copied);
1999 skb_queue_head(&sk->sk_receive_queue, skb);
2001 /* When only partial message is copied to the user, increase
2002 * rwnd by that amount. If all the data in the skb is read,
2003 * rwnd is updated when the event is freed.
2005 if (!sctp_ulpevent_is_notification(event))
2006 sctp_assoc_rwnd_increase(event->asoc, copied);
2007 goto out;
2008 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2009 (event->msg_flags & MSG_EOR))
2010 msg->msg_flags |= MSG_EOR;
2011 else
2012 msg->msg_flags &= ~MSG_EOR;
2014 out_free:
2015 if (flags & MSG_PEEK) {
2016 /* Release the skb reference acquired after peeking the skb in
2017 * sctp_skb_recv_datagram().
2019 kfree_skb(skb);
2020 } else {
2021 /* Free the event which includes releasing the reference to
2022 * the owner of the skb, freeing the skb and updating the
2023 * rwnd.
2025 sctp_ulpevent_free(event);
2027 out:
2028 sctp_release_sock(sk);
2029 return err;
2032 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2034 * This option is a on/off flag. If enabled no SCTP message
2035 * fragmentation will be performed. Instead if a message being sent
2036 * exceeds the current PMTU size, the message will NOT be sent and
2037 * instead a error will be indicated to the user.
2039 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2040 char __user *optval,
2041 unsigned int optlen)
2043 int val;
2045 if (optlen < sizeof(int))
2046 return -EINVAL;
2048 if (get_user(val, (int __user *)optval))
2049 return -EFAULT;
2051 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2053 return 0;
2056 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2057 unsigned int optlen)
2059 if (optlen > sizeof(struct sctp_event_subscribe))
2060 return -EINVAL;
2061 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2062 return -EFAULT;
2063 return 0;
2066 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2068 * This socket option is applicable to the UDP-style socket only. When
2069 * set it will cause associations that are idle for more than the
2070 * specified number of seconds to automatically close. An association
2071 * being idle is defined an association that has NOT sent or received
2072 * user data. The special value of '0' indicates that no automatic
2073 * close of any associations should be performed. The option expects an
2074 * integer defining the number of seconds of idle time before an
2075 * association is closed.
2077 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2078 unsigned int optlen)
2080 struct sctp_sock *sp = sctp_sk(sk);
2082 /* Applicable to UDP-style socket only */
2083 if (sctp_style(sk, TCP))
2084 return -EOPNOTSUPP;
2085 if (optlen != sizeof(int))
2086 return -EINVAL;
2087 if (copy_from_user(&sp->autoclose, optval, optlen))
2088 return -EFAULT;
2090 return 0;
2093 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2095 * Applications can enable or disable heartbeats for any peer address of
2096 * an association, modify an address's heartbeat interval, force a
2097 * heartbeat to be sent immediately, and adjust the address's maximum
2098 * number of retransmissions sent before an address is considered
2099 * unreachable. The following structure is used to access and modify an
2100 * address's parameters:
2102 * struct sctp_paddrparams {
2103 * sctp_assoc_t spp_assoc_id;
2104 * struct sockaddr_storage spp_address;
2105 * uint32_t spp_hbinterval;
2106 * uint16_t spp_pathmaxrxt;
2107 * uint32_t spp_pathmtu;
2108 * uint32_t spp_sackdelay;
2109 * uint32_t spp_flags;
2110 * };
2112 * spp_assoc_id - (one-to-many style socket) This is filled in the
2113 * application, and identifies the association for
2114 * this query.
2115 * spp_address - This specifies which address is of interest.
2116 * spp_hbinterval - This contains the value of the heartbeat interval,
2117 * in milliseconds. If a value of zero
2118 * is present in this field then no changes are to
2119 * be made to this parameter.
2120 * spp_pathmaxrxt - This contains the maximum number of
2121 * retransmissions before this address shall be
2122 * considered unreachable. If a value of zero
2123 * is present in this field then no changes are to
2124 * be made to this parameter.
2125 * spp_pathmtu - When Path MTU discovery is disabled the value
2126 * specified here will be the "fixed" path mtu.
2127 * Note that if the spp_address field is empty
2128 * then all associations on this address will
2129 * have this fixed path mtu set upon them.
2131 * spp_sackdelay - When delayed sack is enabled, this value specifies
2132 * the number of milliseconds that sacks will be delayed
2133 * for. This value will apply to all addresses of an
2134 * association if the spp_address field is empty. Note
2135 * also, that if delayed sack is enabled and this
2136 * value is set to 0, no change is made to the last
2137 * recorded delayed sack timer value.
2139 * spp_flags - These flags are used to control various features
2140 * on an association. The flag field may contain
2141 * zero or more of the following options.
2143 * SPP_HB_ENABLE - Enable heartbeats on the
2144 * specified address. Note that if the address
2145 * field is empty all addresses for the association
2146 * have heartbeats enabled upon them.
2148 * SPP_HB_DISABLE - Disable heartbeats on the
2149 * speicifed address. Note that if the address
2150 * field is empty all addresses for the association
2151 * will have their heartbeats disabled. Note also
2152 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2153 * mutually exclusive, only one of these two should
2154 * be specified. Enabling both fields will have
2155 * undetermined results.
2157 * SPP_HB_DEMAND - Request a user initiated heartbeat
2158 * to be made immediately.
2160 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2161 * heartbeat delayis to be set to the value of 0
2162 * milliseconds.
2164 * SPP_PMTUD_ENABLE - This field will enable PMTU
2165 * discovery upon the specified address. Note that
2166 * if the address feild is empty then all addresses
2167 * on the association are effected.
2169 * SPP_PMTUD_DISABLE - This field will disable PMTU
2170 * discovery upon the specified address. Note that
2171 * if the address feild is empty then all addresses
2172 * on the association are effected. Not also that
2173 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2174 * exclusive. Enabling both will have undetermined
2175 * results.
2177 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2178 * on delayed sack. The time specified in spp_sackdelay
2179 * is used to specify the sack delay for this address. Note
2180 * that if spp_address is empty then all addresses will
2181 * enable delayed sack and take on the sack delay
2182 * value specified in spp_sackdelay.
2183 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2184 * off delayed sack. If the spp_address field is blank then
2185 * delayed sack is disabled for the entire association. Note
2186 * also that this field is mutually exclusive to
2187 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2188 * results.
2190 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2191 struct sctp_transport *trans,
2192 struct sctp_association *asoc,
2193 struct sctp_sock *sp,
2194 int hb_change,
2195 int pmtud_change,
2196 int sackdelay_change)
2198 int error;
2200 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2201 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2202 if (error)
2203 return error;
2206 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2207 * this field is ignored. Note also that a value of zero indicates
2208 * the current setting should be left unchanged.
2210 if (params->spp_flags & SPP_HB_ENABLE) {
2212 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2213 * set. This lets us use 0 value when this flag
2214 * is set.
2216 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2217 params->spp_hbinterval = 0;
2219 if (params->spp_hbinterval ||
2220 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2221 if (trans) {
2222 trans->hbinterval =
2223 msecs_to_jiffies(params->spp_hbinterval);
2224 } else if (asoc) {
2225 asoc->hbinterval =
2226 msecs_to_jiffies(params->spp_hbinterval);
2227 } else {
2228 sp->hbinterval = params->spp_hbinterval;
2233 if (hb_change) {
2234 if (trans) {
2235 trans->param_flags =
2236 (trans->param_flags & ~SPP_HB) | hb_change;
2237 } else if (asoc) {
2238 asoc->param_flags =
2239 (asoc->param_flags & ~SPP_HB) | hb_change;
2240 } else {
2241 sp->param_flags =
2242 (sp->param_flags & ~SPP_HB) | hb_change;
2246 /* When Path MTU discovery is disabled the value specified here will
2247 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2248 * include the flag SPP_PMTUD_DISABLE for this field to have any
2249 * effect).
2251 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2252 if (trans) {
2253 trans->pathmtu = params->spp_pathmtu;
2254 sctp_assoc_sync_pmtu(asoc);
2255 } else if (asoc) {
2256 asoc->pathmtu = params->spp_pathmtu;
2257 sctp_frag_point(asoc, params->spp_pathmtu);
2258 } else {
2259 sp->pathmtu = params->spp_pathmtu;
2263 if (pmtud_change) {
2264 if (trans) {
2265 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2266 (params->spp_flags & SPP_PMTUD_ENABLE);
2267 trans->param_flags =
2268 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2269 if (update) {
2270 sctp_transport_pmtu(trans);
2271 sctp_assoc_sync_pmtu(asoc);
2273 } else if (asoc) {
2274 asoc->param_flags =
2275 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2276 } else {
2277 sp->param_flags =
2278 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2282 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2283 * value of this field is ignored. Note also that a value of zero
2284 * indicates the current setting should be left unchanged.
2286 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2287 if (trans) {
2288 trans->sackdelay =
2289 msecs_to_jiffies(params->spp_sackdelay);
2290 } else if (asoc) {
2291 asoc->sackdelay =
2292 msecs_to_jiffies(params->spp_sackdelay);
2293 } else {
2294 sp->sackdelay = params->spp_sackdelay;
2298 if (sackdelay_change) {
2299 if (trans) {
2300 trans->param_flags =
2301 (trans->param_flags & ~SPP_SACKDELAY) |
2302 sackdelay_change;
2303 } else if (asoc) {
2304 asoc->param_flags =
2305 (asoc->param_flags & ~SPP_SACKDELAY) |
2306 sackdelay_change;
2307 } else {
2308 sp->param_flags =
2309 (sp->param_flags & ~SPP_SACKDELAY) |
2310 sackdelay_change;
2314 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2315 * of this field is ignored. Note also that a value of zero
2316 * indicates the current setting should be left unchanged.
2318 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2319 if (trans) {
2320 trans->pathmaxrxt = params->spp_pathmaxrxt;
2321 } else if (asoc) {
2322 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2323 } else {
2324 sp->pathmaxrxt = params->spp_pathmaxrxt;
2328 return 0;
2331 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2332 char __user *optval,
2333 unsigned int optlen)
2335 struct sctp_paddrparams params;
2336 struct sctp_transport *trans = NULL;
2337 struct sctp_association *asoc = NULL;
2338 struct sctp_sock *sp = sctp_sk(sk);
2339 int error;
2340 int hb_change, pmtud_change, sackdelay_change;
2342 if (optlen != sizeof(struct sctp_paddrparams))
2343 return - EINVAL;
2345 if (copy_from_user(&params, optval, optlen))
2346 return -EFAULT;
2348 /* Validate flags and value parameters. */
2349 hb_change = params.spp_flags & SPP_HB;
2350 pmtud_change = params.spp_flags & SPP_PMTUD;
2351 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2353 if (hb_change == SPP_HB ||
2354 pmtud_change == SPP_PMTUD ||
2355 sackdelay_change == SPP_SACKDELAY ||
2356 params.spp_sackdelay > 500 ||
2357 (params.spp_pathmtu
2358 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2359 return -EINVAL;
2361 /* If an address other than INADDR_ANY is specified, and
2362 * no transport is found, then the request is invalid.
2364 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2365 trans = sctp_addr_id2transport(sk, &params.spp_address,
2366 params.spp_assoc_id);
2367 if (!trans)
2368 return -EINVAL;
2371 /* Get association, if assoc_id != 0 and the socket is a one
2372 * to many style socket, and an association was not found, then
2373 * the id was invalid.
2375 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2376 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2377 return -EINVAL;
2379 /* Heartbeat demand can only be sent on a transport or
2380 * association, but not a socket.
2382 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2383 return -EINVAL;
2385 /* Process parameters. */
2386 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2387 hb_change, pmtud_change,
2388 sackdelay_change);
2390 if (error)
2391 return error;
2393 /* If changes are for association, also apply parameters to each
2394 * transport.
2396 if (!trans && asoc) {
2397 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2398 transports) {
2399 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2400 hb_change, pmtud_change,
2401 sackdelay_change);
2405 return 0;
2409 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2411 * This option will effect the way delayed acks are performed. This
2412 * option allows you to get or set the delayed ack time, in
2413 * milliseconds. It also allows changing the delayed ack frequency.
2414 * Changing the frequency to 1 disables the delayed sack algorithm. If
2415 * the assoc_id is 0, then this sets or gets the endpoints default
2416 * values. If the assoc_id field is non-zero, then the set or get
2417 * effects the specified association for the one to many model (the
2418 * assoc_id field is ignored by the one to one model). Note that if
2419 * sack_delay or sack_freq are 0 when setting this option, then the
2420 * current values will remain unchanged.
2422 * struct sctp_sack_info {
2423 * sctp_assoc_t sack_assoc_id;
2424 * uint32_t sack_delay;
2425 * uint32_t sack_freq;
2426 * };
2428 * sack_assoc_id - This parameter, indicates which association the user
2429 * is performing an action upon. Note that if this field's value is
2430 * zero then the endpoints default value is changed (effecting future
2431 * associations only).
2433 * sack_delay - This parameter contains the number of milliseconds that
2434 * the user is requesting the delayed ACK timer be set to. Note that
2435 * this value is defined in the standard to be between 200 and 500
2436 * milliseconds.
2438 * sack_freq - This parameter contains the number of packets that must
2439 * be received before a sack is sent without waiting for the delay
2440 * timer to expire. The default value for this is 2, setting this
2441 * value to 1 will disable the delayed sack algorithm.
2444 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2445 char __user *optval, unsigned int optlen)
2447 struct sctp_sack_info params;
2448 struct sctp_transport *trans = NULL;
2449 struct sctp_association *asoc = NULL;
2450 struct sctp_sock *sp = sctp_sk(sk);
2452 if (optlen == sizeof(struct sctp_sack_info)) {
2453 if (copy_from_user(&params, optval, optlen))
2454 return -EFAULT;
2456 if (params.sack_delay == 0 && params.sack_freq == 0)
2457 return 0;
2458 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2459 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
2460 "in delayed_ack socket option deprecated\n");
2461 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
2462 if (copy_from_user(&params, optval, optlen))
2463 return -EFAULT;
2465 if (params.sack_delay == 0)
2466 params.sack_freq = 1;
2467 else
2468 params.sack_freq = 0;
2469 } else
2470 return - EINVAL;
2472 /* Validate value parameter. */
2473 if (params.sack_delay > 500)
2474 return -EINVAL;
2476 /* Get association, if sack_assoc_id != 0 and the socket is a one
2477 * to many style socket, and an association was not found, then
2478 * the id was invalid.
2480 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2481 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2482 return -EINVAL;
2484 if (params.sack_delay) {
2485 if (asoc) {
2486 asoc->sackdelay =
2487 msecs_to_jiffies(params.sack_delay);
2488 asoc->param_flags =
2489 (asoc->param_flags & ~SPP_SACKDELAY) |
2490 SPP_SACKDELAY_ENABLE;
2491 } else {
2492 sp->sackdelay = params.sack_delay;
2493 sp->param_flags =
2494 (sp->param_flags & ~SPP_SACKDELAY) |
2495 SPP_SACKDELAY_ENABLE;
2499 if (params.sack_freq == 1) {
2500 if (asoc) {
2501 asoc->param_flags =
2502 (asoc->param_flags & ~SPP_SACKDELAY) |
2503 SPP_SACKDELAY_DISABLE;
2504 } else {
2505 sp->param_flags =
2506 (sp->param_flags & ~SPP_SACKDELAY) |
2507 SPP_SACKDELAY_DISABLE;
2509 } else if (params.sack_freq > 1) {
2510 if (asoc) {
2511 asoc->sackfreq = params.sack_freq;
2512 asoc->param_flags =
2513 (asoc->param_flags & ~SPP_SACKDELAY) |
2514 SPP_SACKDELAY_ENABLE;
2515 } else {
2516 sp->sackfreq = params.sack_freq;
2517 sp->param_flags =
2518 (sp->param_flags & ~SPP_SACKDELAY) |
2519 SPP_SACKDELAY_ENABLE;
2523 /* If change is for association, also apply to each transport. */
2524 if (asoc) {
2525 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2526 transports) {
2527 if (params.sack_delay) {
2528 trans->sackdelay =
2529 msecs_to_jiffies(params.sack_delay);
2530 trans->param_flags =
2531 (trans->param_flags & ~SPP_SACKDELAY) |
2532 SPP_SACKDELAY_ENABLE;
2534 if (params.sack_freq == 1) {
2535 trans->param_flags =
2536 (trans->param_flags & ~SPP_SACKDELAY) |
2537 SPP_SACKDELAY_DISABLE;
2538 } else if (params.sack_freq > 1) {
2539 trans->sackfreq = params.sack_freq;
2540 trans->param_flags =
2541 (trans->param_flags & ~SPP_SACKDELAY) |
2542 SPP_SACKDELAY_ENABLE;
2547 return 0;
2550 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2552 * Applications can specify protocol parameters for the default association
2553 * initialization. The option name argument to setsockopt() and getsockopt()
2554 * is SCTP_INITMSG.
2556 * Setting initialization parameters is effective only on an unconnected
2557 * socket (for UDP-style sockets only future associations are effected
2558 * by the change). With TCP-style sockets, this option is inherited by
2559 * sockets derived from a listener socket.
2561 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2563 struct sctp_initmsg sinit;
2564 struct sctp_sock *sp = sctp_sk(sk);
2566 if (optlen != sizeof(struct sctp_initmsg))
2567 return -EINVAL;
2568 if (copy_from_user(&sinit, optval, optlen))
2569 return -EFAULT;
2571 if (sinit.sinit_num_ostreams)
2572 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2573 if (sinit.sinit_max_instreams)
2574 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2575 if (sinit.sinit_max_attempts)
2576 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2577 if (sinit.sinit_max_init_timeo)
2578 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2580 return 0;
2584 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2586 * Applications that wish to use the sendto() system call may wish to
2587 * specify a default set of parameters that would normally be supplied
2588 * through the inclusion of ancillary data. This socket option allows
2589 * such an application to set the default sctp_sndrcvinfo structure.
2590 * The application that wishes to use this socket option simply passes
2591 * in to this call the sctp_sndrcvinfo structure defined in Section
2592 * 5.2.2) The input parameters accepted by this call include
2593 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2594 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2595 * to this call if the caller is using the UDP model.
2597 static int sctp_setsockopt_default_send_param(struct sock *sk,
2598 char __user *optval,
2599 unsigned int optlen)
2601 struct sctp_sndrcvinfo info;
2602 struct sctp_association *asoc;
2603 struct sctp_sock *sp = sctp_sk(sk);
2605 if (optlen != sizeof(struct sctp_sndrcvinfo))
2606 return -EINVAL;
2607 if (copy_from_user(&info, optval, optlen))
2608 return -EFAULT;
2610 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2611 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2612 return -EINVAL;
2614 if (asoc) {
2615 asoc->default_stream = info.sinfo_stream;
2616 asoc->default_flags = info.sinfo_flags;
2617 asoc->default_ppid = info.sinfo_ppid;
2618 asoc->default_context = info.sinfo_context;
2619 asoc->default_timetolive = info.sinfo_timetolive;
2620 } else {
2621 sp->default_stream = info.sinfo_stream;
2622 sp->default_flags = info.sinfo_flags;
2623 sp->default_ppid = info.sinfo_ppid;
2624 sp->default_context = info.sinfo_context;
2625 sp->default_timetolive = info.sinfo_timetolive;
2628 return 0;
2631 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2633 * Requests that the local SCTP stack use the enclosed peer address as
2634 * the association primary. The enclosed address must be one of the
2635 * association peer's addresses.
2637 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2638 unsigned int optlen)
2640 struct sctp_prim prim;
2641 struct sctp_transport *trans;
2643 if (optlen != sizeof(struct sctp_prim))
2644 return -EINVAL;
2646 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2647 return -EFAULT;
2649 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2650 if (!trans)
2651 return -EINVAL;
2653 sctp_assoc_set_primary(trans->asoc, trans);
2655 return 0;
2659 * 7.1.5 SCTP_NODELAY
2661 * Turn on/off any Nagle-like algorithm. This means that packets are
2662 * generally sent as soon as possible and no unnecessary delays are
2663 * introduced, at the cost of more packets in the network. Expects an
2664 * integer boolean flag.
2666 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2667 unsigned int optlen)
2669 int val;
2671 if (optlen < sizeof(int))
2672 return -EINVAL;
2673 if (get_user(val, (int __user *)optval))
2674 return -EFAULT;
2676 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2677 return 0;
2682 * 7.1.1 SCTP_RTOINFO
2684 * The protocol parameters used to initialize and bound retransmission
2685 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2686 * and modify these parameters.
2687 * All parameters are time values, in milliseconds. A value of 0, when
2688 * modifying the parameters, indicates that the current value should not
2689 * be changed.
2692 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2694 struct sctp_rtoinfo rtoinfo;
2695 struct sctp_association *asoc;
2697 if (optlen != sizeof (struct sctp_rtoinfo))
2698 return -EINVAL;
2700 if (copy_from_user(&rtoinfo, optval, optlen))
2701 return -EFAULT;
2703 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2705 /* Set the values to the specific association */
2706 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2707 return -EINVAL;
2709 if (asoc) {
2710 if (rtoinfo.srto_initial != 0)
2711 asoc->rto_initial =
2712 msecs_to_jiffies(rtoinfo.srto_initial);
2713 if (rtoinfo.srto_max != 0)
2714 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2715 if (rtoinfo.srto_min != 0)
2716 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2717 } else {
2718 /* If there is no association or the association-id = 0
2719 * set the values to the endpoint.
2721 struct sctp_sock *sp = sctp_sk(sk);
2723 if (rtoinfo.srto_initial != 0)
2724 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2725 if (rtoinfo.srto_max != 0)
2726 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2727 if (rtoinfo.srto_min != 0)
2728 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2731 return 0;
2736 * 7.1.2 SCTP_ASSOCINFO
2738 * This option is used to tune the maximum retransmission attempts
2739 * of the association.
2740 * Returns an error if the new association retransmission value is
2741 * greater than the sum of the retransmission value of the peer.
2742 * See [SCTP] for more information.
2745 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2748 struct sctp_assocparams assocparams;
2749 struct sctp_association *asoc;
2751 if (optlen != sizeof(struct sctp_assocparams))
2752 return -EINVAL;
2753 if (copy_from_user(&assocparams, optval, optlen))
2754 return -EFAULT;
2756 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2758 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2759 return -EINVAL;
2761 /* Set the values to the specific association */
2762 if (asoc) {
2763 if (assocparams.sasoc_asocmaxrxt != 0) {
2764 __u32 path_sum = 0;
2765 int paths = 0;
2766 struct sctp_transport *peer_addr;
2768 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2769 transports) {
2770 path_sum += peer_addr->pathmaxrxt;
2771 paths++;
2774 /* Only validate asocmaxrxt if we have more than
2775 * one path/transport. We do this because path
2776 * retransmissions are only counted when we have more
2777 * then one path.
2779 if (paths > 1 &&
2780 assocparams.sasoc_asocmaxrxt > path_sum)
2781 return -EINVAL;
2783 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2786 if (assocparams.sasoc_cookie_life != 0) {
2787 asoc->cookie_life.tv_sec =
2788 assocparams.sasoc_cookie_life / 1000;
2789 asoc->cookie_life.tv_usec =
2790 (assocparams.sasoc_cookie_life % 1000)
2791 * 1000;
2793 } else {
2794 /* Set the values to the endpoint */
2795 struct sctp_sock *sp = sctp_sk(sk);
2797 if (assocparams.sasoc_asocmaxrxt != 0)
2798 sp->assocparams.sasoc_asocmaxrxt =
2799 assocparams.sasoc_asocmaxrxt;
2800 if (assocparams.sasoc_cookie_life != 0)
2801 sp->assocparams.sasoc_cookie_life =
2802 assocparams.sasoc_cookie_life;
2804 return 0;
2808 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2810 * This socket option is a boolean flag which turns on or off mapped V4
2811 * addresses. If this option is turned on and the socket is type
2812 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2813 * If this option is turned off, then no mapping will be done of V4
2814 * addresses and a user will receive both PF_INET6 and PF_INET type
2815 * addresses on the socket.
2817 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2819 int val;
2820 struct sctp_sock *sp = sctp_sk(sk);
2822 if (optlen < sizeof(int))
2823 return -EINVAL;
2824 if (get_user(val, (int __user *)optval))
2825 return -EFAULT;
2826 if (val)
2827 sp->v4mapped = 1;
2828 else
2829 sp->v4mapped = 0;
2831 return 0;
2835 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2836 * This option will get or set the maximum size to put in any outgoing
2837 * SCTP DATA chunk. If a message is larger than this size it will be
2838 * fragmented by SCTP into the specified size. Note that the underlying
2839 * SCTP implementation may fragment into smaller sized chunks when the
2840 * PMTU of the underlying association is smaller than the value set by
2841 * the user. The default value for this option is '0' which indicates
2842 * the user is NOT limiting fragmentation and only the PMTU will effect
2843 * SCTP's choice of DATA chunk size. Note also that values set larger
2844 * than the maximum size of an IP datagram will effectively let SCTP
2845 * control fragmentation (i.e. the same as setting this option to 0).
2847 * The following structure is used to access and modify this parameter:
2849 * struct sctp_assoc_value {
2850 * sctp_assoc_t assoc_id;
2851 * uint32_t assoc_value;
2852 * };
2854 * assoc_id: This parameter is ignored for one-to-one style sockets.
2855 * For one-to-many style sockets this parameter indicates which
2856 * association the user is performing an action upon. Note that if
2857 * this field's value is zero then the endpoints default value is
2858 * changed (effecting future associations only).
2859 * assoc_value: This parameter specifies the maximum size in bytes.
2861 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2863 struct sctp_assoc_value params;
2864 struct sctp_association *asoc;
2865 struct sctp_sock *sp = sctp_sk(sk);
2866 int val;
2868 if (optlen == sizeof(int)) {
2869 printk(KERN_WARNING
2870 "SCTP: Use of int in maxseg socket option deprecated\n");
2871 printk(KERN_WARNING
2872 "SCTP: Use struct sctp_assoc_value instead\n");
2873 if (copy_from_user(&val, optval, optlen))
2874 return -EFAULT;
2875 params.assoc_id = 0;
2876 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2877 if (copy_from_user(&params, optval, optlen))
2878 return -EFAULT;
2879 val = params.assoc_value;
2880 } else
2881 return -EINVAL;
2883 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2884 return -EINVAL;
2886 asoc = sctp_id2assoc(sk, params.assoc_id);
2887 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2888 return -EINVAL;
2890 if (asoc) {
2891 if (val == 0) {
2892 val = asoc->pathmtu;
2893 val -= sp->pf->af->net_header_len;
2894 val -= sizeof(struct sctphdr) +
2895 sizeof(struct sctp_data_chunk);
2897 asoc->user_frag = val;
2898 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2899 } else {
2900 sp->user_frag = val;
2903 return 0;
2908 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2910 * Requests that the peer mark the enclosed address as the association
2911 * primary. The enclosed address must be one of the association's
2912 * locally bound addresses. The following structure is used to make a
2913 * set primary request:
2915 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2916 unsigned int optlen)
2918 struct sctp_sock *sp;
2919 struct sctp_endpoint *ep;
2920 struct sctp_association *asoc = NULL;
2921 struct sctp_setpeerprim prim;
2922 struct sctp_chunk *chunk;
2923 int err;
2925 sp = sctp_sk(sk);
2926 ep = sp->ep;
2928 if (!sctp_addip_enable)
2929 return -EPERM;
2931 if (optlen != sizeof(struct sctp_setpeerprim))
2932 return -EINVAL;
2934 if (copy_from_user(&prim, optval, optlen))
2935 return -EFAULT;
2937 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2938 if (!asoc)
2939 return -EINVAL;
2941 if (!asoc->peer.asconf_capable)
2942 return -EPERM;
2944 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2945 return -EPERM;
2947 if (!sctp_state(asoc, ESTABLISHED))
2948 return -ENOTCONN;
2950 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2951 return -EADDRNOTAVAIL;
2953 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2954 chunk = sctp_make_asconf_set_prim(asoc,
2955 (union sctp_addr *)&prim.sspp_addr);
2956 if (!chunk)
2957 return -ENOMEM;
2959 err = sctp_send_asconf(asoc, chunk);
2961 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2963 return err;
2966 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2967 unsigned int optlen)
2969 struct sctp_setadaptation adaptation;
2971 if (optlen != sizeof(struct sctp_setadaptation))
2972 return -EINVAL;
2973 if (copy_from_user(&adaptation, optval, optlen))
2974 return -EFAULT;
2976 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2978 return 0;
2982 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2984 * The context field in the sctp_sndrcvinfo structure is normally only
2985 * used when a failed message is retrieved holding the value that was
2986 * sent down on the actual send call. This option allows the setting of
2987 * a default context on an association basis that will be received on
2988 * reading messages from the peer. This is especially helpful in the
2989 * one-2-many model for an application to keep some reference to an
2990 * internal state machine that is processing messages on the
2991 * association. Note that the setting of this value only effects
2992 * received messages from the peer and does not effect the value that is
2993 * saved with outbound messages.
2995 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2996 unsigned int optlen)
2998 struct sctp_assoc_value params;
2999 struct sctp_sock *sp;
3000 struct sctp_association *asoc;
3002 if (optlen != sizeof(struct sctp_assoc_value))
3003 return -EINVAL;
3004 if (copy_from_user(&params, optval, optlen))
3005 return -EFAULT;
3007 sp = sctp_sk(sk);
3009 if (params.assoc_id != 0) {
3010 asoc = sctp_id2assoc(sk, params.assoc_id);
3011 if (!asoc)
3012 return -EINVAL;
3013 asoc->default_rcv_context = params.assoc_value;
3014 } else {
3015 sp->default_rcv_context = params.assoc_value;
3018 return 0;
3022 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3024 * This options will at a minimum specify if the implementation is doing
3025 * fragmented interleave. Fragmented interleave, for a one to many
3026 * socket, is when subsequent calls to receive a message may return
3027 * parts of messages from different associations. Some implementations
3028 * may allow you to turn this value on or off. If so, when turned off,
3029 * no fragment interleave will occur (which will cause a head of line
3030 * blocking amongst multiple associations sharing the same one to many
3031 * socket). When this option is turned on, then each receive call may
3032 * come from a different association (thus the user must receive data
3033 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3034 * association each receive belongs to.
3036 * This option takes a boolean value. A non-zero value indicates that
3037 * fragmented interleave is on. A value of zero indicates that
3038 * fragmented interleave is off.
3040 * Note that it is important that an implementation that allows this
3041 * option to be turned on, have it off by default. Otherwise an unaware
3042 * application using the one to many model may become confused and act
3043 * incorrectly.
3045 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3046 char __user *optval,
3047 unsigned int optlen)
3049 int val;
3051 if (optlen != sizeof(int))
3052 return -EINVAL;
3053 if (get_user(val, (int __user *)optval))
3054 return -EFAULT;
3056 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3058 return 0;
3062 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3063 * (SCTP_PARTIAL_DELIVERY_POINT)
3065 * This option will set or get the SCTP partial delivery point. This
3066 * point is the size of a message where the partial delivery API will be
3067 * invoked to help free up rwnd space for the peer. Setting this to a
3068 * lower value will cause partial deliveries to happen more often. The
3069 * calls argument is an integer that sets or gets the partial delivery
3070 * point. Note also that the call will fail if the user attempts to set
3071 * this value larger than the socket receive buffer size.
3073 * Note that any single message having a length smaller than or equal to
3074 * the SCTP partial delivery point will be delivered in one single read
3075 * call as long as the user provided buffer is large enough to hold the
3076 * message.
3078 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3079 char __user *optval,
3080 unsigned int optlen)
3082 u32 val;
3084 if (optlen != sizeof(u32))
3085 return -EINVAL;
3086 if (get_user(val, (int __user *)optval))
3087 return -EFAULT;
3089 /* Note: We double the receive buffer from what the user sets
3090 * it to be, also initial rwnd is based on rcvbuf/2.
3092 if (val > (sk->sk_rcvbuf >> 1))
3093 return -EINVAL;
3095 sctp_sk(sk)->pd_point = val;
3097 return 0; /* is this the right error code? */
3101 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3103 * This option will allow a user to change the maximum burst of packets
3104 * that can be emitted by this association. Note that the default value
3105 * is 4, and some implementations may restrict this setting so that it
3106 * can only be lowered.
3108 * NOTE: This text doesn't seem right. Do this on a socket basis with
3109 * future associations inheriting the socket value.
3111 static int sctp_setsockopt_maxburst(struct sock *sk,
3112 char __user *optval,
3113 unsigned int optlen)
3115 struct sctp_assoc_value params;
3116 struct sctp_sock *sp;
3117 struct sctp_association *asoc;
3118 int val;
3119 int assoc_id = 0;
3121 if (optlen == sizeof(int)) {
3122 printk(KERN_WARNING
3123 "SCTP: Use of int in max_burst socket option deprecated\n");
3124 printk(KERN_WARNING
3125 "SCTP: Use struct sctp_assoc_value instead\n");
3126 if (copy_from_user(&val, optval, optlen))
3127 return -EFAULT;
3128 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3129 if (copy_from_user(&params, optval, optlen))
3130 return -EFAULT;
3131 val = params.assoc_value;
3132 assoc_id = params.assoc_id;
3133 } else
3134 return -EINVAL;
3136 sp = sctp_sk(sk);
3138 if (assoc_id != 0) {
3139 asoc = sctp_id2assoc(sk, assoc_id);
3140 if (!asoc)
3141 return -EINVAL;
3142 asoc->max_burst = val;
3143 } else
3144 sp->max_burst = val;
3146 return 0;
3150 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3152 * This set option adds a chunk type that the user is requesting to be
3153 * received only in an authenticated way. Changes to the list of chunks
3154 * will only effect future associations on the socket.
3156 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3157 char __user *optval,
3158 unsigned int optlen)
3160 struct sctp_authchunk val;
3162 if (!sctp_auth_enable)
3163 return -EACCES;
3165 if (optlen != sizeof(struct sctp_authchunk))
3166 return -EINVAL;
3167 if (copy_from_user(&val, optval, optlen))
3168 return -EFAULT;
3170 switch (val.sauth_chunk) {
3171 case SCTP_CID_INIT:
3172 case SCTP_CID_INIT_ACK:
3173 case SCTP_CID_SHUTDOWN_COMPLETE:
3174 case SCTP_CID_AUTH:
3175 return -EINVAL;
3178 /* add this chunk id to the endpoint */
3179 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3183 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3185 * This option gets or sets the list of HMAC algorithms that the local
3186 * endpoint requires the peer to use.
3188 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3189 char __user *optval,
3190 unsigned int optlen)
3192 struct sctp_hmacalgo *hmacs;
3193 u32 idents;
3194 int err;
3196 if (!sctp_auth_enable)
3197 return -EACCES;
3199 if (optlen < sizeof(struct sctp_hmacalgo))
3200 return -EINVAL;
3202 hmacs = kmalloc(optlen, GFP_KERNEL);
3203 if (!hmacs)
3204 return -ENOMEM;
3206 if (copy_from_user(hmacs, optval, optlen)) {
3207 err = -EFAULT;
3208 goto out;
3211 idents = hmacs->shmac_num_idents;
3212 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3213 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3214 err = -EINVAL;
3215 goto out;
3218 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3219 out:
3220 kfree(hmacs);
3221 return err;
3225 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3227 * This option will set a shared secret key which is used to build an
3228 * association shared key.
3230 static int sctp_setsockopt_auth_key(struct sock *sk,
3231 char __user *optval,
3232 unsigned int optlen)
3234 struct sctp_authkey *authkey;
3235 struct sctp_association *asoc;
3236 int ret;
3238 if (!sctp_auth_enable)
3239 return -EACCES;
3241 if (optlen <= sizeof(struct sctp_authkey))
3242 return -EINVAL;
3244 authkey = kmalloc(optlen, GFP_KERNEL);
3245 if (!authkey)
3246 return -ENOMEM;
3248 if (copy_from_user(authkey, optval, optlen)) {
3249 ret = -EFAULT;
3250 goto out;
3253 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3254 ret = -EINVAL;
3255 goto out;
3258 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3259 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3260 ret = -EINVAL;
3261 goto out;
3264 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3265 out:
3266 kfree(authkey);
3267 return ret;
3271 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3273 * This option will get or set the active shared key to be used to build
3274 * the association shared key.
3276 static int sctp_setsockopt_active_key(struct sock *sk,
3277 char __user *optval,
3278 unsigned int optlen)
3280 struct sctp_authkeyid val;
3281 struct sctp_association *asoc;
3283 if (!sctp_auth_enable)
3284 return -EACCES;
3286 if (optlen != sizeof(struct sctp_authkeyid))
3287 return -EINVAL;
3288 if (copy_from_user(&val, optval, optlen))
3289 return -EFAULT;
3291 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3292 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3293 return -EINVAL;
3295 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3296 val.scact_keynumber);
3300 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3302 * This set option will delete a shared secret key from use.
3304 static int sctp_setsockopt_del_key(struct sock *sk,
3305 char __user *optval,
3306 unsigned int optlen)
3308 struct sctp_authkeyid val;
3309 struct sctp_association *asoc;
3311 if (!sctp_auth_enable)
3312 return -EACCES;
3314 if (optlen != sizeof(struct sctp_authkeyid))
3315 return -EINVAL;
3316 if (copy_from_user(&val, optval, optlen))
3317 return -EFAULT;
3319 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3320 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3321 return -EINVAL;
3323 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3324 val.scact_keynumber);
3329 /* API 6.2 setsockopt(), getsockopt()
3331 * Applications use setsockopt() and getsockopt() to set or retrieve
3332 * socket options. Socket options are used to change the default
3333 * behavior of sockets calls. They are described in Section 7.
3335 * The syntax is:
3337 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3338 * int __user *optlen);
3339 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3340 * int optlen);
3342 * sd - the socket descript.
3343 * level - set to IPPROTO_SCTP for all SCTP options.
3344 * optname - the option name.
3345 * optval - the buffer to store the value of the option.
3346 * optlen - the size of the buffer.
3348 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3349 char __user *optval, unsigned int optlen)
3351 int retval = 0;
3353 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3354 sk, optname);
3356 /* I can hardly begin to describe how wrong this is. This is
3357 * so broken as to be worse than useless. The API draft
3358 * REALLY is NOT helpful here... I am not convinced that the
3359 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3360 * are at all well-founded.
3362 if (level != SOL_SCTP) {
3363 struct sctp_af *af = sctp_sk(sk)->pf->af;
3364 retval = af->setsockopt(sk, level, optname, optval, optlen);
3365 goto out_nounlock;
3368 sctp_lock_sock(sk);
3370 switch (optname) {
3371 case SCTP_SOCKOPT_BINDX_ADD:
3372 /* 'optlen' is the size of the addresses buffer. */
3373 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3374 optlen, SCTP_BINDX_ADD_ADDR);
3375 break;
3377 case SCTP_SOCKOPT_BINDX_REM:
3378 /* 'optlen' is the size of the addresses buffer. */
3379 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3380 optlen, SCTP_BINDX_REM_ADDR);
3381 break;
3383 case SCTP_SOCKOPT_CONNECTX_OLD:
3384 /* 'optlen' is the size of the addresses buffer. */
3385 retval = sctp_setsockopt_connectx_old(sk,
3386 (struct sockaddr __user *)optval,
3387 optlen);
3388 break;
3390 case SCTP_SOCKOPT_CONNECTX:
3391 /* 'optlen' is the size of the addresses buffer. */
3392 retval = sctp_setsockopt_connectx(sk,
3393 (struct sockaddr __user *)optval,
3394 optlen);
3395 break;
3397 case SCTP_DISABLE_FRAGMENTS:
3398 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3399 break;
3401 case SCTP_EVENTS:
3402 retval = sctp_setsockopt_events(sk, optval, optlen);
3403 break;
3405 case SCTP_AUTOCLOSE:
3406 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3407 break;
3409 case SCTP_PEER_ADDR_PARAMS:
3410 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3411 break;
3413 case SCTP_DELAYED_ACK:
3414 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3415 break;
3416 case SCTP_PARTIAL_DELIVERY_POINT:
3417 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3418 break;
3420 case SCTP_INITMSG:
3421 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3422 break;
3423 case SCTP_DEFAULT_SEND_PARAM:
3424 retval = sctp_setsockopt_default_send_param(sk, optval,
3425 optlen);
3426 break;
3427 case SCTP_PRIMARY_ADDR:
3428 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3429 break;
3430 case SCTP_SET_PEER_PRIMARY_ADDR:
3431 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3432 break;
3433 case SCTP_NODELAY:
3434 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3435 break;
3436 case SCTP_RTOINFO:
3437 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3438 break;
3439 case SCTP_ASSOCINFO:
3440 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3441 break;
3442 case SCTP_I_WANT_MAPPED_V4_ADDR:
3443 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3444 break;
3445 case SCTP_MAXSEG:
3446 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3447 break;
3448 case SCTP_ADAPTATION_LAYER:
3449 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3450 break;
3451 case SCTP_CONTEXT:
3452 retval = sctp_setsockopt_context(sk, optval, optlen);
3453 break;
3454 case SCTP_FRAGMENT_INTERLEAVE:
3455 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3456 break;
3457 case SCTP_MAX_BURST:
3458 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3459 break;
3460 case SCTP_AUTH_CHUNK:
3461 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3462 break;
3463 case SCTP_HMAC_IDENT:
3464 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3465 break;
3466 case SCTP_AUTH_KEY:
3467 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3468 break;
3469 case SCTP_AUTH_ACTIVE_KEY:
3470 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3471 break;
3472 case SCTP_AUTH_DELETE_KEY:
3473 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3474 break;
3475 default:
3476 retval = -ENOPROTOOPT;
3477 break;
3480 sctp_release_sock(sk);
3482 out_nounlock:
3483 return retval;
3486 /* API 3.1.6 connect() - UDP Style Syntax
3488 * An application may use the connect() call in the UDP model to initiate an
3489 * association without sending data.
3491 * The syntax is:
3493 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3495 * sd: the socket descriptor to have a new association added to.
3497 * nam: the address structure (either struct sockaddr_in or struct
3498 * sockaddr_in6 defined in RFC2553 [7]).
3500 * len: the size of the address.
3502 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3503 int addr_len)
3505 int err = 0;
3506 struct sctp_af *af;
3508 sctp_lock_sock(sk);
3510 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3511 __func__, sk, addr, addr_len);
3513 /* Validate addr_len before calling common connect/connectx routine. */
3514 af = sctp_get_af_specific(addr->sa_family);
3515 if (!af || addr_len < af->sockaddr_len) {
3516 err = -EINVAL;
3517 } else {
3518 /* Pass correct addr len to common routine (so it knows there
3519 * is only one address being passed.
3521 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3524 sctp_release_sock(sk);
3525 return err;
3528 /* FIXME: Write comments. */
3529 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3531 return -EOPNOTSUPP; /* STUB */
3534 /* 4.1.4 accept() - TCP Style Syntax
3536 * Applications use accept() call to remove an established SCTP
3537 * association from the accept queue of the endpoint. A new socket
3538 * descriptor will be returned from accept() to represent the newly
3539 * formed association.
3541 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3543 struct sctp_sock *sp;
3544 struct sctp_endpoint *ep;
3545 struct sock *newsk = NULL;
3546 struct sctp_association *asoc;
3547 long timeo;
3548 int error = 0;
3550 sctp_lock_sock(sk);
3552 sp = sctp_sk(sk);
3553 ep = sp->ep;
3555 if (!sctp_style(sk, TCP)) {
3556 error = -EOPNOTSUPP;
3557 goto out;
3560 if (!sctp_sstate(sk, LISTENING)) {
3561 error = -EINVAL;
3562 goto out;
3565 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3567 error = sctp_wait_for_accept(sk, timeo);
3568 if (error)
3569 goto out;
3571 /* We treat the list of associations on the endpoint as the accept
3572 * queue and pick the first association on the list.
3574 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3576 newsk = sp->pf->create_accept_sk(sk, asoc);
3577 if (!newsk) {
3578 error = -ENOMEM;
3579 goto out;
3582 /* Populate the fields of the newsk from the oldsk and migrate the
3583 * asoc to the newsk.
3585 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3587 out:
3588 sctp_release_sock(sk);
3589 *err = error;
3590 return newsk;
3593 /* The SCTP ioctl handler. */
3594 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3596 return -ENOIOCTLCMD;
3599 /* This is the function which gets called during socket creation to
3600 * initialized the SCTP-specific portion of the sock.
3601 * The sock structure should already be zero-filled memory.
3603 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3605 struct sctp_endpoint *ep;
3606 struct sctp_sock *sp;
3608 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3610 sp = sctp_sk(sk);
3612 /* Initialize the SCTP per socket area. */
3613 switch (sk->sk_type) {
3614 case SOCK_SEQPACKET:
3615 sp->type = SCTP_SOCKET_UDP;
3616 break;
3617 case SOCK_STREAM:
3618 sp->type = SCTP_SOCKET_TCP;
3619 break;
3620 default:
3621 return -ESOCKTNOSUPPORT;
3624 /* Initialize default send parameters. These parameters can be
3625 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3627 sp->default_stream = 0;
3628 sp->default_ppid = 0;
3629 sp->default_flags = 0;
3630 sp->default_context = 0;
3631 sp->default_timetolive = 0;
3633 sp->default_rcv_context = 0;
3634 sp->max_burst = sctp_max_burst;
3636 /* Initialize default setup parameters. These parameters
3637 * can be modified with the SCTP_INITMSG socket option or
3638 * overridden by the SCTP_INIT CMSG.
3640 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3641 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3642 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3643 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3645 /* Initialize default RTO related parameters. These parameters can
3646 * be modified for with the SCTP_RTOINFO socket option.
3648 sp->rtoinfo.srto_initial = sctp_rto_initial;
3649 sp->rtoinfo.srto_max = sctp_rto_max;
3650 sp->rtoinfo.srto_min = sctp_rto_min;
3652 /* Initialize default association related parameters. These parameters
3653 * can be modified with the SCTP_ASSOCINFO socket option.
3655 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3656 sp->assocparams.sasoc_number_peer_destinations = 0;
3657 sp->assocparams.sasoc_peer_rwnd = 0;
3658 sp->assocparams.sasoc_local_rwnd = 0;
3659 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3661 /* Initialize default event subscriptions. By default, all the
3662 * options are off.
3664 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3666 /* Default Peer Address Parameters. These defaults can
3667 * be modified via SCTP_PEER_ADDR_PARAMS
3669 sp->hbinterval = sctp_hb_interval;
3670 sp->pathmaxrxt = sctp_max_retrans_path;
3671 sp->pathmtu = 0; // allow default discovery
3672 sp->sackdelay = sctp_sack_timeout;
3673 sp->sackfreq = 2;
3674 sp->param_flags = SPP_HB_ENABLE |
3675 SPP_PMTUD_ENABLE |
3676 SPP_SACKDELAY_ENABLE;
3678 /* If enabled no SCTP message fragmentation will be performed.
3679 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3681 sp->disable_fragments = 0;
3683 /* Enable Nagle algorithm by default. */
3684 sp->nodelay = 0;
3686 /* Enable by default. */
3687 sp->v4mapped = 1;
3689 /* Auto-close idle associations after the configured
3690 * number of seconds. A value of 0 disables this
3691 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3692 * for UDP-style sockets only.
3694 sp->autoclose = 0;
3696 /* User specified fragmentation limit. */
3697 sp->user_frag = 0;
3699 sp->adaptation_ind = 0;
3701 sp->pf = sctp_get_pf_specific(sk->sk_family);
3703 /* Control variables for partial data delivery. */
3704 atomic_set(&sp->pd_mode, 0);
3705 skb_queue_head_init(&sp->pd_lobby);
3706 sp->frag_interleave = 0;
3708 /* Create a per socket endpoint structure. Even if we
3709 * change the data structure relationships, this may still
3710 * be useful for storing pre-connect address information.
3712 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3713 if (!ep)
3714 return -ENOMEM;
3716 sp->ep = ep;
3717 sp->hmac = NULL;
3719 SCTP_DBG_OBJCNT_INC(sock);
3720 percpu_counter_inc(&sctp_sockets_allocated);
3722 local_bh_disable();
3723 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3724 local_bh_enable();
3726 return 0;
3729 /* Cleanup any SCTP per socket resources. */
3730 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3732 struct sctp_endpoint *ep;
3734 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3736 /* Release our hold on the endpoint. */
3737 ep = sctp_sk(sk)->ep;
3738 sctp_endpoint_free(ep);
3739 percpu_counter_dec(&sctp_sockets_allocated);
3740 local_bh_disable();
3741 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3742 local_bh_enable();
3745 /* API 4.1.7 shutdown() - TCP Style Syntax
3746 * int shutdown(int socket, int how);
3748 * sd - the socket descriptor of the association to be closed.
3749 * how - Specifies the type of shutdown. The values are
3750 * as follows:
3751 * SHUT_RD
3752 * Disables further receive operations. No SCTP
3753 * protocol action is taken.
3754 * SHUT_WR
3755 * Disables further send operations, and initiates
3756 * the SCTP shutdown sequence.
3757 * SHUT_RDWR
3758 * Disables further send and receive operations
3759 * and initiates the SCTP shutdown sequence.
3761 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3763 struct sctp_endpoint *ep;
3764 struct sctp_association *asoc;
3766 if (!sctp_style(sk, TCP))
3767 return;
3769 if (how & SEND_SHUTDOWN) {
3770 ep = sctp_sk(sk)->ep;
3771 if (!list_empty(&ep->asocs)) {
3772 asoc = list_entry(ep->asocs.next,
3773 struct sctp_association, asocs);
3774 sctp_primitive_SHUTDOWN(asoc, NULL);
3779 /* 7.2.1 Association Status (SCTP_STATUS)
3781 * Applications can retrieve current status information about an
3782 * association, including association state, peer receiver window size,
3783 * number of unacked data chunks, and number of data chunks pending
3784 * receipt. This information is read-only.
3786 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3787 char __user *optval,
3788 int __user *optlen)
3790 struct sctp_status status;
3791 struct sctp_association *asoc = NULL;
3792 struct sctp_transport *transport;
3793 sctp_assoc_t associd;
3794 int retval = 0;
3796 if (len < sizeof(status)) {
3797 retval = -EINVAL;
3798 goto out;
3801 len = sizeof(status);
3802 if (copy_from_user(&status, optval, len)) {
3803 retval = -EFAULT;
3804 goto out;
3807 associd = status.sstat_assoc_id;
3808 asoc = sctp_id2assoc(sk, associd);
3809 if (!asoc) {
3810 retval = -EINVAL;
3811 goto out;
3814 transport = asoc->peer.primary_path;
3816 status.sstat_assoc_id = sctp_assoc2id(asoc);
3817 status.sstat_state = asoc->state;
3818 status.sstat_rwnd = asoc->peer.rwnd;
3819 status.sstat_unackdata = asoc->unack_data;
3821 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3822 status.sstat_instrms = asoc->c.sinit_max_instreams;
3823 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3824 status.sstat_fragmentation_point = asoc->frag_point;
3825 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3826 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3827 transport->af_specific->sockaddr_len);
3828 /* Map ipv4 address into v4-mapped-on-v6 address. */
3829 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3830 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3831 status.sstat_primary.spinfo_state = transport->state;
3832 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3833 status.sstat_primary.spinfo_srtt = transport->srtt;
3834 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3835 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3837 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3838 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3840 if (put_user(len, optlen)) {
3841 retval = -EFAULT;
3842 goto out;
3845 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3846 len, status.sstat_state, status.sstat_rwnd,
3847 status.sstat_assoc_id);
3849 if (copy_to_user(optval, &status, len)) {
3850 retval = -EFAULT;
3851 goto out;
3854 out:
3855 return (retval);
3859 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3861 * Applications can retrieve information about a specific peer address
3862 * of an association, including its reachability state, congestion
3863 * window, and retransmission timer values. This information is
3864 * read-only.
3866 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3867 char __user *optval,
3868 int __user *optlen)
3870 struct sctp_paddrinfo pinfo;
3871 struct sctp_transport *transport;
3872 int retval = 0;
3874 if (len < sizeof(pinfo)) {
3875 retval = -EINVAL;
3876 goto out;
3879 len = sizeof(pinfo);
3880 if (copy_from_user(&pinfo, optval, len)) {
3881 retval = -EFAULT;
3882 goto out;
3885 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3886 pinfo.spinfo_assoc_id);
3887 if (!transport)
3888 return -EINVAL;
3890 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3891 pinfo.spinfo_state = transport->state;
3892 pinfo.spinfo_cwnd = transport->cwnd;
3893 pinfo.spinfo_srtt = transport->srtt;
3894 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3895 pinfo.spinfo_mtu = transport->pathmtu;
3897 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3898 pinfo.spinfo_state = SCTP_ACTIVE;
3900 if (put_user(len, optlen)) {
3901 retval = -EFAULT;
3902 goto out;
3905 if (copy_to_user(optval, &pinfo, len)) {
3906 retval = -EFAULT;
3907 goto out;
3910 out:
3911 return (retval);
3914 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3916 * This option is a on/off flag. If enabled no SCTP message
3917 * fragmentation will be performed. Instead if a message being sent
3918 * exceeds the current PMTU size, the message will NOT be sent and
3919 * instead a error will be indicated to the user.
3921 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3922 char __user *optval, int __user *optlen)
3924 int val;
3926 if (len < sizeof(int))
3927 return -EINVAL;
3929 len = sizeof(int);
3930 val = (sctp_sk(sk)->disable_fragments == 1);
3931 if (put_user(len, optlen))
3932 return -EFAULT;
3933 if (copy_to_user(optval, &val, len))
3934 return -EFAULT;
3935 return 0;
3938 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3940 * This socket option is used to specify various notifications and
3941 * ancillary data the user wishes to receive.
3943 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3944 int __user *optlen)
3946 if (len < sizeof(struct sctp_event_subscribe))
3947 return -EINVAL;
3948 len = sizeof(struct sctp_event_subscribe);
3949 if (put_user(len, optlen))
3950 return -EFAULT;
3951 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3952 return -EFAULT;
3953 return 0;
3956 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3958 * This socket option is applicable to the UDP-style socket only. When
3959 * set it will cause associations that are idle for more than the
3960 * specified number of seconds to automatically close. An association
3961 * being idle is defined an association that has NOT sent or received
3962 * user data. The special value of '0' indicates that no automatic
3963 * close of any associations should be performed. The option expects an
3964 * integer defining the number of seconds of idle time before an
3965 * association is closed.
3967 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3969 /* Applicable to UDP-style socket only */
3970 if (sctp_style(sk, TCP))
3971 return -EOPNOTSUPP;
3972 if (len < sizeof(int))
3973 return -EINVAL;
3974 len = sizeof(int);
3975 if (put_user(len, optlen))
3976 return -EFAULT;
3977 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3978 return -EFAULT;
3979 return 0;
3982 /* Helper routine to branch off an association to a new socket. */
3983 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3984 struct socket **sockp)
3986 struct sock *sk = asoc->base.sk;
3987 struct socket *sock;
3988 struct sctp_af *af;
3989 int err = 0;
3991 /* An association cannot be branched off from an already peeled-off
3992 * socket, nor is this supported for tcp style sockets.
3994 if (!sctp_style(sk, UDP))
3995 return -EINVAL;
3997 /* Create a new socket. */
3998 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3999 if (err < 0)
4000 return err;
4002 sctp_copy_sock(sock->sk, sk, asoc);
4004 /* Make peeled-off sockets more like 1-1 accepted sockets.
4005 * Set the daddr and initialize id to something more random
4007 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4008 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4010 /* Populate the fields of the newsk from the oldsk and migrate the
4011 * asoc to the newsk.
4013 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4015 *sockp = sock;
4017 return err;
4020 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4022 sctp_peeloff_arg_t peeloff;
4023 struct socket *newsock;
4024 int retval = 0;
4025 struct sctp_association *asoc;
4027 if (len < sizeof(sctp_peeloff_arg_t))
4028 return -EINVAL;
4029 len = sizeof(sctp_peeloff_arg_t);
4030 if (copy_from_user(&peeloff, optval, len))
4031 return -EFAULT;
4033 asoc = sctp_id2assoc(sk, peeloff.associd);
4034 if (!asoc) {
4035 retval = -EINVAL;
4036 goto out;
4039 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4041 retval = sctp_do_peeloff(asoc, &newsock);
4042 if (retval < 0)
4043 goto out;
4045 /* Map the socket to an unused fd that can be returned to the user. */
4046 retval = sock_map_fd(newsock, 0);
4047 if (retval < 0) {
4048 sock_release(newsock);
4049 goto out;
4052 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4053 __func__, sk, asoc, newsock->sk, retval);
4055 /* Return the fd mapped to the new socket. */
4056 peeloff.sd = retval;
4057 if (put_user(len, optlen))
4058 return -EFAULT;
4059 if (copy_to_user(optval, &peeloff, len))
4060 retval = -EFAULT;
4062 out:
4063 return retval;
4066 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4068 * Applications can enable or disable heartbeats for any peer address of
4069 * an association, modify an address's heartbeat interval, force a
4070 * heartbeat to be sent immediately, and adjust the address's maximum
4071 * number of retransmissions sent before an address is considered
4072 * unreachable. The following structure is used to access and modify an
4073 * address's parameters:
4075 * struct sctp_paddrparams {
4076 * sctp_assoc_t spp_assoc_id;
4077 * struct sockaddr_storage spp_address;
4078 * uint32_t spp_hbinterval;
4079 * uint16_t spp_pathmaxrxt;
4080 * uint32_t spp_pathmtu;
4081 * uint32_t spp_sackdelay;
4082 * uint32_t spp_flags;
4083 * };
4085 * spp_assoc_id - (one-to-many style socket) This is filled in the
4086 * application, and identifies the association for
4087 * this query.
4088 * spp_address - This specifies which address is of interest.
4089 * spp_hbinterval - This contains the value of the heartbeat interval,
4090 * in milliseconds. If a value of zero
4091 * is present in this field then no changes are to
4092 * be made to this parameter.
4093 * spp_pathmaxrxt - This contains the maximum number of
4094 * retransmissions before this address shall be
4095 * considered unreachable. If a value of zero
4096 * is present in this field then no changes are to
4097 * be made to this parameter.
4098 * spp_pathmtu - When Path MTU discovery is disabled the value
4099 * specified here will be the "fixed" path mtu.
4100 * Note that if the spp_address field is empty
4101 * then all associations on this address will
4102 * have this fixed path mtu set upon them.
4104 * spp_sackdelay - When delayed sack is enabled, this value specifies
4105 * the number of milliseconds that sacks will be delayed
4106 * for. This value will apply to all addresses of an
4107 * association if the spp_address field is empty. Note
4108 * also, that if delayed sack is enabled and this
4109 * value is set to 0, no change is made to the last
4110 * recorded delayed sack timer value.
4112 * spp_flags - These flags are used to control various features
4113 * on an association. The flag field may contain
4114 * zero or more of the following options.
4116 * SPP_HB_ENABLE - Enable heartbeats on the
4117 * specified address. Note that if the address
4118 * field is empty all addresses for the association
4119 * have heartbeats enabled upon them.
4121 * SPP_HB_DISABLE - Disable heartbeats on the
4122 * speicifed address. Note that if the address
4123 * field is empty all addresses for the association
4124 * will have their heartbeats disabled. Note also
4125 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4126 * mutually exclusive, only one of these two should
4127 * be specified. Enabling both fields will have
4128 * undetermined results.
4130 * SPP_HB_DEMAND - Request a user initiated heartbeat
4131 * to be made immediately.
4133 * SPP_PMTUD_ENABLE - This field will enable PMTU
4134 * discovery upon the specified address. Note that
4135 * if the address feild is empty then all addresses
4136 * on the association are effected.
4138 * SPP_PMTUD_DISABLE - This field will disable PMTU
4139 * discovery upon the specified address. Note that
4140 * if the address feild is empty then all addresses
4141 * on the association are effected. Not also that
4142 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4143 * exclusive. Enabling both will have undetermined
4144 * results.
4146 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4147 * on delayed sack. The time specified in spp_sackdelay
4148 * is used to specify the sack delay for this address. Note
4149 * that if spp_address is empty then all addresses will
4150 * enable delayed sack and take on the sack delay
4151 * value specified in spp_sackdelay.
4152 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4153 * off delayed sack. If the spp_address field is blank then
4154 * delayed sack is disabled for the entire association. Note
4155 * also that this field is mutually exclusive to
4156 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4157 * results.
4159 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4160 char __user *optval, int __user *optlen)
4162 struct sctp_paddrparams params;
4163 struct sctp_transport *trans = NULL;
4164 struct sctp_association *asoc = NULL;
4165 struct sctp_sock *sp = sctp_sk(sk);
4167 if (len < sizeof(struct sctp_paddrparams))
4168 return -EINVAL;
4169 len = sizeof(struct sctp_paddrparams);
4170 if (copy_from_user(&params, optval, len))
4171 return -EFAULT;
4173 /* If an address other than INADDR_ANY is specified, and
4174 * no transport is found, then the request is invalid.
4176 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4177 trans = sctp_addr_id2transport(sk, &params.spp_address,
4178 params.spp_assoc_id);
4179 if (!trans) {
4180 SCTP_DEBUG_PRINTK("Failed no transport\n");
4181 return -EINVAL;
4185 /* Get association, if assoc_id != 0 and the socket is a one
4186 * to many style socket, and an association was not found, then
4187 * the id was invalid.
4189 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4190 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4191 SCTP_DEBUG_PRINTK("Failed no association\n");
4192 return -EINVAL;
4195 if (trans) {
4196 /* Fetch transport values. */
4197 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4198 params.spp_pathmtu = trans->pathmtu;
4199 params.spp_pathmaxrxt = trans->pathmaxrxt;
4200 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4202 /*draft-11 doesn't say what to return in spp_flags*/
4203 params.spp_flags = trans->param_flags;
4204 } else if (asoc) {
4205 /* Fetch association values. */
4206 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4207 params.spp_pathmtu = asoc->pathmtu;
4208 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4209 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4211 /*draft-11 doesn't say what to return in spp_flags*/
4212 params.spp_flags = asoc->param_flags;
4213 } else {
4214 /* Fetch socket values. */
4215 params.spp_hbinterval = sp->hbinterval;
4216 params.spp_pathmtu = sp->pathmtu;
4217 params.spp_sackdelay = sp->sackdelay;
4218 params.spp_pathmaxrxt = sp->pathmaxrxt;
4220 /*draft-11 doesn't say what to return in spp_flags*/
4221 params.spp_flags = sp->param_flags;
4224 if (copy_to_user(optval, &params, len))
4225 return -EFAULT;
4227 if (put_user(len, optlen))
4228 return -EFAULT;
4230 return 0;
4234 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4236 * This option will effect the way delayed acks are performed. This
4237 * option allows you to get or set the delayed ack time, in
4238 * milliseconds. It also allows changing the delayed ack frequency.
4239 * Changing the frequency to 1 disables the delayed sack algorithm. If
4240 * the assoc_id is 0, then this sets or gets the endpoints default
4241 * values. If the assoc_id field is non-zero, then the set or get
4242 * effects the specified association for the one to many model (the
4243 * assoc_id field is ignored by the one to one model). Note that if
4244 * sack_delay or sack_freq are 0 when setting this option, then the
4245 * current values will remain unchanged.
4247 * struct sctp_sack_info {
4248 * sctp_assoc_t sack_assoc_id;
4249 * uint32_t sack_delay;
4250 * uint32_t sack_freq;
4251 * };
4253 * sack_assoc_id - This parameter, indicates which association the user
4254 * is performing an action upon. Note that if this field's value is
4255 * zero then the endpoints default value is changed (effecting future
4256 * associations only).
4258 * sack_delay - This parameter contains the number of milliseconds that
4259 * the user is requesting the delayed ACK timer be set to. Note that
4260 * this value is defined in the standard to be between 200 and 500
4261 * milliseconds.
4263 * sack_freq - This parameter contains the number of packets that must
4264 * be received before a sack is sent without waiting for the delay
4265 * timer to expire. The default value for this is 2, setting this
4266 * value to 1 will disable the delayed sack algorithm.
4268 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4269 char __user *optval,
4270 int __user *optlen)
4272 struct sctp_sack_info params;
4273 struct sctp_association *asoc = NULL;
4274 struct sctp_sock *sp = sctp_sk(sk);
4276 if (len >= sizeof(struct sctp_sack_info)) {
4277 len = sizeof(struct sctp_sack_info);
4279 if (copy_from_user(&params, optval, len))
4280 return -EFAULT;
4281 } else if (len == sizeof(struct sctp_assoc_value)) {
4282 printk(KERN_WARNING "SCTP: Use of struct sctp_assoc_value "
4283 "in delayed_ack socket option deprecated\n");
4284 printk(KERN_WARNING "SCTP: Use struct sctp_sack_info instead\n");
4285 if (copy_from_user(&params, optval, len))
4286 return -EFAULT;
4287 } else
4288 return - EINVAL;
4290 /* Get association, if sack_assoc_id != 0 and the socket is a one
4291 * to many style socket, and an association was not found, then
4292 * the id was invalid.
4294 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4295 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4296 return -EINVAL;
4298 if (asoc) {
4299 /* Fetch association values. */
4300 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4301 params.sack_delay = jiffies_to_msecs(
4302 asoc->sackdelay);
4303 params.sack_freq = asoc->sackfreq;
4305 } else {
4306 params.sack_delay = 0;
4307 params.sack_freq = 1;
4309 } else {
4310 /* Fetch socket values. */
4311 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4312 params.sack_delay = sp->sackdelay;
4313 params.sack_freq = sp->sackfreq;
4314 } else {
4315 params.sack_delay = 0;
4316 params.sack_freq = 1;
4320 if (copy_to_user(optval, &params, len))
4321 return -EFAULT;
4323 if (put_user(len, optlen))
4324 return -EFAULT;
4326 return 0;
4329 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4331 * Applications can specify protocol parameters for the default association
4332 * initialization. The option name argument to setsockopt() and getsockopt()
4333 * is SCTP_INITMSG.
4335 * Setting initialization parameters is effective only on an unconnected
4336 * socket (for UDP-style sockets only future associations are effected
4337 * by the change). With TCP-style sockets, this option is inherited by
4338 * sockets derived from a listener socket.
4340 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4342 if (len < sizeof(struct sctp_initmsg))
4343 return -EINVAL;
4344 len = sizeof(struct sctp_initmsg);
4345 if (put_user(len, optlen))
4346 return -EFAULT;
4347 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4348 return -EFAULT;
4349 return 0;
4352 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4353 char __user *optval,
4354 int __user *optlen)
4356 sctp_assoc_t id;
4357 struct sctp_association *asoc;
4358 struct list_head *pos;
4359 int cnt = 0;
4361 if (len < sizeof(sctp_assoc_t))
4362 return -EINVAL;
4364 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4365 return -EFAULT;
4367 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4368 "socket option deprecated\n");
4369 /* For UDP-style sockets, id specifies the association to query. */
4370 asoc = sctp_id2assoc(sk, id);
4371 if (!asoc)
4372 return -EINVAL;
4374 list_for_each(pos, &asoc->peer.transport_addr_list) {
4375 cnt ++;
4378 return cnt;
4382 * Old API for getting list of peer addresses. Does not work for 32-bit
4383 * programs running on a 64-bit kernel
4385 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4386 char __user *optval,
4387 int __user *optlen)
4389 struct sctp_association *asoc;
4390 int cnt = 0;
4391 struct sctp_getaddrs_old getaddrs;
4392 struct sctp_transport *from;
4393 void __user *to;
4394 union sctp_addr temp;
4395 struct sctp_sock *sp = sctp_sk(sk);
4396 int addrlen;
4398 if (len < sizeof(struct sctp_getaddrs_old))
4399 return -EINVAL;
4401 len = sizeof(struct sctp_getaddrs_old);
4403 if (copy_from_user(&getaddrs, optval, len))
4404 return -EFAULT;
4406 if (getaddrs.addr_num <= 0) return -EINVAL;
4408 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4409 "socket option deprecated\n");
4411 /* For UDP-style sockets, id specifies the association to query. */
4412 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4413 if (!asoc)
4414 return -EINVAL;
4416 to = (void __user *)getaddrs.addrs;
4417 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4418 transports) {
4419 memcpy(&temp, &from->ipaddr, sizeof(temp));
4420 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4421 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4422 if (copy_to_user(to, &temp, addrlen))
4423 return -EFAULT;
4424 to += addrlen ;
4425 cnt ++;
4426 if (cnt >= getaddrs.addr_num) break;
4428 getaddrs.addr_num = cnt;
4429 if (put_user(len, optlen))
4430 return -EFAULT;
4431 if (copy_to_user(optval, &getaddrs, len))
4432 return -EFAULT;
4434 return 0;
4437 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4438 char __user *optval, int __user *optlen)
4440 struct sctp_association *asoc;
4441 int cnt = 0;
4442 struct sctp_getaddrs getaddrs;
4443 struct sctp_transport *from;
4444 void __user *to;
4445 union sctp_addr temp;
4446 struct sctp_sock *sp = sctp_sk(sk);
4447 int addrlen;
4448 size_t space_left;
4449 int bytes_copied;
4451 if (len < sizeof(struct sctp_getaddrs))
4452 return -EINVAL;
4454 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4455 return -EFAULT;
4457 /* For UDP-style sockets, id specifies the association to query. */
4458 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4459 if (!asoc)
4460 return -EINVAL;
4462 to = optval + offsetof(struct sctp_getaddrs,addrs);
4463 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4465 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4466 transports) {
4467 memcpy(&temp, &from->ipaddr, sizeof(temp));
4468 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4469 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4470 if (space_left < addrlen)
4471 return -ENOMEM;
4472 if (copy_to_user(to, &temp, addrlen))
4473 return -EFAULT;
4474 to += addrlen;
4475 cnt++;
4476 space_left -= addrlen;
4479 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4480 return -EFAULT;
4481 bytes_copied = ((char __user *)to) - optval;
4482 if (put_user(bytes_copied, optlen))
4483 return -EFAULT;
4485 return 0;
4488 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4489 char __user *optval,
4490 int __user *optlen)
4492 sctp_assoc_t id;
4493 struct sctp_bind_addr *bp;
4494 struct sctp_association *asoc;
4495 struct sctp_sockaddr_entry *addr;
4496 int cnt = 0;
4498 if (len < sizeof(sctp_assoc_t))
4499 return -EINVAL;
4501 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4502 return -EFAULT;
4504 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4505 "socket option deprecated\n");
4508 * For UDP-style sockets, id specifies the association to query.
4509 * If the id field is set to the value '0' then the locally bound
4510 * addresses are returned without regard to any particular
4511 * association.
4513 if (0 == id) {
4514 bp = &sctp_sk(sk)->ep->base.bind_addr;
4515 } else {
4516 asoc = sctp_id2assoc(sk, id);
4517 if (!asoc)
4518 return -EINVAL;
4519 bp = &asoc->base.bind_addr;
4522 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4523 * addresses from the global local address list.
4525 if (sctp_list_single_entry(&bp->address_list)) {
4526 addr = list_entry(bp->address_list.next,
4527 struct sctp_sockaddr_entry, list);
4528 if (sctp_is_any(sk, &addr->a)) {
4529 rcu_read_lock();
4530 list_for_each_entry_rcu(addr,
4531 &sctp_local_addr_list, list) {
4532 if (!addr->valid)
4533 continue;
4535 if ((PF_INET == sk->sk_family) &&
4536 (AF_INET6 == addr->a.sa.sa_family))
4537 continue;
4539 if ((PF_INET6 == sk->sk_family) &&
4540 inet_v6_ipv6only(sk) &&
4541 (AF_INET == addr->a.sa.sa_family))
4542 continue;
4544 cnt++;
4546 rcu_read_unlock();
4547 } else {
4548 cnt = 1;
4550 goto done;
4553 /* Protection on the bound address list is not needed,
4554 * since in the socket option context we hold the socket lock,
4555 * so there is no way that the bound address list can change.
4557 list_for_each_entry(addr, &bp->address_list, list) {
4558 cnt ++;
4560 done:
4561 return cnt;
4564 /* Helper function that copies local addresses to user and returns the number
4565 * of addresses copied.
4567 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4568 int max_addrs, void *to,
4569 int *bytes_copied)
4571 struct sctp_sockaddr_entry *addr;
4572 union sctp_addr temp;
4573 int cnt = 0;
4574 int addrlen;
4576 rcu_read_lock();
4577 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4578 if (!addr->valid)
4579 continue;
4581 if ((PF_INET == sk->sk_family) &&
4582 (AF_INET6 == addr->a.sa.sa_family))
4583 continue;
4584 if ((PF_INET6 == sk->sk_family) &&
4585 inet_v6_ipv6only(sk) &&
4586 (AF_INET == addr->a.sa.sa_family))
4587 continue;
4588 memcpy(&temp, &addr->a, sizeof(temp));
4589 if (!temp.v4.sin_port)
4590 temp.v4.sin_port = htons(port);
4592 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4593 &temp);
4594 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4595 memcpy(to, &temp, addrlen);
4597 to += addrlen;
4598 *bytes_copied += addrlen;
4599 cnt ++;
4600 if (cnt >= max_addrs) break;
4602 rcu_read_unlock();
4604 return cnt;
4607 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4608 size_t space_left, int *bytes_copied)
4610 struct sctp_sockaddr_entry *addr;
4611 union sctp_addr temp;
4612 int cnt = 0;
4613 int addrlen;
4615 rcu_read_lock();
4616 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4617 if (!addr->valid)
4618 continue;
4620 if ((PF_INET == sk->sk_family) &&
4621 (AF_INET6 == addr->a.sa.sa_family))
4622 continue;
4623 if ((PF_INET6 == sk->sk_family) &&
4624 inet_v6_ipv6only(sk) &&
4625 (AF_INET == addr->a.sa.sa_family))
4626 continue;
4627 memcpy(&temp, &addr->a, sizeof(temp));
4628 if (!temp.v4.sin_port)
4629 temp.v4.sin_port = htons(port);
4631 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4632 &temp);
4633 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4634 if (space_left < addrlen) {
4635 cnt = -ENOMEM;
4636 break;
4638 memcpy(to, &temp, addrlen);
4640 to += addrlen;
4641 cnt ++;
4642 space_left -= addrlen;
4643 *bytes_copied += addrlen;
4645 rcu_read_unlock();
4647 return cnt;
4650 /* Old API for getting list of local addresses. Does not work for 32-bit
4651 * programs running on a 64-bit kernel
4653 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4654 char __user *optval, int __user *optlen)
4656 struct sctp_bind_addr *bp;
4657 struct sctp_association *asoc;
4658 int cnt = 0;
4659 struct sctp_getaddrs_old getaddrs;
4660 struct sctp_sockaddr_entry *addr;
4661 void __user *to;
4662 union sctp_addr temp;
4663 struct sctp_sock *sp = sctp_sk(sk);
4664 int addrlen;
4665 int err = 0;
4666 void *addrs;
4667 void *buf;
4668 int bytes_copied = 0;
4670 if (len < sizeof(struct sctp_getaddrs_old))
4671 return -EINVAL;
4673 len = sizeof(struct sctp_getaddrs_old);
4674 if (copy_from_user(&getaddrs, optval, len))
4675 return -EFAULT;
4677 if (getaddrs.addr_num <= 0 ||
4678 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4679 return -EINVAL;
4681 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4682 "socket option deprecated\n");
4685 * For UDP-style sockets, id specifies the association to query.
4686 * If the id field is set to the value '0' then the locally bound
4687 * addresses are returned without regard to any particular
4688 * association.
4690 if (0 == getaddrs.assoc_id) {
4691 bp = &sctp_sk(sk)->ep->base.bind_addr;
4692 } else {
4693 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4694 if (!asoc)
4695 return -EINVAL;
4696 bp = &asoc->base.bind_addr;
4699 to = getaddrs.addrs;
4701 /* Allocate space for a local instance of packed array to hold all
4702 * the data. We store addresses here first and then put write them
4703 * to the user in one shot.
4705 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4706 GFP_KERNEL);
4707 if (!addrs)
4708 return -ENOMEM;
4710 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4711 * addresses from the global local address list.
4713 if (sctp_list_single_entry(&bp->address_list)) {
4714 addr = list_entry(bp->address_list.next,
4715 struct sctp_sockaddr_entry, list);
4716 if (sctp_is_any(sk, &addr->a)) {
4717 cnt = sctp_copy_laddrs_old(sk, bp->port,
4718 getaddrs.addr_num,
4719 addrs, &bytes_copied);
4720 goto copy_getaddrs;
4724 buf = addrs;
4725 /* Protection on the bound address list is not needed since
4726 * in the socket option context we hold a socket lock and
4727 * thus the bound address list can't change.
4729 list_for_each_entry(addr, &bp->address_list, list) {
4730 memcpy(&temp, &addr->a, sizeof(temp));
4731 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4732 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4733 memcpy(buf, &temp, addrlen);
4734 buf += addrlen;
4735 bytes_copied += addrlen;
4736 cnt ++;
4737 if (cnt >= getaddrs.addr_num) break;
4740 copy_getaddrs:
4741 /* copy the entire address list into the user provided space */
4742 if (copy_to_user(to, addrs, bytes_copied)) {
4743 err = -EFAULT;
4744 goto error;
4747 /* copy the leading structure back to user */
4748 getaddrs.addr_num = cnt;
4749 if (copy_to_user(optval, &getaddrs, len))
4750 err = -EFAULT;
4752 error:
4753 kfree(addrs);
4754 return err;
4757 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4758 char __user *optval, int __user *optlen)
4760 struct sctp_bind_addr *bp;
4761 struct sctp_association *asoc;
4762 int cnt = 0;
4763 struct sctp_getaddrs getaddrs;
4764 struct sctp_sockaddr_entry *addr;
4765 void __user *to;
4766 union sctp_addr temp;
4767 struct sctp_sock *sp = sctp_sk(sk);
4768 int addrlen;
4769 int err = 0;
4770 size_t space_left;
4771 int bytes_copied = 0;
4772 void *addrs;
4773 void *buf;
4775 if (len < sizeof(struct sctp_getaddrs))
4776 return -EINVAL;
4778 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4779 return -EFAULT;
4782 * For UDP-style sockets, id specifies the association to query.
4783 * If the id field is set to the value '0' then the locally bound
4784 * addresses are returned without regard to any particular
4785 * association.
4787 if (0 == getaddrs.assoc_id) {
4788 bp = &sctp_sk(sk)->ep->base.bind_addr;
4789 } else {
4790 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4791 if (!asoc)
4792 return -EINVAL;
4793 bp = &asoc->base.bind_addr;
4796 to = optval + offsetof(struct sctp_getaddrs,addrs);
4797 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4799 addrs = kmalloc(space_left, GFP_KERNEL);
4800 if (!addrs)
4801 return -ENOMEM;
4803 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4804 * addresses from the global local address list.
4806 if (sctp_list_single_entry(&bp->address_list)) {
4807 addr = list_entry(bp->address_list.next,
4808 struct sctp_sockaddr_entry, list);
4809 if (sctp_is_any(sk, &addr->a)) {
4810 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4811 space_left, &bytes_copied);
4812 if (cnt < 0) {
4813 err = cnt;
4814 goto out;
4816 goto copy_getaddrs;
4820 buf = addrs;
4821 /* Protection on the bound address list is not needed since
4822 * in the socket option context we hold a socket lock and
4823 * thus the bound address list can't change.
4825 list_for_each_entry(addr, &bp->address_list, list) {
4826 memcpy(&temp, &addr->a, sizeof(temp));
4827 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4828 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4829 if (space_left < addrlen) {
4830 err = -ENOMEM; /*fixme: right error?*/
4831 goto out;
4833 memcpy(buf, &temp, addrlen);
4834 buf += addrlen;
4835 bytes_copied += addrlen;
4836 cnt ++;
4837 space_left -= addrlen;
4840 copy_getaddrs:
4841 if (copy_to_user(to, addrs, bytes_copied)) {
4842 err = -EFAULT;
4843 goto out;
4845 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4846 err = -EFAULT;
4847 goto out;
4849 if (put_user(bytes_copied, optlen))
4850 err = -EFAULT;
4851 out:
4852 kfree(addrs);
4853 return err;
4856 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4858 * Requests that the local SCTP stack use the enclosed peer address as
4859 * the association primary. The enclosed address must be one of the
4860 * association peer's addresses.
4862 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4863 char __user *optval, int __user *optlen)
4865 struct sctp_prim prim;
4866 struct sctp_association *asoc;
4867 struct sctp_sock *sp = sctp_sk(sk);
4869 if (len < sizeof(struct sctp_prim))
4870 return -EINVAL;
4872 len = sizeof(struct sctp_prim);
4874 if (copy_from_user(&prim, optval, len))
4875 return -EFAULT;
4877 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4878 if (!asoc)
4879 return -EINVAL;
4881 if (!asoc->peer.primary_path)
4882 return -ENOTCONN;
4884 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4885 asoc->peer.primary_path->af_specific->sockaddr_len);
4887 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4888 (union sctp_addr *)&prim.ssp_addr);
4890 if (put_user(len, optlen))
4891 return -EFAULT;
4892 if (copy_to_user(optval, &prim, len))
4893 return -EFAULT;
4895 return 0;
4899 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4901 * Requests that the local endpoint set the specified Adaptation Layer
4902 * Indication parameter for all future INIT and INIT-ACK exchanges.
4904 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4905 char __user *optval, int __user *optlen)
4907 struct sctp_setadaptation adaptation;
4909 if (len < sizeof(struct sctp_setadaptation))
4910 return -EINVAL;
4912 len = sizeof(struct sctp_setadaptation);
4914 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4916 if (put_user(len, optlen))
4917 return -EFAULT;
4918 if (copy_to_user(optval, &adaptation, len))
4919 return -EFAULT;
4921 return 0;
4926 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4928 * Applications that wish to use the sendto() system call may wish to
4929 * specify a default set of parameters that would normally be supplied
4930 * through the inclusion of ancillary data. This socket option allows
4931 * such an application to set the default sctp_sndrcvinfo structure.
4934 * The application that wishes to use this socket option simply passes
4935 * in to this call the sctp_sndrcvinfo structure defined in Section
4936 * 5.2.2) The input parameters accepted by this call include
4937 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4938 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4939 * to this call if the caller is using the UDP model.
4941 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4943 static int sctp_getsockopt_default_send_param(struct sock *sk,
4944 int len, char __user *optval,
4945 int __user *optlen)
4947 struct sctp_sndrcvinfo info;
4948 struct sctp_association *asoc;
4949 struct sctp_sock *sp = sctp_sk(sk);
4951 if (len < sizeof(struct sctp_sndrcvinfo))
4952 return -EINVAL;
4954 len = sizeof(struct sctp_sndrcvinfo);
4956 if (copy_from_user(&info, optval, len))
4957 return -EFAULT;
4959 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4960 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4961 return -EINVAL;
4963 if (asoc) {
4964 info.sinfo_stream = asoc->default_stream;
4965 info.sinfo_flags = asoc->default_flags;
4966 info.sinfo_ppid = asoc->default_ppid;
4967 info.sinfo_context = asoc->default_context;
4968 info.sinfo_timetolive = asoc->default_timetolive;
4969 } else {
4970 info.sinfo_stream = sp->default_stream;
4971 info.sinfo_flags = sp->default_flags;
4972 info.sinfo_ppid = sp->default_ppid;
4973 info.sinfo_context = sp->default_context;
4974 info.sinfo_timetolive = sp->default_timetolive;
4977 if (put_user(len, optlen))
4978 return -EFAULT;
4979 if (copy_to_user(optval, &info, len))
4980 return -EFAULT;
4982 return 0;
4987 * 7.1.5 SCTP_NODELAY
4989 * Turn on/off any Nagle-like algorithm. This means that packets are
4990 * generally sent as soon as possible and no unnecessary delays are
4991 * introduced, at the cost of more packets in the network. Expects an
4992 * integer boolean flag.
4995 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4996 char __user *optval, int __user *optlen)
4998 int val;
5000 if (len < sizeof(int))
5001 return -EINVAL;
5003 len = sizeof(int);
5004 val = (sctp_sk(sk)->nodelay == 1);
5005 if (put_user(len, optlen))
5006 return -EFAULT;
5007 if (copy_to_user(optval, &val, len))
5008 return -EFAULT;
5009 return 0;
5014 * 7.1.1 SCTP_RTOINFO
5016 * The protocol parameters used to initialize and bound retransmission
5017 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5018 * and modify these parameters.
5019 * All parameters are time values, in milliseconds. A value of 0, when
5020 * modifying the parameters, indicates that the current value should not
5021 * be changed.
5024 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5025 char __user *optval,
5026 int __user *optlen) {
5027 struct sctp_rtoinfo rtoinfo;
5028 struct sctp_association *asoc;
5030 if (len < sizeof (struct sctp_rtoinfo))
5031 return -EINVAL;
5033 len = sizeof(struct sctp_rtoinfo);
5035 if (copy_from_user(&rtoinfo, optval, len))
5036 return -EFAULT;
5038 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5040 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5041 return -EINVAL;
5043 /* Values corresponding to the specific association. */
5044 if (asoc) {
5045 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5046 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5047 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5048 } else {
5049 /* Values corresponding to the endpoint. */
5050 struct sctp_sock *sp = sctp_sk(sk);
5052 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5053 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5054 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5057 if (put_user(len, optlen))
5058 return -EFAULT;
5060 if (copy_to_user(optval, &rtoinfo, len))
5061 return -EFAULT;
5063 return 0;
5068 * 7.1.2 SCTP_ASSOCINFO
5070 * This option is used to tune the maximum retransmission attempts
5071 * of the association.
5072 * Returns an error if the new association retransmission value is
5073 * greater than the sum of the retransmission value of the peer.
5074 * See [SCTP] for more information.
5077 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5078 char __user *optval,
5079 int __user *optlen)
5082 struct sctp_assocparams assocparams;
5083 struct sctp_association *asoc;
5084 struct list_head *pos;
5085 int cnt = 0;
5087 if (len < sizeof (struct sctp_assocparams))
5088 return -EINVAL;
5090 len = sizeof(struct sctp_assocparams);
5092 if (copy_from_user(&assocparams, optval, len))
5093 return -EFAULT;
5095 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5097 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5098 return -EINVAL;
5100 /* Values correspoinding to the specific association */
5101 if (asoc) {
5102 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5103 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5104 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5105 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5106 * 1000) +
5107 (asoc->cookie_life.tv_usec
5108 / 1000);
5110 list_for_each(pos, &asoc->peer.transport_addr_list) {
5111 cnt ++;
5114 assocparams.sasoc_number_peer_destinations = cnt;
5115 } else {
5116 /* Values corresponding to the endpoint */
5117 struct sctp_sock *sp = sctp_sk(sk);
5119 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5120 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5121 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5122 assocparams.sasoc_cookie_life =
5123 sp->assocparams.sasoc_cookie_life;
5124 assocparams.sasoc_number_peer_destinations =
5125 sp->assocparams.
5126 sasoc_number_peer_destinations;
5129 if (put_user(len, optlen))
5130 return -EFAULT;
5132 if (copy_to_user(optval, &assocparams, len))
5133 return -EFAULT;
5135 return 0;
5139 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5141 * This socket option is a boolean flag which turns on or off mapped V4
5142 * addresses. If this option is turned on and the socket is type
5143 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5144 * If this option is turned off, then no mapping will be done of V4
5145 * addresses and a user will receive both PF_INET6 and PF_INET type
5146 * addresses on the socket.
5148 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5149 char __user *optval, int __user *optlen)
5151 int val;
5152 struct sctp_sock *sp = sctp_sk(sk);
5154 if (len < sizeof(int))
5155 return -EINVAL;
5157 len = sizeof(int);
5158 val = sp->v4mapped;
5159 if (put_user(len, optlen))
5160 return -EFAULT;
5161 if (copy_to_user(optval, &val, len))
5162 return -EFAULT;
5164 return 0;
5168 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5169 * (chapter and verse is quoted at sctp_setsockopt_context())
5171 static int sctp_getsockopt_context(struct sock *sk, int len,
5172 char __user *optval, int __user *optlen)
5174 struct sctp_assoc_value params;
5175 struct sctp_sock *sp;
5176 struct sctp_association *asoc;
5178 if (len < sizeof(struct sctp_assoc_value))
5179 return -EINVAL;
5181 len = sizeof(struct sctp_assoc_value);
5183 if (copy_from_user(&params, optval, len))
5184 return -EFAULT;
5186 sp = sctp_sk(sk);
5188 if (params.assoc_id != 0) {
5189 asoc = sctp_id2assoc(sk, params.assoc_id);
5190 if (!asoc)
5191 return -EINVAL;
5192 params.assoc_value = asoc->default_rcv_context;
5193 } else {
5194 params.assoc_value = sp->default_rcv_context;
5197 if (put_user(len, optlen))
5198 return -EFAULT;
5199 if (copy_to_user(optval, &params, len))
5200 return -EFAULT;
5202 return 0;
5206 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5207 * This option will get or set the maximum size to put in any outgoing
5208 * SCTP DATA chunk. If a message is larger than this size it will be
5209 * fragmented by SCTP into the specified size. Note that the underlying
5210 * SCTP implementation may fragment into smaller sized chunks when the
5211 * PMTU of the underlying association is smaller than the value set by
5212 * the user. The default value for this option is '0' which indicates
5213 * the user is NOT limiting fragmentation and only the PMTU will effect
5214 * SCTP's choice of DATA chunk size. Note also that values set larger
5215 * than the maximum size of an IP datagram will effectively let SCTP
5216 * control fragmentation (i.e. the same as setting this option to 0).
5218 * The following structure is used to access and modify this parameter:
5220 * struct sctp_assoc_value {
5221 * sctp_assoc_t assoc_id;
5222 * uint32_t assoc_value;
5223 * };
5225 * assoc_id: This parameter is ignored for one-to-one style sockets.
5226 * For one-to-many style sockets this parameter indicates which
5227 * association the user is performing an action upon. Note that if
5228 * this field's value is zero then the endpoints default value is
5229 * changed (effecting future associations only).
5230 * assoc_value: This parameter specifies the maximum size in bytes.
5232 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5233 char __user *optval, int __user *optlen)
5235 struct sctp_assoc_value params;
5236 struct sctp_association *asoc;
5238 if (len == sizeof(int)) {
5239 printk(KERN_WARNING
5240 "SCTP: Use of int in maxseg socket option deprecated\n");
5241 printk(KERN_WARNING
5242 "SCTP: Use struct sctp_assoc_value instead\n");
5243 params.assoc_id = 0;
5244 } else if (len >= sizeof(struct sctp_assoc_value)) {
5245 len = sizeof(struct sctp_assoc_value);
5246 if (copy_from_user(&params, optval, sizeof(params)))
5247 return -EFAULT;
5248 } else
5249 return -EINVAL;
5251 asoc = sctp_id2assoc(sk, params.assoc_id);
5252 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5253 return -EINVAL;
5255 if (asoc)
5256 params.assoc_value = asoc->frag_point;
5257 else
5258 params.assoc_value = sctp_sk(sk)->user_frag;
5260 if (put_user(len, optlen))
5261 return -EFAULT;
5262 if (len == sizeof(int)) {
5263 if (copy_to_user(optval, &params.assoc_value, len))
5264 return -EFAULT;
5265 } else {
5266 if (copy_to_user(optval, &params, len))
5267 return -EFAULT;
5270 return 0;
5274 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5275 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5277 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5278 char __user *optval, int __user *optlen)
5280 int val;
5282 if (len < sizeof(int))
5283 return -EINVAL;
5285 len = sizeof(int);
5287 val = sctp_sk(sk)->frag_interleave;
5288 if (put_user(len, optlen))
5289 return -EFAULT;
5290 if (copy_to_user(optval, &val, len))
5291 return -EFAULT;
5293 return 0;
5297 * 7.1.25. Set or Get the sctp partial delivery point
5298 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5300 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5301 char __user *optval,
5302 int __user *optlen)
5304 u32 val;
5306 if (len < sizeof(u32))
5307 return -EINVAL;
5309 len = sizeof(u32);
5311 val = sctp_sk(sk)->pd_point;
5312 if (put_user(len, optlen))
5313 return -EFAULT;
5314 if (copy_to_user(optval, &val, len))
5315 return -EFAULT;
5317 return -ENOTSUPP;
5321 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5322 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5324 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5325 char __user *optval,
5326 int __user *optlen)
5328 struct sctp_assoc_value params;
5329 struct sctp_sock *sp;
5330 struct sctp_association *asoc;
5332 if (len == sizeof(int)) {
5333 printk(KERN_WARNING
5334 "SCTP: Use of int in max_burst socket option deprecated\n");
5335 printk(KERN_WARNING
5336 "SCTP: Use struct sctp_assoc_value instead\n");
5337 params.assoc_id = 0;
5338 } else if (len >= sizeof(struct sctp_assoc_value)) {
5339 len = sizeof(struct sctp_assoc_value);
5340 if (copy_from_user(&params, optval, len))
5341 return -EFAULT;
5342 } else
5343 return -EINVAL;
5345 sp = sctp_sk(sk);
5347 if (params.assoc_id != 0) {
5348 asoc = sctp_id2assoc(sk, params.assoc_id);
5349 if (!asoc)
5350 return -EINVAL;
5351 params.assoc_value = asoc->max_burst;
5352 } else
5353 params.assoc_value = sp->max_burst;
5355 if (len == sizeof(int)) {
5356 if (copy_to_user(optval, &params.assoc_value, len))
5357 return -EFAULT;
5358 } else {
5359 if (copy_to_user(optval, &params, len))
5360 return -EFAULT;
5363 return 0;
5367 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5368 char __user *optval, int __user *optlen)
5370 struct sctp_hmacalgo __user *p = (void __user *)optval;
5371 struct sctp_hmac_algo_param *hmacs;
5372 __u16 data_len = 0;
5373 u32 num_idents;
5375 if (!sctp_auth_enable)
5376 return -EACCES;
5378 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5379 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5381 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5382 return -EINVAL;
5384 len = sizeof(struct sctp_hmacalgo) + data_len;
5385 num_idents = data_len / sizeof(u16);
5387 if (put_user(len, optlen))
5388 return -EFAULT;
5389 if (put_user(num_idents, &p->shmac_num_idents))
5390 return -EFAULT;
5391 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5392 return -EFAULT;
5393 return 0;
5396 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5397 char __user *optval, int __user *optlen)
5399 struct sctp_authkeyid val;
5400 struct sctp_association *asoc;
5402 if (!sctp_auth_enable)
5403 return -EACCES;
5405 if (len < sizeof(struct sctp_authkeyid))
5406 return -EINVAL;
5407 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5408 return -EFAULT;
5410 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5411 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5412 return -EINVAL;
5414 if (asoc)
5415 val.scact_keynumber = asoc->active_key_id;
5416 else
5417 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5419 len = sizeof(struct sctp_authkeyid);
5420 if (put_user(len, optlen))
5421 return -EFAULT;
5422 if (copy_to_user(optval, &val, len))
5423 return -EFAULT;
5425 return 0;
5428 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5429 char __user *optval, int __user *optlen)
5431 struct sctp_authchunks __user *p = (void __user *)optval;
5432 struct sctp_authchunks val;
5433 struct sctp_association *asoc;
5434 struct sctp_chunks_param *ch;
5435 u32 num_chunks = 0;
5436 char __user *to;
5438 if (!sctp_auth_enable)
5439 return -EACCES;
5441 if (len < sizeof(struct sctp_authchunks))
5442 return -EINVAL;
5444 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5445 return -EFAULT;
5447 to = p->gauth_chunks;
5448 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5449 if (!asoc)
5450 return -EINVAL;
5452 ch = asoc->peer.peer_chunks;
5453 if (!ch)
5454 goto num;
5456 /* See if the user provided enough room for all the data */
5457 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5458 if (len < num_chunks)
5459 return -EINVAL;
5461 if (copy_to_user(to, ch->chunks, num_chunks))
5462 return -EFAULT;
5463 num:
5464 len = sizeof(struct sctp_authchunks) + num_chunks;
5465 if (put_user(len, optlen)) return -EFAULT;
5466 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5467 return -EFAULT;
5468 return 0;
5471 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5472 char __user *optval, int __user *optlen)
5474 struct sctp_authchunks __user *p = (void __user *)optval;
5475 struct sctp_authchunks val;
5476 struct sctp_association *asoc;
5477 struct sctp_chunks_param *ch;
5478 u32 num_chunks = 0;
5479 char __user *to;
5481 if (!sctp_auth_enable)
5482 return -EACCES;
5484 if (len < sizeof(struct sctp_authchunks))
5485 return -EINVAL;
5487 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5488 return -EFAULT;
5490 to = p->gauth_chunks;
5491 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5492 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5493 return -EINVAL;
5495 if (asoc)
5496 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5497 else
5498 ch = sctp_sk(sk)->ep->auth_chunk_list;
5500 if (!ch)
5501 goto num;
5503 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5504 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5505 return -EINVAL;
5507 if (copy_to_user(to, ch->chunks, num_chunks))
5508 return -EFAULT;
5509 num:
5510 len = sizeof(struct sctp_authchunks) + num_chunks;
5511 if (put_user(len, optlen))
5512 return -EFAULT;
5513 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5514 return -EFAULT;
5516 return 0;
5520 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5521 * This option gets the current number of associations that are attached
5522 * to a one-to-many style socket. The option value is an uint32_t.
5524 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5525 char __user *optval, int __user *optlen)
5527 struct sctp_sock *sp = sctp_sk(sk);
5528 struct sctp_association *asoc;
5529 u32 val = 0;
5531 if (sctp_style(sk, TCP))
5532 return -EOPNOTSUPP;
5534 if (len < sizeof(u32))
5535 return -EINVAL;
5537 len = sizeof(u32);
5539 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5540 val++;
5543 if (put_user(len, optlen))
5544 return -EFAULT;
5545 if (copy_to_user(optval, &val, len))
5546 return -EFAULT;
5548 return 0;
5551 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5552 char __user *optval, int __user *optlen)
5554 int retval = 0;
5555 int len;
5557 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5558 sk, optname);
5560 /* I can hardly begin to describe how wrong this is. This is
5561 * so broken as to be worse than useless. The API draft
5562 * REALLY is NOT helpful here... I am not convinced that the
5563 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5564 * are at all well-founded.
5566 if (level != SOL_SCTP) {
5567 struct sctp_af *af = sctp_sk(sk)->pf->af;
5569 retval = af->getsockopt(sk, level, optname, optval, optlen);
5570 return retval;
5573 if (get_user(len, optlen))
5574 return -EFAULT;
5576 sctp_lock_sock(sk);
5578 switch (optname) {
5579 case SCTP_STATUS:
5580 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5581 break;
5582 case SCTP_DISABLE_FRAGMENTS:
5583 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5584 optlen);
5585 break;
5586 case SCTP_EVENTS:
5587 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5588 break;
5589 case SCTP_AUTOCLOSE:
5590 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5591 break;
5592 case SCTP_SOCKOPT_PEELOFF:
5593 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5594 break;
5595 case SCTP_PEER_ADDR_PARAMS:
5596 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5597 optlen);
5598 break;
5599 case SCTP_DELAYED_ACK:
5600 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5601 optlen);
5602 break;
5603 case SCTP_INITMSG:
5604 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5605 break;
5606 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5607 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5608 optlen);
5609 break;
5610 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5611 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5612 optlen);
5613 break;
5614 case SCTP_GET_PEER_ADDRS_OLD:
5615 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5616 optlen);
5617 break;
5618 case SCTP_GET_LOCAL_ADDRS_OLD:
5619 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5620 optlen);
5621 break;
5622 case SCTP_GET_PEER_ADDRS:
5623 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5624 optlen);
5625 break;
5626 case SCTP_GET_LOCAL_ADDRS:
5627 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5628 optlen);
5629 break;
5630 case SCTP_SOCKOPT_CONNECTX3:
5631 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5632 break;
5633 case SCTP_DEFAULT_SEND_PARAM:
5634 retval = sctp_getsockopt_default_send_param(sk, len,
5635 optval, optlen);
5636 break;
5637 case SCTP_PRIMARY_ADDR:
5638 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5639 break;
5640 case SCTP_NODELAY:
5641 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5642 break;
5643 case SCTP_RTOINFO:
5644 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5645 break;
5646 case SCTP_ASSOCINFO:
5647 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5648 break;
5649 case SCTP_I_WANT_MAPPED_V4_ADDR:
5650 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5651 break;
5652 case SCTP_MAXSEG:
5653 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5654 break;
5655 case SCTP_GET_PEER_ADDR_INFO:
5656 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5657 optlen);
5658 break;
5659 case SCTP_ADAPTATION_LAYER:
5660 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5661 optlen);
5662 break;
5663 case SCTP_CONTEXT:
5664 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5665 break;
5666 case SCTP_FRAGMENT_INTERLEAVE:
5667 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5668 optlen);
5669 break;
5670 case SCTP_PARTIAL_DELIVERY_POINT:
5671 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5672 optlen);
5673 break;
5674 case SCTP_MAX_BURST:
5675 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5676 break;
5677 case SCTP_AUTH_KEY:
5678 case SCTP_AUTH_CHUNK:
5679 case SCTP_AUTH_DELETE_KEY:
5680 retval = -EOPNOTSUPP;
5681 break;
5682 case SCTP_HMAC_IDENT:
5683 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5684 break;
5685 case SCTP_AUTH_ACTIVE_KEY:
5686 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5687 break;
5688 case SCTP_PEER_AUTH_CHUNKS:
5689 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5690 optlen);
5691 break;
5692 case SCTP_LOCAL_AUTH_CHUNKS:
5693 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5694 optlen);
5695 break;
5696 case SCTP_GET_ASSOC_NUMBER:
5697 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5698 break;
5699 default:
5700 retval = -ENOPROTOOPT;
5701 break;
5704 sctp_release_sock(sk);
5705 return retval;
5708 static void sctp_hash(struct sock *sk)
5710 /* STUB */
5713 static void sctp_unhash(struct sock *sk)
5715 /* STUB */
5718 /* Check if port is acceptable. Possibly find first available port.
5720 * The port hash table (contained in the 'global' SCTP protocol storage
5721 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5722 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5723 * list (the list number is the port number hashed out, so as you
5724 * would expect from a hash function, all the ports in a given list have
5725 * such a number that hashes out to the same list number; you were
5726 * expecting that, right?); so each list has a set of ports, with a
5727 * link to the socket (struct sock) that uses it, the port number and
5728 * a fastreuse flag (FIXME: NPI ipg).
5730 static struct sctp_bind_bucket *sctp_bucket_create(
5731 struct sctp_bind_hashbucket *head, unsigned short snum);
5733 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5735 struct sctp_bind_hashbucket *head; /* hash list */
5736 struct sctp_bind_bucket *pp; /* hash list port iterator */
5737 struct hlist_node *node;
5738 unsigned short snum;
5739 int ret;
5741 snum = ntohs(addr->v4.sin_port);
5743 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5744 sctp_local_bh_disable();
5746 if (snum == 0) {
5747 /* Search for an available port. */
5748 int low, high, remaining, index;
5749 unsigned int rover;
5751 inet_get_local_port_range(&low, &high);
5752 remaining = (high - low) + 1;
5753 rover = net_random() % remaining + low;
5755 do {
5756 rover++;
5757 if ((rover < low) || (rover > high))
5758 rover = low;
5759 index = sctp_phashfn(rover);
5760 head = &sctp_port_hashtable[index];
5761 sctp_spin_lock(&head->lock);
5762 sctp_for_each_hentry(pp, node, &head->chain)
5763 if (pp->port == rover)
5764 goto next;
5765 break;
5766 next:
5767 sctp_spin_unlock(&head->lock);
5768 } while (--remaining > 0);
5770 /* Exhausted local port range during search? */
5771 ret = 1;
5772 if (remaining <= 0)
5773 goto fail;
5775 /* OK, here is the one we will use. HEAD (the port
5776 * hash table list entry) is non-NULL and we hold it's
5777 * mutex.
5779 snum = rover;
5780 } else {
5781 /* We are given an specific port number; we verify
5782 * that it is not being used. If it is used, we will
5783 * exahust the search in the hash list corresponding
5784 * to the port number (snum) - we detect that with the
5785 * port iterator, pp being NULL.
5787 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5788 sctp_spin_lock(&head->lock);
5789 sctp_for_each_hentry(pp, node, &head->chain) {
5790 if (pp->port == snum)
5791 goto pp_found;
5794 pp = NULL;
5795 goto pp_not_found;
5796 pp_found:
5797 if (!hlist_empty(&pp->owner)) {
5798 /* We had a port hash table hit - there is an
5799 * available port (pp != NULL) and it is being
5800 * used by other socket (pp->owner not empty); that other
5801 * socket is going to be sk2.
5803 int reuse = sk->sk_reuse;
5804 struct sock *sk2;
5805 struct hlist_node *node;
5807 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5808 if (pp->fastreuse && sk->sk_reuse &&
5809 sk->sk_state != SCTP_SS_LISTENING)
5810 goto success;
5812 /* Run through the list of sockets bound to the port
5813 * (pp->port) [via the pointers bind_next and
5814 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5815 * we get the endpoint they describe and run through
5816 * the endpoint's list of IP (v4 or v6) addresses,
5817 * comparing each of the addresses with the address of
5818 * the socket sk. If we find a match, then that means
5819 * that this port/socket (sk) combination are already
5820 * in an endpoint.
5822 sk_for_each_bound(sk2, node, &pp->owner) {
5823 struct sctp_endpoint *ep2;
5824 ep2 = sctp_sk(sk2)->ep;
5826 if (sk == sk2 ||
5827 (reuse && sk2->sk_reuse &&
5828 sk2->sk_state != SCTP_SS_LISTENING))
5829 continue;
5831 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5832 sctp_sk(sk2), sctp_sk(sk))) {
5833 ret = (long)sk2;
5834 goto fail_unlock;
5837 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5839 pp_not_found:
5840 /* If there was a hash table miss, create a new port. */
5841 ret = 1;
5842 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5843 goto fail_unlock;
5845 /* In either case (hit or miss), make sure fastreuse is 1 only
5846 * if sk->sk_reuse is too (that is, if the caller requested
5847 * SO_REUSEADDR on this socket -sk-).
5849 if (hlist_empty(&pp->owner)) {
5850 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5851 pp->fastreuse = 1;
5852 else
5853 pp->fastreuse = 0;
5854 } else if (pp->fastreuse &&
5855 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5856 pp->fastreuse = 0;
5858 /* We are set, so fill up all the data in the hash table
5859 * entry, tie the socket list information with the rest of the
5860 * sockets FIXME: Blurry, NPI (ipg).
5862 success:
5863 if (!sctp_sk(sk)->bind_hash) {
5864 inet_sk(sk)->num = snum;
5865 sk_add_bind_node(sk, &pp->owner);
5866 sctp_sk(sk)->bind_hash = pp;
5868 ret = 0;
5870 fail_unlock:
5871 sctp_spin_unlock(&head->lock);
5873 fail:
5874 sctp_local_bh_enable();
5875 return ret;
5878 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5879 * port is requested.
5881 static int sctp_get_port(struct sock *sk, unsigned short snum)
5883 long ret;
5884 union sctp_addr addr;
5885 struct sctp_af *af = sctp_sk(sk)->pf->af;
5887 /* Set up a dummy address struct from the sk. */
5888 af->from_sk(&addr, sk);
5889 addr.v4.sin_port = htons(snum);
5891 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5892 ret = sctp_get_port_local(sk, &addr);
5894 return (ret ? 1 : 0);
5898 * Move a socket to LISTENING state.
5900 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5902 struct sctp_sock *sp = sctp_sk(sk);
5903 struct sctp_endpoint *ep = sp->ep;
5904 struct crypto_hash *tfm = NULL;
5906 /* Allocate HMAC for generating cookie. */
5907 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5908 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5909 if (IS_ERR(tfm)) {
5910 if (net_ratelimit()) {
5911 printk(KERN_INFO
5912 "SCTP: failed to load transform for %s: %ld\n",
5913 sctp_hmac_alg, PTR_ERR(tfm));
5915 return -ENOSYS;
5917 sctp_sk(sk)->hmac = tfm;
5921 * If a bind() or sctp_bindx() is not called prior to a listen()
5922 * call that allows new associations to be accepted, the system
5923 * picks an ephemeral port and will choose an address set equivalent
5924 * to binding with a wildcard address.
5926 * This is not currently spelled out in the SCTP sockets
5927 * extensions draft, but follows the practice as seen in TCP
5928 * sockets.
5931 sk->sk_state = SCTP_SS_LISTENING;
5932 if (!ep->base.bind_addr.port) {
5933 if (sctp_autobind(sk))
5934 return -EAGAIN;
5935 } else {
5936 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5937 sk->sk_state = SCTP_SS_CLOSED;
5938 return -EADDRINUSE;
5942 sk->sk_max_ack_backlog = backlog;
5943 sctp_hash_endpoint(ep);
5944 return 0;
5948 * 4.1.3 / 5.1.3 listen()
5950 * By default, new associations are not accepted for UDP style sockets.
5951 * An application uses listen() to mark a socket as being able to
5952 * accept new associations.
5954 * On TCP style sockets, applications use listen() to ready the SCTP
5955 * endpoint for accepting inbound associations.
5957 * On both types of endpoints a backlog of '0' disables listening.
5959 * Move a socket to LISTENING state.
5961 int sctp_inet_listen(struct socket *sock, int backlog)
5963 struct sock *sk = sock->sk;
5964 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5965 int err = -EINVAL;
5967 if (unlikely(backlog < 0))
5968 return err;
5970 sctp_lock_sock(sk);
5972 /* Peeled-off sockets are not allowed to listen(). */
5973 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5974 goto out;
5976 if (sock->state != SS_UNCONNECTED)
5977 goto out;
5979 /* If backlog is zero, disable listening. */
5980 if (!backlog) {
5981 if (sctp_sstate(sk, CLOSED))
5982 goto out;
5984 err = 0;
5985 sctp_unhash_endpoint(ep);
5986 sk->sk_state = SCTP_SS_CLOSED;
5987 if (sk->sk_reuse)
5988 sctp_sk(sk)->bind_hash->fastreuse = 1;
5989 goto out;
5992 /* If we are already listening, just update the backlog */
5993 if (sctp_sstate(sk, LISTENING))
5994 sk->sk_max_ack_backlog = backlog;
5995 else {
5996 err = sctp_listen_start(sk, backlog);
5997 if (err)
5998 goto out;
6001 err = 0;
6002 out:
6003 sctp_release_sock(sk);
6004 return err;
6008 * This function is done by modeling the current datagram_poll() and the
6009 * tcp_poll(). Note that, based on these implementations, we don't
6010 * lock the socket in this function, even though it seems that,
6011 * ideally, locking or some other mechanisms can be used to ensure
6012 * the integrity of the counters (sndbuf and wmem_alloc) used
6013 * in this place. We assume that we don't need locks either until proven
6014 * otherwise.
6016 * Another thing to note is that we include the Async I/O support
6017 * here, again, by modeling the current TCP/UDP code. We don't have
6018 * a good way to test with it yet.
6020 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6022 struct sock *sk = sock->sk;
6023 struct sctp_sock *sp = sctp_sk(sk);
6024 unsigned int mask;
6026 poll_wait(file, sk->sk_sleep, wait);
6028 /* A TCP-style listening socket becomes readable when the accept queue
6029 * is not empty.
6031 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6032 return (!list_empty(&sp->ep->asocs)) ?
6033 (POLLIN | POLLRDNORM) : 0;
6035 mask = 0;
6037 /* Is there any exceptional events? */
6038 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6039 mask |= POLLERR;
6040 if (sk->sk_shutdown & RCV_SHUTDOWN)
6041 mask |= POLLRDHUP;
6042 if (sk->sk_shutdown == SHUTDOWN_MASK)
6043 mask |= POLLHUP;
6045 /* Is it readable? Reconsider this code with TCP-style support. */
6046 if (!skb_queue_empty(&sk->sk_receive_queue) ||
6047 (sk->sk_shutdown & RCV_SHUTDOWN))
6048 mask |= POLLIN | POLLRDNORM;
6050 /* The association is either gone or not ready. */
6051 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6052 return mask;
6054 /* Is it writable? */
6055 if (sctp_writeable(sk)) {
6056 mask |= POLLOUT | POLLWRNORM;
6057 } else {
6058 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6060 * Since the socket is not locked, the buffer
6061 * might be made available after the writeable check and
6062 * before the bit is set. This could cause a lost I/O
6063 * signal. tcp_poll() has a race breaker for this race
6064 * condition. Based on their implementation, we put
6065 * in the following code to cover it as well.
6067 if (sctp_writeable(sk))
6068 mask |= POLLOUT | POLLWRNORM;
6070 return mask;
6073 /********************************************************************
6074 * 2nd Level Abstractions
6075 ********************************************************************/
6077 static struct sctp_bind_bucket *sctp_bucket_create(
6078 struct sctp_bind_hashbucket *head, unsigned short snum)
6080 struct sctp_bind_bucket *pp;
6082 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6083 if (pp) {
6084 SCTP_DBG_OBJCNT_INC(bind_bucket);
6085 pp->port = snum;
6086 pp->fastreuse = 0;
6087 INIT_HLIST_HEAD(&pp->owner);
6088 hlist_add_head(&pp->node, &head->chain);
6090 return pp;
6093 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6094 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6096 if (pp && hlist_empty(&pp->owner)) {
6097 __hlist_del(&pp->node);
6098 kmem_cache_free(sctp_bucket_cachep, pp);
6099 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6103 /* Release this socket's reference to a local port. */
6104 static inline void __sctp_put_port(struct sock *sk)
6106 struct sctp_bind_hashbucket *head =
6107 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
6108 struct sctp_bind_bucket *pp;
6110 sctp_spin_lock(&head->lock);
6111 pp = sctp_sk(sk)->bind_hash;
6112 __sk_del_bind_node(sk);
6113 sctp_sk(sk)->bind_hash = NULL;
6114 inet_sk(sk)->num = 0;
6115 sctp_bucket_destroy(pp);
6116 sctp_spin_unlock(&head->lock);
6119 void sctp_put_port(struct sock *sk)
6121 sctp_local_bh_disable();
6122 __sctp_put_port(sk);
6123 sctp_local_bh_enable();
6127 * The system picks an ephemeral port and choose an address set equivalent
6128 * to binding with a wildcard address.
6129 * One of those addresses will be the primary address for the association.
6130 * This automatically enables the multihoming capability of SCTP.
6132 static int sctp_autobind(struct sock *sk)
6134 union sctp_addr autoaddr;
6135 struct sctp_af *af;
6136 __be16 port;
6138 /* Initialize a local sockaddr structure to INADDR_ANY. */
6139 af = sctp_sk(sk)->pf->af;
6141 port = htons(inet_sk(sk)->num);
6142 af->inaddr_any(&autoaddr, port);
6144 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6147 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6149 * From RFC 2292
6150 * 4.2 The cmsghdr Structure *
6152 * When ancillary data is sent or received, any number of ancillary data
6153 * objects can be specified by the msg_control and msg_controllen members of
6154 * the msghdr structure, because each object is preceded by
6155 * a cmsghdr structure defining the object's length (the cmsg_len member).
6156 * Historically Berkeley-derived implementations have passed only one object
6157 * at a time, but this API allows multiple objects to be
6158 * passed in a single call to sendmsg() or recvmsg(). The following example
6159 * shows two ancillary data objects in a control buffer.
6161 * |<--------------------------- msg_controllen -------------------------->|
6162 * | |
6164 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6166 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6167 * | | |
6169 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6171 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6172 * | | | | |
6174 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6175 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6177 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6179 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6183 * msg_control
6184 * points here
6186 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6187 sctp_cmsgs_t *cmsgs)
6189 struct cmsghdr *cmsg;
6190 struct msghdr *my_msg = (struct msghdr *)msg;
6192 for (cmsg = CMSG_FIRSTHDR(msg);
6193 cmsg != NULL;
6194 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6195 if (!CMSG_OK(my_msg, cmsg))
6196 return -EINVAL;
6198 /* Should we parse this header or ignore? */
6199 if (cmsg->cmsg_level != IPPROTO_SCTP)
6200 continue;
6202 /* Strictly check lengths following example in SCM code. */
6203 switch (cmsg->cmsg_type) {
6204 case SCTP_INIT:
6205 /* SCTP Socket API Extension
6206 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6208 * This cmsghdr structure provides information for
6209 * initializing new SCTP associations with sendmsg().
6210 * The SCTP_INITMSG socket option uses this same data
6211 * structure. This structure is not used for
6212 * recvmsg().
6214 * cmsg_level cmsg_type cmsg_data[]
6215 * ------------ ------------ ----------------------
6216 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6218 if (cmsg->cmsg_len !=
6219 CMSG_LEN(sizeof(struct sctp_initmsg)))
6220 return -EINVAL;
6221 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6222 break;
6224 case SCTP_SNDRCV:
6225 /* SCTP Socket API Extension
6226 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6228 * This cmsghdr structure specifies SCTP options for
6229 * sendmsg() and describes SCTP header information
6230 * about a received message through recvmsg().
6232 * cmsg_level cmsg_type cmsg_data[]
6233 * ------------ ------------ ----------------------
6234 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6236 if (cmsg->cmsg_len !=
6237 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6238 return -EINVAL;
6240 cmsgs->info =
6241 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6243 /* Minimally, validate the sinfo_flags. */
6244 if (cmsgs->info->sinfo_flags &
6245 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6246 SCTP_ABORT | SCTP_EOF))
6247 return -EINVAL;
6248 break;
6250 default:
6251 return -EINVAL;
6254 return 0;
6258 * Wait for a packet..
6259 * Note: This function is the same function as in core/datagram.c
6260 * with a few modifications to make lksctp work.
6262 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6264 int error;
6265 DEFINE_WAIT(wait);
6267 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6269 /* Socket errors? */
6270 error = sock_error(sk);
6271 if (error)
6272 goto out;
6274 if (!skb_queue_empty(&sk->sk_receive_queue))
6275 goto ready;
6277 /* Socket shut down? */
6278 if (sk->sk_shutdown & RCV_SHUTDOWN)
6279 goto out;
6281 /* Sequenced packets can come disconnected. If so we report the
6282 * problem.
6284 error = -ENOTCONN;
6286 /* Is there a good reason to think that we may receive some data? */
6287 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6288 goto out;
6290 /* Handle signals. */
6291 if (signal_pending(current))
6292 goto interrupted;
6294 /* Let another process have a go. Since we are going to sleep
6295 * anyway. Note: This may cause odd behaviors if the message
6296 * does not fit in the user's buffer, but this seems to be the
6297 * only way to honor MSG_DONTWAIT realistically.
6299 sctp_release_sock(sk);
6300 *timeo_p = schedule_timeout(*timeo_p);
6301 sctp_lock_sock(sk);
6303 ready:
6304 finish_wait(sk->sk_sleep, &wait);
6305 return 0;
6307 interrupted:
6308 error = sock_intr_errno(*timeo_p);
6310 out:
6311 finish_wait(sk->sk_sleep, &wait);
6312 *err = error;
6313 return error;
6316 /* Receive a datagram.
6317 * Note: This is pretty much the same routine as in core/datagram.c
6318 * with a few changes to make lksctp work.
6320 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6321 int noblock, int *err)
6323 int error;
6324 struct sk_buff *skb;
6325 long timeo;
6327 timeo = sock_rcvtimeo(sk, noblock);
6329 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6330 timeo, MAX_SCHEDULE_TIMEOUT);
6332 do {
6333 /* Again only user level code calls this function,
6334 * so nothing interrupt level
6335 * will suddenly eat the receive_queue.
6337 * Look at current nfs client by the way...
6338 * However, this function was corrent in any case. 8)
6340 if (flags & MSG_PEEK) {
6341 spin_lock_bh(&sk->sk_receive_queue.lock);
6342 skb = skb_peek(&sk->sk_receive_queue);
6343 if (skb)
6344 atomic_inc(&skb->users);
6345 spin_unlock_bh(&sk->sk_receive_queue.lock);
6346 } else {
6347 skb = skb_dequeue(&sk->sk_receive_queue);
6350 if (skb)
6351 return skb;
6353 /* Caller is allowed not to check sk->sk_err before calling. */
6354 error = sock_error(sk);
6355 if (error)
6356 goto no_packet;
6358 if (sk->sk_shutdown & RCV_SHUTDOWN)
6359 break;
6361 /* User doesn't want to wait. */
6362 error = -EAGAIN;
6363 if (!timeo)
6364 goto no_packet;
6365 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6367 return NULL;
6369 no_packet:
6370 *err = error;
6371 return NULL;
6374 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6375 static void __sctp_write_space(struct sctp_association *asoc)
6377 struct sock *sk = asoc->base.sk;
6378 struct socket *sock = sk->sk_socket;
6380 if ((sctp_wspace(asoc) > 0) && sock) {
6381 if (waitqueue_active(&asoc->wait))
6382 wake_up_interruptible(&asoc->wait);
6384 if (sctp_writeable(sk)) {
6385 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6386 wake_up_interruptible(sk->sk_sleep);
6388 /* Note that we try to include the Async I/O support
6389 * here by modeling from the current TCP/UDP code.
6390 * We have not tested with it yet.
6392 if (sock->fasync_list &&
6393 !(sk->sk_shutdown & SEND_SHUTDOWN))
6394 sock_wake_async(sock,
6395 SOCK_WAKE_SPACE, POLL_OUT);
6400 /* Do accounting for the sndbuf space.
6401 * Decrement the used sndbuf space of the corresponding association by the
6402 * data size which was just transmitted(freed).
6404 static void sctp_wfree(struct sk_buff *skb)
6406 struct sctp_association *asoc;
6407 struct sctp_chunk *chunk;
6408 struct sock *sk;
6410 /* Get the saved chunk pointer. */
6411 chunk = *((struct sctp_chunk **)(skb->cb));
6412 asoc = chunk->asoc;
6413 sk = asoc->base.sk;
6414 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6415 sizeof(struct sk_buff) +
6416 sizeof(struct sctp_chunk);
6418 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6421 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6423 sk->sk_wmem_queued -= skb->truesize;
6424 sk_mem_uncharge(sk, skb->truesize);
6426 sock_wfree(skb);
6427 __sctp_write_space(asoc);
6429 sctp_association_put(asoc);
6432 /* Do accounting for the receive space on the socket.
6433 * Accounting for the association is done in ulpevent.c
6434 * We set this as a destructor for the cloned data skbs so that
6435 * accounting is done at the correct time.
6437 void sctp_sock_rfree(struct sk_buff *skb)
6439 struct sock *sk = skb->sk;
6440 struct sctp_ulpevent *event = sctp_skb2event(skb);
6442 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6445 * Mimic the behavior of sock_rfree
6447 sk_mem_uncharge(sk, event->rmem_len);
6451 /* Helper function to wait for space in the sndbuf. */
6452 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6453 size_t msg_len)
6455 struct sock *sk = asoc->base.sk;
6456 int err = 0;
6457 long current_timeo = *timeo_p;
6458 DEFINE_WAIT(wait);
6460 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6461 asoc, (long)(*timeo_p), msg_len);
6463 /* Increment the association's refcnt. */
6464 sctp_association_hold(asoc);
6466 /* Wait on the association specific sndbuf space. */
6467 for (;;) {
6468 prepare_to_wait_exclusive(&asoc->wait, &wait,
6469 TASK_INTERRUPTIBLE);
6470 if (!*timeo_p)
6471 goto do_nonblock;
6472 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6473 asoc->base.dead)
6474 goto do_error;
6475 if (signal_pending(current))
6476 goto do_interrupted;
6477 if (msg_len <= sctp_wspace(asoc))
6478 break;
6480 /* Let another process have a go. Since we are going
6481 * to sleep anyway.
6483 sctp_release_sock(sk);
6484 current_timeo = schedule_timeout(current_timeo);
6485 BUG_ON(sk != asoc->base.sk);
6486 sctp_lock_sock(sk);
6488 *timeo_p = current_timeo;
6491 out:
6492 finish_wait(&asoc->wait, &wait);
6494 /* Release the association's refcnt. */
6495 sctp_association_put(asoc);
6497 return err;
6499 do_error:
6500 err = -EPIPE;
6501 goto out;
6503 do_interrupted:
6504 err = sock_intr_errno(*timeo_p);
6505 goto out;
6507 do_nonblock:
6508 err = -EAGAIN;
6509 goto out;
6512 /* If socket sndbuf has changed, wake up all per association waiters. */
6513 void sctp_write_space(struct sock *sk)
6515 struct sctp_association *asoc;
6517 /* Wake up the tasks in each wait queue. */
6518 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6519 __sctp_write_space(asoc);
6523 /* Is there any sndbuf space available on the socket?
6525 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6526 * associations on the same socket. For a UDP-style socket with
6527 * multiple associations, it is possible for it to be "unwriteable"
6528 * prematurely. I assume that this is acceptable because
6529 * a premature "unwriteable" is better than an accidental "writeable" which
6530 * would cause an unwanted block under certain circumstances. For the 1-1
6531 * UDP-style sockets or TCP-style sockets, this code should work.
6532 * - Daisy
6534 static int sctp_writeable(struct sock *sk)
6536 int amt = 0;
6538 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6539 if (amt < 0)
6540 amt = 0;
6541 return amt;
6544 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6545 * returns immediately with EINPROGRESS.
6547 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6549 struct sock *sk = asoc->base.sk;
6550 int err = 0;
6551 long current_timeo = *timeo_p;
6552 DEFINE_WAIT(wait);
6554 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6555 (long)(*timeo_p));
6557 /* Increment the association's refcnt. */
6558 sctp_association_hold(asoc);
6560 for (;;) {
6561 prepare_to_wait_exclusive(&asoc->wait, &wait,
6562 TASK_INTERRUPTIBLE);
6563 if (!*timeo_p)
6564 goto do_nonblock;
6565 if (sk->sk_shutdown & RCV_SHUTDOWN)
6566 break;
6567 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6568 asoc->base.dead)
6569 goto do_error;
6570 if (signal_pending(current))
6571 goto do_interrupted;
6573 if (sctp_state(asoc, ESTABLISHED))
6574 break;
6576 /* Let another process have a go. Since we are going
6577 * to sleep anyway.
6579 sctp_release_sock(sk);
6580 current_timeo = schedule_timeout(current_timeo);
6581 sctp_lock_sock(sk);
6583 *timeo_p = current_timeo;
6586 out:
6587 finish_wait(&asoc->wait, &wait);
6589 /* Release the association's refcnt. */
6590 sctp_association_put(asoc);
6592 return err;
6594 do_error:
6595 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6596 err = -ETIMEDOUT;
6597 else
6598 err = -ECONNREFUSED;
6599 goto out;
6601 do_interrupted:
6602 err = sock_intr_errno(*timeo_p);
6603 goto out;
6605 do_nonblock:
6606 err = -EINPROGRESS;
6607 goto out;
6610 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6612 struct sctp_endpoint *ep;
6613 int err = 0;
6614 DEFINE_WAIT(wait);
6616 ep = sctp_sk(sk)->ep;
6619 for (;;) {
6620 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6621 TASK_INTERRUPTIBLE);
6623 if (list_empty(&ep->asocs)) {
6624 sctp_release_sock(sk);
6625 timeo = schedule_timeout(timeo);
6626 sctp_lock_sock(sk);
6629 err = -EINVAL;
6630 if (!sctp_sstate(sk, LISTENING))
6631 break;
6633 err = 0;
6634 if (!list_empty(&ep->asocs))
6635 break;
6637 err = sock_intr_errno(timeo);
6638 if (signal_pending(current))
6639 break;
6641 err = -EAGAIN;
6642 if (!timeo)
6643 break;
6646 finish_wait(sk->sk_sleep, &wait);
6648 return err;
6651 static void sctp_wait_for_close(struct sock *sk, long timeout)
6653 DEFINE_WAIT(wait);
6655 do {
6656 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6657 if (list_empty(&sctp_sk(sk)->ep->asocs))
6658 break;
6659 sctp_release_sock(sk);
6660 timeout = schedule_timeout(timeout);
6661 sctp_lock_sock(sk);
6662 } while (!signal_pending(current) && timeout);
6664 finish_wait(sk->sk_sleep, &wait);
6667 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6669 struct sk_buff *frag;
6671 if (!skb->data_len)
6672 goto done;
6674 /* Don't forget the fragments. */
6675 skb_walk_frags(skb, frag)
6676 sctp_skb_set_owner_r_frag(frag, sk);
6678 done:
6679 sctp_skb_set_owner_r(skb, sk);
6682 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6683 struct sctp_association *asoc)
6685 struct inet_sock *inet = inet_sk(sk);
6686 struct inet_sock *newinet = inet_sk(newsk);
6688 newsk->sk_type = sk->sk_type;
6689 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6690 newsk->sk_flags = sk->sk_flags;
6691 newsk->sk_no_check = sk->sk_no_check;
6692 newsk->sk_reuse = sk->sk_reuse;
6694 newsk->sk_shutdown = sk->sk_shutdown;
6695 newsk->sk_destruct = inet_sock_destruct;
6696 newsk->sk_family = sk->sk_family;
6697 newsk->sk_protocol = IPPROTO_SCTP;
6698 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6699 newsk->sk_sndbuf = sk->sk_sndbuf;
6700 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6701 newsk->sk_lingertime = sk->sk_lingertime;
6702 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6703 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6705 newinet = inet_sk(newsk);
6707 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6708 * getsockname() and getpeername()
6710 newinet->sport = inet->sport;
6711 newinet->saddr = inet->saddr;
6712 newinet->rcv_saddr = inet->rcv_saddr;
6713 newinet->dport = htons(asoc->peer.port);
6714 newinet->pmtudisc = inet->pmtudisc;
6715 newinet->id = asoc->next_tsn ^ jiffies;
6717 newinet->uc_ttl = inet->uc_ttl;
6718 newinet->mc_loop = 1;
6719 newinet->mc_ttl = 1;
6720 newinet->mc_index = 0;
6721 newinet->mc_list = NULL;
6724 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6725 * and its messages to the newsk.
6727 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6728 struct sctp_association *assoc,
6729 sctp_socket_type_t type)
6731 struct sctp_sock *oldsp = sctp_sk(oldsk);
6732 struct sctp_sock *newsp = sctp_sk(newsk);
6733 struct sctp_bind_bucket *pp; /* hash list port iterator */
6734 struct sctp_endpoint *newep = newsp->ep;
6735 struct sk_buff *skb, *tmp;
6736 struct sctp_ulpevent *event;
6737 struct sctp_bind_hashbucket *head;
6739 /* Migrate socket buffer sizes and all the socket level options to the
6740 * new socket.
6742 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6743 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6744 /* Brute force copy old sctp opt. */
6745 inet_sk_copy_descendant(newsk, oldsk);
6747 /* Restore the ep value that was overwritten with the above structure
6748 * copy.
6750 newsp->ep = newep;
6751 newsp->hmac = NULL;
6753 /* Hook this new socket in to the bind_hash list. */
6754 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6755 sctp_local_bh_disable();
6756 sctp_spin_lock(&head->lock);
6757 pp = sctp_sk(oldsk)->bind_hash;
6758 sk_add_bind_node(newsk, &pp->owner);
6759 sctp_sk(newsk)->bind_hash = pp;
6760 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6761 sctp_spin_unlock(&head->lock);
6762 sctp_local_bh_enable();
6764 /* Copy the bind_addr list from the original endpoint to the new
6765 * endpoint so that we can handle restarts properly
6767 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6768 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6770 /* Move any messages in the old socket's receive queue that are for the
6771 * peeled off association to the new socket's receive queue.
6773 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6774 event = sctp_skb2event(skb);
6775 if (event->asoc == assoc) {
6776 __skb_unlink(skb, &oldsk->sk_receive_queue);
6777 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6778 sctp_skb_set_owner_r_frag(skb, newsk);
6782 /* Clean up any messages pending delivery due to partial
6783 * delivery. Three cases:
6784 * 1) No partial deliver; no work.
6785 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6786 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6788 skb_queue_head_init(&newsp->pd_lobby);
6789 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6791 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6792 struct sk_buff_head *queue;
6794 /* Decide which queue to move pd_lobby skbs to. */
6795 if (assoc->ulpq.pd_mode) {
6796 queue = &newsp->pd_lobby;
6797 } else
6798 queue = &newsk->sk_receive_queue;
6800 /* Walk through the pd_lobby, looking for skbs that
6801 * need moved to the new socket.
6803 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6804 event = sctp_skb2event(skb);
6805 if (event->asoc == assoc) {
6806 __skb_unlink(skb, &oldsp->pd_lobby);
6807 __skb_queue_tail(queue, skb);
6808 sctp_skb_set_owner_r_frag(skb, newsk);
6812 /* Clear up any skbs waiting for the partial
6813 * delivery to finish.
6815 if (assoc->ulpq.pd_mode)
6816 sctp_clear_pd(oldsk, NULL);
6820 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6821 sctp_skb_set_owner_r_frag(skb, newsk);
6823 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6824 sctp_skb_set_owner_r_frag(skb, newsk);
6826 /* Set the type of socket to indicate that it is peeled off from the
6827 * original UDP-style socket or created with the accept() call on a
6828 * TCP-style socket..
6830 newsp->type = type;
6832 /* Mark the new socket "in-use" by the user so that any packets
6833 * that may arrive on the association after we've moved it are
6834 * queued to the backlog. This prevents a potential race between
6835 * backlog processing on the old socket and new-packet processing
6836 * on the new socket.
6838 * The caller has just allocated newsk so we can guarantee that other
6839 * paths won't try to lock it and then oldsk.
6841 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6842 sctp_assoc_migrate(assoc, newsk);
6844 /* If the association on the newsk is already closed before accept()
6845 * is called, set RCV_SHUTDOWN flag.
6847 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6848 newsk->sk_shutdown |= RCV_SHUTDOWN;
6850 newsk->sk_state = SCTP_SS_ESTABLISHED;
6851 sctp_release_sock(newsk);
6855 /* This proto struct describes the ULP interface for SCTP. */
6856 struct proto sctp_prot = {
6857 .name = "SCTP",
6858 .owner = THIS_MODULE,
6859 .close = sctp_close,
6860 .connect = sctp_connect,
6861 .disconnect = sctp_disconnect,
6862 .accept = sctp_accept,
6863 .ioctl = sctp_ioctl,
6864 .init = sctp_init_sock,
6865 .destroy = sctp_destroy_sock,
6866 .shutdown = sctp_shutdown,
6867 .setsockopt = sctp_setsockopt,
6868 .getsockopt = sctp_getsockopt,
6869 .sendmsg = sctp_sendmsg,
6870 .recvmsg = sctp_recvmsg,
6871 .bind = sctp_bind,
6872 .backlog_rcv = sctp_backlog_rcv,
6873 .hash = sctp_hash,
6874 .unhash = sctp_unhash,
6875 .get_port = sctp_get_port,
6876 .obj_size = sizeof(struct sctp_sock),
6877 .sysctl_mem = sysctl_sctp_mem,
6878 .sysctl_rmem = sysctl_sctp_rmem,
6879 .sysctl_wmem = sysctl_sctp_wmem,
6880 .memory_pressure = &sctp_memory_pressure,
6881 .enter_memory_pressure = sctp_enter_memory_pressure,
6882 .memory_allocated = &sctp_memory_allocated,
6883 .sockets_allocated = &sctp_sockets_allocated,
6886 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6888 struct proto sctpv6_prot = {
6889 .name = "SCTPv6",
6890 .owner = THIS_MODULE,
6891 .close = sctp_close,
6892 .connect = sctp_connect,
6893 .disconnect = sctp_disconnect,
6894 .accept = sctp_accept,
6895 .ioctl = sctp_ioctl,
6896 .init = sctp_init_sock,
6897 .destroy = sctp_destroy_sock,
6898 .shutdown = sctp_shutdown,
6899 .setsockopt = sctp_setsockopt,
6900 .getsockopt = sctp_getsockopt,
6901 .sendmsg = sctp_sendmsg,
6902 .recvmsg = sctp_recvmsg,
6903 .bind = sctp_bind,
6904 .backlog_rcv = sctp_backlog_rcv,
6905 .hash = sctp_hash,
6906 .unhash = sctp_unhash,
6907 .get_port = sctp_get_port,
6908 .obj_size = sizeof(struct sctp6_sock),
6909 .sysctl_mem = sysctl_sctp_mem,
6910 .sysctl_rmem = sysctl_sctp_rmem,
6911 .sysctl_wmem = sysctl_sctp_wmem,
6912 .memory_pressure = &sctp_memory_pressure,
6913 .enter_memory_pressure = sctp_enter_memory_pressure,
6914 .memory_allocated = &sctp_memory_allocated,
6915 .sockets_allocated = &sctp_sockets_allocated,
6917 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */