net: Make sure BHs are disabled in sock_prot_inuse_add()
[linux-2.6/mini2440.git] / net / sctp / socket.c
blobf03af84edf63c2073efcfcc3f488d1d2cd15048b
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 static atomic_t 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 = atomic_read(&asoc->base.sk->sk_wmem_alloc);
135 if (amt >= asoc->base.sk->sk_sndbuf) {
136 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
137 amt = 0;
138 else {
139 amt = sk_stream_wspace(asoc->base.sk);
140 if (amt < 0)
141 amt = 0;
143 } else {
144 amt = asoc->base.sk->sk_sndbuf - amt;
146 return amt;
149 /* Increment the used sndbuf space count of the corresponding association by
150 * the size of the outgoing data chunk.
151 * Also, set the skb destructor for sndbuf accounting later.
153 * Since it is always 1-1 between chunk and skb, and also a new skb is always
154 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
155 * destructor in the data chunk skb for the purpose of the sndbuf space
156 * tracking.
158 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
160 struct sctp_association *asoc = chunk->asoc;
161 struct sock *sk = asoc->base.sk;
163 /* The sndbuf space is tracked per association. */
164 sctp_association_hold(asoc);
166 skb_set_owner_w(chunk->skb, sk);
168 chunk->skb->destructor = sctp_wfree;
169 /* Save the chunk pointer in skb for sctp_wfree to use later. */
170 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
172 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
173 sizeof(struct sk_buff) +
174 sizeof(struct sctp_chunk);
176 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
177 sk->sk_wmem_queued += chunk->skb->truesize;
178 sk_mem_charge(sk, chunk->skb->truesize);
181 /* Verify that this is a valid address. */
182 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
183 int len)
185 struct sctp_af *af;
187 /* Verify basic sockaddr. */
188 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
189 if (!af)
190 return -EINVAL;
192 /* Is this a valid SCTP address? */
193 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
194 return -EINVAL;
196 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
197 return -EINVAL;
199 return 0;
202 /* Look up the association by its id. If this is not a UDP-style
203 * socket, the ID field is always ignored.
205 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
207 struct sctp_association *asoc = NULL;
209 /* If this is not a UDP-style socket, assoc id should be ignored. */
210 if (!sctp_style(sk, UDP)) {
211 /* Return NULL if the socket state is not ESTABLISHED. It
212 * could be a TCP-style listening socket or a socket which
213 * hasn't yet called connect() to establish an association.
215 if (!sctp_sstate(sk, ESTABLISHED))
216 return NULL;
218 /* Get the first and the only association from the list. */
219 if (!list_empty(&sctp_sk(sk)->ep->asocs))
220 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
221 struct sctp_association, asocs);
222 return asoc;
225 /* Otherwise this is a UDP-style socket. */
226 if (!id || (id == (sctp_assoc_t)-1))
227 return NULL;
229 spin_lock_bh(&sctp_assocs_id_lock);
230 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
231 spin_unlock_bh(&sctp_assocs_id_lock);
233 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
234 return NULL;
236 return asoc;
239 /* Look up the transport from an address and an assoc id. If both address and
240 * id are specified, the associations matching the address and the id should be
241 * the same.
243 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
244 struct sockaddr_storage *addr,
245 sctp_assoc_t id)
247 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
248 struct sctp_transport *transport;
249 union sctp_addr *laddr = (union sctp_addr *)addr;
251 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
252 laddr,
253 &transport);
255 if (!addr_asoc)
256 return NULL;
258 id_asoc = sctp_id2assoc(sk, id);
259 if (id_asoc && (id_asoc != addr_asoc))
260 return NULL;
262 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
263 (union sctp_addr *)addr);
265 return transport;
268 /* API 3.1.2 bind() - UDP Style Syntax
269 * The syntax of bind() is,
271 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
273 * sd - the socket descriptor returned by socket().
274 * addr - the address structure (struct sockaddr_in or struct
275 * sockaddr_in6 [RFC 2553]),
276 * addr_len - the size of the address structure.
278 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
280 int retval = 0;
282 sctp_lock_sock(sk);
284 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
285 sk, addr, addr_len);
287 /* Disallow binding twice. */
288 if (!sctp_sk(sk)->ep->base.bind_addr.port)
289 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
290 addr_len);
291 else
292 retval = -EINVAL;
294 sctp_release_sock(sk);
296 return retval;
299 static long sctp_get_port_local(struct sock *, union sctp_addr *);
301 /* Verify this is a valid sockaddr. */
302 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
303 union sctp_addr *addr, int len)
305 struct sctp_af *af;
307 /* Check minimum size. */
308 if (len < sizeof (struct sockaddr))
309 return NULL;
311 /* V4 mapped address are really of AF_INET family */
312 if (addr->sa.sa_family == AF_INET6 &&
313 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
314 if (!opt->pf->af_supported(AF_INET, opt))
315 return NULL;
316 } else {
317 /* Does this PF support this AF? */
318 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
319 return NULL;
322 /* If we get this far, af is valid. */
323 af = sctp_get_af_specific(addr->sa.sa_family);
325 if (len < af->sockaddr_len)
326 return NULL;
328 return af;
331 /* Bind a local address either to an endpoint or to an association. */
332 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
334 struct sctp_sock *sp = sctp_sk(sk);
335 struct sctp_endpoint *ep = sp->ep;
336 struct sctp_bind_addr *bp = &ep->base.bind_addr;
337 struct sctp_af *af;
338 unsigned short snum;
339 int ret = 0;
341 /* Common sockaddr verification. */
342 af = sctp_sockaddr_af(sp, addr, len);
343 if (!af) {
344 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
345 sk, addr, len);
346 return -EINVAL;
349 snum = ntohs(addr->v4.sin_port);
351 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
352 ", port: %d, new port: %d, len: %d)\n",
354 addr,
355 bp->port, snum,
356 len);
358 /* PF specific bind() address verification. */
359 if (!sp->pf->bind_verify(sp, addr))
360 return -EADDRNOTAVAIL;
362 /* We must either be unbound, or bind to the same port.
363 * It's OK to allow 0 ports if we are already bound.
364 * We'll just inhert an already bound port in this case
366 if (bp->port) {
367 if (!snum)
368 snum = bp->port;
369 else if (snum != bp->port) {
370 SCTP_DEBUG_PRINTK("sctp_do_bind:"
371 " New port %d does not match existing port "
372 "%d.\n", snum, bp->port);
373 return -EINVAL;
377 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
378 return -EACCES;
380 /* See if the address matches any of the addresses we may have
381 * already bound before checking against other endpoints.
383 if (sctp_bind_addr_match(bp, addr, sp))
384 return -EINVAL;
386 /* Make sure we are allowed to bind here.
387 * The function sctp_get_port_local() does duplicate address
388 * detection.
390 addr->v4.sin_port = htons(snum);
391 if ((ret = sctp_get_port_local(sk, addr))) {
392 return -EADDRINUSE;
395 /* Refresh ephemeral port. */
396 if (!bp->port)
397 bp->port = inet_sk(sk)->num;
399 /* Add the address to the bind address list.
400 * Use GFP_ATOMIC since BHs will be disabled.
402 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
404 /* Copy back into socket for getsockname() use. */
405 if (!ret) {
406 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
407 af->to_sk_saddr(addr, sk);
410 return ret;
413 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
415 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
416 * at any one time. If a sender, after sending an ASCONF chunk, decides
417 * it needs to transfer another ASCONF Chunk, it MUST wait until the
418 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
419 * subsequent ASCONF. Note this restriction binds each side, so at any
420 * time two ASCONF may be in-transit on any given association (one sent
421 * from each endpoint).
423 static int sctp_send_asconf(struct sctp_association *asoc,
424 struct sctp_chunk *chunk)
426 int retval = 0;
428 /* If there is an outstanding ASCONF chunk, queue it for later
429 * transmission.
431 if (asoc->addip_last_asconf) {
432 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
433 goto out;
436 /* Hold the chunk until an ASCONF_ACK is received. */
437 sctp_chunk_hold(chunk);
438 retval = sctp_primitive_ASCONF(asoc, chunk);
439 if (retval)
440 sctp_chunk_free(chunk);
441 else
442 asoc->addip_last_asconf = chunk;
444 out:
445 return retval;
448 /* Add a list of addresses as bind addresses to local endpoint or
449 * association.
451 * Basically run through each address specified in the addrs/addrcnt
452 * array/length pair, determine if it is IPv6 or IPv4 and call
453 * sctp_do_bind() on it.
455 * If any of them fails, then the operation will be reversed and the
456 * ones that were added will be removed.
458 * Only sctp_setsockopt_bindx() is supposed to call this function.
460 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
462 int cnt;
463 int retval = 0;
464 void *addr_buf;
465 struct sockaddr *sa_addr;
466 struct sctp_af *af;
468 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
469 sk, addrs, addrcnt);
471 addr_buf = addrs;
472 for (cnt = 0; cnt < addrcnt; cnt++) {
473 /* The list may contain either IPv4 or IPv6 address;
474 * determine the address length for walking thru the list.
476 sa_addr = (struct sockaddr *)addr_buf;
477 af = sctp_get_af_specific(sa_addr->sa_family);
478 if (!af) {
479 retval = -EINVAL;
480 goto err_bindx_add;
483 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
484 af->sockaddr_len);
486 addr_buf += af->sockaddr_len;
488 err_bindx_add:
489 if (retval < 0) {
490 /* Failed. Cleanup the ones that have been added */
491 if (cnt > 0)
492 sctp_bindx_rem(sk, addrs, cnt);
493 return retval;
497 return retval;
500 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
501 * associations that are part of the endpoint indicating that a list of local
502 * addresses are added to the endpoint.
504 * If any of the addresses is already in the bind address list of the
505 * association, we do not send the chunk for that association. But it will not
506 * affect other associations.
508 * Only sctp_setsockopt_bindx() is supposed to call this function.
510 static int sctp_send_asconf_add_ip(struct sock *sk,
511 struct sockaddr *addrs,
512 int addrcnt)
514 struct sctp_sock *sp;
515 struct sctp_endpoint *ep;
516 struct sctp_association *asoc;
517 struct sctp_bind_addr *bp;
518 struct sctp_chunk *chunk;
519 struct sctp_sockaddr_entry *laddr;
520 union sctp_addr *addr;
521 union sctp_addr saveaddr;
522 void *addr_buf;
523 struct sctp_af *af;
524 struct list_head *p;
525 int i;
526 int retval = 0;
528 if (!sctp_addip_enable)
529 return retval;
531 sp = sctp_sk(sk);
532 ep = sp->ep;
534 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
535 __func__, sk, addrs, addrcnt);
537 list_for_each_entry(asoc, &ep->asocs, asocs) {
539 if (!asoc->peer.asconf_capable)
540 continue;
542 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
543 continue;
545 if (!sctp_state(asoc, ESTABLISHED))
546 continue;
548 /* Check if any address in the packed array of addresses is
549 * in the bind address list of the association. If so,
550 * do not send the asconf chunk to its peer, but continue with
551 * other associations.
553 addr_buf = addrs;
554 for (i = 0; i < addrcnt; i++) {
555 addr = (union sctp_addr *)addr_buf;
556 af = sctp_get_af_specific(addr->v4.sin_family);
557 if (!af) {
558 retval = -EINVAL;
559 goto out;
562 if (sctp_assoc_lookup_laddr(asoc, addr))
563 break;
565 addr_buf += af->sockaddr_len;
567 if (i < addrcnt)
568 continue;
570 /* Use the first valid address in bind addr list of
571 * association as Address Parameter of ASCONF CHUNK.
573 bp = &asoc->base.bind_addr;
574 p = bp->address_list.next;
575 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
576 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
577 addrcnt, SCTP_PARAM_ADD_IP);
578 if (!chunk) {
579 retval = -ENOMEM;
580 goto out;
583 retval = sctp_send_asconf(asoc, chunk);
584 if (retval)
585 goto out;
587 /* Add the new addresses to the bind address list with
588 * use_as_src set to 0.
590 addr_buf = addrs;
591 for (i = 0; i < addrcnt; i++) {
592 addr = (union sctp_addr *)addr_buf;
593 af = sctp_get_af_specific(addr->v4.sin_family);
594 memcpy(&saveaddr, addr, af->sockaddr_len);
595 retval = sctp_add_bind_addr(bp, &saveaddr,
596 SCTP_ADDR_NEW, GFP_ATOMIC);
597 addr_buf += af->sockaddr_len;
601 out:
602 return retval;
605 /* Remove a list of addresses from bind addresses list. Do not remove the
606 * last address.
608 * Basically run through each address specified in the addrs/addrcnt
609 * array/length pair, determine if it is IPv6 or IPv4 and call
610 * sctp_del_bind() on it.
612 * If any of them fails, then the operation will be reversed and the
613 * ones that were removed will be added back.
615 * At least one address has to be left; if only one address is
616 * available, the operation will return -EBUSY.
618 * Only sctp_setsockopt_bindx() is supposed to call this function.
620 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
622 struct sctp_sock *sp = sctp_sk(sk);
623 struct sctp_endpoint *ep = sp->ep;
624 int cnt;
625 struct sctp_bind_addr *bp = &ep->base.bind_addr;
626 int retval = 0;
627 void *addr_buf;
628 union sctp_addr *sa_addr;
629 struct sctp_af *af;
631 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
632 sk, addrs, addrcnt);
634 addr_buf = addrs;
635 for (cnt = 0; cnt < addrcnt; cnt++) {
636 /* If the bind address list is empty or if there is only one
637 * bind address, there is nothing more to be removed (we need
638 * at least one address here).
640 if (list_empty(&bp->address_list) ||
641 (sctp_list_single_entry(&bp->address_list))) {
642 retval = -EBUSY;
643 goto err_bindx_rem;
646 sa_addr = (union sctp_addr *)addr_buf;
647 af = sctp_get_af_specific(sa_addr->sa.sa_family);
648 if (!af) {
649 retval = -EINVAL;
650 goto err_bindx_rem;
653 if (!af->addr_valid(sa_addr, sp, NULL)) {
654 retval = -EADDRNOTAVAIL;
655 goto err_bindx_rem;
658 if (sa_addr->v4.sin_port != htons(bp->port)) {
659 retval = -EINVAL;
660 goto err_bindx_rem;
663 /* FIXME - There is probably a need to check if sk->sk_saddr and
664 * sk->sk_rcv_addr are currently set to one of the addresses to
665 * be removed. This is something which needs to be looked into
666 * when we are fixing the outstanding issues with multi-homing
667 * socket routing and failover schemes. Refer to comments in
668 * sctp_do_bind(). -daisy
670 retval = sctp_del_bind_addr(bp, sa_addr);
672 addr_buf += af->sockaddr_len;
673 err_bindx_rem:
674 if (retval < 0) {
675 /* Failed. Add the ones that has been removed back */
676 if (cnt > 0)
677 sctp_bindx_add(sk, addrs, cnt);
678 return retval;
682 return retval;
685 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
686 * the associations that are part of the endpoint indicating that a list of
687 * local addresses are removed from the endpoint.
689 * If any of the addresses is already in the bind address list of the
690 * association, we do not send the chunk for that association. But it will not
691 * affect other associations.
693 * Only sctp_setsockopt_bindx() is supposed to call this function.
695 static int sctp_send_asconf_del_ip(struct sock *sk,
696 struct sockaddr *addrs,
697 int addrcnt)
699 struct sctp_sock *sp;
700 struct sctp_endpoint *ep;
701 struct sctp_association *asoc;
702 struct sctp_transport *transport;
703 struct sctp_bind_addr *bp;
704 struct sctp_chunk *chunk;
705 union sctp_addr *laddr;
706 void *addr_buf;
707 struct sctp_af *af;
708 struct sctp_sockaddr_entry *saddr;
709 int i;
710 int retval = 0;
712 if (!sctp_addip_enable)
713 return retval;
715 sp = sctp_sk(sk);
716 ep = sp->ep;
718 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
719 __func__, sk, addrs, addrcnt);
721 list_for_each_entry(asoc, &ep->asocs, asocs) {
723 if (!asoc->peer.asconf_capable)
724 continue;
726 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
727 continue;
729 if (!sctp_state(asoc, ESTABLISHED))
730 continue;
732 /* Check if any address in the packed array of addresses is
733 * not present in the bind address list of the association.
734 * If so, do not send the asconf chunk to its peer, but
735 * continue with other associations.
737 addr_buf = addrs;
738 for (i = 0; i < addrcnt; i++) {
739 laddr = (union sctp_addr *)addr_buf;
740 af = sctp_get_af_specific(laddr->v4.sin_family);
741 if (!af) {
742 retval = -EINVAL;
743 goto out;
746 if (!sctp_assoc_lookup_laddr(asoc, laddr))
747 break;
749 addr_buf += af->sockaddr_len;
751 if (i < addrcnt)
752 continue;
754 /* Find one address in the association's bind address list
755 * that is not in the packed array of addresses. This is to
756 * make sure that we do not delete all the addresses in the
757 * association.
759 bp = &asoc->base.bind_addr;
760 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
761 addrcnt, sp);
762 if (!laddr)
763 continue;
765 /* We do not need RCU protection throughout this loop
766 * because this is done under a socket lock from the
767 * setsockopt call.
769 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
770 SCTP_PARAM_DEL_IP);
771 if (!chunk) {
772 retval = -ENOMEM;
773 goto out;
776 /* Reset use_as_src flag for the addresses in the bind address
777 * list that are to be deleted.
779 addr_buf = addrs;
780 for (i = 0; i < addrcnt; i++) {
781 laddr = (union sctp_addr *)addr_buf;
782 af = sctp_get_af_specific(laddr->v4.sin_family);
783 list_for_each_entry(saddr, &bp->address_list, list) {
784 if (sctp_cmp_addr_exact(&saddr->a, laddr))
785 saddr->state = SCTP_ADDR_DEL;
787 addr_buf += af->sockaddr_len;
790 /* Update the route and saddr entries for all the transports
791 * as some of the addresses in the bind address list are
792 * about to be deleted and cannot be used as source addresses.
794 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
795 transports) {
796 dst_release(transport->dst);
797 sctp_transport_route(transport, NULL,
798 sctp_sk(asoc->base.sk));
801 retval = sctp_send_asconf(asoc, chunk);
803 out:
804 return retval;
807 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
809 * API 8.1
810 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
811 * int flags);
813 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
814 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
815 * or IPv6 addresses.
817 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
818 * Section 3.1.2 for this usage.
820 * addrs is a pointer to an array of one or more socket addresses. Each
821 * address is contained in its appropriate structure (i.e. struct
822 * sockaddr_in or struct sockaddr_in6) the family of the address type
823 * must be used to distinguish the address length (note that this
824 * representation is termed a "packed array" of addresses). The caller
825 * specifies the number of addresses in the array with addrcnt.
827 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
828 * -1, and sets errno to the appropriate error code.
830 * For SCTP, the port given in each socket address must be the same, or
831 * sctp_bindx() will fail, setting errno to EINVAL.
833 * The flags parameter is formed from the bitwise OR of zero or more of
834 * the following currently defined flags:
836 * SCTP_BINDX_ADD_ADDR
838 * SCTP_BINDX_REM_ADDR
840 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
841 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
842 * addresses from the association. The two flags are mutually exclusive;
843 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
844 * not remove all addresses from an association; sctp_bindx() will
845 * reject such an attempt with EINVAL.
847 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
848 * additional addresses with an endpoint after calling bind(). Or use
849 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
850 * socket is associated with so that no new association accepted will be
851 * associated with those addresses. If the endpoint supports dynamic
852 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
853 * endpoint to send the appropriate message to the peer to change the
854 * peers address lists.
856 * Adding and removing addresses from a connected association is
857 * optional functionality. Implementations that do not support this
858 * functionality should return EOPNOTSUPP.
860 * Basically do nothing but copying the addresses from user to kernel
861 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
862 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
863 * from userspace.
865 * We don't use copy_from_user() for optimization: we first do the
866 * sanity checks (buffer size -fast- and access check-healthy
867 * pointer); if all of those succeed, then we can alloc the memory
868 * (expensive operation) needed to copy the data to kernel. Then we do
869 * the copying without checking the user space area
870 * (__copy_from_user()).
872 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
873 * it.
875 * sk The sk of the socket
876 * addrs The pointer to the addresses in user land
877 * addrssize Size of the addrs buffer
878 * op Operation to perform (add or remove, see the flags of
879 * sctp_bindx)
881 * Returns 0 if ok, <0 errno code on error.
883 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
884 struct sockaddr __user *addrs,
885 int addrs_size, int op)
887 struct sockaddr *kaddrs;
888 int err;
889 int addrcnt = 0;
890 int walk_size = 0;
891 struct sockaddr *sa_addr;
892 void *addr_buf;
893 struct sctp_af *af;
895 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
896 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
898 if (unlikely(addrs_size <= 0))
899 return -EINVAL;
901 /* Check the user passed a healthy pointer. */
902 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
903 return -EFAULT;
905 /* Alloc space for the address array in kernel memory. */
906 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
907 if (unlikely(!kaddrs))
908 return -ENOMEM;
910 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
911 kfree(kaddrs);
912 return -EFAULT;
915 /* Walk through the addrs buffer and count the number of addresses. */
916 addr_buf = kaddrs;
917 while (walk_size < addrs_size) {
918 sa_addr = (struct sockaddr *)addr_buf;
919 af = sctp_get_af_specific(sa_addr->sa_family);
921 /* If the address family is not supported or if this address
922 * causes the address buffer to overflow return EINVAL.
924 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
925 kfree(kaddrs);
926 return -EINVAL;
928 addrcnt++;
929 addr_buf += af->sockaddr_len;
930 walk_size += af->sockaddr_len;
933 /* Do the work. */
934 switch (op) {
935 case SCTP_BINDX_ADD_ADDR:
936 err = sctp_bindx_add(sk, kaddrs, addrcnt);
937 if (err)
938 goto out;
939 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
940 break;
942 case SCTP_BINDX_REM_ADDR:
943 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
944 if (err)
945 goto out;
946 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
947 break;
949 default:
950 err = -EINVAL;
951 break;
954 out:
955 kfree(kaddrs);
957 return err;
960 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
962 * Common routine for handling connect() and sctp_connectx().
963 * Connect will come in with just a single address.
965 static int __sctp_connect(struct sock* sk,
966 struct sockaddr *kaddrs,
967 int addrs_size,
968 sctp_assoc_t *assoc_id)
970 struct sctp_sock *sp;
971 struct sctp_endpoint *ep;
972 struct sctp_association *asoc = NULL;
973 struct sctp_association *asoc2;
974 struct sctp_transport *transport;
975 union sctp_addr to;
976 struct sctp_af *af;
977 sctp_scope_t scope;
978 long timeo;
979 int err = 0;
980 int addrcnt = 0;
981 int walk_size = 0;
982 union sctp_addr *sa_addr = NULL;
983 void *addr_buf;
984 unsigned short port;
985 unsigned int f_flags = 0;
987 sp = sctp_sk(sk);
988 ep = sp->ep;
990 /* connect() cannot be done on a socket that is already in ESTABLISHED
991 * state - UDP-style peeled off socket or a TCP-style socket that
992 * is already connected.
993 * It cannot be done even on a TCP-style listening socket.
995 if (sctp_sstate(sk, ESTABLISHED) ||
996 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
997 err = -EISCONN;
998 goto out_free;
1001 /* Walk through the addrs buffer and count the number of addresses. */
1002 addr_buf = kaddrs;
1003 while (walk_size < addrs_size) {
1004 sa_addr = (union sctp_addr *)addr_buf;
1005 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1006 port = ntohs(sa_addr->v4.sin_port);
1008 /* If the address family is not supported or if this address
1009 * causes the address buffer to overflow return EINVAL.
1011 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1012 err = -EINVAL;
1013 goto out_free;
1016 /* Save current address so we can work with it */
1017 memcpy(&to, sa_addr, af->sockaddr_len);
1019 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1020 if (err)
1021 goto out_free;
1023 /* Make sure the destination port is correctly set
1024 * in all addresses.
1026 if (asoc && asoc->peer.port && asoc->peer.port != port)
1027 goto out_free;
1030 /* Check if there already is a matching association on the
1031 * endpoint (other than the one created here).
1033 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1034 if (asoc2 && asoc2 != asoc) {
1035 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1036 err = -EISCONN;
1037 else
1038 err = -EALREADY;
1039 goto out_free;
1042 /* If we could not find a matching association on the endpoint,
1043 * make sure that there is no peeled-off association matching
1044 * the peer address even on another socket.
1046 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1047 err = -EADDRNOTAVAIL;
1048 goto out_free;
1051 if (!asoc) {
1052 /* If a bind() or sctp_bindx() is not called prior to
1053 * an sctp_connectx() call, the system picks an
1054 * ephemeral port and will choose an address set
1055 * equivalent to binding with a wildcard address.
1057 if (!ep->base.bind_addr.port) {
1058 if (sctp_autobind(sk)) {
1059 err = -EAGAIN;
1060 goto out_free;
1062 } else {
1064 * If an unprivileged user inherits a 1-many
1065 * style socket with open associations on a
1066 * privileged port, it MAY be permitted to
1067 * accept new associations, but it SHOULD NOT
1068 * be permitted to open new associations.
1070 if (ep->base.bind_addr.port < PROT_SOCK &&
1071 !capable(CAP_NET_BIND_SERVICE)) {
1072 err = -EACCES;
1073 goto out_free;
1077 scope = sctp_scope(&to);
1078 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1079 if (!asoc) {
1080 err = -ENOMEM;
1081 goto out_free;
1085 /* Prime the peer's transport structures. */
1086 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1087 SCTP_UNKNOWN);
1088 if (!transport) {
1089 err = -ENOMEM;
1090 goto out_free;
1093 addrcnt++;
1094 addr_buf += af->sockaddr_len;
1095 walk_size += af->sockaddr_len;
1098 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1099 if (err < 0) {
1100 goto out_free;
1103 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1104 if (err < 0) {
1105 goto out_free;
1108 /* Initialize sk's dport and daddr for getpeername() */
1109 inet_sk(sk)->dport = htons(asoc->peer.port);
1110 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1111 af->to_sk_daddr(sa_addr, sk);
1112 sk->sk_err = 0;
1114 /* in-kernel sockets don't generally have a file allocated to them
1115 * if all they do is call sock_create_kern().
1117 if (sk->sk_socket->file)
1118 f_flags = sk->sk_socket->file->f_flags;
1120 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1122 err = sctp_wait_for_connect(asoc, &timeo);
1123 if (!err && assoc_id)
1124 *assoc_id = asoc->assoc_id;
1126 /* Don't free association on exit. */
1127 asoc = NULL;
1129 out_free:
1131 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1132 " kaddrs: %p err: %d\n",
1133 asoc, kaddrs, err);
1134 if (asoc)
1135 sctp_association_free(asoc);
1136 return err;
1139 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1141 * API 8.9
1142 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1143 * sctp_assoc_t *asoc);
1145 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1146 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1147 * or IPv6 addresses.
1149 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1150 * Section 3.1.2 for this usage.
1152 * addrs is a pointer to an array of one or more socket addresses. Each
1153 * address is contained in its appropriate structure (i.e. struct
1154 * sockaddr_in or struct sockaddr_in6) the family of the address type
1155 * must be used to distengish the address length (note that this
1156 * representation is termed a "packed array" of addresses). The caller
1157 * specifies the number of addresses in the array with addrcnt.
1159 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1160 * the association id of the new association. On failure, sctp_connectx()
1161 * returns -1, and sets errno to the appropriate error code. The assoc_id
1162 * is not touched by the kernel.
1164 * For SCTP, the port given in each socket address must be the same, or
1165 * sctp_connectx() will fail, setting errno to EINVAL.
1167 * An application can use sctp_connectx to initiate an association with
1168 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1169 * allows a caller to specify multiple addresses at which a peer can be
1170 * reached. The way the SCTP stack uses the list of addresses to set up
1171 * the association is implementation dependant. This function only
1172 * specifies that the stack will try to make use of all the addresses in
1173 * the list when needed.
1175 * Note that the list of addresses passed in is only used for setting up
1176 * the association. It does not necessarily equal the set of addresses
1177 * the peer uses for the resulting association. If the caller wants to
1178 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1179 * retrieve them after the association has been set up.
1181 * Basically do nothing but copying the addresses from user to kernel
1182 * land and invoking either sctp_connectx(). This is used for tunneling
1183 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1185 * We don't use copy_from_user() for optimization: we first do the
1186 * sanity checks (buffer size -fast- and access check-healthy
1187 * pointer); if all of those succeed, then we can alloc the memory
1188 * (expensive operation) needed to copy the data to kernel. Then we do
1189 * the copying without checking the user space area
1190 * (__copy_from_user()).
1192 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1193 * it.
1195 * sk The sk of the socket
1196 * addrs The pointer to the addresses in user land
1197 * addrssize Size of the addrs buffer
1199 * Returns >=0 if ok, <0 errno code on error.
1201 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1202 struct sockaddr __user *addrs,
1203 int addrs_size,
1204 sctp_assoc_t *assoc_id)
1206 int err = 0;
1207 struct sockaddr *kaddrs;
1209 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1210 __func__, sk, addrs, addrs_size);
1212 if (unlikely(addrs_size <= 0))
1213 return -EINVAL;
1215 /* Check the user passed a healthy pointer. */
1216 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1217 return -EFAULT;
1219 /* Alloc space for the address array in kernel memory. */
1220 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1221 if (unlikely(!kaddrs))
1222 return -ENOMEM;
1224 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1225 err = -EFAULT;
1226 } else {
1227 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1230 kfree(kaddrs);
1232 return err;
1236 * This is an older interface. It's kept for backward compatibility
1237 * to the option that doesn't provide association id.
1239 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1240 struct sockaddr __user *addrs,
1241 int addrs_size)
1243 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1247 * New interface for the API. The since the API is done with a socket
1248 * option, to make it simple we feed back the association id is as a return
1249 * indication to the call. Error is always negative and association id is
1250 * always positive.
1252 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1253 struct sockaddr __user *addrs,
1254 int addrs_size)
1256 sctp_assoc_t assoc_id = 0;
1257 int err = 0;
1259 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1261 if (err)
1262 return err;
1263 else
1264 return assoc_id;
1267 /* API 3.1.4 close() - UDP Style Syntax
1268 * Applications use close() to perform graceful shutdown (as described in
1269 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1270 * by a UDP-style socket.
1272 * The syntax is
1274 * ret = close(int sd);
1276 * sd - the socket descriptor of the associations to be closed.
1278 * To gracefully shutdown a specific association represented by the
1279 * UDP-style socket, an application should use the sendmsg() call,
1280 * passing no user data, but including the appropriate flag in the
1281 * ancillary data (see Section xxxx).
1283 * If sd in the close() call is a branched-off socket representing only
1284 * one association, the shutdown is performed on that association only.
1286 * 4.1.6 close() - TCP Style Syntax
1288 * Applications use close() to gracefully close down an association.
1290 * The syntax is:
1292 * int close(int sd);
1294 * sd - the socket descriptor of the association to be closed.
1296 * After an application calls close() on a socket descriptor, no further
1297 * socket operations will succeed on that descriptor.
1299 * API 7.1.4 SO_LINGER
1301 * An application using the TCP-style socket can use this option to
1302 * perform the SCTP ABORT primitive. The linger option structure is:
1304 * struct linger {
1305 * int l_onoff; // option on/off
1306 * int l_linger; // linger time
1307 * };
1309 * To enable the option, set l_onoff to 1. If the l_linger value is set
1310 * to 0, calling close() is the same as the ABORT primitive. If the
1311 * value is set to a negative value, the setsockopt() call will return
1312 * an error. If the value is set to a positive value linger_time, the
1313 * close() can be blocked for at most linger_time ms. If the graceful
1314 * shutdown phase does not finish during this period, close() will
1315 * return but the graceful shutdown phase continues in the system.
1317 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1319 struct sctp_endpoint *ep;
1320 struct sctp_association *asoc;
1321 struct list_head *pos, *temp;
1323 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1325 sctp_lock_sock(sk);
1326 sk->sk_shutdown = SHUTDOWN_MASK;
1328 ep = sctp_sk(sk)->ep;
1330 /* Walk all associations on an endpoint. */
1331 list_for_each_safe(pos, temp, &ep->asocs) {
1332 asoc = list_entry(pos, struct sctp_association, asocs);
1334 if (sctp_style(sk, TCP)) {
1335 /* A closed association can still be in the list if
1336 * it belongs to a TCP-style listening socket that is
1337 * not yet accepted. If so, free it. If not, send an
1338 * ABORT or SHUTDOWN based on the linger options.
1340 if (sctp_state(asoc, CLOSED)) {
1341 sctp_unhash_established(asoc);
1342 sctp_association_free(asoc);
1343 continue;
1347 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1348 struct sctp_chunk *chunk;
1350 chunk = sctp_make_abort_user(asoc, NULL, 0);
1351 if (chunk)
1352 sctp_primitive_ABORT(asoc, chunk);
1353 } else
1354 sctp_primitive_SHUTDOWN(asoc, NULL);
1357 /* Clean up any skbs sitting on the receive queue. */
1358 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1359 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1361 /* On a TCP-style socket, block for at most linger_time if set. */
1362 if (sctp_style(sk, TCP) && timeout)
1363 sctp_wait_for_close(sk, timeout);
1365 /* This will run the backlog queue. */
1366 sctp_release_sock(sk);
1368 /* Supposedly, no process has access to the socket, but
1369 * the net layers still may.
1371 sctp_local_bh_disable();
1372 sctp_bh_lock_sock(sk);
1374 /* Hold the sock, since sk_common_release() will put sock_put()
1375 * and we have just a little more cleanup.
1377 sock_hold(sk);
1378 sk_common_release(sk);
1380 sctp_bh_unlock_sock(sk);
1381 sctp_local_bh_enable();
1383 sock_put(sk);
1385 SCTP_DBG_OBJCNT_DEC(sock);
1388 /* Handle EPIPE error. */
1389 static int sctp_error(struct sock *sk, int flags, int err)
1391 if (err == -EPIPE)
1392 err = sock_error(sk) ? : -EPIPE;
1393 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1394 send_sig(SIGPIPE, current, 0);
1395 return err;
1398 /* API 3.1.3 sendmsg() - UDP Style Syntax
1400 * An application uses sendmsg() and recvmsg() calls to transmit data to
1401 * and receive data from its peer.
1403 * ssize_t sendmsg(int socket, const struct msghdr *message,
1404 * int flags);
1406 * socket - the socket descriptor of the endpoint.
1407 * message - pointer to the msghdr structure which contains a single
1408 * user message and possibly some ancillary data.
1410 * See Section 5 for complete description of the data
1411 * structures.
1413 * flags - flags sent or received with the user message, see Section
1414 * 5 for complete description of the flags.
1416 * Note: This function could use a rewrite especially when explicit
1417 * connect support comes in.
1419 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1421 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1423 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1424 struct msghdr *msg, size_t msg_len)
1426 struct sctp_sock *sp;
1427 struct sctp_endpoint *ep;
1428 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1429 struct sctp_transport *transport, *chunk_tp;
1430 struct sctp_chunk *chunk;
1431 union sctp_addr to;
1432 struct sockaddr *msg_name = NULL;
1433 struct sctp_sndrcvinfo default_sinfo = { 0 };
1434 struct sctp_sndrcvinfo *sinfo;
1435 struct sctp_initmsg *sinit;
1436 sctp_assoc_t associd = 0;
1437 sctp_cmsgs_t cmsgs = { NULL };
1438 int err;
1439 sctp_scope_t scope;
1440 long timeo;
1441 __u16 sinfo_flags = 0;
1442 struct sctp_datamsg *datamsg;
1443 int msg_flags = msg->msg_flags;
1445 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1446 sk, msg, msg_len);
1448 err = 0;
1449 sp = sctp_sk(sk);
1450 ep = sp->ep;
1452 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1454 /* We cannot send a message over a TCP-style listening socket. */
1455 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1456 err = -EPIPE;
1457 goto out_nounlock;
1460 /* Parse out the SCTP CMSGs. */
1461 err = sctp_msghdr_parse(msg, &cmsgs);
1463 if (err) {
1464 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1465 goto out_nounlock;
1468 /* Fetch the destination address for this packet. This
1469 * address only selects the association--it is not necessarily
1470 * the address we will send to.
1471 * For a peeled-off socket, msg_name is ignored.
1473 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1474 int msg_namelen = msg->msg_namelen;
1476 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1477 msg_namelen);
1478 if (err)
1479 return err;
1481 if (msg_namelen > sizeof(to))
1482 msg_namelen = sizeof(to);
1483 memcpy(&to, msg->msg_name, msg_namelen);
1484 msg_name = msg->msg_name;
1487 sinfo = cmsgs.info;
1488 sinit = cmsgs.init;
1490 /* Did the user specify SNDRCVINFO? */
1491 if (sinfo) {
1492 sinfo_flags = sinfo->sinfo_flags;
1493 associd = sinfo->sinfo_assoc_id;
1496 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1497 msg_len, sinfo_flags);
1499 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1500 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1501 err = -EINVAL;
1502 goto out_nounlock;
1505 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1506 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1507 * If SCTP_ABORT is set, the message length could be non zero with
1508 * the msg_iov set to the user abort reason.
1510 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1511 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1512 err = -EINVAL;
1513 goto out_nounlock;
1516 /* If SCTP_ADDR_OVER is set, there must be an address
1517 * specified in msg_name.
1519 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1520 err = -EINVAL;
1521 goto out_nounlock;
1524 transport = NULL;
1526 SCTP_DEBUG_PRINTK("About to look up association.\n");
1528 sctp_lock_sock(sk);
1530 /* If a msg_name has been specified, assume this is to be used. */
1531 if (msg_name) {
1532 /* Look for a matching association on the endpoint. */
1533 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1534 if (!asoc) {
1535 /* If we could not find a matching association on the
1536 * endpoint, make sure that it is not a TCP-style
1537 * socket that already has an association or there is
1538 * no peeled-off association on another socket.
1540 if ((sctp_style(sk, TCP) &&
1541 sctp_sstate(sk, ESTABLISHED)) ||
1542 sctp_endpoint_is_peeled_off(ep, &to)) {
1543 err = -EADDRNOTAVAIL;
1544 goto out_unlock;
1547 } else {
1548 asoc = sctp_id2assoc(sk, associd);
1549 if (!asoc) {
1550 err = -EPIPE;
1551 goto out_unlock;
1555 if (asoc) {
1556 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1558 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1559 * socket that has an association in CLOSED state. This can
1560 * happen when an accepted socket has an association that is
1561 * already CLOSED.
1563 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1564 err = -EPIPE;
1565 goto out_unlock;
1568 if (sinfo_flags & SCTP_EOF) {
1569 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1570 asoc);
1571 sctp_primitive_SHUTDOWN(asoc, NULL);
1572 err = 0;
1573 goto out_unlock;
1575 if (sinfo_flags & SCTP_ABORT) {
1577 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1578 if (!chunk) {
1579 err = -ENOMEM;
1580 goto out_unlock;
1583 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1584 sctp_primitive_ABORT(asoc, chunk);
1585 err = 0;
1586 goto out_unlock;
1590 /* Do we need to create the association? */
1591 if (!asoc) {
1592 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1594 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1595 err = -EINVAL;
1596 goto out_unlock;
1599 /* Check for invalid stream against the stream counts,
1600 * either the default or the user specified stream counts.
1602 if (sinfo) {
1603 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1604 /* Check against the defaults. */
1605 if (sinfo->sinfo_stream >=
1606 sp->initmsg.sinit_num_ostreams) {
1607 err = -EINVAL;
1608 goto out_unlock;
1610 } else {
1611 /* Check against the requested. */
1612 if (sinfo->sinfo_stream >=
1613 sinit->sinit_num_ostreams) {
1614 err = -EINVAL;
1615 goto out_unlock;
1621 * API 3.1.2 bind() - UDP Style Syntax
1622 * If a bind() or sctp_bindx() is not called prior to a
1623 * sendmsg() call that initiates a new association, the
1624 * system picks an ephemeral port and will choose an address
1625 * set equivalent to binding with a wildcard address.
1627 if (!ep->base.bind_addr.port) {
1628 if (sctp_autobind(sk)) {
1629 err = -EAGAIN;
1630 goto out_unlock;
1632 } else {
1634 * If an unprivileged user inherits a one-to-many
1635 * style socket with open associations on a privileged
1636 * port, it MAY be permitted to accept new associations,
1637 * but it SHOULD NOT be permitted to open new
1638 * associations.
1640 if (ep->base.bind_addr.port < PROT_SOCK &&
1641 !capable(CAP_NET_BIND_SERVICE)) {
1642 err = -EACCES;
1643 goto out_unlock;
1647 scope = sctp_scope(&to);
1648 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1649 if (!new_asoc) {
1650 err = -ENOMEM;
1651 goto out_unlock;
1653 asoc = new_asoc;
1655 /* If the SCTP_INIT ancillary data is specified, set all
1656 * the association init values accordingly.
1658 if (sinit) {
1659 if (sinit->sinit_num_ostreams) {
1660 asoc->c.sinit_num_ostreams =
1661 sinit->sinit_num_ostreams;
1663 if (sinit->sinit_max_instreams) {
1664 asoc->c.sinit_max_instreams =
1665 sinit->sinit_max_instreams;
1667 if (sinit->sinit_max_attempts) {
1668 asoc->max_init_attempts
1669 = sinit->sinit_max_attempts;
1671 if (sinit->sinit_max_init_timeo) {
1672 asoc->max_init_timeo =
1673 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1677 /* Prime the peer's transport structures. */
1678 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1679 if (!transport) {
1680 err = -ENOMEM;
1681 goto out_free;
1683 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1684 if (err < 0) {
1685 err = -ENOMEM;
1686 goto out_free;
1690 /* ASSERT: we have a valid association at this point. */
1691 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1693 if (!sinfo) {
1694 /* If the user didn't specify SNDRCVINFO, make up one with
1695 * some defaults.
1697 default_sinfo.sinfo_stream = asoc->default_stream;
1698 default_sinfo.sinfo_flags = asoc->default_flags;
1699 default_sinfo.sinfo_ppid = asoc->default_ppid;
1700 default_sinfo.sinfo_context = asoc->default_context;
1701 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1702 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1703 sinfo = &default_sinfo;
1706 /* API 7.1.7, the sndbuf size per association bounds the
1707 * maximum size of data that can be sent in a single send call.
1709 if (msg_len > sk->sk_sndbuf) {
1710 err = -EMSGSIZE;
1711 goto out_free;
1714 if (asoc->pmtu_pending)
1715 sctp_assoc_pending_pmtu(asoc);
1717 /* If fragmentation is disabled and the message length exceeds the
1718 * association fragmentation point, return EMSGSIZE. The I-D
1719 * does not specify what this error is, but this looks like
1720 * a great fit.
1722 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1723 err = -EMSGSIZE;
1724 goto out_free;
1727 if (sinfo) {
1728 /* Check for invalid stream. */
1729 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1730 err = -EINVAL;
1731 goto out_free;
1735 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1736 if (!sctp_wspace(asoc)) {
1737 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1738 if (err)
1739 goto out_free;
1742 /* If an address is passed with the sendto/sendmsg call, it is used
1743 * to override the primary destination address in the TCP model, or
1744 * when SCTP_ADDR_OVER flag is set in the UDP model.
1746 if ((sctp_style(sk, TCP) && msg_name) ||
1747 (sinfo_flags & SCTP_ADDR_OVER)) {
1748 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1749 if (!chunk_tp) {
1750 err = -EINVAL;
1751 goto out_free;
1753 } else
1754 chunk_tp = NULL;
1756 /* Auto-connect, if we aren't connected already. */
1757 if (sctp_state(asoc, CLOSED)) {
1758 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1759 if (err < 0)
1760 goto out_free;
1761 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1764 /* Break the message into multiple chunks of maximum size. */
1765 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1766 if (!datamsg) {
1767 err = -ENOMEM;
1768 goto out_free;
1771 /* Now send the (possibly) fragmented message. */
1772 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1773 sctp_chunk_hold(chunk);
1775 /* Do accounting for the write space. */
1776 sctp_set_owner_w(chunk);
1778 chunk->transport = chunk_tp;
1780 /* Send it to the lower layers. Note: all chunks
1781 * must either fail or succeed. The lower layer
1782 * works that way today. Keep it that way or this
1783 * breaks.
1785 err = sctp_primitive_SEND(asoc, chunk);
1786 /* Did the lower layer accept the chunk? */
1787 if (err)
1788 sctp_chunk_free(chunk);
1789 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1792 sctp_datamsg_put(datamsg);
1793 if (err)
1794 goto out_free;
1795 else
1796 err = msg_len;
1798 /* If we are already past ASSOCIATE, the lower
1799 * layers are responsible for association cleanup.
1801 goto out_unlock;
1803 out_free:
1804 if (new_asoc)
1805 sctp_association_free(asoc);
1806 out_unlock:
1807 sctp_release_sock(sk);
1809 out_nounlock:
1810 return sctp_error(sk, msg_flags, err);
1812 #if 0
1813 do_sock_err:
1814 if (msg_len)
1815 err = msg_len;
1816 else
1817 err = sock_error(sk);
1818 goto out;
1820 do_interrupted:
1821 if (msg_len)
1822 err = msg_len;
1823 goto out;
1824 #endif /* 0 */
1827 /* This is an extended version of skb_pull() that removes the data from the
1828 * start of a skb even when data is spread across the list of skb's in the
1829 * frag_list. len specifies the total amount of data that needs to be removed.
1830 * when 'len' bytes could be removed from the skb, it returns 0.
1831 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1832 * could not be removed.
1834 static int sctp_skb_pull(struct sk_buff *skb, int len)
1836 struct sk_buff *list;
1837 int skb_len = skb_headlen(skb);
1838 int rlen;
1840 if (len <= skb_len) {
1841 __skb_pull(skb, len);
1842 return 0;
1844 len -= skb_len;
1845 __skb_pull(skb, skb_len);
1847 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1848 rlen = sctp_skb_pull(list, len);
1849 skb->len -= (len-rlen);
1850 skb->data_len -= (len-rlen);
1852 if (!rlen)
1853 return 0;
1855 len = rlen;
1858 return len;
1861 /* API 3.1.3 recvmsg() - UDP Style Syntax
1863 * ssize_t recvmsg(int socket, struct msghdr *message,
1864 * int flags);
1866 * socket - the socket descriptor of the endpoint.
1867 * message - pointer to the msghdr structure which contains a single
1868 * user message and possibly some ancillary data.
1870 * See Section 5 for complete description of the data
1871 * structures.
1873 * flags - flags sent or received with the user message, see Section
1874 * 5 for complete description of the flags.
1876 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1878 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1879 struct msghdr *msg, size_t len, int noblock,
1880 int flags, int *addr_len)
1882 struct sctp_ulpevent *event = NULL;
1883 struct sctp_sock *sp = sctp_sk(sk);
1884 struct sk_buff *skb;
1885 int copied;
1886 int err = 0;
1887 int skb_len;
1889 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1890 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1891 "len", len, "knoblauch", noblock,
1892 "flags", flags, "addr_len", addr_len);
1894 sctp_lock_sock(sk);
1896 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1897 err = -ENOTCONN;
1898 goto out;
1901 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1902 if (!skb)
1903 goto out;
1905 /* Get the total length of the skb including any skb's in the
1906 * frag_list.
1908 skb_len = skb->len;
1910 copied = skb_len;
1911 if (copied > len)
1912 copied = len;
1914 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1916 event = sctp_skb2event(skb);
1918 if (err)
1919 goto out_free;
1921 sock_recv_timestamp(msg, sk, skb);
1922 if (sctp_ulpevent_is_notification(event)) {
1923 msg->msg_flags |= MSG_NOTIFICATION;
1924 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1925 } else {
1926 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1929 /* Check if we allow SCTP_SNDRCVINFO. */
1930 if (sp->subscribe.sctp_data_io_event)
1931 sctp_ulpevent_read_sndrcvinfo(event, msg);
1932 #if 0
1933 /* FIXME: we should be calling IP/IPv6 layers. */
1934 if (sk->sk_protinfo.af_inet.cmsg_flags)
1935 ip_cmsg_recv(msg, skb);
1936 #endif
1938 err = copied;
1940 /* If skb's length exceeds the user's buffer, update the skb and
1941 * push it back to the receive_queue so that the next call to
1942 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1944 if (skb_len > copied) {
1945 msg->msg_flags &= ~MSG_EOR;
1946 if (flags & MSG_PEEK)
1947 goto out_free;
1948 sctp_skb_pull(skb, copied);
1949 skb_queue_head(&sk->sk_receive_queue, skb);
1951 /* When only partial message is copied to the user, increase
1952 * rwnd by that amount. If all the data in the skb is read,
1953 * rwnd is updated when the event is freed.
1955 if (!sctp_ulpevent_is_notification(event))
1956 sctp_assoc_rwnd_increase(event->asoc, copied);
1957 goto out;
1958 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1959 (event->msg_flags & MSG_EOR))
1960 msg->msg_flags |= MSG_EOR;
1961 else
1962 msg->msg_flags &= ~MSG_EOR;
1964 out_free:
1965 if (flags & MSG_PEEK) {
1966 /* Release the skb reference acquired after peeking the skb in
1967 * sctp_skb_recv_datagram().
1969 kfree_skb(skb);
1970 } else {
1971 /* Free the event which includes releasing the reference to
1972 * the owner of the skb, freeing the skb and updating the
1973 * rwnd.
1975 sctp_ulpevent_free(event);
1977 out:
1978 sctp_release_sock(sk);
1979 return err;
1982 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1984 * This option is a on/off flag. If enabled no SCTP message
1985 * fragmentation will be performed. Instead if a message being sent
1986 * exceeds the current PMTU size, the message will NOT be sent and
1987 * instead a error will be indicated to the user.
1989 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1990 char __user *optval, int optlen)
1992 int val;
1994 if (optlen < sizeof(int))
1995 return -EINVAL;
1997 if (get_user(val, (int __user *)optval))
1998 return -EFAULT;
2000 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2002 return 0;
2005 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2006 int optlen)
2008 if (optlen > sizeof(struct sctp_event_subscribe))
2009 return -EINVAL;
2010 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2011 return -EFAULT;
2012 return 0;
2015 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2017 * This socket option is applicable to the UDP-style socket only. When
2018 * set it will cause associations that are idle for more than the
2019 * specified number of seconds to automatically close. An association
2020 * being idle is defined an association that has NOT sent or received
2021 * user data. The special value of '0' indicates that no automatic
2022 * close of any associations should be performed. The option expects an
2023 * integer defining the number of seconds of idle time before an
2024 * association is closed.
2026 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2027 int optlen)
2029 struct sctp_sock *sp = sctp_sk(sk);
2031 /* Applicable to UDP-style socket only */
2032 if (sctp_style(sk, TCP))
2033 return -EOPNOTSUPP;
2034 if (optlen != sizeof(int))
2035 return -EINVAL;
2036 if (copy_from_user(&sp->autoclose, optval, optlen))
2037 return -EFAULT;
2039 return 0;
2042 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2044 * Applications can enable or disable heartbeats for any peer address of
2045 * an association, modify an address's heartbeat interval, force a
2046 * heartbeat to be sent immediately, and adjust the address's maximum
2047 * number of retransmissions sent before an address is considered
2048 * unreachable. The following structure is used to access and modify an
2049 * address's parameters:
2051 * struct sctp_paddrparams {
2052 * sctp_assoc_t spp_assoc_id;
2053 * struct sockaddr_storage spp_address;
2054 * uint32_t spp_hbinterval;
2055 * uint16_t spp_pathmaxrxt;
2056 * uint32_t spp_pathmtu;
2057 * uint32_t spp_sackdelay;
2058 * uint32_t spp_flags;
2059 * };
2061 * spp_assoc_id - (one-to-many style socket) This is filled in the
2062 * application, and identifies the association for
2063 * this query.
2064 * spp_address - This specifies which address is of interest.
2065 * spp_hbinterval - This contains the value of the heartbeat interval,
2066 * in milliseconds. If a value of zero
2067 * is present in this field then no changes are to
2068 * be made to this parameter.
2069 * spp_pathmaxrxt - This contains the maximum number of
2070 * retransmissions before this address shall be
2071 * considered unreachable. If a value of zero
2072 * is present in this field then no changes are to
2073 * be made to this parameter.
2074 * spp_pathmtu - When Path MTU discovery is disabled the value
2075 * specified here will be the "fixed" path mtu.
2076 * Note that if the spp_address field is empty
2077 * then all associations on this address will
2078 * have this fixed path mtu set upon them.
2080 * spp_sackdelay - When delayed sack is enabled, this value specifies
2081 * the number of milliseconds that sacks will be delayed
2082 * for. This value will apply to all addresses of an
2083 * association if the spp_address field is empty. Note
2084 * also, that if delayed sack is enabled and this
2085 * value is set to 0, no change is made to the last
2086 * recorded delayed sack timer value.
2088 * spp_flags - These flags are used to control various features
2089 * on an association. The flag field may contain
2090 * zero or more of the following options.
2092 * SPP_HB_ENABLE - Enable heartbeats on the
2093 * specified address. Note that if the address
2094 * field is empty all addresses for the association
2095 * have heartbeats enabled upon them.
2097 * SPP_HB_DISABLE - Disable heartbeats on the
2098 * speicifed address. Note that if the address
2099 * field is empty all addresses for the association
2100 * will have their heartbeats disabled. Note also
2101 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2102 * mutually exclusive, only one of these two should
2103 * be specified. Enabling both fields will have
2104 * undetermined results.
2106 * SPP_HB_DEMAND - Request a user initiated heartbeat
2107 * to be made immediately.
2109 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2110 * heartbeat delayis to be set to the value of 0
2111 * milliseconds.
2113 * SPP_PMTUD_ENABLE - This field will enable PMTU
2114 * discovery upon the specified address. Note that
2115 * if the address feild is empty then all addresses
2116 * on the association are effected.
2118 * SPP_PMTUD_DISABLE - This field will disable PMTU
2119 * discovery upon the specified address. Note that
2120 * if the address feild is empty then all addresses
2121 * on the association are effected. Not also that
2122 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2123 * exclusive. Enabling both will have undetermined
2124 * results.
2126 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2127 * on delayed sack. The time specified in spp_sackdelay
2128 * is used to specify the sack delay for this address. Note
2129 * that if spp_address is empty then all addresses will
2130 * enable delayed sack and take on the sack delay
2131 * value specified in spp_sackdelay.
2132 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2133 * off delayed sack. If the spp_address field is blank then
2134 * delayed sack is disabled for the entire association. Note
2135 * also that this field is mutually exclusive to
2136 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2137 * results.
2139 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2140 struct sctp_transport *trans,
2141 struct sctp_association *asoc,
2142 struct sctp_sock *sp,
2143 int hb_change,
2144 int pmtud_change,
2145 int sackdelay_change)
2147 int error;
2149 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2150 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2151 if (error)
2152 return error;
2155 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2156 * this field is ignored. Note also that a value of zero indicates
2157 * the current setting should be left unchanged.
2159 if (params->spp_flags & SPP_HB_ENABLE) {
2161 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2162 * set. This lets us use 0 value when this flag
2163 * is set.
2165 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2166 params->spp_hbinterval = 0;
2168 if (params->spp_hbinterval ||
2169 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2170 if (trans) {
2171 trans->hbinterval =
2172 msecs_to_jiffies(params->spp_hbinterval);
2173 } else if (asoc) {
2174 asoc->hbinterval =
2175 msecs_to_jiffies(params->spp_hbinterval);
2176 } else {
2177 sp->hbinterval = params->spp_hbinterval;
2182 if (hb_change) {
2183 if (trans) {
2184 trans->param_flags =
2185 (trans->param_flags & ~SPP_HB) | hb_change;
2186 } else if (asoc) {
2187 asoc->param_flags =
2188 (asoc->param_flags & ~SPP_HB) | hb_change;
2189 } else {
2190 sp->param_flags =
2191 (sp->param_flags & ~SPP_HB) | hb_change;
2195 /* When Path MTU discovery is disabled the value specified here will
2196 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2197 * include the flag SPP_PMTUD_DISABLE for this field to have any
2198 * effect).
2200 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2201 if (trans) {
2202 trans->pathmtu = params->spp_pathmtu;
2203 sctp_assoc_sync_pmtu(asoc);
2204 } else if (asoc) {
2205 asoc->pathmtu = params->spp_pathmtu;
2206 sctp_frag_point(sp, params->spp_pathmtu);
2207 } else {
2208 sp->pathmtu = params->spp_pathmtu;
2212 if (pmtud_change) {
2213 if (trans) {
2214 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2215 (params->spp_flags & SPP_PMTUD_ENABLE);
2216 trans->param_flags =
2217 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2218 if (update) {
2219 sctp_transport_pmtu(trans);
2220 sctp_assoc_sync_pmtu(asoc);
2222 } else if (asoc) {
2223 asoc->param_flags =
2224 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2225 } else {
2226 sp->param_flags =
2227 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2231 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2232 * value of this field is ignored. Note also that a value of zero
2233 * indicates the current setting should be left unchanged.
2235 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2236 if (trans) {
2237 trans->sackdelay =
2238 msecs_to_jiffies(params->spp_sackdelay);
2239 } else if (asoc) {
2240 asoc->sackdelay =
2241 msecs_to_jiffies(params->spp_sackdelay);
2242 } else {
2243 sp->sackdelay = params->spp_sackdelay;
2247 if (sackdelay_change) {
2248 if (trans) {
2249 trans->param_flags =
2250 (trans->param_flags & ~SPP_SACKDELAY) |
2251 sackdelay_change;
2252 } else if (asoc) {
2253 asoc->param_flags =
2254 (asoc->param_flags & ~SPP_SACKDELAY) |
2255 sackdelay_change;
2256 } else {
2257 sp->param_flags =
2258 (sp->param_flags & ~SPP_SACKDELAY) |
2259 sackdelay_change;
2263 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2264 * of this field is ignored. Note also that a value of zero
2265 * indicates the current setting should be left unchanged.
2267 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2268 if (trans) {
2269 trans->pathmaxrxt = params->spp_pathmaxrxt;
2270 } else if (asoc) {
2271 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2272 } else {
2273 sp->pathmaxrxt = params->spp_pathmaxrxt;
2277 return 0;
2280 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2281 char __user *optval, int optlen)
2283 struct sctp_paddrparams params;
2284 struct sctp_transport *trans = NULL;
2285 struct sctp_association *asoc = NULL;
2286 struct sctp_sock *sp = sctp_sk(sk);
2287 int error;
2288 int hb_change, pmtud_change, sackdelay_change;
2290 if (optlen != sizeof(struct sctp_paddrparams))
2291 return - EINVAL;
2293 if (copy_from_user(&params, optval, optlen))
2294 return -EFAULT;
2296 /* Validate flags and value parameters. */
2297 hb_change = params.spp_flags & SPP_HB;
2298 pmtud_change = params.spp_flags & SPP_PMTUD;
2299 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2301 if (hb_change == SPP_HB ||
2302 pmtud_change == SPP_PMTUD ||
2303 sackdelay_change == SPP_SACKDELAY ||
2304 params.spp_sackdelay > 500 ||
2305 (params.spp_pathmtu
2306 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2307 return -EINVAL;
2309 /* If an address other than INADDR_ANY is specified, and
2310 * no transport is found, then the request is invalid.
2312 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2313 trans = sctp_addr_id2transport(sk, &params.spp_address,
2314 params.spp_assoc_id);
2315 if (!trans)
2316 return -EINVAL;
2319 /* Get association, if assoc_id != 0 and the socket is a one
2320 * to many style socket, and an association was not found, then
2321 * the id was invalid.
2323 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2324 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2325 return -EINVAL;
2327 /* Heartbeat demand can only be sent on a transport or
2328 * association, but not a socket.
2330 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2331 return -EINVAL;
2333 /* Process parameters. */
2334 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2335 hb_change, pmtud_change,
2336 sackdelay_change);
2338 if (error)
2339 return error;
2341 /* If changes are for association, also apply parameters to each
2342 * transport.
2344 if (!trans && asoc) {
2345 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2346 transports) {
2347 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2348 hb_change, pmtud_change,
2349 sackdelay_change);
2353 return 0;
2357 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2359 * This option will effect the way delayed acks are performed. This
2360 * option allows you to get or set the delayed ack time, in
2361 * milliseconds. It also allows changing the delayed ack frequency.
2362 * Changing the frequency to 1 disables the delayed sack algorithm. If
2363 * the assoc_id is 0, then this sets or gets the endpoints default
2364 * values. If the assoc_id field is non-zero, then the set or get
2365 * effects the specified association for the one to many model (the
2366 * assoc_id field is ignored by the one to one model). Note that if
2367 * sack_delay or sack_freq are 0 when setting this option, then the
2368 * current values will remain unchanged.
2370 * struct sctp_sack_info {
2371 * sctp_assoc_t sack_assoc_id;
2372 * uint32_t sack_delay;
2373 * uint32_t sack_freq;
2374 * };
2376 * sack_assoc_id - This parameter, indicates which association the user
2377 * is performing an action upon. Note that if this field's value is
2378 * zero then the endpoints default value is changed (effecting future
2379 * associations only).
2381 * sack_delay - This parameter contains the number of milliseconds that
2382 * the user is requesting the delayed ACK timer be set to. Note that
2383 * this value is defined in the standard to be between 200 and 500
2384 * milliseconds.
2386 * sack_freq - This parameter contains the number of packets that must
2387 * be received before a sack is sent without waiting for the delay
2388 * timer to expire. The default value for this is 2, setting this
2389 * value to 1 will disable the delayed sack algorithm.
2392 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2393 char __user *optval, int optlen)
2395 struct sctp_sack_info params;
2396 struct sctp_transport *trans = NULL;
2397 struct sctp_association *asoc = NULL;
2398 struct sctp_sock *sp = sctp_sk(sk);
2400 if (optlen == sizeof(struct sctp_sack_info)) {
2401 if (copy_from_user(&params, optval, optlen))
2402 return -EFAULT;
2404 if (params.sack_delay == 0 && params.sack_freq == 0)
2405 return 0;
2406 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2407 printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
2408 "in delayed_ack socket option deprecated\n");
2409 printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
2410 if (copy_from_user(&params, optval, optlen))
2411 return -EFAULT;
2413 if (params.sack_delay == 0)
2414 params.sack_freq = 1;
2415 else
2416 params.sack_freq = 0;
2417 } else
2418 return - EINVAL;
2420 /* Validate value parameter. */
2421 if (params.sack_delay > 500)
2422 return -EINVAL;
2424 /* Get association, if sack_assoc_id != 0 and the socket is a one
2425 * to many style socket, and an association was not found, then
2426 * the id was invalid.
2428 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2429 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2430 return -EINVAL;
2432 if (params.sack_delay) {
2433 if (asoc) {
2434 asoc->sackdelay =
2435 msecs_to_jiffies(params.sack_delay);
2436 asoc->param_flags =
2437 (asoc->param_flags & ~SPP_SACKDELAY) |
2438 SPP_SACKDELAY_ENABLE;
2439 } else {
2440 sp->sackdelay = params.sack_delay;
2441 sp->param_flags =
2442 (sp->param_flags & ~SPP_SACKDELAY) |
2443 SPP_SACKDELAY_ENABLE;
2447 if (params.sack_freq == 1) {
2448 if (asoc) {
2449 asoc->param_flags =
2450 (asoc->param_flags & ~SPP_SACKDELAY) |
2451 SPP_SACKDELAY_DISABLE;
2452 } else {
2453 sp->param_flags =
2454 (sp->param_flags & ~SPP_SACKDELAY) |
2455 SPP_SACKDELAY_DISABLE;
2457 } else if (params.sack_freq > 1) {
2458 if (asoc) {
2459 asoc->sackfreq = params.sack_freq;
2460 asoc->param_flags =
2461 (asoc->param_flags & ~SPP_SACKDELAY) |
2462 SPP_SACKDELAY_ENABLE;
2463 } else {
2464 sp->sackfreq = params.sack_freq;
2465 sp->param_flags =
2466 (sp->param_flags & ~SPP_SACKDELAY) |
2467 SPP_SACKDELAY_ENABLE;
2471 /* If change is for association, also apply to each transport. */
2472 if (asoc) {
2473 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2474 transports) {
2475 if (params.sack_delay) {
2476 trans->sackdelay =
2477 msecs_to_jiffies(params.sack_delay);
2478 trans->param_flags =
2479 (trans->param_flags & ~SPP_SACKDELAY) |
2480 SPP_SACKDELAY_ENABLE;
2482 if (params.sack_freq == 1) {
2483 trans->param_flags =
2484 (trans->param_flags & ~SPP_SACKDELAY) |
2485 SPP_SACKDELAY_DISABLE;
2486 } else if (params.sack_freq > 1) {
2487 trans->sackfreq = params.sack_freq;
2488 trans->param_flags =
2489 (trans->param_flags & ~SPP_SACKDELAY) |
2490 SPP_SACKDELAY_ENABLE;
2495 return 0;
2498 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2500 * Applications can specify protocol parameters for the default association
2501 * initialization. The option name argument to setsockopt() and getsockopt()
2502 * is SCTP_INITMSG.
2504 * Setting initialization parameters is effective only on an unconnected
2505 * socket (for UDP-style sockets only future associations are effected
2506 * by the change). With TCP-style sockets, this option is inherited by
2507 * sockets derived from a listener socket.
2509 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2511 struct sctp_initmsg sinit;
2512 struct sctp_sock *sp = sctp_sk(sk);
2514 if (optlen != sizeof(struct sctp_initmsg))
2515 return -EINVAL;
2516 if (copy_from_user(&sinit, optval, optlen))
2517 return -EFAULT;
2519 if (sinit.sinit_num_ostreams)
2520 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2521 if (sinit.sinit_max_instreams)
2522 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2523 if (sinit.sinit_max_attempts)
2524 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2525 if (sinit.sinit_max_init_timeo)
2526 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2528 return 0;
2532 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2534 * Applications that wish to use the sendto() system call may wish to
2535 * specify a default set of parameters that would normally be supplied
2536 * through the inclusion of ancillary data. This socket option allows
2537 * such an application to set the default sctp_sndrcvinfo structure.
2538 * The application that wishes to use this socket option simply passes
2539 * in to this call the sctp_sndrcvinfo structure defined in Section
2540 * 5.2.2) The input parameters accepted by this call include
2541 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2542 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2543 * to this call if the caller is using the UDP model.
2545 static int sctp_setsockopt_default_send_param(struct sock *sk,
2546 char __user *optval, int optlen)
2548 struct sctp_sndrcvinfo info;
2549 struct sctp_association *asoc;
2550 struct sctp_sock *sp = sctp_sk(sk);
2552 if (optlen != sizeof(struct sctp_sndrcvinfo))
2553 return -EINVAL;
2554 if (copy_from_user(&info, optval, optlen))
2555 return -EFAULT;
2557 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2558 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2559 return -EINVAL;
2561 if (asoc) {
2562 asoc->default_stream = info.sinfo_stream;
2563 asoc->default_flags = info.sinfo_flags;
2564 asoc->default_ppid = info.sinfo_ppid;
2565 asoc->default_context = info.sinfo_context;
2566 asoc->default_timetolive = info.sinfo_timetolive;
2567 } else {
2568 sp->default_stream = info.sinfo_stream;
2569 sp->default_flags = info.sinfo_flags;
2570 sp->default_ppid = info.sinfo_ppid;
2571 sp->default_context = info.sinfo_context;
2572 sp->default_timetolive = info.sinfo_timetolive;
2575 return 0;
2578 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2580 * Requests that the local SCTP stack use the enclosed peer address as
2581 * the association primary. The enclosed address must be one of the
2582 * association peer's addresses.
2584 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2585 int optlen)
2587 struct sctp_prim prim;
2588 struct sctp_transport *trans;
2590 if (optlen != sizeof(struct sctp_prim))
2591 return -EINVAL;
2593 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2594 return -EFAULT;
2596 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2597 if (!trans)
2598 return -EINVAL;
2600 sctp_assoc_set_primary(trans->asoc, trans);
2602 return 0;
2606 * 7.1.5 SCTP_NODELAY
2608 * Turn on/off any Nagle-like algorithm. This means that packets are
2609 * generally sent as soon as possible and no unnecessary delays are
2610 * introduced, at the cost of more packets in the network. Expects an
2611 * integer boolean flag.
2613 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2614 int optlen)
2616 int val;
2618 if (optlen < sizeof(int))
2619 return -EINVAL;
2620 if (get_user(val, (int __user *)optval))
2621 return -EFAULT;
2623 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2624 return 0;
2629 * 7.1.1 SCTP_RTOINFO
2631 * The protocol parameters used to initialize and bound retransmission
2632 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2633 * and modify these parameters.
2634 * All parameters are time values, in milliseconds. A value of 0, when
2635 * modifying the parameters, indicates that the current value should not
2636 * be changed.
2639 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2640 struct sctp_rtoinfo rtoinfo;
2641 struct sctp_association *asoc;
2643 if (optlen != sizeof (struct sctp_rtoinfo))
2644 return -EINVAL;
2646 if (copy_from_user(&rtoinfo, optval, optlen))
2647 return -EFAULT;
2649 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2651 /* Set the values to the specific association */
2652 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2653 return -EINVAL;
2655 if (asoc) {
2656 if (rtoinfo.srto_initial != 0)
2657 asoc->rto_initial =
2658 msecs_to_jiffies(rtoinfo.srto_initial);
2659 if (rtoinfo.srto_max != 0)
2660 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2661 if (rtoinfo.srto_min != 0)
2662 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2663 } else {
2664 /* If there is no association or the association-id = 0
2665 * set the values to the endpoint.
2667 struct sctp_sock *sp = sctp_sk(sk);
2669 if (rtoinfo.srto_initial != 0)
2670 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2671 if (rtoinfo.srto_max != 0)
2672 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2673 if (rtoinfo.srto_min != 0)
2674 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2677 return 0;
2682 * 7.1.2 SCTP_ASSOCINFO
2684 * This option is used to tune the maximum retransmission attempts
2685 * of the association.
2686 * Returns an error if the new association retransmission value is
2687 * greater than the sum of the retransmission value of the peer.
2688 * See [SCTP] for more information.
2691 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2694 struct sctp_assocparams assocparams;
2695 struct sctp_association *asoc;
2697 if (optlen != sizeof(struct sctp_assocparams))
2698 return -EINVAL;
2699 if (copy_from_user(&assocparams, optval, optlen))
2700 return -EFAULT;
2702 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2704 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2705 return -EINVAL;
2707 /* Set the values to the specific association */
2708 if (asoc) {
2709 if (assocparams.sasoc_asocmaxrxt != 0) {
2710 __u32 path_sum = 0;
2711 int paths = 0;
2712 struct sctp_transport *peer_addr;
2714 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2715 transports) {
2716 path_sum += peer_addr->pathmaxrxt;
2717 paths++;
2720 /* Only validate asocmaxrxt if we have more then
2721 * one path/transport. We do this because path
2722 * retransmissions are only counted when we have more
2723 * then one path.
2725 if (paths > 1 &&
2726 assocparams.sasoc_asocmaxrxt > path_sum)
2727 return -EINVAL;
2729 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2732 if (assocparams.sasoc_cookie_life != 0) {
2733 asoc->cookie_life.tv_sec =
2734 assocparams.sasoc_cookie_life / 1000;
2735 asoc->cookie_life.tv_usec =
2736 (assocparams.sasoc_cookie_life % 1000)
2737 * 1000;
2739 } else {
2740 /* Set the values to the endpoint */
2741 struct sctp_sock *sp = sctp_sk(sk);
2743 if (assocparams.sasoc_asocmaxrxt != 0)
2744 sp->assocparams.sasoc_asocmaxrxt =
2745 assocparams.sasoc_asocmaxrxt;
2746 if (assocparams.sasoc_cookie_life != 0)
2747 sp->assocparams.sasoc_cookie_life =
2748 assocparams.sasoc_cookie_life;
2750 return 0;
2754 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2756 * This socket option is a boolean flag which turns on or off mapped V4
2757 * addresses. If this option is turned on and the socket is type
2758 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2759 * If this option is turned off, then no mapping will be done of V4
2760 * addresses and a user will receive both PF_INET6 and PF_INET type
2761 * addresses on the socket.
2763 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2765 int val;
2766 struct sctp_sock *sp = sctp_sk(sk);
2768 if (optlen < sizeof(int))
2769 return -EINVAL;
2770 if (get_user(val, (int __user *)optval))
2771 return -EFAULT;
2772 if (val)
2773 sp->v4mapped = 1;
2774 else
2775 sp->v4mapped = 0;
2777 return 0;
2781 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2783 * This socket option specifies the maximum size to put in any outgoing
2784 * SCTP chunk. If a message is larger than this size it will be
2785 * fragmented by SCTP into the specified size. Note that the underlying
2786 * SCTP implementation may fragment into smaller sized chunks when the
2787 * PMTU of the underlying association is smaller than the value set by
2788 * the user.
2790 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2792 struct sctp_association *asoc;
2793 struct sctp_sock *sp = sctp_sk(sk);
2794 int val;
2796 if (optlen < sizeof(int))
2797 return -EINVAL;
2798 if (get_user(val, (int __user *)optval))
2799 return -EFAULT;
2800 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2801 return -EINVAL;
2802 sp->user_frag = val;
2804 /* Update the frag_point of the existing associations. */
2805 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2806 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2809 return 0;
2814 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2816 * Requests that the peer mark the enclosed address as the association
2817 * primary. The enclosed address must be one of the association's
2818 * locally bound addresses. The following structure is used to make a
2819 * set primary request:
2821 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2822 int optlen)
2824 struct sctp_sock *sp;
2825 struct sctp_endpoint *ep;
2826 struct sctp_association *asoc = NULL;
2827 struct sctp_setpeerprim prim;
2828 struct sctp_chunk *chunk;
2829 int err;
2831 sp = sctp_sk(sk);
2832 ep = sp->ep;
2834 if (!sctp_addip_enable)
2835 return -EPERM;
2837 if (optlen != sizeof(struct sctp_setpeerprim))
2838 return -EINVAL;
2840 if (copy_from_user(&prim, optval, optlen))
2841 return -EFAULT;
2843 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2844 if (!asoc)
2845 return -EINVAL;
2847 if (!asoc->peer.asconf_capable)
2848 return -EPERM;
2850 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2851 return -EPERM;
2853 if (!sctp_state(asoc, ESTABLISHED))
2854 return -ENOTCONN;
2856 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2857 return -EADDRNOTAVAIL;
2859 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2860 chunk = sctp_make_asconf_set_prim(asoc,
2861 (union sctp_addr *)&prim.sspp_addr);
2862 if (!chunk)
2863 return -ENOMEM;
2865 err = sctp_send_asconf(asoc, chunk);
2867 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2869 return err;
2872 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2873 int optlen)
2875 struct sctp_setadaptation adaptation;
2877 if (optlen != sizeof(struct sctp_setadaptation))
2878 return -EINVAL;
2879 if (copy_from_user(&adaptation, optval, optlen))
2880 return -EFAULT;
2882 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2884 return 0;
2888 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2890 * The context field in the sctp_sndrcvinfo structure is normally only
2891 * used when a failed message is retrieved holding the value that was
2892 * sent down on the actual send call. This option allows the setting of
2893 * a default context on an association basis that will be received on
2894 * reading messages from the peer. This is especially helpful in the
2895 * one-2-many model for an application to keep some reference to an
2896 * internal state machine that is processing messages on the
2897 * association. Note that the setting of this value only effects
2898 * received messages from the peer and does not effect the value that is
2899 * saved with outbound messages.
2901 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2902 int optlen)
2904 struct sctp_assoc_value params;
2905 struct sctp_sock *sp;
2906 struct sctp_association *asoc;
2908 if (optlen != sizeof(struct sctp_assoc_value))
2909 return -EINVAL;
2910 if (copy_from_user(&params, optval, optlen))
2911 return -EFAULT;
2913 sp = sctp_sk(sk);
2915 if (params.assoc_id != 0) {
2916 asoc = sctp_id2assoc(sk, params.assoc_id);
2917 if (!asoc)
2918 return -EINVAL;
2919 asoc->default_rcv_context = params.assoc_value;
2920 } else {
2921 sp->default_rcv_context = params.assoc_value;
2924 return 0;
2928 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2930 * This options will at a minimum specify if the implementation is doing
2931 * fragmented interleave. Fragmented interleave, for a one to many
2932 * socket, is when subsequent calls to receive a message may return
2933 * parts of messages from different associations. Some implementations
2934 * may allow you to turn this value on or off. If so, when turned off,
2935 * no fragment interleave will occur (which will cause a head of line
2936 * blocking amongst multiple associations sharing the same one to many
2937 * socket). When this option is turned on, then each receive call may
2938 * come from a different association (thus the user must receive data
2939 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2940 * association each receive belongs to.
2942 * This option takes a boolean value. A non-zero value indicates that
2943 * fragmented interleave is on. A value of zero indicates that
2944 * fragmented interleave is off.
2946 * Note that it is important that an implementation that allows this
2947 * option to be turned on, have it off by default. Otherwise an unaware
2948 * application using the one to many model may become confused and act
2949 * incorrectly.
2951 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2952 char __user *optval,
2953 int optlen)
2955 int val;
2957 if (optlen != sizeof(int))
2958 return -EINVAL;
2959 if (get_user(val, (int __user *)optval))
2960 return -EFAULT;
2962 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2964 return 0;
2968 * 7.1.25. Set or Get the sctp partial delivery point
2969 * (SCTP_PARTIAL_DELIVERY_POINT)
2970 * This option will set or get the SCTP partial delivery point. This
2971 * point is the size of a message where the partial delivery API will be
2972 * invoked to help free up rwnd space for the peer. Setting this to a
2973 * lower value will cause partial delivery's to happen more often. The
2974 * calls argument is an integer that sets or gets the partial delivery
2975 * point.
2977 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2978 char __user *optval,
2979 int optlen)
2981 u32 val;
2983 if (optlen != sizeof(u32))
2984 return -EINVAL;
2985 if (get_user(val, (int __user *)optval))
2986 return -EFAULT;
2988 sctp_sk(sk)->pd_point = val;
2990 return 0; /* is this the right error code? */
2994 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2996 * This option will allow a user to change the maximum burst of packets
2997 * that can be emitted by this association. Note that the default value
2998 * is 4, and some implementations may restrict this setting so that it
2999 * can only be lowered.
3001 * NOTE: This text doesn't seem right. Do this on a socket basis with
3002 * future associations inheriting the socket value.
3004 static int sctp_setsockopt_maxburst(struct sock *sk,
3005 char __user *optval,
3006 int optlen)
3008 struct sctp_assoc_value params;
3009 struct sctp_sock *sp;
3010 struct sctp_association *asoc;
3011 int val;
3012 int assoc_id = 0;
3014 if (optlen < sizeof(int))
3015 return -EINVAL;
3017 if (optlen == sizeof(int)) {
3018 printk(KERN_WARNING
3019 "SCTP: Use of int in max_burst socket option deprecated\n");
3020 printk(KERN_WARNING
3021 "SCTP: Use struct sctp_assoc_value instead\n");
3022 if (copy_from_user(&val, optval, optlen))
3023 return -EFAULT;
3024 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3025 if (copy_from_user(&params, optval, optlen))
3026 return -EFAULT;
3027 val = params.assoc_value;
3028 assoc_id = params.assoc_id;
3029 } else
3030 return -EINVAL;
3032 sp = sctp_sk(sk);
3034 if (assoc_id != 0) {
3035 asoc = sctp_id2assoc(sk, assoc_id);
3036 if (!asoc)
3037 return -EINVAL;
3038 asoc->max_burst = val;
3039 } else
3040 sp->max_burst = val;
3042 return 0;
3046 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3048 * This set option adds a chunk type that the user is requesting to be
3049 * received only in an authenticated way. Changes to the list of chunks
3050 * will only effect future associations on the socket.
3052 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3053 char __user *optval,
3054 int optlen)
3056 struct sctp_authchunk val;
3058 if (!sctp_auth_enable)
3059 return -EACCES;
3061 if (optlen != sizeof(struct sctp_authchunk))
3062 return -EINVAL;
3063 if (copy_from_user(&val, optval, optlen))
3064 return -EFAULT;
3066 switch (val.sauth_chunk) {
3067 case SCTP_CID_INIT:
3068 case SCTP_CID_INIT_ACK:
3069 case SCTP_CID_SHUTDOWN_COMPLETE:
3070 case SCTP_CID_AUTH:
3071 return -EINVAL;
3074 /* add this chunk id to the endpoint */
3075 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3079 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3081 * This option gets or sets the list of HMAC algorithms that the local
3082 * endpoint requires the peer to use.
3084 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3085 char __user *optval,
3086 int optlen)
3088 struct sctp_hmacalgo *hmacs;
3089 u32 idents;
3090 int err;
3092 if (!sctp_auth_enable)
3093 return -EACCES;
3095 if (optlen < sizeof(struct sctp_hmacalgo))
3096 return -EINVAL;
3098 hmacs = kmalloc(optlen, GFP_KERNEL);
3099 if (!hmacs)
3100 return -ENOMEM;
3102 if (copy_from_user(hmacs, optval, optlen)) {
3103 err = -EFAULT;
3104 goto out;
3107 idents = hmacs->shmac_num_idents;
3108 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3109 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3110 err = -EINVAL;
3111 goto out;
3114 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3115 out:
3116 kfree(hmacs);
3117 return err;
3121 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3123 * This option will set a shared secret key which is used to build an
3124 * association shared key.
3126 static int sctp_setsockopt_auth_key(struct sock *sk,
3127 char __user *optval,
3128 int optlen)
3130 struct sctp_authkey *authkey;
3131 struct sctp_association *asoc;
3132 int ret;
3134 if (!sctp_auth_enable)
3135 return -EACCES;
3137 if (optlen <= sizeof(struct sctp_authkey))
3138 return -EINVAL;
3140 authkey = kmalloc(optlen, GFP_KERNEL);
3141 if (!authkey)
3142 return -ENOMEM;
3144 if (copy_from_user(authkey, optval, optlen)) {
3145 ret = -EFAULT;
3146 goto out;
3149 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3150 ret = -EINVAL;
3151 goto out;
3154 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3155 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3156 ret = -EINVAL;
3157 goto out;
3160 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3161 out:
3162 kfree(authkey);
3163 return ret;
3167 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3169 * This option will get or set the active shared key to be used to build
3170 * the association shared key.
3172 static int sctp_setsockopt_active_key(struct sock *sk,
3173 char __user *optval,
3174 int optlen)
3176 struct sctp_authkeyid val;
3177 struct sctp_association *asoc;
3179 if (!sctp_auth_enable)
3180 return -EACCES;
3182 if (optlen != sizeof(struct sctp_authkeyid))
3183 return -EINVAL;
3184 if (copy_from_user(&val, optval, optlen))
3185 return -EFAULT;
3187 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3188 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3189 return -EINVAL;
3191 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3192 val.scact_keynumber);
3196 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3198 * This set option will delete a shared secret key from use.
3200 static int sctp_setsockopt_del_key(struct sock *sk,
3201 char __user *optval,
3202 int optlen)
3204 struct sctp_authkeyid val;
3205 struct sctp_association *asoc;
3207 if (!sctp_auth_enable)
3208 return -EACCES;
3210 if (optlen != sizeof(struct sctp_authkeyid))
3211 return -EINVAL;
3212 if (copy_from_user(&val, optval, optlen))
3213 return -EFAULT;
3215 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3216 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3217 return -EINVAL;
3219 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3220 val.scact_keynumber);
3225 /* API 6.2 setsockopt(), getsockopt()
3227 * Applications use setsockopt() and getsockopt() to set or retrieve
3228 * socket options. Socket options are used to change the default
3229 * behavior of sockets calls. They are described in Section 7.
3231 * The syntax is:
3233 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3234 * int __user *optlen);
3235 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3236 * int optlen);
3238 * sd - the socket descript.
3239 * level - set to IPPROTO_SCTP for all SCTP options.
3240 * optname - the option name.
3241 * optval - the buffer to store the value of the option.
3242 * optlen - the size of the buffer.
3244 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3245 char __user *optval, int optlen)
3247 int retval = 0;
3249 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3250 sk, optname);
3252 /* I can hardly begin to describe how wrong this is. This is
3253 * so broken as to be worse than useless. The API draft
3254 * REALLY is NOT helpful here... I am not convinced that the
3255 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3256 * are at all well-founded.
3258 if (level != SOL_SCTP) {
3259 struct sctp_af *af = sctp_sk(sk)->pf->af;
3260 retval = af->setsockopt(sk, level, optname, optval, optlen);
3261 goto out_nounlock;
3264 sctp_lock_sock(sk);
3266 switch (optname) {
3267 case SCTP_SOCKOPT_BINDX_ADD:
3268 /* 'optlen' is the size of the addresses buffer. */
3269 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3270 optlen, SCTP_BINDX_ADD_ADDR);
3271 break;
3273 case SCTP_SOCKOPT_BINDX_REM:
3274 /* 'optlen' is the size of the addresses buffer. */
3275 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3276 optlen, SCTP_BINDX_REM_ADDR);
3277 break;
3279 case SCTP_SOCKOPT_CONNECTX_OLD:
3280 /* 'optlen' is the size of the addresses buffer. */
3281 retval = sctp_setsockopt_connectx_old(sk,
3282 (struct sockaddr __user *)optval,
3283 optlen);
3284 break;
3286 case SCTP_SOCKOPT_CONNECTX:
3287 /* 'optlen' is the size of the addresses buffer. */
3288 retval = sctp_setsockopt_connectx(sk,
3289 (struct sockaddr __user *)optval,
3290 optlen);
3291 break;
3293 case SCTP_DISABLE_FRAGMENTS:
3294 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3295 break;
3297 case SCTP_EVENTS:
3298 retval = sctp_setsockopt_events(sk, optval, optlen);
3299 break;
3301 case SCTP_AUTOCLOSE:
3302 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3303 break;
3305 case SCTP_PEER_ADDR_PARAMS:
3306 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3307 break;
3309 case SCTP_DELAYED_ACK:
3310 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3311 break;
3312 case SCTP_PARTIAL_DELIVERY_POINT:
3313 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3314 break;
3316 case SCTP_INITMSG:
3317 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3318 break;
3319 case SCTP_DEFAULT_SEND_PARAM:
3320 retval = sctp_setsockopt_default_send_param(sk, optval,
3321 optlen);
3322 break;
3323 case SCTP_PRIMARY_ADDR:
3324 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3325 break;
3326 case SCTP_SET_PEER_PRIMARY_ADDR:
3327 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3328 break;
3329 case SCTP_NODELAY:
3330 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3331 break;
3332 case SCTP_RTOINFO:
3333 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3334 break;
3335 case SCTP_ASSOCINFO:
3336 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3337 break;
3338 case SCTP_I_WANT_MAPPED_V4_ADDR:
3339 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3340 break;
3341 case SCTP_MAXSEG:
3342 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3343 break;
3344 case SCTP_ADAPTATION_LAYER:
3345 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3346 break;
3347 case SCTP_CONTEXT:
3348 retval = sctp_setsockopt_context(sk, optval, optlen);
3349 break;
3350 case SCTP_FRAGMENT_INTERLEAVE:
3351 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3352 break;
3353 case SCTP_MAX_BURST:
3354 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3355 break;
3356 case SCTP_AUTH_CHUNK:
3357 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3358 break;
3359 case SCTP_HMAC_IDENT:
3360 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3361 break;
3362 case SCTP_AUTH_KEY:
3363 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3364 break;
3365 case SCTP_AUTH_ACTIVE_KEY:
3366 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3367 break;
3368 case SCTP_AUTH_DELETE_KEY:
3369 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3370 break;
3371 default:
3372 retval = -ENOPROTOOPT;
3373 break;
3376 sctp_release_sock(sk);
3378 out_nounlock:
3379 return retval;
3382 /* API 3.1.6 connect() - UDP Style Syntax
3384 * An application may use the connect() call in the UDP model to initiate an
3385 * association without sending data.
3387 * The syntax is:
3389 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3391 * sd: the socket descriptor to have a new association added to.
3393 * nam: the address structure (either struct sockaddr_in or struct
3394 * sockaddr_in6 defined in RFC2553 [7]).
3396 * len: the size of the address.
3398 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3399 int addr_len)
3401 int err = 0;
3402 struct sctp_af *af;
3404 sctp_lock_sock(sk);
3406 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3407 __func__, sk, addr, addr_len);
3409 /* Validate addr_len before calling common connect/connectx routine. */
3410 af = sctp_get_af_specific(addr->sa_family);
3411 if (!af || addr_len < af->sockaddr_len) {
3412 err = -EINVAL;
3413 } else {
3414 /* Pass correct addr len to common routine (so it knows there
3415 * is only one address being passed.
3417 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3420 sctp_release_sock(sk);
3421 return err;
3424 /* FIXME: Write comments. */
3425 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3427 return -EOPNOTSUPP; /* STUB */
3430 /* 4.1.4 accept() - TCP Style Syntax
3432 * Applications use accept() call to remove an established SCTP
3433 * association from the accept queue of the endpoint. A new socket
3434 * descriptor will be returned from accept() to represent the newly
3435 * formed association.
3437 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3439 struct sctp_sock *sp;
3440 struct sctp_endpoint *ep;
3441 struct sock *newsk = NULL;
3442 struct sctp_association *asoc;
3443 long timeo;
3444 int error = 0;
3446 sctp_lock_sock(sk);
3448 sp = sctp_sk(sk);
3449 ep = sp->ep;
3451 if (!sctp_style(sk, TCP)) {
3452 error = -EOPNOTSUPP;
3453 goto out;
3456 if (!sctp_sstate(sk, LISTENING)) {
3457 error = -EINVAL;
3458 goto out;
3461 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3463 error = sctp_wait_for_accept(sk, timeo);
3464 if (error)
3465 goto out;
3467 /* We treat the list of associations on the endpoint as the accept
3468 * queue and pick the first association on the list.
3470 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3472 newsk = sp->pf->create_accept_sk(sk, asoc);
3473 if (!newsk) {
3474 error = -ENOMEM;
3475 goto out;
3478 /* Populate the fields of the newsk from the oldsk and migrate the
3479 * asoc to the newsk.
3481 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3483 out:
3484 sctp_release_sock(sk);
3485 *err = error;
3486 return newsk;
3489 /* The SCTP ioctl handler. */
3490 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3492 return -ENOIOCTLCMD;
3495 /* This is the function which gets called during socket creation to
3496 * initialized the SCTP-specific portion of the sock.
3497 * The sock structure should already be zero-filled memory.
3499 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3501 struct sctp_endpoint *ep;
3502 struct sctp_sock *sp;
3504 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3506 sp = sctp_sk(sk);
3508 /* Initialize the SCTP per socket area. */
3509 switch (sk->sk_type) {
3510 case SOCK_SEQPACKET:
3511 sp->type = SCTP_SOCKET_UDP;
3512 break;
3513 case SOCK_STREAM:
3514 sp->type = SCTP_SOCKET_TCP;
3515 break;
3516 default:
3517 return -ESOCKTNOSUPPORT;
3520 /* Initialize default send parameters. These parameters can be
3521 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3523 sp->default_stream = 0;
3524 sp->default_ppid = 0;
3525 sp->default_flags = 0;
3526 sp->default_context = 0;
3527 sp->default_timetolive = 0;
3529 sp->default_rcv_context = 0;
3530 sp->max_burst = sctp_max_burst;
3532 /* Initialize default setup parameters. These parameters
3533 * can be modified with the SCTP_INITMSG socket option or
3534 * overridden by the SCTP_INIT CMSG.
3536 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3537 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3538 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3539 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3541 /* Initialize default RTO related parameters. These parameters can
3542 * be modified for with the SCTP_RTOINFO socket option.
3544 sp->rtoinfo.srto_initial = sctp_rto_initial;
3545 sp->rtoinfo.srto_max = sctp_rto_max;
3546 sp->rtoinfo.srto_min = sctp_rto_min;
3548 /* Initialize default association related parameters. These parameters
3549 * can be modified with the SCTP_ASSOCINFO socket option.
3551 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3552 sp->assocparams.sasoc_number_peer_destinations = 0;
3553 sp->assocparams.sasoc_peer_rwnd = 0;
3554 sp->assocparams.sasoc_local_rwnd = 0;
3555 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3557 /* Initialize default event subscriptions. By default, all the
3558 * options are off.
3560 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3562 /* Default Peer Address Parameters. These defaults can
3563 * be modified via SCTP_PEER_ADDR_PARAMS
3565 sp->hbinterval = sctp_hb_interval;
3566 sp->pathmaxrxt = sctp_max_retrans_path;
3567 sp->pathmtu = 0; // allow default discovery
3568 sp->sackdelay = sctp_sack_timeout;
3569 sp->sackfreq = 2;
3570 sp->param_flags = SPP_HB_ENABLE |
3571 SPP_PMTUD_ENABLE |
3572 SPP_SACKDELAY_ENABLE;
3574 /* If enabled no SCTP message fragmentation will be performed.
3575 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3577 sp->disable_fragments = 0;
3579 /* Enable Nagle algorithm by default. */
3580 sp->nodelay = 0;
3582 /* Enable by default. */
3583 sp->v4mapped = 1;
3585 /* Auto-close idle associations after the configured
3586 * number of seconds. A value of 0 disables this
3587 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3588 * for UDP-style sockets only.
3590 sp->autoclose = 0;
3592 /* User specified fragmentation limit. */
3593 sp->user_frag = 0;
3595 sp->adaptation_ind = 0;
3597 sp->pf = sctp_get_pf_specific(sk->sk_family);
3599 /* Control variables for partial data delivery. */
3600 atomic_set(&sp->pd_mode, 0);
3601 skb_queue_head_init(&sp->pd_lobby);
3602 sp->frag_interleave = 0;
3604 /* Create a per socket endpoint structure. Even if we
3605 * change the data structure relationships, this may still
3606 * be useful for storing pre-connect address information.
3608 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3609 if (!ep)
3610 return -ENOMEM;
3612 sp->ep = ep;
3613 sp->hmac = NULL;
3615 SCTP_DBG_OBJCNT_INC(sock);
3616 atomic_inc(&sctp_sockets_allocated);
3618 local_bh_disable();
3619 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3620 local_bh_enable();
3622 return 0;
3625 /* Cleanup any SCTP per socket resources. */
3626 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3628 struct sctp_endpoint *ep;
3630 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3632 /* Release our hold on the endpoint. */
3633 ep = sctp_sk(sk)->ep;
3634 sctp_endpoint_free(ep);
3635 atomic_dec(&sctp_sockets_allocated);
3636 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3639 /* API 4.1.7 shutdown() - TCP Style Syntax
3640 * int shutdown(int socket, int how);
3642 * sd - the socket descriptor of the association to be closed.
3643 * how - Specifies the type of shutdown. The values are
3644 * as follows:
3645 * SHUT_RD
3646 * Disables further receive operations. No SCTP
3647 * protocol action is taken.
3648 * SHUT_WR
3649 * Disables further send operations, and initiates
3650 * the SCTP shutdown sequence.
3651 * SHUT_RDWR
3652 * Disables further send and receive operations
3653 * and initiates the SCTP shutdown sequence.
3655 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3657 struct sctp_endpoint *ep;
3658 struct sctp_association *asoc;
3660 if (!sctp_style(sk, TCP))
3661 return;
3663 if (how & SEND_SHUTDOWN) {
3664 ep = sctp_sk(sk)->ep;
3665 if (!list_empty(&ep->asocs)) {
3666 asoc = list_entry(ep->asocs.next,
3667 struct sctp_association, asocs);
3668 sctp_primitive_SHUTDOWN(asoc, NULL);
3673 /* 7.2.1 Association Status (SCTP_STATUS)
3675 * Applications can retrieve current status information about an
3676 * association, including association state, peer receiver window size,
3677 * number of unacked data chunks, and number of data chunks pending
3678 * receipt. This information is read-only.
3680 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3681 char __user *optval,
3682 int __user *optlen)
3684 struct sctp_status status;
3685 struct sctp_association *asoc = NULL;
3686 struct sctp_transport *transport;
3687 sctp_assoc_t associd;
3688 int retval = 0;
3690 if (len < sizeof(status)) {
3691 retval = -EINVAL;
3692 goto out;
3695 len = sizeof(status);
3696 if (copy_from_user(&status, optval, len)) {
3697 retval = -EFAULT;
3698 goto out;
3701 associd = status.sstat_assoc_id;
3702 asoc = sctp_id2assoc(sk, associd);
3703 if (!asoc) {
3704 retval = -EINVAL;
3705 goto out;
3708 transport = asoc->peer.primary_path;
3710 status.sstat_assoc_id = sctp_assoc2id(asoc);
3711 status.sstat_state = asoc->state;
3712 status.sstat_rwnd = asoc->peer.rwnd;
3713 status.sstat_unackdata = asoc->unack_data;
3715 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3716 status.sstat_instrms = asoc->c.sinit_max_instreams;
3717 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3718 status.sstat_fragmentation_point = asoc->frag_point;
3719 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3720 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3721 transport->af_specific->sockaddr_len);
3722 /* Map ipv4 address into v4-mapped-on-v6 address. */
3723 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3724 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3725 status.sstat_primary.spinfo_state = transport->state;
3726 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3727 status.sstat_primary.spinfo_srtt = transport->srtt;
3728 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3729 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3731 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3732 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3734 if (put_user(len, optlen)) {
3735 retval = -EFAULT;
3736 goto out;
3739 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3740 len, status.sstat_state, status.sstat_rwnd,
3741 status.sstat_assoc_id);
3743 if (copy_to_user(optval, &status, len)) {
3744 retval = -EFAULT;
3745 goto out;
3748 out:
3749 return (retval);
3753 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3755 * Applications can retrieve information about a specific peer address
3756 * of an association, including its reachability state, congestion
3757 * window, and retransmission timer values. This information is
3758 * read-only.
3760 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3761 char __user *optval,
3762 int __user *optlen)
3764 struct sctp_paddrinfo pinfo;
3765 struct sctp_transport *transport;
3766 int retval = 0;
3768 if (len < sizeof(pinfo)) {
3769 retval = -EINVAL;
3770 goto out;
3773 len = sizeof(pinfo);
3774 if (copy_from_user(&pinfo, optval, len)) {
3775 retval = -EFAULT;
3776 goto out;
3779 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3780 pinfo.spinfo_assoc_id);
3781 if (!transport)
3782 return -EINVAL;
3784 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3785 pinfo.spinfo_state = transport->state;
3786 pinfo.spinfo_cwnd = transport->cwnd;
3787 pinfo.spinfo_srtt = transport->srtt;
3788 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3789 pinfo.spinfo_mtu = transport->pathmtu;
3791 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3792 pinfo.spinfo_state = SCTP_ACTIVE;
3794 if (put_user(len, optlen)) {
3795 retval = -EFAULT;
3796 goto out;
3799 if (copy_to_user(optval, &pinfo, len)) {
3800 retval = -EFAULT;
3801 goto out;
3804 out:
3805 return (retval);
3808 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3810 * This option is a on/off flag. If enabled no SCTP message
3811 * fragmentation will be performed. Instead if a message being sent
3812 * exceeds the current PMTU size, the message will NOT be sent and
3813 * instead a error will be indicated to the user.
3815 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3816 char __user *optval, int __user *optlen)
3818 int val;
3820 if (len < sizeof(int))
3821 return -EINVAL;
3823 len = sizeof(int);
3824 val = (sctp_sk(sk)->disable_fragments == 1);
3825 if (put_user(len, optlen))
3826 return -EFAULT;
3827 if (copy_to_user(optval, &val, len))
3828 return -EFAULT;
3829 return 0;
3832 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3834 * This socket option is used to specify various notifications and
3835 * ancillary data the user wishes to receive.
3837 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3838 int __user *optlen)
3840 if (len < sizeof(struct sctp_event_subscribe))
3841 return -EINVAL;
3842 len = sizeof(struct sctp_event_subscribe);
3843 if (put_user(len, optlen))
3844 return -EFAULT;
3845 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3846 return -EFAULT;
3847 return 0;
3850 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3852 * This socket option is applicable to the UDP-style socket only. When
3853 * set it will cause associations that are idle for more than the
3854 * specified number of seconds to automatically close. An association
3855 * being idle is defined an association that has NOT sent or received
3856 * user data. The special value of '0' indicates that no automatic
3857 * close of any associations should be performed. The option expects an
3858 * integer defining the number of seconds of idle time before an
3859 * association is closed.
3861 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3863 /* Applicable to UDP-style socket only */
3864 if (sctp_style(sk, TCP))
3865 return -EOPNOTSUPP;
3866 if (len < sizeof(int))
3867 return -EINVAL;
3868 len = sizeof(int);
3869 if (put_user(len, optlen))
3870 return -EFAULT;
3871 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3872 return -EFAULT;
3873 return 0;
3876 /* Helper routine to branch off an association to a new socket. */
3877 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3878 struct socket **sockp)
3880 struct sock *sk = asoc->base.sk;
3881 struct socket *sock;
3882 struct inet_sock *inetsk;
3883 struct sctp_af *af;
3884 int err = 0;
3886 /* An association cannot be branched off from an already peeled-off
3887 * socket, nor is this supported for tcp style sockets.
3889 if (!sctp_style(sk, UDP))
3890 return -EINVAL;
3892 /* Create a new socket. */
3893 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3894 if (err < 0)
3895 return err;
3897 /* Populate the fields of the newsk from the oldsk and migrate the
3898 * asoc to the newsk.
3900 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3902 /* Make peeled-off sockets more like 1-1 accepted sockets.
3903 * Set the daddr and initialize id to something more random
3905 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3906 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3907 inetsk = inet_sk(sock->sk);
3908 inetsk->id = asoc->next_tsn ^ jiffies;
3910 *sockp = sock;
3912 return err;
3915 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3917 sctp_peeloff_arg_t peeloff;
3918 struct socket *newsock;
3919 int retval = 0;
3920 struct sctp_association *asoc;
3922 if (len < sizeof(sctp_peeloff_arg_t))
3923 return -EINVAL;
3924 len = sizeof(sctp_peeloff_arg_t);
3925 if (copy_from_user(&peeloff, optval, len))
3926 return -EFAULT;
3928 asoc = sctp_id2assoc(sk, peeloff.associd);
3929 if (!asoc) {
3930 retval = -EINVAL;
3931 goto out;
3934 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
3936 retval = sctp_do_peeloff(asoc, &newsock);
3937 if (retval < 0)
3938 goto out;
3940 /* Map the socket to an unused fd that can be returned to the user. */
3941 retval = sock_map_fd(newsock, 0);
3942 if (retval < 0) {
3943 sock_release(newsock);
3944 goto out;
3947 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3948 __func__, sk, asoc, newsock->sk, retval);
3950 /* Return the fd mapped to the new socket. */
3951 peeloff.sd = retval;
3952 if (put_user(len, optlen))
3953 return -EFAULT;
3954 if (copy_to_user(optval, &peeloff, len))
3955 retval = -EFAULT;
3957 out:
3958 return retval;
3961 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3963 * Applications can enable or disable heartbeats for any peer address of
3964 * an association, modify an address's heartbeat interval, force a
3965 * heartbeat to be sent immediately, and adjust the address's maximum
3966 * number of retransmissions sent before an address is considered
3967 * unreachable. The following structure is used to access and modify an
3968 * address's parameters:
3970 * struct sctp_paddrparams {
3971 * sctp_assoc_t spp_assoc_id;
3972 * struct sockaddr_storage spp_address;
3973 * uint32_t spp_hbinterval;
3974 * uint16_t spp_pathmaxrxt;
3975 * uint32_t spp_pathmtu;
3976 * uint32_t spp_sackdelay;
3977 * uint32_t spp_flags;
3978 * };
3980 * spp_assoc_id - (one-to-many style socket) This is filled in the
3981 * application, and identifies the association for
3982 * this query.
3983 * spp_address - This specifies which address is of interest.
3984 * spp_hbinterval - This contains the value of the heartbeat interval,
3985 * in milliseconds. If a value of zero
3986 * is present in this field then no changes are to
3987 * be made to this parameter.
3988 * spp_pathmaxrxt - This contains the maximum number of
3989 * retransmissions before this address shall be
3990 * considered unreachable. If a value of zero
3991 * is present in this field then no changes are to
3992 * be made to this parameter.
3993 * spp_pathmtu - When Path MTU discovery is disabled the value
3994 * specified here will be the "fixed" path mtu.
3995 * Note that if the spp_address field is empty
3996 * then all associations on this address will
3997 * have this fixed path mtu set upon them.
3999 * spp_sackdelay - When delayed sack is enabled, this value specifies
4000 * the number of milliseconds that sacks will be delayed
4001 * for. This value will apply to all addresses of an
4002 * association if the spp_address field is empty. Note
4003 * also, that if delayed sack is enabled and this
4004 * value is set to 0, no change is made to the last
4005 * recorded delayed sack timer value.
4007 * spp_flags - These flags are used to control various features
4008 * on an association. The flag field may contain
4009 * zero or more of the following options.
4011 * SPP_HB_ENABLE - Enable heartbeats on the
4012 * specified address. Note that if the address
4013 * field is empty all addresses for the association
4014 * have heartbeats enabled upon them.
4016 * SPP_HB_DISABLE - Disable heartbeats on the
4017 * speicifed address. Note that if the address
4018 * field is empty all addresses for the association
4019 * will have their heartbeats disabled. Note also
4020 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4021 * mutually exclusive, only one of these two should
4022 * be specified. Enabling both fields will have
4023 * undetermined results.
4025 * SPP_HB_DEMAND - Request a user initiated heartbeat
4026 * to be made immediately.
4028 * SPP_PMTUD_ENABLE - This field will enable PMTU
4029 * discovery upon the specified address. Note that
4030 * if the address feild is empty then all addresses
4031 * on the association are effected.
4033 * SPP_PMTUD_DISABLE - This field will disable PMTU
4034 * discovery upon the specified address. Note that
4035 * if the address feild is empty then all addresses
4036 * on the association are effected. Not also that
4037 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4038 * exclusive. Enabling both will have undetermined
4039 * results.
4041 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4042 * on delayed sack. The time specified in spp_sackdelay
4043 * is used to specify the sack delay for this address. Note
4044 * that if spp_address is empty then all addresses will
4045 * enable delayed sack and take on the sack delay
4046 * value specified in spp_sackdelay.
4047 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4048 * off delayed sack. If the spp_address field is blank then
4049 * delayed sack is disabled for the entire association. Note
4050 * also that this field is mutually exclusive to
4051 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4052 * results.
4054 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4055 char __user *optval, int __user *optlen)
4057 struct sctp_paddrparams params;
4058 struct sctp_transport *trans = NULL;
4059 struct sctp_association *asoc = NULL;
4060 struct sctp_sock *sp = sctp_sk(sk);
4062 if (len < sizeof(struct sctp_paddrparams))
4063 return -EINVAL;
4064 len = sizeof(struct sctp_paddrparams);
4065 if (copy_from_user(&params, optval, len))
4066 return -EFAULT;
4068 /* If an address other than INADDR_ANY is specified, and
4069 * no transport is found, then the request is invalid.
4071 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4072 trans = sctp_addr_id2transport(sk, &params.spp_address,
4073 params.spp_assoc_id);
4074 if (!trans) {
4075 SCTP_DEBUG_PRINTK("Failed no transport\n");
4076 return -EINVAL;
4080 /* Get association, if assoc_id != 0 and the socket is a one
4081 * to many style socket, and an association was not found, then
4082 * the id was invalid.
4084 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4085 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4086 SCTP_DEBUG_PRINTK("Failed no association\n");
4087 return -EINVAL;
4090 if (trans) {
4091 /* Fetch transport values. */
4092 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4093 params.spp_pathmtu = trans->pathmtu;
4094 params.spp_pathmaxrxt = trans->pathmaxrxt;
4095 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4097 /*draft-11 doesn't say what to return in spp_flags*/
4098 params.spp_flags = trans->param_flags;
4099 } else if (asoc) {
4100 /* Fetch association values. */
4101 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4102 params.spp_pathmtu = asoc->pathmtu;
4103 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4104 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4106 /*draft-11 doesn't say what to return in spp_flags*/
4107 params.spp_flags = asoc->param_flags;
4108 } else {
4109 /* Fetch socket values. */
4110 params.spp_hbinterval = sp->hbinterval;
4111 params.spp_pathmtu = sp->pathmtu;
4112 params.spp_sackdelay = sp->sackdelay;
4113 params.spp_pathmaxrxt = sp->pathmaxrxt;
4115 /*draft-11 doesn't say what to return in spp_flags*/
4116 params.spp_flags = sp->param_flags;
4119 if (copy_to_user(optval, &params, len))
4120 return -EFAULT;
4122 if (put_user(len, optlen))
4123 return -EFAULT;
4125 return 0;
4129 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4131 * This option will effect the way delayed acks are performed. This
4132 * option allows you to get or set the delayed ack time, in
4133 * milliseconds. It also allows changing the delayed ack frequency.
4134 * Changing the frequency to 1 disables the delayed sack algorithm. If
4135 * the assoc_id is 0, then this sets or gets the endpoints default
4136 * values. If the assoc_id field is non-zero, then the set or get
4137 * effects the specified association for the one to many model (the
4138 * assoc_id field is ignored by the one to one model). Note that if
4139 * sack_delay or sack_freq are 0 when setting this option, then the
4140 * current values will remain unchanged.
4142 * struct sctp_sack_info {
4143 * sctp_assoc_t sack_assoc_id;
4144 * uint32_t sack_delay;
4145 * uint32_t sack_freq;
4146 * };
4148 * sack_assoc_id - This parameter, indicates which association the user
4149 * is performing an action upon. Note that if this field's value is
4150 * zero then the endpoints default value is changed (effecting future
4151 * associations only).
4153 * sack_delay - This parameter contains the number of milliseconds that
4154 * the user is requesting the delayed ACK timer be set to. Note that
4155 * this value is defined in the standard to be between 200 and 500
4156 * milliseconds.
4158 * sack_freq - This parameter contains the number of packets that must
4159 * be received before a sack is sent without waiting for the delay
4160 * timer to expire. The default value for this is 2, setting this
4161 * value to 1 will disable the delayed sack algorithm.
4163 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4164 char __user *optval,
4165 int __user *optlen)
4167 struct sctp_sack_info params;
4168 struct sctp_association *asoc = NULL;
4169 struct sctp_sock *sp = sctp_sk(sk);
4171 if (len >= sizeof(struct sctp_sack_info)) {
4172 len = sizeof(struct sctp_sack_info);
4174 if (copy_from_user(&params, optval, len))
4175 return -EFAULT;
4176 } else if (len == sizeof(struct sctp_assoc_value)) {
4177 printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
4178 "in delayed_ack socket option deprecated\n");
4179 printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
4180 if (copy_from_user(&params, optval, len))
4181 return -EFAULT;
4182 } else
4183 return - EINVAL;
4185 /* Get association, if sack_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.sack_assoc_id);
4190 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4191 return -EINVAL;
4193 if (asoc) {
4194 /* Fetch association values. */
4195 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4196 params.sack_delay = jiffies_to_msecs(
4197 asoc->sackdelay);
4198 params.sack_freq = asoc->sackfreq;
4200 } else {
4201 params.sack_delay = 0;
4202 params.sack_freq = 1;
4204 } else {
4205 /* Fetch socket values. */
4206 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4207 params.sack_delay = sp->sackdelay;
4208 params.sack_freq = sp->sackfreq;
4209 } else {
4210 params.sack_delay = 0;
4211 params.sack_freq = 1;
4215 if (copy_to_user(optval, &params, len))
4216 return -EFAULT;
4218 if (put_user(len, optlen))
4219 return -EFAULT;
4221 return 0;
4224 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4226 * Applications can specify protocol parameters for the default association
4227 * initialization. The option name argument to setsockopt() and getsockopt()
4228 * is SCTP_INITMSG.
4230 * Setting initialization parameters is effective only on an unconnected
4231 * socket (for UDP-style sockets only future associations are effected
4232 * by the change). With TCP-style sockets, this option is inherited by
4233 * sockets derived from a listener socket.
4235 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4237 if (len < sizeof(struct sctp_initmsg))
4238 return -EINVAL;
4239 len = sizeof(struct sctp_initmsg);
4240 if (put_user(len, optlen))
4241 return -EFAULT;
4242 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4243 return -EFAULT;
4244 return 0;
4247 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4248 char __user *optval,
4249 int __user *optlen)
4251 sctp_assoc_t id;
4252 struct sctp_association *asoc;
4253 struct list_head *pos;
4254 int cnt = 0;
4256 if (len < sizeof(sctp_assoc_t))
4257 return -EINVAL;
4259 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4260 return -EFAULT;
4262 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_NUM_OLD "
4263 "socket option deprecated\n");
4264 /* For UDP-style sockets, id specifies the association to query. */
4265 asoc = sctp_id2assoc(sk, id);
4266 if (!asoc)
4267 return -EINVAL;
4269 list_for_each(pos, &asoc->peer.transport_addr_list) {
4270 cnt ++;
4273 return cnt;
4277 * Old API for getting list of peer addresses. Does not work for 32-bit
4278 * programs running on a 64-bit kernel
4280 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4281 char __user *optval,
4282 int __user *optlen)
4284 struct sctp_association *asoc;
4285 int cnt = 0;
4286 struct sctp_getaddrs_old getaddrs;
4287 struct sctp_transport *from;
4288 void __user *to;
4289 union sctp_addr temp;
4290 struct sctp_sock *sp = sctp_sk(sk);
4291 int addrlen;
4293 if (len < sizeof(struct sctp_getaddrs_old))
4294 return -EINVAL;
4296 len = sizeof(struct sctp_getaddrs_old);
4298 if (copy_from_user(&getaddrs, optval, len))
4299 return -EFAULT;
4301 if (getaddrs.addr_num <= 0) return -EINVAL;
4303 printk(KERN_WARNING "SCTP: Use of SCTP_GET_PEER_ADDRS_OLD "
4304 "socket option deprecated\n");
4306 /* For UDP-style sockets, id specifies the association to query. */
4307 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4308 if (!asoc)
4309 return -EINVAL;
4311 to = (void __user *)getaddrs.addrs;
4312 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4313 transports) {
4314 memcpy(&temp, &from->ipaddr, sizeof(temp));
4315 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4316 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4317 if (copy_to_user(to, &temp, addrlen))
4318 return -EFAULT;
4319 to += addrlen ;
4320 cnt ++;
4321 if (cnt >= getaddrs.addr_num) break;
4323 getaddrs.addr_num = cnt;
4324 if (put_user(len, optlen))
4325 return -EFAULT;
4326 if (copy_to_user(optval, &getaddrs, len))
4327 return -EFAULT;
4329 return 0;
4332 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4333 char __user *optval, int __user *optlen)
4335 struct sctp_association *asoc;
4336 int cnt = 0;
4337 struct sctp_getaddrs getaddrs;
4338 struct sctp_transport *from;
4339 void __user *to;
4340 union sctp_addr temp;
4341 struct sctp_sock *sp = sctp_sk(sk);
4342 int addrlen;
4343 size_t space_left;
4344 int bytes_copied;
4346 if (len < sizeof(struct sctp_getaddrs))
4347 return -EINVAL;
4349 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4350 return -EFAULT;
4352 /* For UDP-style sockets, id specifies the association to query. */
4353 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4354 if (!asoc)
4355 return -EINVAL;
4357 to = optval + offsetof(struct sctp_getaddrs,addrs);
4358 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4360 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4361 transports) {
4362 memcpy(&temp, &from->ipaddr, sizeof(temp));
4363 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4364 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4365 if (space_left < addrlen)
4366 return -ENOMEM;
4367 if (copy_to_user(to, &temp, addrlen))
4368 return -EFAULT;
4369 to += addrlen;
4370 cnt++;
4371 space_left -= addrlen;
4374 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4375 return -EFAULT;
4376 bytes_copied = ((char __user *)to) - optval;
4377 if (put_user(bytes_copied, optlen))
4378 return -EFAULT;
4380 return 0;
4383 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4384 char __user *optval,
4385 int __user *optlen)
4387 sctp_assoc_t id;
4388 struct sctp_bind_addr *bp;
4389 struct sctp_association *asoc;
4390 struct sctp_sockaddr_entry *addr;
4391 int cnt = 0;
4393 if (len < sizeof(sctp_assoc_t))
4394 return -EINVAL;
4396 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4397 return -EFAULT;
4399 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_NUM_OLD "
4400 "socket option deprecated\n");
4403 * For UDP-style sockets, id specifies the association to query.
4404 * If the id field is set to the value '0' then the locally bound
4405 * addresses are returned without regard to any particular
4406 * association.
4408 if (0 == id) {
4409 bp = &sctp_sk(sk)->ep->base.bind_addr;
4410 } else {
4411 asoc = sctp_id2assoc(sk, id);
4412 if (!asoc)
4413 return -EINVAL;
4414 bp = &asoc->base.bind_addr;
4417 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4418 * addresses from the global local address list.
4420 if (sctp_list_single_entry(&bp->address_list)) {
4421 addr = list_entry(bp->address_list.next,
4422 struct sctp_sockaddr_entry, list);
4423 if (sctp_is_any(sk, &addr->a)) {
4424 rcu_read_lock();
4425 list_for_each_entry_rcu(addr,
4426 &sctp_local_addr_list, list) {
4427 if (!addr->valid)
4428 continue;
4430 if ((PF_INET == sk->sk_family) &&
4431 (AF_INET6 == addr->a.sa.sa_family))
4432 continue;
4434 if ((PF_INET6 == sk->sk_family) &&
4435 inet_v6_ipv6only(sk) &&
4436 (AF_INET == addr->a.sa.sa_family))
4437 continue;
4439 cnt++;
4441 rcu_read_unlock();
4442 } else {
4443 cnt = 1;
4445 goto done;
4448 /* Protection on the bound address list is not needed,
4449 * since in the socket option context we hold the socket lock,
4450 * so there is no way that the bound address list can change.
4452 list_for_each_entry(addr, &bp->address_list, list) {
4453 cnt ++;
4455 done:
4456 return cnt;
4459 /* Helper function that copies local addresses to user and returns the number
4460 * of addresses copied.
4462 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4463 int max_addrs, void *to,
4464 int *bytes_copied)
4466 struct sctp_sockaddr_entry *addr;
4467 union sctp_addr temp;
4468 int cnt = 0;
4469 int addrlen;
4471 rcu_read_lock();
4472 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4473 if (!addr->valid)
4474 continue;
4476 if ((PF_INET == sk->sk_family) &&
4477 (AF_INET6 == addr->a.sa.sa_family))
4478 continue;
4479 if ((PF_INET6 == sk->sk_family) &&
4480 inet_v6_ipv6only(sk) &&
4481 (AF_INET == addr->a.sa.sa_family))
4482 continue;
4483 memcpy(&temp, &addr->a, sizeof(temp));
4484 if (!temp.v4.sin_port)
4485 temp.v4.sin_port = htons(port);
4487 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4488 &temp);
4489 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4490 memcpy(to, &temp, addrlen);
4492 to += addrlen;
4493 *bytes_copied += addrlen;
4494 cnt ++;
4495 if (cnt >= max_addrs) break;
4497 rcu_read_unlock();
4499 return cnt;
4502 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4503 size_t space_left, int *bytes_copied)
4505 struct sctp_sockaddr_entry *addr;
4506 union sctp_addr temp;
4507 int cnt = 0;
4508 int addrlen;
4510 rcu_read_lock();
4511 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4512 if (!addr->valid)
4513 continue;
4515 if ((PF_INET == sk->sk_family) &&
4516 (AF_INET6 == addr->a.sa.sa_family))
4517 continue;
4518 if ((PF_INET6 == sk->sk_family) &&
4519 inet_v6_ipv6only(sk) &&
4520 (AF_INET == addr->a.sa.sa_family))
4521 continue;
4522 memcpy(&temp, &addr->a, sizeof(temp));
4523 if (!temp.v4.sin_port)
4524 temp.v4.sin_port = htons(port);
4526 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4527 &temp);
4528 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4529 if (space_left < addrlen) {
4530 cnt = -ENOMEM;
4531 break;
4533 memcpy(to, &temp, addrlen);
4535 to += addrlen;
4536 cnt ++;
4537 space_left -= addrlen;
4538 *bytes_copied += addrlen;
4540 rcu_read_unlock();
4542 return cnt;
4545 /* Old API for getting list of local addresses. Does not work for 32-bit
4546 * programs running on a 64-bit kernel
4548 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4549 char __user *optval, int __user *optlen)
4551 struct sctp_bind_addr *bp;
4552 struct sctp_association *asoc;
4553 int cnt = 0;
4554 struct sctp_getaddrs_old getaddrs;
4555 struct sctp_sockaddr_entry *addr;
4556 void __user *to;
4557 union sctp_addr temp;
4558 struct sctp_sock *sp = sctp_sk(sk);
4559 int addrlen;
4560 int err = 0;
4561 void *addrs;
4562 void *buf;
4563 int bytes_copied = 0;
4565 if (len < sizeof(struct sctp_getaddrs_old))
4566 return -EINVAL;
4568 len = sizeof(struct sctp_getaddrs_old);
4569 if (copy_from_user(&getaddrs, optval, len))
4570 return -EFAULT;
4572 if (getaddrs.addr_num <= 0 ||
4573 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4574 return -EINVAL;
4576 printk(KERN_WARNING "SCTP: Use of SCTP_GET_LOCAL_ADDRS_OLD "
4577 "socket option deprecated\n");
4580 * For UDP-style sockets, id specifies the association to query.
4581 * If the id field is set to the value '0' then the locally bound
4582 * addresses are returned without regard to any particular
4583 * association.
4585 if (0 == getaddrs.assoc_id) {
4586 bp = &sctp_sk(sk)->ep->base.bind_addr;
4587 } else {
4588 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4589 if (!asoc)
4590 return -EINVAL;
4591 bp = &asoc->base.bind_addr;
4594 to = getaddrs.addrs;
4596 /* Allocate space for a local instance of packed array to hold all
4597 * the data. We store addresses here first and then put write them
4598 * to the user in one shot.
4600 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4601 GFP_KERNEL);
4602 if (!addrs)
4603 return -ENOMEM;
4605 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4606 * addresses from the global local address list.
4608 if (sctp_list_single_entry(&bp->address_list)) {
4609 addr = list_entry(bp->address_list.next,
4610 struct sctp_sockaddr_entry, list);
4611 if (sctp_is_any(sk, &addr->a)) {
4612 cnt = sctp_copy_laddrs_old(sk, bp->port,
4613 getaddrs.addr_num,
4614 addrs, &bytes_copied);
4615 goto copy_getaddrs;
4619 buf = addrs;
4620 /* Protection on the bound address list is not needed since
4621 * in the socket option context we hold a socket lock and
4622 * thus the bound address list can't change.
4624 list_for_each_entry(addr, &bp->address_list, list) {
4625 memcpy(&temp, &addr->a, sizeof(temp));
4626 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4627 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4628 memcpy(buf, &temp, addrlen);
4629 buf += addrlen;
4630 bytes_copied += addrlen;
4631 cnt ++;
4632 if (cnt >= getaddrs.addr_num) break;
4635 copy_getaddrs:
4636 /* copy the entire address list into the user provided space */
4637 if (copy_to_user(to, addrs, bytes_copied)) {
4638 err = -EFAULT;
4639 goto error;
4642 /* copy the leading structure back to user */
4643 getaddrs.addr_num = cnt;
4644 if (copy_to_user(optval, &getaddrs, len))
4645 err = -EFAULT;
4647 error:
4648 kfree(addrs);
4649 return err;
4652 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4653 char __user *optval, int __user *optlen)
4655 struct sctp_bind_addr *bp;
4656 struct sctp_association *asoc;
4657 int cnt = 0;
4658 struct sctp_getaddrs getaddrs;
4659 struct sctp_sockaddr_entry *addr;
4660 void __user *to;
4661 union sctp_addr temp;
4662 struct sctp_sock *sp = sctp_sk(sk);
4663 int addrlen;
4664 int err = 0;
4665 size_t space_left;
4666 int bytes_copied = 0;
4667 void *addrs;
4668 void *buf;
4670 if (len < sizeof(struct sctp_getaddrs))
4671 return -EINVAL;
4673 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4674 return -EFAULT;
4677 * For UDP-style sockets, id specifies the association to query.
4678 * If the id field is set to the value '0' then the locally bound
4679 * addresses are returned without regard to any particular
4680 * association.
4682 if (0 == getaddrs.assoc_id) {
4683 bp = &sctp_sk(sk)->ep->base.bind_addr;
4684 } else {
4685 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4686 if (!asoc)
4687 return -EINVAL;
4688 bp = &asoc->base.bind_addr;
4691 to = optval + offsetof(struct sctp_getaddrs,addrs);
4692 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4694 addrs = kmalloc(space_left, GFP_KERNEL);
4695 if (!addrs)
4696 return -ENOMEM;
4698 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4699 * addresses from the global local address list.
4701 if (sctp_list_single_entry(&bp->address_list)) {
4702 addr = list_entry(bp->address_list.next,
4703 struct sctp_sockaddr_entry, list);
4704 if (sctp_is_any(sk, &addr->a)) {
4705 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4706 space_left, &bytes_copied);
4707 if (cnt < 0) {
4708 err = cnt;
4709 goto out;
4711 goto copy_getaddrs;
4715 buf = addrs;
4716 /* Protection on the bound address list is not needed since
4717 * in the socket option context we hold a socket lock and
4718 * thus the bound address list can't change.
4720 list_for_each_entry(addr, &bp->address_list, list) {
4721 memcpy(&temp, &addr->a, sizeof(temp));
4722 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4723 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4724 if (space_left < addrlen) {
4725 err = -ENOMEM; /*fixme: right error?*/
4726 goto out;
4728 memcpy(buf, &temp, addrlen);
4729 buf += addrlen;
4730 bytes_copied += addrlen;
4731 cnt ++;
4732 space_left -= addrlen;
4735 copy_getaddrs:
4736 if (copy_to_user(to, addrs, bytes_copied)) {
4737 err = -EFAULT;
4738 goto out;
4740 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4741 err = -EFAULT;
4742 goto out;
4744 if (put_user(bytes_copied, optlen))
4745 err = -EFAULT;
4746 out:
4747 kfree(addrs);
4748 return err;
4751 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4753 * Requests that the local SCTP stack use the enclosed peer address as
4754 * the association primary. The enclosed address must be one of the
4755 * association peer's addresses.
4757 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4758 char __user *optval, int __user *optlen)
4760 struct sctp_prim prim;
4761 struct sctp_association *asoc;
4762 struct sctp_sock *sp = sctp_sk(sk);
4764 if (len < sizeof(struct sctp_prim))
4765 return -EINVAL;
4767 len = sizeof(struct sctp_prim);
4769 if (copy_from_user(&prim, optval, len))
4770 return -EFAULT;
4772 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4773 if (!asoc)
4774 return -EINVAL;
4776 if (!asoc->peer.primary_path)
4777 return -ENOTCONN;
4779 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4780 asoc->peer.primary_path->af_specific->sockaddr_len);
4782 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4783 (union sctp_addr *)&prim.ssp_addr);
4785 if (put_user(len, optlen))
4786 return -EFAULT;
4787 if (copy_to_user(optval, &prim, len))
4788 return -EFAULT;
4790 return 0;
4794 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4796 * Requests that the local endpoint set the specified Adaptation Layer
4797 * Indication parameter for all future INIT and INIT-ACK exchanges.
4799 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4800 char __user *optval, int __user *optlen)
4802 struct sctp_setadaptation adaptation;
4804 if (len < sizeof(struct sctp_setadaptation))
4805 return -EINVAL;
4807 len = sizeof(struct sctp_setadaptation);
4809 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4811 if (put_user(len, optlen))
4812 return -EFAULT;
4813 if (copy_to_user(optval, &adaptation, len))
4814 return -EFAULT;
4816 return 0;
4821 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4823 * Applications that wish to use the sendto() system call may wish to
4824 * specify a default set of parameters that would normally be supplied
4825 * through the inclusion of ancillary data. This socket option allows
4826 * such an application to set the default sctp_sndrcvinfo structure.
4829 * The application that wishes to use this socket option simply passes
4830 * in to this call the sctp_sndrcvinfo structure defined in Section
4831 * 5.2.2) The input parameters accepted by this call include
4832 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4833 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4834 * to this call if the caller is using the UDP model.
4836 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4838 static int sctp_getsockopt_default_send_param(struct sock *sk,
4839 int len, char __user *optval,
4840 int __user *optlen)
4842 struct sctp_sndrcvinfo info;
4843 struct sctp_association *asoc;
4844 struct sctp_sock *sp = sctp_sk(sk);
4846 if (len < sizeof(struct sctp_sndrcvinfo))
4847 return -EINVAL;
4849 len = sizeof(struct sctp_sndrcvinfo);
4851 if (copy_from_user(&info, optval, len))
4852 return -EFAULT;
4854 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4855 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4856 return -EINVAL;
4858 if (asoc) {
4859 info.sinfo_stream = asoc->default_stream;
4860 info.sinfo_flags = asoc->default_flags;
4861 info.sinfo_ppid = asoc->default_ppid;
4862 info.sinfo_context = asoc->default_context;
4863 info.sinfo_timetolive = asoc->default_timetolive;
4864 } else {
4865 info.sinfo_stream = sp->default_stream;
4866 info.sinfo_flags = sp->default_flags;
4867 info.sinfo_ppid = sp->default_ppid;
4868 info.sinfo_context = sp->default_context;
4869 info.sinfo_timetolive = sp->default_timetolive;
4872 if (put_user(len, optlen))
4873 return -EFAULT;
4874 if (copy_to_user(optval, &info, len))
4875 return -EFAULT;
4877 return 0;
4882 * 7.1.5 SCTP_NODELAY
4884 * Turn on/off any Nagle-like algorithm. This means that packets are
4885 * generally sent as soon as possible and no unnecessary delays are
4886 * introduced, at the cost of more packets in the network. Expects an
4887 * integer boolean flag.
4890 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4891 char __user *optval, int __user *optlen)
4893 int val;
4895 if (len < sizeof(int))
4896 return -EINVAL;
4898 len = sizeof(int);
4899 val = (sctp_sk(sk)->nodelay == 1);
4900 if (put_user(len, optlen))
4901 return -EFAULT;
4902 if (copy_to_user(optval, &val, len))
4903 return -EFAULT;
4904 return 0;
4909 * 7.1.1 SCTP_RTOINFO
4911 * The protocol parameters used to initialize and bound retransmission
4912 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4913 * and modify these parameters.
4914 * All parameters are time values, in milliseconds. A value of 0, when
4915 * modifying the parameters, indicates that the current value should not
4916 * be changed.
4919 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4920 char __user *optval,
4921 int __user *optlen) {
4922 struct sctp_rtoinfo rtoinfo;
4923 struct sctp_association *asoc;
4925 if (len < sizeof (struct sctp_rtoinfo))
4926 return -EINVAL;
4928 len = sizeof(struct sctp_rtoinfo);
4930 if (copy_from_user(&rtoinfo, optval, len))
4931 return -EFAULT;
4933 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4935 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4936 return -EINVAL;
4938 /* Values corresponding to the specific association. */
4939 if (asoc) {
4940 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4941 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4942 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4943 } else {
4944 /* Values corresponding to the endpoint. */
4945 struct sctp_sock *sp = sctp_sk(sk);
4947 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4948 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4949 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4952 if (put_user(len, optlen))
4953 return -EFAULT;
4955 if (copy_to_user(optval, &rtoinfo, len))
4956 return -EFAULT;
4958 return 0;
4963 * 7.1.2 SCTP_ASSOCINFO
4965 * This option is used to tune the maximum retransmission attempts
4966 * of the association.
4967 * Returns an error if the new association retransmission value is
4968 * greater than the sum of the retransmission value of the peer.
4969 * See [SCTP] for more information.
4972 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4973 char __user *optval,
4974 int __user *optlen)
4977 struct sctp_assocparams assocparams;
4978 struct sctp_association *asoc;
4979 struct list_head *pos;
4980 int cnt = 0;
4982 if (len < sizeof (struct sctp_assocparams))
4983 return -EINVAL;
4985 len = sizeof(struct sctp_assocparams);
4987 if (copy_from_user(&assocparams, optval, len))
4988 return -EFAULT;
4990 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4992 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4993 return -EINVAL;
4995 /* Values correspoinding to the specific association */
4996 if (asoc) {
4997 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4998 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4999 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5000 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
5001 * 1000) +
5002 (asoc->cookie_life.tv_usec
5003 / 1000);
5005 list_for_each(pos, &asoc->peer.transport_addr_list) {
5006 cnt ++;
5009 assocparams.sasoc_number_peer_destinations = cnt;
5010 } else {
5011 /* Values corresponding to the endpoint */
5012 struct sctp_sock *sp = sctp_sk(sk);
5014 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5015 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5016 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5017 assocparams.sasoc_cookie_life =
5018 sp->assocparams.sasoc_cookie_life;
5019 assocparams.sasoc_number_peer_destinations =
5020 sp->assocparams.
5021 sasoc_number_peer_destinations;
5024 if (put_user(len, optlen))
5025 return -EFAULT;
5027 if (copy_to_user(optval, &assocparams, len))
5028 return -EFAULT;
5030 return 0;
5034 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5036 * This socket option is a boolean flag which turns on or off mapped V4
5037 * addresses. If this option is turned on and the socket is type
5038 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5039 * If this option is turned off, then no mapping will be done of V4
5040 * addresses and a user will receive both PF_INET6 and PF_INET type
5041 * addresses on the socket.
5043 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5044 char __user *optval, int __user *optlen)
5046 int val;
5047 struct sctp_sock *sp = sctp_sk(sk);
5049 if (len < sizeof(int))
5050 return -EINVAL;
5052 len = sizeof(int);
5053 val = sp->v4mapped;
5054 if (put_user(len, optlen))
5055 return -EFAULT;
5056 if (copy_to_user(optval, &val, len))
5057 return -EFAULT;
5059 return 0;
5063 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5064 * (chapter and verse is quoted at sctp_setsockopt_context())
5066 static int sctp_getsockopt_context(struct sock *sk, int len,
5067 char __user *optval, int __user *optlen)
5069 struct sctp_assoc_value params;
5070 struct sctp_sock *sp;
5071 struct sctp_association *asoc;
5073 if (len < sizeof(struct sctp_assoc_value))
5074 return -EINVAL;
5076 len = sizeof(struct sctp_assoc_value);
5078 if (copy_from_user(&params, optval, len))
5079 return -EFAULT;
5081 sp = sctp_sk(sk);
5083 if (params.assoc_id != 0) {
5084 asoc = sctp_id2assoc(sk, params.assoc_id);
5085 if (!asoc)
5086 return -EINVAL;
5087 params.assoc_value = asoc->default_rcv_context;
5088 } else {
5089 params.assoc_value = sp->default_rcv_context;
5092 if (put_user(len, optlen))
5093 return -EFAULT;
5094 if (copy_to_user(optval, &params, len))
5095 return -EFAULT;
5097 return 0;
5101 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
5103 * This socket option specifies the maximum size to put in any outgoing
5104 * SCTP chunk. If a message is larger than this size it will be
5105 * fragmented by SCTP into the specified size. Note that the underlying
5106 * SCTP implementation may fragment into smaller sized chunks when the
5107 * PMTU of the underlying association is smaller than the value set by
5108 * the user.
5110 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5111 char __user *optval, int __user *optlen)
5113 int val;
5115 if (len < sizeof(int))
5116 return -EINVAL;
5118 len = sizeof(int);
5120 val = sctp_sk(sk)->user_frag;
5121 if (put_user(len, optlen))
5122 return -EFAULT;
5123 if (copy_to_user(optval, &val, len))
5124 return -EFAULT;
5126 return 0;
5130 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5131 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5133 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5134 char __user *optval, int __user *optlen)
5136 int val;
5138 if (len < sizeof(int))
5139 return -EINVAL;
5141 len = sizeof(int);
5143 val = sctp_sk(sk)->frag_interleave;
5144 if (put_user(len, optlen))
5145 return -EFAULT;
5146 if (copy_to_user(optval, &val, len))
5147 return -EFAULT;
5149 return 0;
5153 * 7.1.25. Set or Get the sctp partial delivery point
5154 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5156 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5157 char __user *optval,
5158 int __user *optlen)
5160 u32 val;
5162 if (len < sizeof(u32))
5163 return -EINVAL;
5165 len = sizeof(u32);
5167 val = sctp_sk(sk)->pd_point;
5168 if (put_user(len, optlen))
5169 return -EFAULT;
5170 if (copy_to_user(optval, &val, len))
5171 return -EFAULT;
5173 return -ENOTSUPP;
5177 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5178 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5180 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5181 char __user *optval,
5182 int __user *optlen)
5184 struct sctp_assoc_value params;
5185 struct sctp_sock *sp;
5186 struct sctp_association *asoc;
5188 if (len < sizeof(int))
5189 return -EINVAL;
5191 if (len == sizeof(int)) {
5192 printk(KERN_WARNING
5193 "SCTP: Use of int in max_burst socket option deprecated\n");
5194 printk(KERN_WARNING
5195 "SCTP: Use struct sctp_assoc_value instead\n");
5196 params.assoc_id = 0;
5197 } else if (len == sizeof (struct sctp_assoc_value)) {
5198 if (copy_from_user(&params, optval, len))
5199 return -EFAULT;
5200 } else
5201 return -EINVAL;
5203 sp = sctp_sk(sk);
5205 if (params.assoc_id != 0) {
5206 asoc = sctp_id2assoc(sk, params.assoc_id);
5207 if (!asoc)
5208 return -EINVAL;
5209 params.assoc_value = asoc->max_burst;
5210 } else
5211 params.assoc_value = sp->max_burst;
5213 if (len == sizeof(int)) {
5214 if (copy_to_user(optval, &params.assoc_value, len))
5215 return -EFAULT;
5216 } else {
5217 if (copy_to_user(optval, &params, len))
5218 return -EFAULT;
5221 return 0;
5225 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5226 char __user *optval, int __user *optlen)
5228 struct sctp_hmacalgo __user *p = (void __user *)optval;
5229 struct sctp_hmac_algo_param *hmacs;
5230 __u16 data_len = 0;
5231 u32 num_idents;
5233 if (!sctp_auth_enable)
5234 return -EACCES;
5236 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5237 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5239 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5240 return -EINVAL;
5242 len = sizeof(struct sctp_hmacalgo) + data_len;
5243 num_idents = data_len / sizeof(u16);
5245 if (put_user(len, optlen))
5246 return -EFAULT;
5247 if (put_user(num_idents, &p->shmac_num_idents))
5248 return -EFAULT;
5249 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5250 return -EFAULT;
5251 return 0;
5254 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5255 char __user *optval, int __user *optlen)
5257 struct sctp_authkeyid val;
5258 struct sctp_association *asoc;
5260 if (!sctp_auth_enable)
5261 return -EACCES;
5263 if (len < sizeof(struct sctp_authkeyid))
5264 return -EINVAL;
5265 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5266 return -EFAULT;
5268 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5269 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5270 return -EINVAL;
5272 if (asoc)
5273 val.scact_keynumber = asoc->active_key_id;
5274 else
5275 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5277 len = sizeof(struct sctp_authkeyid);
5278 if (put_user(len, optlen))
5279 return -EFAULT;
5280 if (copy_to_user(optval, &val, len))
5281 return -EFAULT;
5283 return 0;
5286 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5287 char __user *optval, int __user *optlen)
5289 struct sctp_authchunks __user *p = (void __user *)optval;
5290 struct sctp_authchunks val;
5291 struct sctp_association *asoc;
5292 struct sctp_chunks_param *ch;
5293 u32 num_chunks = 0;
5294 char __user *to;
5296 if (!sctp_auth_enable)
5297 return -EACCES;
5299 if (len < sizeof(struct sctp_authchunks))
5300 return -EINVAL;
5302 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5303 return -EFAULT;
5305 to = p->gauth_chunks;
5306 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5307 if (!asoc)
5308 return -EINVAL;
5310 ch = asoc->peer.peer_chunks;
5311 if (!ch)
5312 goto num;
5314 /* See if the user provided enough room for all the data */
5315 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5316 if (len < num_chunks)
5317 return -EINVAL;
5319 if (copy_to_user(to, ch->chunks, num_chunks))
5320 return -EFAULT;
5321 num:
5322 len = sizeof(struct sctp_authchunks) + num_chunks;
5323 if (put_user(len, optlen)) return -EFAULT;
5324 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5325 return -EFAULT;
5326 return 0;
5329 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5330 char __user *optval, int __user *optlen)
5332 struct sctp_authchunks __user *p = (void __user *)optval;
5333 struct sctp_authchunks val;
5334 struct sctp_association *asoc;
5335 struct sctp_chunks_param *ch;
5336 u32 num_chunks = 0;
5337 char __user *to;
5339 if (!sctp_auth_enable)
5340 return -EACCES;
5342 if (len < sizeof(struct sctp_authchunks))
5343 return -EINVAL;
5345 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5346 return -EFAULT;
5348 to = p->gauth_chunks;
5349 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5350 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5351 return -EINVAL;
5353 if (asoc)
5354 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5355 else
5356 ch = sctp_sk(sk)->ep->auth_chunk_list;
5358 if (!ch)
5359 goto num;
5361 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5362 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5363 return -EINVAL;
5365 if (copy_to_user(to, ch->chunks, num_chunks))
5366 return -EFAULT;
5367 num:
5368 len = sizeof(struct sctp_authchunks) + num_chunks;
5369 if (put_user(len, optlen))
5370 return -EFAULT;
5371 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5372 return -EFAULT;
5374 return 0;
5377 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5378 char __user *optval, int __user *optlen)
5380 int retval = 0;
5381 int len;
5383 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5384 sk, optname);
5386 /* I can hardly begin to describe how wrong this is. This is
5387 * so broken as to be worse than useless. The API draft
5388 * REALLY is NOT helpful here... I am not convinced that the
5389 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5390 * are at all well-founded.
5392 if (level != SOL_SCTP) {
5393 struct sctp_af *af = sctp_sk(sk)->pf->af;
5395 retval = af->getsockopt(sk, level, optname, optval, optlen);
5396 return retval;
5399 if (get_user(len, optlen))
5400 return -EFAULT;
5402 sctp_lock_sock(sk);
5404 switch (optname) {
5405 case SCTP_STATUS:
5406 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5407 break;
5408 case SCTP_DISABLE_FRAGMENTS:
5409 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5410 optlen);
5411 break;
5412 case SCTP_EVENTS:
5413 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5414 break;
5415 case SCTP_AUTOCLOSE:
5416 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5417 break;
5418 case SCTP_SOCKOPT_PEELOFF:
5419 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5420 break;
5421 case SCTP_PEER_ADDR_PARAMS:
5422 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5423 optlen);
5424 break;
5425 case SCTP_DELAYED_ACK:
5426 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5427 optlen);
5428 break;
5429 case SCTP_INITMSG:
5430 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5431 break;
5432 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5433 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5434 optlen);
5435 break;
5436 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5437 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5438 optlen);
5439 break;
5440 case SCTP_GET_PEER_ADDRS_OLD:
5441 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5442 optlen);
5443 break;
5444 case SCTP_GET_LOCAL_ADDRS_OLD:
5445 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5446 optlen);
5447 break;
5448 case SCTP_GET_PEER_ADDRS:
5449 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5450 optlen);
5451 break;
5452 case SCTP_GET_LOCAL_ADDRS:
5453 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5454 optlen);
5455 break;
5456 case SCTP_DEFAULT_SEND_PARAM:
5457 retval = sctp_getsockopt_default_send_param(sk, len,
5458 optval, optlen);
5459 break;
5460 case SCTP_PRIMARY_ADDR:
5461 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5462 break;
5463 case SCTP_NODELAY:
5464 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5465 break;
5466 case SCTP_RTOINFO:
5467 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5468 break;
5469 case SCTP_ASSOCINFO:
5470 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5471 break;
5472 case SCTP_I_WANT_MAPPED_V4_ADDR:
5473 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5474 break;
5475 case SCTP_MAXSEG:
5476 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5477 break;
5478 case SCTP_GET_PEER_ADDR_INFO:
5479 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5480 optlen);
5481 break;
5482 case SCTP_ADAPTATION_LAYER:
5483 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5484 optlen);
5485 break;
5486 case SCTP_CONTEXT:
5487 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5488 break;
5489 case SCTP_FRAGMENT_INTERLEAVE:
5490 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5491 optlen);
5492 break;
5493 case SCTP_PARTIAL_DELIVERY_POINT:
5494 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5495 optlen);
5496 break;
5497 case SCTP_MAX_BURST:
5498 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5499 break;
5500 case SCTP_AUTH_KEY:
5501 case SCTP_AUTH_CHUNK:
5502 case SCTP_AUTH_DELETE_KEY:
5503 retval = -EOPNOTSUPP;
5504 break;
5505 case SCTP_HMAC_IDENT:
5506 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5507 break;
5508 case SCTP_AUTH_ACTIVE_KEY:
5509 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5510 break;
5511 case SCTP_PEER_AUTH_CHUNKS:
5512 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5513 optlen);
5514 break;
5515 case SCTP_LOCAL_AUTH_CHUNKS:
5516 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5517 optlen);
5518 break;
5519 default:
5520 retval = -ENOPROTOOPT;
5521 break;
5524 sctp_release_sock(sk);
5525 return retval;
5528 static void sctp_hash(struct sock *sk)
5530 /* STUB */
5533 static void sctp_unhash(struct sock *sk)
5535 /* STUB */
5538 /* Check if port is acceptable. Possibly find first available port.
5540 * The port hash table (contained in the 'global' SCTP protocol storage
5541 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5542 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5543 * list (the list number is the port number hashed out, so as you
5544 * would expect from a hash function, all the ports in a given list have
5545 * such a number that hashes out to the same list number; you were
5546 * expecting that, right?); so each list has a set of ports, with a
5547 * link to the socket (struct sock) that uses it, the port number and
5548 * a fastreuse flag (FIXME: NPI ipg).
5550 static struct sctp_bind_bucket *sctp_bucket_create(
5551 struct sctp_bind_hashbucket *head, unsigned short snum);
5553 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5555 struct sctp_bind_hashbucket *head; /* hash list */
5556 struct sctp_bind_bucket *pp; /* hash list port iterator */
5557 struct hlist_node *node;
5558 unsigned short snum;
5559 int ret;
5561 snum = ntohs(addr->v4.sin_port);
5563 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5564 sctp_local_bh_disable();
5566 if (snum == 0) {
5567 /* Search for an available port. */
5568 int low, high, remaining, index;
5569 unsigned int rover;
5571 inet_get_local_port_range(&low, &high);
5572 remaining = (high - low) + 1;
5573 rover = net_random() % remaining + low;
5575 do {
5576 rover++;
5577 if ((rover < low) || (rover > high))
5578 rover = low;
5579 index = sctp_phashfn(rover);
5580 head = &sctp_port_hashtable[index];
5581 sctp_spin_lock(&head->lock);
5582 sctp_for_each_hentry(pp, node, &head->chain)
5583 if (pp->port == rover)
5584 goto next;
5585 break;
5586 next:
5587 sctp_spin_unlock(&head->lock);
5588 } while (--remaining > 0);
5590 /* Exhausted local port range during search? */
5591 ret = 1;
5592 if (remaining <= 0)
5593 goto fail;
5595 /* OK, here is the one we will use. HEAD (the port
5596 * hash table list entry) is non-NULL and we hold it's
5597 * mutex.
5599 snum = rover;
5600 } else {
5601 /* We are given an specific port number; we verify
5602 * that it is not being used. If it is used, we will
5603 * exahust the search in the hash list corresponding
5604 * to the port number (snum) - we detect that with the
5605 * port iterator, pp being NULL.
5607 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5608 sctp_spin_lock(&head->lock);
5609 sctp_for_each_hentry(pp, node, &head->chain) {
5610 if (pp->port == snum)
5611 goto pp_found;
5614 pp = NULL;
5615 goto pp_not_found;
5616 pp_found:
5617 if (!hlist_empty(&pp->owner)) {
5618 /* We had a port hash table hit - there is an
5619 * available port (pp != NULL) and it is being
5620 * used by other socket (pp->owner not empty); that other
5621 * socket is going to be sk2.
5623 int reuse = sk->sk_reuse;
5624 struct sock *sk2;
5625 struct hlist_node *node;
5627 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5628 if (pp->fastreuse && sk->sk_reuse &&
5629 sk->sk_state != SCTP_SS_LISTENING)
5630 goto success;
5632 /* Run through the list of sockets bound to the port
5633 * (pp->port) [via the pointers bind_next and
5634 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5635 * we get the endpoint they describe and run through
5636 * the endpoint's list of IP (v4 or v6) addresses,
5637 * comparing each of the addresses with the address of
5638 * the socket sk. If we find a match, then that means
5639 * that this port/socket (sk) combination are already
5640 * in an endpoint.
5642 sk_for_each_bound(sk2, node, &pp->owner) {
5643 struct sctp_endpoint *ep2;
5644 ep2 = sctp_sk(sk2)->ep;
5646 if (sk == sk2 ||
5647 (reuse && sk2->sk_reuse &&
5648 sk2->sk_state != SCTP_SS_LISTENING))
5649 continue;
5651 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5652 sctp_sk(sk2), sctp_sk(sk))) {
5653 ret = (long)sk2;
5654 goto fail_unlock;
5657 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5659 pp_not_found:
5660 /* If there was a hash table miss, create a new port. */
5661 ret = 1;
5662 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5663 goto fail_unlock;
5665 /* In either case (hit or miss), make sure fastreuse is 1 only
5666 * if sk->sk_reuse is too (that is, if the caller requested
5667 * SO_REUSEADDR on this socket -sk-).
5669 if (hlist_empty(&pp->owner)) {
5670 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5671 pp->fastreuse = 1;
5672 else
5673 pp->fastreuse = 0;
5674 } else if (pp->fastreuse &&
5675 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5676 pp->fastreuse = 0;
5678 /* We are set, so fill up all the data in the hash table
5679 * entry, tie the socket list information with the rest of the
5680 * sockets FIXME: Blurry, NPI (ipg).
5682 success:
5683 if (!sctp_sk(sk)->bind_hash) {
5684 inet_sk(sk)->num = snum;
5685 sk_add_bind_node(sk, &pp->owner);
5686 sctp_sk(sk)->bind_hash = pp;
5688 ret = 0;
5690 fail_unlock:
5691 sctp_spin_unlock(&head->lock);
5693 fail:
5694 sctp_local_bh_enable();
5695 return ret;
5698 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5699 * port is requested.
5701 static int sctp_get_port(struct sock *sk, unsigned short snum)
5703 long ret;
5704 union sctp_addr addr;
5705 struct sctp_af *af = sctp_sk(sk)->pf->af;
5707 /* Set up a dummy address struct from the sk. */
5708 af->from_sk(&addr, sk);
5709 addr.v4.sin_port = htons(snum);
5711 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5712 ret = sctp_get_port_local(sk, &addr);
5714 return (ret ? 1 : 0);
5718 * 3.1.3 listen() - UDP Style Syntax
5720 * By default, new associations are not accepted for UDP style sockets.
5721 * An application uses listen() to mark a socket as being able to
5722 * accept new associations.
5724 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5726 struct sctp_sock *sp = sctp_sk(sk);
5727 struct sctp_endpoint *ep = sp->ep;
5729 /* Only UDP style sockets that are not peeled off are allowed to
5730 * listen().
5732 if (!sctp_style(sk, UDP))
5733 return -EINVAL;
5735 /* If backlog is zero, disable listening. */
5736 if (!backlog) {
5737 if (sctp_sstate(sk, CLOSED))
5738 return 0;
5740 sctp_unhash_endpoint(ep);
5741 sk->sk_state = SCTP_SS_CLOSED;
5742 return 0;
5745 /* Return if we are already listening. */
5746 if (sctp_sstate(sk, LISTENING))
5747 return 0;
5750 * If a bind() or sctp_bindx() is not called prior to a listen()
5751 * call that allows new associations to be accepted, the system
5752 * picks an ephemeral port and will choose an address set equivalent
5753 * to binding with a wildcard address.
5755 * This is not currently spelled out in the SCTP sockets
5756 * extensions draft, but follows the practice as seen in TCP
5757 * sockets.
5759 * Additionally, turn off fastreuse flag since we are not listening
5761 sk->sk_state = SCTP_SS_LISTENING;
5762 if (!ep->base.bind_addr.port) {
5763 if (sctp_autobind(sk))
5764 return -EAGAIN;
5765 } else {
5766 if (sctp_get_port(sk, inet_sk(sk)->num)) {
5767 sk->sk_state = SCTP_SS_CLOSED;
5768 return -EADDRINUSE;
5770 sctp_sk(sk)->bind_hash->fastreuse = 0;
5773 sctp_hash_endpoint(ep);
5774 return 0;
5778 * 4.1.3 listen() - TCP Style Syntax
5780 * Applications uses listen() to ready the SCTP endpoint for accepting
5781 * inbound associations.
5783 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5785 struct sctp_sock *sp = sctp_sk(sk);
5786 struct sctp_endpoint *ep = sp->ep;
5788 /* If backlog is zero, disable listening. */
5789 if (!backlog) {
5790 if (sctp_sstate(sk, CLOSED))
5791 return 0;
5793 sctp_unhash_endpoint(ep);
5794 sk->sk_state = SCTP_SS_CLOSED;
5795 return 0;
5798 if (sctp_sstate(sk, LISTENING))
5799 return 0;
5802 * If a bind() or sctp_bindx() is not called prior to a listen()
5803 * call that allows new associations to be accepted, the system
5804 * picks an ephemeral port and will choose an address set equivalent
5805 * to binding with a wildcard address.
5807 * This is not currently spelled out in the SCTP sockets
5808 * extensions draft, but follows the practice as seen in TCP
5809 * sockets.
5811 sk->sk_state = SCTP_SS_LISTENING;
5812 if (!ep->base.bind_addr.port) {
5813 if (sctp_autobind(sk))
5814 return -EAGAIN;
5815 } else
5816 sctp_sk(sk)->bind_hash->fastreuse = 0;
5818 sk->sk_max_ack_backlog = backlog;
5819 sctp_hash_endpoint(ep);
5820 return 0;
5824 * Move a socket to LISTENING state.
5826 int sctp_inet_listen(struct socket *sock, int backlog)
5828 struct sock *sk = sock->sk;
5829 struct crypto_hash *tfm = NULL;
5830 int err = -EINVAL;
5832 if (unlikely(backlog < 0))
5833 goto out;
5835 sctp_lock_sock(sk);
5837 if (sock->state != SS_UNCONNECTED)
5838 goto out;
5840 /* Allocate HMAC for generating cookie. */
5841 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5842 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5843 if (IS_ERR(tfm)) {
5844 if (net_ratelimit()) {
5845 printk(KERN_INFO
5846 "SCTP: failed to load transform for %s: %ld\n",
5847 sctp_hmac_alg, PTR_ERR(tfm));
5849 err = -ENOSYS;
5850 goto out;
5854 switch (sock->type) {
5855 case SOCK_SEQPACKET:
5856 err = sctp_seqpacket_listen(sk, backlog);
5857 break;
5858 case SOCK_STREAM:
5859 err = sctp_stream_listen(sk, backlog);
5860 break;
5861 default:
5862 break;
5865 if (err)
5866 goto cleanup;
5868 /* Store away the transform reference. */
5869 if (!sctp_sk(sk)->hmac)
5870 sctp_sk(sk)->hmac = tfm;
5871 out:
5872 sctp_release_sock(sk);
5873 return err;
5874 cleanup:
5875 crypto_free_hash(tfm);
5876 goto out;
5880 * This function is done by modeling the current datagram_poll() and the
5881 * tcp_poll(). Note that, based on these implementations, we don't
5882 * lock the socket in this function, even though it seems that,
5883 * ideally, locking or some other mechanisms can be used to ensure
5884 * the integrity of the counters (sndbuf and wmem_alloc) used
5885 * in this place. We assume that we don't need locks either until proven
5886 * otherwise.
5888 * Another thing to note is that we include the Async I/O support
5889 * here, again, by modeling the current TCP/UDP code. We don't have
5890 * a good way to test with it yet.
5892 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5894 struct sock *sk = sock->sk;
5895 struct sctp_sock *sp = sctp_sk(sk);
5896 unsigned int mask;
5898 poll_wait(file, sk->sk_sleep, wait);
5900 /* A TCP-style listening socket becomes readable when the accept queue
5901 * is not empty.
5903 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5904 return (!list_empty(&sp->ep->asocs)) ?
5905 (POLLIN | POLLRDNORM) : 0;
5907 mask = 0;
5909 /* Is there any exceptional events? */
5910 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5911 mask |= POLLERR;
5912 if (sk->sk_shutdown & RCV_SHUTDOWN)
5913 mask |= POLLRDHUP;
5914 if (sk->sk_shutdown == SHUTDOWN_MASK)
5915 mask |= POLLHUP;
5917 /* Is it readable? Reconsider this code with TCP-style support. */
5918 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5919 (sk->sk_shutdown & RCV_SHUTDOWN))
5920 mask |= POLLIN | POLLRDNORM;
5922 /* The association is either gone or not ready. */
5923 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5924 return mask;
5926 /* Is it writable? */
5927 if (sctp_writeable(sk)) {
5928 mask |= POLLOUT | POLLWRNORM;
5929 } else {
5930 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5932 * Since the socket is not locked, the buffer
5933 * might be made available after the writeable check and
5934 * before the bit is set. This could cause a lost I/O
5935 * signal. tcp_poll() has a race breaker for this race
5936 * condition. Based on their implementation, we put
5937 * in the following code to cover it as well.
5939 if (sctp_writeable(sk))
5940 mask |= POLLOUT | POLLWRNORM;
5942 return mask;
5945 /********************************************************************
5946 * 2nd Level Abstractions
5947 ********************************************************************/
5949 static struct sctp_bind_bucket *sctp_bucket_create(
5950 struct sctp_bind_hashbucket *head, unsigned short snum)
5952 struct sctp_bind_bucket *pp;
5954 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5955 if (pp) {
5956 SCTP_DBG_OBJCNT_INC(bind_bucket);
5957 pp->port = snum;
5958 pp->fastreuse = 0;
5959 INIT_HLIST_HEAD(&pp->owner);
5960 hlist_add_head(&pp->node, &head->chain);
5962 return pp;
5965 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5966 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5968 if (pp && hlist_empty(&pp->owner)) {
5969 __hlist_del(&pp->node);
5970 kmem_cache_free(sctp_bucket_cachep, pp);
5971 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5975 /* Release this socket's reference to a local port. */
5976 static inline void __sctp_put_port(struct sock *sk)
5978 struct sctp_bind_hashbucket *head =
5979 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5980 struct sctp_bind_bucket *pp;
5982 sctp_spin_lock(&head->lock);
5983 pp = sctp_sk(sk)->bind_hash;
5984 __sk_del_bind_node(sk);
5985 sctp_sk(sk)->bind_hash = NULL;
5986 inet_sk(sk)->num = 0;
5987 sctp_bucket_destroy(pp);
5988 sctp_spin_unlock(&head->lock);
5991 void sctp_put_port(struct sock *sk)
5993 sctp_local_bh_disable();
5994 __sctp_put_port(sk);
5995 sctp_local_bh_enable();
5999 * The system picks an ephemeral port and choose an address set equivalent
6000 * to binding with a wildcard address.
6001 * One of those addresses will be the primary address for the association.
6002 * This automatically enables the multihoming capability of SCTP.
6004 static int sctp_autobind(struct sock *sk)
6006 union sctp_addr autoaddr;
6007 struct sctp_af *af;
6008 __be16 port;
6010 /* Initialize a local sockaddr structure to INADDR_ANY. */
6011 af = sctp_sk(sk)->pf->af;
6013 port = htons(inet_sk(sk)->num);
6014 af->inaddr_any(&autoaddr, port);
6016 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6019 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6021 * From RFC 2292
6022 * 4.2 The cmsghdr Structure *
6024 * When ancillary data is sent or received, any number of ancillary data
6025 * objects can be specified by the msg_control and msg_controllen members of
6026 * the msghdr structure, because each object is preceded by
6027 * a cmsghdr structure defining the object's length (the cmsg_len member).
6028 * Historically Berkeley-derived implementations have passed only one object
6029 * at a time, but this API allows multiple objects to be
6030 * passed in a single call to sendmsg() or recvmsg(). The following example
6031 * shows two ancillary data objects in a control buffer.
6033 * |<--------------------------- msg_controllen -------------------------->|
6034 * | |
6036 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6038 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6039 * | | |
6041 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6043 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6044 * | | | | |
6046 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6047 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6049 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6051 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6055 * msg_control
6056 * points here
6058 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6059 sctp_cmsgs_t *cmsgs)
6061 struct cmsghdr *cmsg;
6062 struct msghdr *my_msg = (struct msghdr *)msg;
6064 for (cmsg = CMSG_FIRSTHDR(msg);
6065 cmsg != NULL;
6066 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6067 if (!CMSG_OK(my_msg, cmsg))
6068 return -EINVAL;
6070 /* Should we parse this header or ignore? */
6071 if (cmsg->cmsg_level != IPPROTO_SCTP)
6072 continue;
6074 /* Strictly check lengths following example in SCM code. */
6075 switch (cmsg->cmsg_type) {
6076 case SCTP_INIT:
6077 /* SCTP Socket API Extension
6078 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6080 * This cmsghdr structure provides information for
6081 * initializing new SCTP associations with sendmsg().
6082 * The SCTP_INITMSG socket option uses this same data
6083 * structure. This structure is not used for
6084 * recvmsg().
6086 * cmsg_level cmsg_type cmsg_data[]
6087 * ------------ ------------ ----------------------
6088 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6090 if (cmsg->cmsg_len !=
6091 CMSG_LEN(sizeof(struct sctp_initmsg)))
6092 return -EINVAL;
6093 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6094 break;
6096 case SCTP_SNDRCV:
6097 /* SCTP Socket API Extension
6098 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6100 * This cmsghdr structure specifies SCTP options for
6101 * sendmsg() and describes SCTP header information
6102 * about a received message through recvmsg().
6104 * cmsg_level cmsg_type cmsg_data[]
6105 * ------------ ------------ ----------------------
6106 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6108 if (cmsg->cmsg_len !=
6109 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6110 return -EINVAL;
6112 cmsgs->info =
6113 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6115 /* Minimally, validate the sinfo_flags. */
6116 if (cmsgs->info->sinfo_flags &
6117 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6118 SCTP_ABORT | SCTP_EOF))
6119 return -EINVAL;
6120 break;
6122 default:
6123 return -EINVAL;
6126 return 0;
6130 * Wait for a packet..
6131 * Note: This function is the same function as in core/datagram.c
6132 * with a few modifications to make lksctp work.
6134 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6136 int error;
6137 DEFINE_WAIT(wait);
6139 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6141 /* Socket errors? */
6142 error = sock_error(sk);
6143 if (error)
6144 goto out;
6146 if (!skb_queue_empty(&sk->sk_receive_queue))
6147 goto ready;
6149 /* Socket shut down? */
6150 if (sk->sk_shutdown & RCV_SHUTDOWN)
6151 goto out;
6153 /* Sequenced packets can come disconnected. If so we report the
6154 * problem.
6156 error = -ENOTCONN;
6158 /* Is there a good reason to think that we may receive some data? */
6159 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6160 goto out;
6162 /* Handle signals. */
6163 if (signal_pending(current))
6164 goto interrupted;
6166 /* Let another process have a go. Since we are going to sleep
6167 * anyway. Note: This may cause odd behaviors if the message
6168 * does not fit in the user's buffer, but this seems to be the
6169 * only way to honor MSG_DONTWAIT realistically.
6171 sctp_release_sock(sk);
6172 *timeo_p = schedule_timeout(*timeo_p);
6173 sctp_lock_sock(sk);
6175 ready:
6176 finish_wait(sk->sk_sleep, &wait);
6177 return 0;
6179 interrupted:
6180 error = sock_intr_errno(*timeo_p);
6182 out:
6183 finish_wait(sk->sk_sleep, &wait);
6184 *err = error;
6185 return error;
6188 /* Receive a datagram.
6189 * Note: This is pretty much the same routine as in core/datagram.c
6190 * with a few changes to make lksctp work.
6192 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6193 int noblock, int *err)
6195 int error;
6196 struct sk_buff *skb;
6197 long timeo;
6199 timeo = sock_rcvtimeo(sk, noblock);
6201 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6202 timeo, MAX_SCHEDULE_TIMEOUT);
6204 do {
6205 /* Again only user level code calls this function,
6206 * so nothing interrupt level
6207 * will suddenly eat the receive_queue.
6209 * Look at current nfs client by the way...
6210 * However, this function was corrent in any case. 8)
6212 if (flags & MSG_PEEK) {
6213 spin_lock_bh(&sk->sk_receive_queue.lock);
6214 skb = skb_peek(&sk->sk_receive_queue);
6215 if (skb)
6216 atomic_inc(&skb->users);
6217 spin_unlock_bh(&sk->sk_receive_queue.lock);
6218 } else {
6219 skb = skb_dequeue(&sk->sk_receive_queue);
6222 if (skb)
6223 return skb;
6225 /* Caller is allowed not to check sk->sk_err before calling. */
6226 error = sock_error(sk);
6227 if (error)
6228 goto no_packet;
6230 if (sk->sk_shutdown & RCV_SHUTDOWN)
6231 break;
6233 /* User doesn't want to wait. */
6234 error = -EAGAIN;
6235 if (!timeo)
6236 goto no_packet;
6237 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6239 return NULL;
6241 no_packet:
6242 *err = error;
6243 return NULL;
6246 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6247 static void __sctp_write_space(struct sctp_association *asoc)
6249 struct sock *sk = asoc->base.sk;
6250 struct socket *sock = sk->sk_socket;
6252 if ((sctp_wspace(asoc) > 0) && sock) {
6253 if (waitqueue_active(&asoc->wait))
6254 wake_up_interruptible(&asoc->wait);
6256 if (sctp_writeable(sk)) {
6257 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6258 wake_up_interruptible(sk->sk_sleep);
6260 /* Note that we try to include the Async I/O support
6261 * here by modeling from the current TCP/UDP code.
6262 * We have not tested with it yet.
6264 if (sock->fasync_list &&
6265 !(sk->sk_shutdown & SEND_SHUTDOWN))
6266 sock_wake_async(sock,
6267 SOCK_WAKE_SPACE, POLL_OUT);
6272 /* Do accounting for the sndbuf space.
6273 * Decrement the used sndbuf space of the corresponding association by the
6274 * data size which was just transmitted(freed).
6276 static void sctp_wfree(struct sk_buff *skb)
6278 struct sctp_association *asoc;
6279 struct sctp_chunk *chunk;
6280 struct sock *sk;
6282 /* Get the saved chunk pointer. */
6283 chunk = *((struct sctp_chunk **)(skb->cb));
6284 asoc = chunk->asoc;
6285 sk = asoc->base.sk;
6286 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6287 sizeof(struct sk_buff) +
6288 sizeof(struct sctp_chunk);
6290 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6293 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6295 sk->sk_wmem_queued -= skb->truesize;
6296 sk_mem_uncharge(sk, skb->truesize);
6298 sock_wfree(skb);
6299 __sctp_write_space(asoc);
6301 sctp_association_put(asoc);
6304 /* Do accounting for the receive space on the socket.
6305 * Accounting for the association is done in ulpevent.c
6306 * We set this as a destructor for the cloned data skbs so that
6307 * accounting is done at the correct time.
6309 void sctp_sock_rfree(struct sk_buff *skb)
6311 struct sock *sk = skb->sk;
6312 struct sctp_ulpevent *event = sctp_skb2event(skb);
6314 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6317 * Mimic the behavior of sock_rfree
6319 sk_mem_uncharge(sk, event->rmem_len);
6323 /* Helper function to wait for space in the sndbuf. */
6324 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6325 size_t msg_len)
6327 struct sock *sk = asoc->base.sk;
6328 int err = 0;
6329 long current_timeo = *timeo_p;
6330 DEFINE_WAIT(wait);
6332 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6333 asoc, (long)(*timeo_p), msg_len);
6335 /* Increment the association's refcnt. */
6336 sctp_association_hold(asoc);
6338 /* Wait on the association specific sndbuf space. */
6339 for (;;) {
6340 prepare_to_wait_exclusive(&asoc->wait, &wait,
6341 TASK_INTERRUPTIBLE);
6342 if (!*timeo_p)
6343 goto do_nonblock;
6344 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6345 asoc->base.dead)
6346 goto do_error;
6347 if (signal_pending(current))
6348 goto do_interrupted;
6349 if (msg_len <= sctp_wspace(asoc))
6350 break;
6352 /* Let another process have a go. Since we are going
6353 * to sleep anyway.
6355 sctp_release_sock(sk);
6356 current_timeo = schedule_timeout(current_timeo);
6357 BUG_ON(sk != asoc->base.sk);
6358 sctp_lock_sock(sk);
6360 *timeo_p = current_timeo;
6363 out:
6364 finish_wait(&asoc->wait, &wait);
6366 /* Release the association's refcnt. */
6367 sctp_association_put(asoc);
6369 return err;
6371 do_error:
6372 err = -EPIPE;
6373 goto out;
6375 do_interrupted:
6376 err = sock_intr_errno(*timeo_p);
6377 goto out;
6379 do_nonblock:
6380 err = -EAGAIN;
6381 goto out;
6384 /* If socket sndbuf has changed, wake up all per association waiters. */
6385 void sctp_write_space(struct sock *sk)
6387 struct sctp_association *asoc;
6389 /* Wake up the tasks in each wait queue. */
6390 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6391 __sctp_write_space(asoc);
6395 /* Is there any sndbuf space available on the socket?
6397 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6398 * associations on the same socket. For a UDP-style socket with
6399 * multiple associations, it is possible for it to be "unwriteable"
6400 * prematurely. I assume that this is acceptable because
6401 * a premature "unwriteable" is better than an accidental "writeable" which
6402 * would cause an unwanted block under certain circumstances. For the 1-1
6403 * UDP-style sockets or TCP-style sockets, this code should work.
6404 * - Daisy
6406 static int sctp_writeable(struct sock *sk)
6408 int amt = 0;
6410 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
6411 if (amt < 0)
6412 amt = 0;
6413 return amt;
6416 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6417 * returns immediately with EINPROGRESS.
6419 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6421 struct sock *sk = asoc->base.sk;
6422 int err = 0;
6423 long current_timeo = *timeo_p;
6424 DEFINE_WAIT(wait);
6426 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6427 (long)(*timeo_p));
6429 /* Increment the association's refcnt. */
6430 sctp_association_hold(asoc);
6432 for (;;) {
6433 prepare_to_wait_exclusive(&asoc->wait, &wait,
6434 TASK_INTERRUPTIBLE);
6435 if (!*timeo_p)
6436 goto do_nonblock;
6437 if (sk->sk_shutdown & RCV_SHUTDOWN)
6438 break;
6439 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6440 asoc->base.dead)
6441 goto do_error;
6442 if (signal_pending(current))
6443 goto do_interrupted;
6445 if (sctp_state(asoc, ESTABLISHED))
6446 break;
6448 /* Let another process have a go. Since we are going
6449 * to sleep anyway.
6451 sctp_release_sock(sk);
6452 current_timeo = schedule_timeout(current_timeo);
6453 sctp_lock_sock(sk);
6455 *timeo_p = current_timeo;
6458 out:
6459 finish_wait(&asoc->wait, &wait);
6461 /* Release the association's refcnt. */
6462 sctp_association_put(asoc);
6464 return err;
6466 do_error:
6467 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6468 err = -ETIMEDOUT;
6469 else
6470 err = -ECONNREFUSED;
6471 goto out;
6473 do_interrupted:
6474 err = sock_intr_errno(*timeo_p);
6475 goto out;
6477 do_nonblock:
6478 err = -EINPROGRESS;
6479 goto out;
6482 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6484 struct sctp_endpoint *ep;
6485 int err = 0;
6486 DEFINE_WAIT(wait);
6488 ep = sctp_sk(sk)->ep;
6491 for (;;) {
6492 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6493 TASK_INTERRUPTIBLE);
6495 if (list_empty(&ep->asocs)) {
6496 sctp_release_sock(sk);
6497 timeo = schedule_timeout(timeo);
6498 sctp_lock_sock(sk);
6501 err = -EINVAL;
6502 if (!sctp_sstate(sk, LISTENING))
6503 break;
6505 err = 0;
6506 if (!list_empty(&ep->asocs))
6507 break;
6509 err = sock_intr_errno(timeo);
6510 if (signal_pending(current))
6511 break;
6513 err = -EAGAIN;
6514 if (!timeo)
6515 break;
6518 finish_wait(sk->sk_sleep, &wait);
6520 return err;
6523 static void sctp_wait_for_close(struct sock *sk, long timeout)
6525 DEFINE_WAIT(wait);
6527 do {
6528 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6529 if (list_empty(&sctp_sk(sk)->ep->asocs))
6530 break;
6531 sctp_release_sock(sk);
6532 timeout = schedule_timeout(timeout);
6533 sctp_lock_sock(sk);
6534 } while (!signal_pending(current) && timeout);
6536 finish_wait(sk->sk_sleep, &wait);
6539 static void sctp_sock_rfree_frag(struct sk_buff *skb)
6541 struct sk_buff *frag;
6543 if (!skb->data_len)
6544 goto done;
6546 /* Don't forget the fragments. */
6547 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6548 sctp_sock_rfree_frag(frag);
6550 done:
6551 sctp_sock_rfree(skb);
6554 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6556 struct sk_buff *frag;
6558 if (!skb->data_len)
6559 goto done;
6561 /* Don't forget the fragments. */
6562 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6563 sctp_skb_set_owner_r_frag(frag, sk);
6565 done:
6566 sctp_skb_set_owner_r(skb, sk);
6569 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6570 * and its messages to the newsk.
6572 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6573 struct sctp_association *assoc,
6574 sctp_socket_type_t type)
6576 struct sctp_sock *oldsp = sctp_sk(oldsk);
6577 struct sctp_sock *newsp = sctp_sk(newsk);
6578 struct sctp_bind_bucket *pp; /* hash list port iterator */
6579 struct sctp_endpoint *newep = newsp->ep;
6580 struct sk_buff *skb, *tmp;
6581 struct sctp_ulpevent *event;
6582 struct sctp_bind_hashbucket *head;
6584 /* Migrate socket buffer sizes and all the socket level options to the
6585 * new socket.
6587 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6588 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6589 /* Brute force copy old sctp opt. */
6590 inet_sk_copy_descendant(newsk, oldsk);
6592 /* Restore the ep value that was overwritten with the above structure
6593 * copy.
6595 newsp->ep = newep;
6596 newsp->hmac = NULL;
6598 /* Hook this new socket in to the bind_hash list. */
6599 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6600 sctp_local_bh_disable();
6601 sctp_spin_lock(&head->lock);
6602 pp = sctp_sk(oldsk)->bind_hash;
6603 sk_add_bind_node(newsk, &pp->owner);
6604 sctp_sk(newsk)->bind_hash = pp;
6605 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6606 sctp_spin_unlock(&head->lock);
6607 sctp_local_bh_enable();
6609 /* Copy the bind_addr list from the original endpoint to the new
6610 * endpoint so that we can handle restarts properly
6612 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6613 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6615 /* Move any messages in the old socket's receive queue that are for the
6616 * peeled off association to the new socket's receive queue.
6618 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6619 event = sctp_skb2event(skb);
6620 if (event->asoc == assoc) {
6621 sctp_sock_rfree_frag(skb);
6622 __skb_unlink(skb, &oldsk->sk_receive_queue);
6623 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6624 sctp_skb_set_owner_r_frag(skb, newsk);
6628 /* Clean up any messages pending delivery due to partial
6629 * delivery. Three cases:
6630 * 1) No partial deliver; no work.
6631 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6632 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6634 skb_queue_head_init(&newsp->pd_lobby);
6635 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6637 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6638 struct sk_buff_head *queue;
6640 /* Decide which queue to move pd_lobby skbs to. */
6641 if (assoc->ulpq.pd_mode) {
6642 queue = &newsp->pd_lobby;
6643 } else
6644 queue = &newsk->sk_receive_queue;
6646 /* Walk through the pd_lobby, looking for skbs that
6647 * need moved to the new socket.
6649 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6650 event = sctp_skb2event(skb);
6651 if (event->asoc == assoc) {
6652 sctp_sock_rfree_frag(skb);
6653 __skb_unlink(skb, &oldsp->pd_lobby);
6654 __skb_queue_tail(queue, skb);
6655 sctp_skb_set_owner_r_frag(skb, newsk);
6659 /* Clear up any skbs waiting for the partial
6660 * delivery to finish.
6662 if (assoc->ulpq.pd_mode)
6663 sctp_clear_pd(oldsk, NULL);
6667 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6668 sctp_sock_rfree_frag(skb);
6669 sctp_skb_set_owner_r_frag(skb, newsk);
6672 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6673 sctp_sock_rfree_frag(skb);
6674 sctp_skb_set_owner_r_frag(skb, newsk);
6677 /* Set the type of socket to indicate that it is peeled off from the
6678 * original UDP-style socket or created with the accept() call on a
6679 * TCP-style socket..
6681 newsp->type = type;
6683 /* Mark the new socket "in-use" by the user so that any packets
6684 * that may arrive on the association after we've moved it are
6685 * queued to the backlog. This prevents a potential race between
6686 * backlog processing on the old socket and new-packet processing
6687 * on the new socket.
6689 * The caller has just allocated newsk so we can guarantee that other
6690 * paths won't try to lock it and then oldsk.
6692 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6693 sctp_assoc_migrate(assoc, newsk);
6695 /* If the association on the newsk is already closed before accept()
6696 * is called, set RCV_SHUTDOWN flag.
6698 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6699 newsk->sk_shutdown |= RCV_SHUTDOWN;
6701 newsk->sk_state = SCTP_SS_ESTABLISHED;
6702 sctp_release_sock(newsk);
6706 /* This proto struct describes the ULP interface for SCTP. */
6707 struct proto sctp_prot = {
6708 .name = "SCTP",
6709 .owner = THIS_MODULE,
6710 .close = sctp_close,
6711 .connect = sctp_connect,
6712 .disconnect = sctp_disconnect,
6713 .accept = sctp_accept,
6714 .ioctl = sctp_ioctl,
6715 .init = sctp_init_sock,
6716 .destroy = sctp_destroy_sock,
6717 .shutdown = sctp_shutdown,
6718 .setsockopt = sctp_setsockopt,
6719 .getsockopt = sctp_getsockopt,
6720 .sendmsg = sctp_sendmsg,
6721 .recvmsg = sctp_recvmsg,
6722 .bind = sctp_bind,
6723 .backlog_rcv = sctp_backlog_rcv,
6724 .hash = sctp_hash,
6725 .unhash = sctp_unhash,
6726 .get_port = sctp_get_port,
6727 .obj_size = sizeof(struct sctp_sock),
6728 .sysctl_mem = sysctl_sctp_mem,
6729 .sysctl_rmem = sysctl_sctp_rmem,
6730 .sysctl_wmem = sysctl_sctp_wmem,
6731 .memory_pressure = &sctp_memory_pressure,
6732 .enter_memory_pressure = sctp_enter_memory_pressure,
6733 .memory_allocated = &sctp_memory_allocated,
6734 .sockets_allocated = &sctp_sockets_allocated,
6737 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6739 struct proto sctpv6_prot = {
6740 .name = "SCTPv6",
6741 .owner = THIS_MODULE,
6742 .close = sctp_close,
6743 .connect = sctp_connect,
6744 .disconnect = sctp_disconnect,
6745 .accept = sctp_accept,
6746 .ioctl = sctp_ioctl,
6747 .init = sctp_init_sock,
6748 .destroy = sctp_destroy_sock,
6749 .shutdown = sctp_shutdown,
6750 .setsockopt = sctp_setsockopt,
6751 .getsockopt = sctp_getsockopt,
6752 .sendmsg = sctp_sendmsg,
6753 .recvmsg = sctp_recvmsg,
6754 .bind = sctp_bind,
6755 .backlog_rcv = sctp_backlog_rcv,
6756 .hash = sctp_hash,
6757 .unhash = sctp_unhash,
6758 .get_port = sctp_get_port,
6759 .obj_size = sizeof(struct sctp6_sock),
6760 .sysctl_mem = sysctl_sctp_mem,
6761 .sysctl_rmem = sysctl_sctp_rmem,
6762 .sysctl_wmem = sysctl_sctp_wmem,
6763 .memory_pressure = &sctp_memory_pressure,
6764 .enter_memory_pressure = sctp_enter_memory_pressure,
6765 .memory_allocated = &sctp_memory_allocated,
6766 .sockets_allocated = &sctp_sockets_allocated,
6768 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */