x86: change copy_e820_map to append_e820_map
[linux-2.6/mini2440.git] / net / sctp / socket.c
blob0dbcde6758ea3158216e37af50df35ea56c37135
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(void)
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 /* Does this PF support this AF? */
312 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
313 return NULL;
315 /* If we get this far, af is valid. */
316 af = sctp_get_af_specific(addr->sa.sa_family);
318 if (len < af->sockaddr_len)
319 return NULL;
321 return af;
324 /* Bind a local address either to an endpoint or to an association. */
325 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
327 struct sctp_sock *sp = sctp_sk(sk);
328 struct sctp_endpoint *ep = sp->ep;
329 struct sctp_bind_addr *bp = &ep->base.bind_addr;
330 struct sctp_af *af;
331 unsigned short snum;
332 int ret = 0;
334 /* Common sockaddr verification. */
335 af = sctp_sockaddr_af(sp, addr, len);
336 if (!af) {
337 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
338 sk, addr, len);
339 return -EINVAL;
342 snum = ntohs(addr->v4.sin_port);
344 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
345 ", port: %d, new port: %d, len: %d)\n",
347 addr,
348 bp->port, snum,
349 len);
351 /* PF specific bind() address verification. */
352 if (!sp->pf->bind_verify(sp, addr))
353 return -EADDRNOTAVAIL;
355 /* We must either be unbound, or bind to the same port.
356 * It's OK to allow 0 ports if we are already bound.
357 * We'll just inhert an already bound port in this case
359 if (bp->port) {
360 if (!snum)
361 snum = bp->port;
362 else if (snum != bp->port) {
363 SCTP_DEBUG_PRINTK("sctp_do_bind:"
364 " New port %d does not match existing port "
365 "%d.\n", snum, bp->port);
366 return -EINVAL;
370 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
371 return -EACCES;
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
375 * detection.
377 addr->v4.sin_port = htons(snum);
378 if ((ret = sctp_get_port_local(sk, addr))) {
379 if (ret == (long) sk) {
380 /* This endpoint has a conflicting address. */
381 return -EINVAL;
382 } else {
383 return -EADDRINUSE;
387 /* Refresh ephemeral port. */
388 if (!bp->port)
389 bp->port = inet_sk(sk)->num;
391 /* Add the address to the bind address list.
392 * Use GFP_ATOMIC since BHs will be disabled.
394 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
396 /* Copy back into socket for getsockname() use. */
397 if (!ret) {
398 inet_sk(sk)->sport = htons(inet_sk(sk)->num);
399 af->to_sk_saddr(addr, sk);
402 return ret;
405 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
407 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408 * at any one time. If a sender, after sending an ASCONF chunk, decides
409 * it needs to transfer another ASCONF Chunk, it MUST wait until the
410 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411 * subsequent ASCONF. Note this restriction binds each side, so at any
412 * time two ASCONF may be in-transit on any given association (one sent
413 * from each endpoint).
415 static int sctp_send_asconf(struct sctp_association *asoc,
416 struct sctp_chunk *chunk)
418 int retval = 0;
420 /* If there is an outstanding ASCONF chunk, queue it for later
421 * transmission.
423 if (asoc->addip_last_asconf) {
424 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
425 goto out;
428 /* Hold the chunk until an ASCONF_ACK is received. */
429 sctp_chunk_hold(chunk);
430 retval = sctp_primitive_ASCONF(asoc, chunk);
431 if (retval)
432 sctp_chunk_free(chunk);
433 else
434 asoc->addip_last_asconf = chunk;
436 out:
437 return retval;
440 /* Add a list of addresses as bind addresses to local endpoint or
441 * association.
443 * Basically run through each address specified in the addrs/addrcnt
444 * array/length pair, determine if it is IPv6 or IPv4 and call
445 * sctp_do_bind() on it.
447 * If any of them fails, then the operation will be reversed and the
448 * ones that were added will be removed.
450 * Only sctp_setsockopt_bindx() is supposed to call this function.
452 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
454 int cnt;
455 int retval = 0;
456 void *addr_buf;
457 struct sockaddr *sa_addr;
458 struct sctp_af *af;
460 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
461 sk, addrs, addrcnt);
463 addr_buf = addrs;
464 for (cnt = 0; cnt < addrcnt; cnt++) {
465 /* The list may contain either IPv4 or IPv6 address;
466 * determine the address length for walking thru the list.
468 sa_addr = (struct sockaddr *)addr_buf;
469 af = sctp_get_af_specific(sa_addr->sa_family);
470 if (!af) {
471 retval = -EINVAL;
472 goto err_bindx_add;
475 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
476 af->sockaddr_len);
478 addr_buf += af->sockaddr_len;
480 err_bindx_add:
481 if (retval < 0) {
482 /* Failed. Cleanup the ones that have been added */
483 if (cnt > 0)
484 sctp_bindx_rem(sk, addrs, cnt);
485 return retval;
489 return retval;
492 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
493 * associations that are part of the endpoint indicating that a list of local
494 * addresses are added to the endpoint.
496 * If any of the addresses is already in the bind address list of the
497 * association, we do not send the chunk for that association. But it will not
498 * affect other associations.
500 * Only sctp_setsockopt_bindx() is supposed to call this function.
502 static int sctp_send_asconf_add_ip(struct sock *sk,
503 struct sockaddr *addrs,
504 int addrcnt)
506 struct sctp_sock *sp;
507 struct sctp_endpoint *ep;
508 struct sctp_association *asoc;
509 struct sctp_bind_addr *bp;
510 struct sctp_chunk *chunk;
511 struct sctp_sockaddr_entry *laddr;
512 union sctp_addr *addr;
513 union sctp_addr saveaddr;
514 void *addr_buf;
515 struct sctp_af *af;
516 struct list_head *p;
517 int i;
518 int retval = 0;
520 if (!sctp_addip_enable)
521 return retval;
523 sp = sctp_sk(sk);
524 ep = sp->ep;
526 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
527 __func__, sk, addrs, addrcnt);
529 list_for_each_entry(asoc, &ep->asocs, asocs) {
531 if (!asoc->peer.asconf_capable)
532 continue;
534 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
535 continue;
537 if (!sctp_state(asoc, ESTABLISHED))
538 continue;
540 /* Check if any address in the packed array of addresses is
541 * in the bind address list of the association. If so,
542 * do not send the asconf chunk to its peer, but continue with
543 * other associations.
545 addr_buf = addrs;
546 for (i = 0; i < addrcnt; i++) {
547 addr = (union sctp_addr *)addr_buf;
548 af = sctp_get_af_specific(addr->v4.sin_family);
549 if (!af) {
550 retval = -EINVAL;
551 goto out;
554 if (sctp_assoc_lookup_laddr(asoc, addr))
555 break;
557 addr_buf += af->sockaddr_len;
559 if (i < addrcnt)
560 continue;
562 /* Use the first valid address in bind addr list of
563 * association as Address Parameter of ASCONF CHUNK.
565 bp = &asoc->base.bind_addr;
566 p = bp->address_list.next;
567 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
568 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
569 addrcnt, SCTP_PARAM_ADD_IP);
570 if (!chunk) {
571 retval = -ENOMEM;
572 goto out;
575 retval = sctp_send_asconf(asoc, chunk);
576 if (retval)
577 goto out;
579 /* Add the new addresses to the bind address list with
580 * use_as_src set to 0.
582 addr_buf = addrs;
583 for (i = 0; i < addrcnt; i++) {
584 addr = (union sctp_addr *)addr_buf;
585 af = sctp_get_af_specific(addr->v4.sin_family);
586 memcpy(&saveaddr, addr, af->sockaddr_len);
587 retval = sctp_add_bind_addr(bp, &saveaddr,
588 SCTP_ADDR_NEW, GFP_ATOMIC);
589 addr_buf += af->sockaddr_len;
593 out:
594 return retval;
597 /* Remove a list of addresses from bind addresses list. Do not remove the
598 * last address.
600 * Basically run through each address specified in the addrs/addrcnt
601 * array/length pair, determine if it is IPv6 or IPv4 and call
602 * sctp_del_bind() on it.
604 * If any of them fails, then the operation will be reversed and the
605 * ones that were removed will be added back.
607 * At least one address has to be left; if only one address is
608 * available, the operation will return -EBUSY.
610 * Only sctp_setsockopt_bindx() is supposed to call this function.
612 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
614 struct sctp_sock *sp = sctp_sk(sk);
615 struct sctp_endpoint *ep = sp->ep;
616 int cnt;
617 struct sctp_bind_addr *bp = &ep->base.bind_addr;
618 int retval = 0;
619 void *addr_buf;
620 union sctp_addr *sa_addr;
621 struct sctp_af *af;
623 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
624 sk, addrs, addrcnt);
626 addr_buf = addrs;
627 for (cnt = 0; cnt < addrcnt; cnt++) {
628 /* If the bind address list is empty or if there is only one
629 * bind address, there is nothing more to be removed (we need
630 * at least one address here).
632 if (list_empty(&bp->address_list) ||
633 (sctp_list_single_entry(&bp->address_list))) {
634 retval = -EBUSY;
635 goto err_bindx_rem;
638 sa_addr = (union sctp_addr *)addr_buf;
639 af = sctp_get_af_specific(sa_addr->sa.sa_family);
640 if (!af) {
641 retval = -EINVAL;
642 goto err_bindx_rem;
645 if (!af->addr_valid(sa_addr, sp, NULL)) {
646 retval = -EADDRNOTAVAIL;
647 goto err_bindx_rem;
650 if (sa_addr->v4.sin_port != htons(bp->port)) {
651 retval = -EINVAL;
652 goto err_bindx_rem;
655 /* FIXME - There is probably a need to check if sk->sk_saddr and
656 * sk->sk_rcv_addr are currently set to one of the addresses to
657 * be removed. This is something which needs to be looked into
658 * when we are fixing the outstanding issues with multi-homing
659 * socket routing and failover schemes. Refer to comments in
660 * sctp_do_bind(). -daisy
662 retval = sctp_del_bind_addr(bp, sa_addr);
664 addr_buf += af->sockaddr_len;
665 err_bindx_rem:
666 if (retval < 0) {
667 /* Failed. Add the ones that has been removed back */
668 if (cnt > 0)
669 sctp_bindx_add(sk, addrs, cnt);
670 return retval;
674 return retval;
677 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
678 * the associations that are part of the endpoint indicating that a list of
679 * local addresses are removed from the endpoint.
681 * If any of the addresses is already in the bind address list of the
682 * association, we do not send the chunk for that association. But it will not
683 * affect other associations.
685 * Only sctp_setsockopt_bindx() is supposed to call this function.
687 static int sctp_send_asconf_del_ip(struct sock *sk,
688 struct sockaddr *addrs,
689 int addrcnt)
691 struct sctp_sock *sp;
692 struct sctp_endpoint *ep;
693 struct sctp_association *asoc;
694 struct sctp_transport *transport;
695 struct sctp_bind_addr *bp;
696 struct sctp_chunk *chunk;
697 union sctp_addr *laddr;
698 void *addr_buf;
699 struct sctp_af *af;
700 struct sctp_sockaddr_entry *saddr;
701 int i;
702 int retval = 0;
704 if (!sctp_addip_enable)
705 return retval;
707 sp = sctp_sk(sk);
708 ep = sp->ep;
710 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
711 __func__, sk, addrs, addrcnt);
713 list_for_each_entry(asoc, &ep->asocs, asocs) {
715 if (!asoc->peer.asconf_capable)
716 continue;
718 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
719 continue;
721 if (!sctp_state(asoc, ESTABLISHED))
722 continue;
724 /* Check if any address in the packed array of addresses is
725 * not present in the bind address list of the association.
726 * If so, do not send the asconf chunk to its peer, but
727 * continue with other associations.
729 addr_buf = addrs;
730 for (i = 0; i < addrcnt; i++) {
731 laddr = (union sctp_addr *)addr_buf;
732 af = sctp_get_af_specific(laddr->v4.sin_family);
733 if (!af) {
734 retval = -EINVAL;
735 goto out;
738 if (!sctp_assoc_lookup_laddr(asoc, laddr))
739 break;
741 addr_buf += af->sockaddr_len;
743 if (i < addrcnt)
744 continue;
746 /* Find one address in the association's bind address list
747 * that is not in the packed array of addresses. This is to
748 * make sure that we do not delete all the addresses in the
749 * association.
751 bp = &asoc->base.bind_addr;
752 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
753 addrcnt, sp);
754 if (!laddr)
755 continue;
757 /* We do not need RCU protection throughout this loop
758 * because this is done under a socket lock from the
759 * setsockopt call.
761 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
762 SCTP_PARAM_DEL_IP);
763 if (!chunk) {
764 retval = -ENOMEM;
765 goto out;
768 /* Reset use_as_src flag for the addresses in the bind address
769 * list that are to be deleted.
771 addr_buf = addrs;
772 for (i = 0; i < addrcnt; i++) {
773 laddr = (union sctp_addr *)addr_buf;
774 af = sctp_get_af_specific(laddr->v4.sin_family);
775 list_for_each_entry(saddr, &bp->address_list, list) {
776 if (sctp_cmp_addr_exact(&saddr->a, laddr))
777 saddr->state = SCTP_ADDR_DEL;
779 addr_buf += af->sockaddr_len;
782 /* Update the route and saddr entries for all the transports
783 * as some of the addresses in the bind address list are
784 * about to be deleted and cannot be used as source addresses.
786 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
787 transports) {
788 dst_release(transport->dst);
789 sctp_transport_route(transport, NULL,
790 sctp_sk(asoc->base.sk));
793 retval = sctp_send_asconf(asoc, chunk);
795 out:
796 return retval;
799 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
801 * API 8.1
802 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
803 * int flags);
805 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
806 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
807 * or IPv6 addresses.
809 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
810 * Section 3.1.2 for this usage.
812 * addrs is a pointer to an array of one or more socket addresses. Each
813 * address is contained in its appropriate structure (i.e. struct
814 * sockaddr_in or struct sockaddr_in6) the family of the address type
815 * must be used to distinguish the address length (note that this
816 * representation is termed a "packed array" of addresses). The caller
817 * specifies the number of addresses in the array with addrcnt.
819 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
820 * -1, and sets errno to the appropriate error code.
822 * For SCTP, the port given in each socket address must be the same, or
823 * sctp_bindx() will fail, setting errno to EINVAL.
825 * The flags parameter is formed from the bitwise OR of zero or more of
826 * the following currently defined flags:
828 * SCTP_BINDX_ADD_ADDR
830 * SCTP_BINDX_REM_ADDR
832 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
833 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
834 * addresses from the association. The two flags are mutually exclusive;
835 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
836 * not remove all addresses from an association; sctp_bindx() will
837 * reject such an attempt with EINVAL.
839 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
840 * additional addresses with an endpoint after calling bind(). Or use
841 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
842 * socket is associated with so that no new association accepted will be
843 * associated with those addresses. If the endpoint supports dynamic
844 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
845 * endpoint to send the appropriate message to the peer to change the
846 * peers address lists.
848 * Adding and removing addresses from a connected association is
849 * optional functionality. Implementations that do not support this
850 * functionality should return EOPNOTSUPP.
852 * Basically do nothing but copying the addresses from user to kernel
853 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
854 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
855 * from userspace.
857 * We don't use copy_from_user() for optimization: we first do the
858 * sanity checks (buffer size -fast- and access check-healthy
859 * pointer); if all of those succeed, then we can alloc the memory
860 * (expensive operation) needed to copy the data to kernel. Then we do
861 * the copying without checking the user space area
862 * (__copy_from_user()).
864 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
865 * it.
867 * sk The sk of the socket
868 * addrs The pointer to the addresses in user land
869 * addrssize Size of the addrs buffer
870 * op Operation to perform (add or remove, see the flags of
871 * sctp_bindx)
873 * Returns 0 if ok, <0 errno code on error.
875 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
876 struct sockaddr __user *addrs,
877 int addrs_size, int op)
879 struct sockaddr *kaddrs;
880 int err;
881 int addrcnt = 0;
882 int walk_size = 0;
883 struct sockaddr *sa_addr;
884 void *addr_buf;
885 struct sctp_af *af;
887 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
888 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
890 if (unlikely(addrs_size <= 0))
891 return -EINVAL;
893 /* Check the user passed a healthy pointer. */
894 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
895 return -EFAULT;
897 /* Alloc space for the address array in kernel memory. */
898 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
899 if (unlikely(!kaddrs))
900 return -ENOMEM;
902 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
903 kfree(kaddrs);
904 return -EFAULT;
907 /* Walk through the addrs buffer and count the number of addresses. */
908 addr_buf = kaddrs;
909 while (walk_size < addrs_size) {
910 sa_addr = (struct sockaddr *)addr_buf;
911 af = sctp_get_af_specific(sa_addr->sa_family);
913 /* If the address family is not supported or if this address
914 * causes the address buffer to overflow return EINVAL.
916 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
917 kfree(kaddrs);
918 return -EINVAL;
920 addrcnt++;
921 addr_buf += af->sockaddr_len;
922 walk_size += af->sockaddr_len;
925 /* Do the work. */
926 switch (op) {
927 case SCTP_BINDX_ADD_ADDR:
928 err = sctp_bindx_add(sk, kaddrs, addrcnt);
929 if (err)
930 goto out;
931 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
932 break;
934 case SCTP_BINDX_REM_ADDR:
935 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
936 if (err)
937 goto out;
938 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
939 break;
941 default:
942 err = -EINVAL;
943 break;
946 out:
947 kfree(kaddrs);
949 return err;
952 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
954 * Common routine for handling connect() and sctp_connectx().
955 * Connect will come in with just a single address.
957 static int __sctp_connect(struct sock* sk,
958 struct sockaddr *kaddrs,
959 int addrs_size)
961 struct sctp_sock *sp;
962 struct sctp_endpoint *ep;
963 struct sctp_association *asoc = NULL;
964 struct sctp_association *asoc2;
965 struct sctp_transport *transport;
966 union sctp_addr to;
967 struct sctp_af *af;
968 sctp_scope_t scope;
969 long timeo;
970 int err = 0;
971 int addrcnt = 0;
972 int walk_size = 0;
973 union sctp_addr *sa_addr = NULL;
974 void *addr_buf;
975 unsigned short port;
976 unsigned int f_flags = 0;
978 sp = sctp_sk(sk);
979 ep = sp->ep;
981 /* connect() cannot be done on a socket that is already in ESTABLISHED
982 * state - UDP-style peeled off socket or a TCP-style socket that
983 * is already connected.
984 * It cannot be done even on a TCP-style listening socket.
986 if (sctp_sstate(sk, ESTABLISHED) ||
987 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
988 err = -EISCONN;
989 goto out_free;
992 /* Walk through the addrs buffer and count the number of addresses. */
993 addr_buf = kaddrs;
994 while (walk_size < addrs_size) {
995 sa_addr = (union sctp_addr *)addr_buf;
996 af = sctp_get_af_specific(sa_addr->sa.sa_family);
997 port = ntohs(sa_addr->v4.sin_port);
999 /* If the address family is not supported or if this address
1000 * causes the address buffer to overflow return EINVAL.
1002 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1003 err = -EINVAL;
1004 goto out_free;
1007 /* Save current address so we can work with it */
1008 memcpy(&to, sa_addr, af->sockaddr_len);
1010 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1011 if (err)
1012 goto out_free;
1014 /* Make sure the destination port is correctly set
1015 * in all addresses.
1017 if (asoc && asoc->peer.port && asoc->peer.port != port)
1018 goto out_free;
1021 /* Check if there already is a matching association on the
1022 * endpoint (other than the one created here).
1024 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1025 if (asoc2 && asoc2 != asoc) {
1026 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1027 err = -EISCONN;
1028 else
1029 err = -EALREADY;
1030 goto out_free;
1033 /* If we could not find a matching association on the endpoint,
1034 * make sure that there is no peeled-off association matching
1035 * the peer address even on another socket.
1037 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1038 err = -EADDRNOTAVAIL;
1039 goto out_free;
1042 if (!asoc) {
1043 /* If a bind() or sctp_bindx() is not called prior to
1044 * an sctp_connectx() call, the system picks an
1045 * ephemeral port and will choose an address set
1046 * equivalent to binding with a wildcard address.
1048 if (!ep->base.bind_addr.port) {
1049 if (sctp_autobind(sk)) {
1050 err = -EAGAIN;
1051 goto out_free;
1053 } else {
1055 * If an unprivileged user inherits a 1-many
1056 * style socket with open associations on a
1057 * privileged port, it MAY be permitted to
1058 * accept new associations, but it SHOULD NOT
1059 * be permitted to open new associations.
1061 if (ep->base.bind_addr.port < PROT_SOCK &&
1062 !capable(CAP_NET_BIND_SERVICE)) {
1063 err = -EACCES;
1064 goto out_free;
1068 scope = sctp_scope(&to);
1069 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1070 if (!asoc) {
1071 err = -ENOMEM;
1072 goto out_free;
1076 /* Prime the peer's transport structures. */
1077 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1078 SCTP_UNKNOWN);
1079 if (!transport) {
1080 err = -ENOMEM;
1081 goto out_free;
1084 addrcnt++;
1085 addr_buf += af->sockaddr_len;
1086 walk_size += af->sockaddr_len;
1089 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1090 if (err < 0) {
1091 goto out_free;
1094 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1095 if (err < 0) {
1096 goto out_free;
1099 /* Initialize sk's dport and daddr for getpeername() */
1100 inet_sk(sk)->dport = htons(asoc->peer.port);
1101 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1102 af->to_sk_daddr(sa_addr, sk);
1103 sk->sk_err = 0;
1105 /* in-kernel sockets don't generally have a file allocated to them
1106 * if all they do is call sock_create_kern().
1108 if (sk->sk_socket->file)
1109 f_flags = sk->sk_socket->file->f_flags;
1111 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1113 err = sctp_wait_for_connect(asoc, &timeo);
1115 /* Don't free association on exit. */
1116 asoc = NULL;
1118 out_free:
1120 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1121 " kaddrs: %p err: %d\n",
1122 asoc, kaddrs, err);
1123 if (asoc)
1124 sctp_association_free(asoc);
1125 return err;
1128 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1130 * API 8.9
1131 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1133 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1134 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1135 * or IPv6 addresses.
1137 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1138 * Section 3.1.2 for this usage.
1140 * addrs is a pointer to an array of one or more socket addresses. Each
1141 * address is contained in its appropriate structure (i.e. struct
1142 * sockaddr_in or struct sockaddr_in6) the family of the address type
1143 * must be used to distengish the address length (note that this
1144 * representation is termed a "packed array" of addresses). The caller
1145 * specifies the number of addresses in the array with addrcnt.
1147 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1148 * -1, and sets errno to the appropriate error code.
1150 * For SCTP, the port given in each socket address must be the same, or
1151 * sctp_connectx() will fail, setting errno to EINVAL.
1153 * An application can use sctp_connectx to initiate an association with
1154 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1155 * allows a caller to specify multiple addresses at which a peer can be
1156 * reached. The way the SCTP stack uses the list of addresses to set up
1157 * the association is implementation dependant. This function only
1158 * specifies that the stack will try to make use of all the addresses in
1159 * the list when needed.
1161 * Note that the list of addresses passed in is only used for setting up
1162 * the association. It does not necessarily equal the set of addresses
1163 * the peer uses for the resulting association. If the caller wants to
1164 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1165 * retrieve them after the association has been set up.
1167 * Basically do nothing but copying the addresses from user to kernel
1168 * land and invoking either sctp_connectx(). This is used for tunneling
1169 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1171 * We don't use copy_from_user() for optimization: we first do the
1172 * sanity checks (buffer size -fast- and access check-healthy
1173 * pointer); if all of those succeed, then we can alloc the memory
1174 * (expensive operation) needed to copy the data to kernel. Then we do
1175 * the copying without checking the user space area
1176 * (__copy_from_user()).
1178 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1179 * it.
1181 * sk The sk of the socket
1182 * addrs The pointer to the addresses in user land
1183 * addrssize Size of the addrs buffer
1185 * Returns 0 if ok, <0 errno code on error.
1187 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1188 struct sockaddr __user *addrs,
1189 int addrs_size)
1191 int err = 0;
1192 struct sockaddr *kaddrs;
1194 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1195 __func__, sk, addrs, addrs_size);
1197 if (unlikely(addrs_size <= 0))
1198 return -EINVAL;
1200 /* Check the user passed a healthy pointer. */
1201 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1202 return -EFAULT;
1204 /* Alloc space for the address array in kernel memory. */
1205 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1206 if (unlikely(!kaddrs))
1207 return -ENOMEM;
1209 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1210 err = -EFAULT;
1211 } else {
1212 err = __sctp_connect(sk, kaddrs, addrs_size);
1215 kfree(kaddrs);
1216 return err;
1219 /* API 3.1.4 close() - UDP Style Syntax
1220 * Applications use close() to perform graceful shutdown (as described in
1221 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1222 * by a UDP-style socket.
1224 * The syntax is
1226 * ret = close(int sd);
1228 * sd - the socket descriptor of the associations to be closed.
1230 * To gracefully shutdown a specific association represented by the
1231 * UDP-style socket, an application should use the sendmsg() call,
1232 * passing no user data, but including the appropriate flag in the
1233 * ancillary data (see Section xxxx).
1235 * If sd in the close() call is a branched-off socket representing only
1236 * one association, the shutdown is performed on that association only.
1238 * 4.1.6 close() - TCP Style Syntax
1240 * Applications use close() to gracefully close down an association.
1242 * The syntax is:
1244 * int close(int sd);
1246 * sd - the socket descriptor of the association to be closed.
1248 * After an application calls close() on a socket descriptor, no further
1249 * socket operations will succeed on that descriptor.
1251 * API 7.1.4 SO_LINGER
1253 * An application using the TCP-style socket can use this option to
1254 * perform the SCTP ABORT primitive. The linger option structure is:
1256 * struct linger {
1257 * int l_onoff; // option on/off
1258 * int l_linger; // linger time
1259 * };
1261 * To enable the option, set l_onoff to 1. If the l_linger value is set
1262 * to 0, calling close() is the same as the ABORT primitive. If the
1263 * value is set to a negative value, the setsockopt() call will return
1264 * an error. If the value is set to a positive value linger_time, the
1265 * close() can be blocked for at most linger_time ms. If the graceful
1266 * shutdown phase does not finish during this period, close() will
1267 * return but the graceful shutdown phase continues in the system.
1269 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1271 struct sctp_endpoint *ep;
1272 struct sctp_association *asoc;
1273 struct list_head *pos, *temp;
1275 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1277 sctp_lock_sock(sk);
1278 sk->sk_shutdown = SHUTDOWN_MASK;
1280 ep = sctp_sk(sk)->ep;
1282 /* Walk all associations on an endpoint. */
1283 list_for_each_safe(pos, temp, &ep->asocs) {
1284 asoc = list_entry(pos, struct sctp_association, asocs);
1286 if (sctp_style(sk, TCP)) {
1287 /* A closed association can still be in the list if
1288 * it belongs to a TCP-style listening socket that is
1289 * not yet accepted. If so, free it. If not, send an
1290 * ABORT or SHUTDOWN based on the linger options.
1292 if (sctp_state(asoc, CLOSED)) {
1293 sctp_unhash_established(asoc);
1294 sctp_association_free(asoc);
1295 continue;
1299 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1300 struct sctp_chunk *chunk;
1302 chunk = sctp_make_abort_user(asoc, NULL, 0);
1303 if (chunk)
1304 sctp_primitive_ABORT(asoc, chunk);
1305 } else
1306 sctp_primitive_SHUTDOWN(asoc, NULL);
1309 /* Clean up any skbs sitting on the receive queue. */
1310 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1311 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1313 /* On a TCP-style socket, block for at most linger_time if set. */
1314 if (sctp_style(sk, TCP) && timeout)
1315 sctp_wait_for_close(sk, timeout);
1317 /* This will run the backlog queue. */
1318 sctp_release_sock(sk);
1320 /* Supposedly, no process has access to the socket, but
1321 * the net layers still may.
1323 sctp_local_bh_disable();
1324 sctp_bh_lock_sock(sk);
1326 /* Hold the sock, since sk_common_release() will put sock_put()
1327 * and we have just a little more cleanup.
1329 sock_hold(sk);
1330 sk_common_release(sk);
1332 sctp_bh_unlock_sock(sk);
1333 sctp_local_bh_enable();
1335 sock_put(sk);
1337 SCTP_DBG_OBJCNT_DEC(sock);
1340 /* Handle EPIPE error. */
1341 static int sctp_error(struct sock *sk, int flags, int err)
1343 if (err == -EPIPE)
1344 err = sock_error(sk) ? : -EPIPE;
1345 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1346 send_sig(SIGPIPE, current, 0);
1347 return err;
1350 /* API 3.1.3 sendmsg() - UDP Style Syntax
1352 * An application uses sendmsg() and recvmsg() calls to transmit data to
1353 * and receive data from its peer.
1355 * ssize_t sendmsg(int socket, const struct msghdr *message,
1356 * int flags);
1358 * socket - the socket descriptor of the endpoint.
1359 * message - pointer to the msghdr structure which contains a single
1360 * user message and possibly some ancillary data.
1362 * See Section 5 for complete description of the data
1363 * structures.
1365 * flags - flags sent or received with the user message, see Section
1366 * 5 for complete description of the flags.
1368 * Note: This function could use a rewrite especially when explicit
1369 * connect support comes in.
1371 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1373 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1375 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1376 struct msghdr *msg, size_t msg_len)
1378 struct sctp_sock *sp;
1379 struct sctp_endpoint *ep;
1380 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1381 struct sctp_transport *transport, *chunk_tp;
1382 struct sctp_chunk *chunk;
1383 union sctp_addr to;
1384 struct sockaddr *msg_name = NULL;
1385 struct sctp_sndrcvinfo default_sinfo = { 0 };
1386 struct sctp_sndrcvinfo *sinfo;
1387 struct sctp_initmsg *sinit;
1388 sctp_assoc_t associd = 0;
1389 sctp_cmsgs_t cmsgs = { NULL };
1390 int err;
1391 sctp_scope_t scope;
1392 long timeo;
1393 __u16 sinfo_flags = 0;
1394 struct sctp_datamsg *datamsg;
1395 int msg_flags = msg->msg_flags;
1397 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1398 sk, msg, msg_len);
1400 err = 0;
1401 sp = sctp_sk(sk);
1402 ep = sp->ep;
1404 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1406 /* We cannot send a message over a TCP-style listening socket. */
1407 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1408 err = -EPIPE;
1409 goto out_nounlock;
1412 /* Parse out the SCTP CMSGs. */
1413 err = sctp_msghdr_parse(msg, &cmsgs);
1415 if (err) {
1416 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1417 goto out_nounlock;
1420 /* Fetch the destination address for this packet. This
1421 * address only selects the association--it is not necessarily
1422 * the address we will send to.
1423 * For a peeled-off socket, msg_name is ignored.
1425 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1426 int msg_namelen = msg->msg_namelen;
1428 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1429 msg_namelen);
1430 if (err)
1431 return err;
1433 if (msg_namelen > sizeof(to))
1434 msg_namelen = sizeof(to);
1435 memcpy(&to, msg->msg_name, msg_namelen);
1436 msg_name = msg->msg_name;
1439 sinfo = cmsgs.info;
1440 sinit = cmsgs.init;
1442 /* Did the user specify SNDRCVINFO? */
1443 if (sinfo) {
1444 sinfo_flags = sinfo->sinfo_flags;
1445 associd = sinfo->sinfo_assoc_id;
1448 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1449 msg_len, sinfo_flags);
1451 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1452 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1453 err = -EINVAL;
1454 goto out_nounlock;
1457 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1458 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1459 * If SCTP_ABORT is set, the message length could be non zero with
1460 * the msg_iov set to the user abort reason.
1462 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1463 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1464 err = -EINVAL;
1465 goto out_nounlock;
1468 /* If SCTP_ADDR_OVER is set, there must be an address
1469 * specified in msg_name.
1471 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1472 err = -EINVAL;
1473 goto out_nounlock;
1476 transport = NULL;
1478 SCTP_DEBUG_PRINTK("About to look up association.\n");
1480 sctp_lock_sock(sk);
1482 /* If a msg_name has been specified, assume this is to be used. */
1483 if (msg_name) {
1484 /* Look for a matching association on the endpoint. */
1485 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1486 if (!asoc) {
1487 /* If we could not find a matching association on the
1488 * endpoint, make sure that it is not a TCP-style
1489 * socket that already has an association or there is
1490 * no peeled-off association on another socket.
1492 if ((sctp_style(sk, TCP) &&
1493 sctp_sstate(sk, ESTABLISHED)) ||
1494 sctp_endpoint_is_peeled_off(ep, &to)) {
1495 err = -EADDRNOTAVAIL;
1496 goto out_unlock;
1499 } else {
1500 asoc = sctp_id2assoc(sk, associd);
1501 if (!asoc) {
1502 err = -EPIPE;
1503 goto out_unlock;
1507 if (asoc) {
1508 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1510 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1511 * socket that has an association in CLOSED state. This can
1512 * happen when an accepted socket has an association that is
1513 * already CLOSED.
1515 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1516 err = -EPIPE;
1517 goto out_unlock;
1520 if (sinfo_flags & SCTP_EOF) {
1521 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1522 asoc);
1523 sctp_primitive_SHUTDOWN(asoc, NULL);
1524 err = 0;
1525 goto out_unlock;
1527 if (sinfo_flags & SCTP_ABORT) {
1529 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1530 if (!chunk) {
1531 err = -ENOMEM;
1532 goto out_unlock;
1535 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1536 sctp_primitive_ABORT(asoc, chunk);
1537 err = 0;
1538 goto out_unlock;
1542 /* Do we need to create the association? */
1543 if (!asoc) {
1544 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1546 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1547 err = -EINVAL;
1548 goto out_unlock;
1551 /* Check for invalid stream against the stream counts,
1552 * either the default or the user specified stream counts.
1554 if (sinfo) {
1555 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1556 /* Check against the defaults. */
1557 if (sinfo->sinfo_stream >=
1558 sp->initmsg.sinit_num_ostreams) {
1559 err = -EINVAL;
1560 goto out_unlock;
1562 } else {
1563 /* Check against the requested. */
1564 if (sinfo->sinfo_stream >=
1565 sinit->sinit_num_ostreams) {
1566 err = -EINVAL;
1567 goto out_unlock;
1573 * API 3.1.2 bind() - UDP Style Syntax
1574 * If a bind() or sctp_bindx() is not called prior to a
1575 * sendmsg() call that initiates a new association, the
1576 * system picks an ephemeral port and will choose an address
1577 * set equivalent to binding with a wildcard address.
1579 if (!ep->base.bind_addr.port) {
1580 if (sctp_autobind(sk)) {
1581 err = -EAGAIN;
1582 goto out_unlock;
1584 } else {
1586 * If an unprivileged user inherits a one-to-many
1587 * style socket with open associations on a privileged
1588 * port, it MAY be permitted to accept new associations,
1589 * but it SHOULD NOT be permitted to open new
1590 * associations.
1592 if (ep->base.bind_addr.port < PROT_SOCK &&
1593 !capable(CAP_NET_BIND_SERVICE)) {
1594 err = -EACCES;
1595 goto out_unlock;
1599 scope = sctp_scope(&to);
1600 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1601 if (!new_asoc) {
1602 err = -ENOMEM;
1603 goto out_unlock;
1605 asoc = new_asoc;
1607 /* If the SCTP_INIT ancillary data is specified, set all
1608 * the association init values accordingly.
1610 if (sinit) {
1611 if (sinit->sinit_num_ostreams) {
1612 asoc->c.sinit_num_ostreams =
1613 sinit->sinit_num_ostreams;
1615 if (sinit->sinit_max_instreams) {
1616 asoc->c.sinit_max_instreams =
1617 sinit->sinit_max_instreams;
1619 if (sinit->sinit_max_attempts) {
1620 asoc->max_init_attempts
1621 = sinit->sinit_max_attempts;
1623 if (sinit->sinit_max_init_timeo) {
1624 asoc->max_init_timeo =
1625 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1629 /* Prime the peer's transport structures. */
1630 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1631 if (!transport) {
1632 err = -ENOMEM;
1633 goto out_free;
1635 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1636 if (err < 0) {
1637 err = -ENOMEM;
1638 goto out_free;
1642 /* ASSERT: we have a valid association at this point. */
1643 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1645 if (!sinfo) {
1646 /* If the user didn't specify SNDRCVINFO, make up one with
1647 * some defaults.
1649 default_sinfo.sinfo_stream = asoc->default_stream;
1650 default_sinfo.sinfo_flags = asoc->default_flags;
1651 default_sinfo.sinfo_ppid = asoc->default_ppid;
1652 default_sinfo.sinfo_context = asoc->default_context;
1653 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1654 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1655 sinfo = &default_sinfo;
1658 /* API 7.1.7, the sndbuf size per association bounds the
1659 * maximum size of data that can be sent in a single send call.
1661 if (msg_len > sk->sk_sndbuf) {
1662 err = -EMSGSIZE;
1663 goto out_free;
1666 if (asoc->pmtu_pending)
1667 sctp_assoc_pending_pmtu(asoc);
1669 /* If fragmentation is disabled and the message length exceeds the
1670 * association fragmentation point, return EMSGSIZE. The I-D
1671 * does not specify what this error is, but this looks like
1672 * a great fit.
1674 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1675 err = -EMSGSIZE;
1676 goto out_free;
1679 if (sinfo) {
1680 /* Check for invalid stream. */
1681 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1682 err = -EINVAL;
1683 goto out_free;
1687 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1688 if (!sctp_wspace(asoc)) {
1689 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1690 if (err)
1691 goto out_free;
1694 /* If an address is passed with the sendto/sendmsg call, it is used
1695 * to override the primary destination address in the TCP model, or
1696 * when SCTP_ADDR_OVER flag is set in the UDP model.
1698 if ((sctp_style(sk, TCP) && msg_name) ||
1699 (sinfo_flags & SCTP_ADDR_OVER)) {
1700 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1701 if (!chunk_tp) {
1702 err = -EINVAL;
1703 goto out_free;
1705 } else
1706 chunk_tp = NULL;
1708 /* Auto-connect, if we aren't connected already. */
1709 if (sctp_state(asoc, CLOSED)) {
1710 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1711 if (err < 0)
1712 goto out_free;
1713 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1716 /* Break the message into multiple chunks of maximum size. */
1717 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1718 if (!datamsg) {
1719 err = -ENOMEM;
1720 goto out_free;
1723 /* Now send the (possibly) fragmented message. */
1724 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1725 sctp_chunk_hold(chunk);
1727 /* Do accounting for the write space. */
1728 sctp_set_owner_w(chunk);
1730 chunk->transport = chunk_tp;
1732 /* Send it to the lower layers. Note: all chunks
1733 * must either fail or succeed. The lower layer
1734 * works that way today. Keep it that way or this
1735 * breaks.
1737 err = sctp_primitive_SEND(asoc, chunk);
1738 /* Did the lower layer accept the chunk? */
1739 if (err)
1740 sctp_chunk_free(chunk);
1741 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1744 sctp_datamsg_put(datamsg);
1745 if (err)
1746 goto out_free;
1747 else
1748 err = msg_len;
1750 /* If we are already past ASSOCIATE, the lower
1751 * layers are responsible for association cleanup.
1753 goto out_unlock;
1755 out_free:
1756 if (new_asoc)
1757 sctp_association_free(asoc);
1758 out_unlock:
1759 sctp_release_sock(sk);
1761 out_nounlock:
1762 return sctp_error(sk, msg_flags, err);
1764 #if 0
1765 do_sock_err:
1766 if (msg_len)
1767 err = msg_len;
1768 else
1769 err = sock_error(sk);
1770 goto out;
1772 do_interrupted:
1773 if (msg_len)
1774 err = msg_len;
1775 goto out;
1776 #endif /* 0 */
1779 /* This is an extended version of skb_pull() that removes the data from the
1780 * start of a skb even when data is spread across the list of skb's in the
1781 * frag_list. len specifies the total amount of data that needs to be removed.
1782 * when 'len' bytes could be removed from the skb, it returns 0.
1783 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1784 * could not be removed.
1786 static int sctp_skb_pull(struct sk_buff *skb, int len)
1788 struct sk_buff *list;
1789 int skb_len = skb_headlen(skb);
1790 int rlen;
1792 if (len <= skb_len) {
1793 __skb_pull(skb, len);
1794 return 0;
1796 len -= skb_len;
1797 __skb_pull(skb, skb_len);
1799 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1800 rlen = sctp_skb_pull(list, len);
1801 skb->len -= (len-rlen);
1802 skb->data_len -= (len-rlen);
1804 if (!rlen)
1805 return 0;
1807 len = rlen;
1810 return len;
1813 /* API 3.1.3 recvmsg() - UDP Style Syntax
1815 * ssize_t recvmsg(int socket, struct msghdr *message,
1816 * int flags);
1818 * socket - the socket descriptor of the endpoint.
1819 * message - pointer to the msghdr structure which contains a single
1820 * user message and possibly some ancillary data.
1822 * See Section 5 for complete description of the data
1823 * structures.
1825 * flags - flags sent or received with the user message, see Section
1826 * 5 for complete description of the flags.
1828 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1830 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1831 struct msghdr *msg, size_t len, int noblock,
1832 int flags, int *addr_len)
1834 struct sctp_ulpevent *event = NULL;
1835 struct sctp_sock *sp = sctp_sk(sk);
1836 struct sk_buff *skb;
1837 int copied;
1838 int err = 0;
1839 int skb_len;
1841 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1842 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1843 "len", len, "knoblauch", noblock,
1844 "flags", flags, "addr_len", addr_len);
1846 sctp_lock_sock(sk);
1848 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1849 err = -ENOTCONN;
1850 goto out;
1853 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1854 if (!skb)
1855 goto out;
1857 /* Get the total length of the skb including any skb's in the
1858 * frag_list.
1860 skb_len = skb->len;
1862 copied = skb_len;
1863 if (copied > len)
1864 copied = len;
1866 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1868 event = sctp_skb2event(skb);
1870 if (err)
1871 goto out_free;
1873 sock_recv_timestamp(msg, sk, skb);
1874 if (sctp_ulpevent_is_notification(event)) {
1875 msg->msg_flags |= MSG_NOTIFICATION;
1876 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1877 } else {
1878 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1881 /* Check if we allow SCTP_SNDRCVINFO. */
1882 if (sp->subscribe.sctp_data_io_event)
1883 sctp_ulpevent_read_sndrcvinfo(event, msg);
1884 #if 0
1885 /* FIXME: we should be calling IP/IPv6 layers. */
1886 if (sk->sk_protinfo.af_inet.cmsg_flags)
1887 ip_cmsg_recv(msg, skb);
1888 #endif
1890 err = copied;
1892 /* If skb's length exceeds the user's buffer, update the skb and
1893 * push it back to the receive_queue so that the next call to
1894 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1896 if (skb_len > copied) {
1897 msg->msg_flags &= ~MSG_EOR;
1898 if (flags & MSG_PEEK)
1899 goto out_free;
1900 sctp_skb_pull(skb, copied);
1901 skb_queue_head(&sk->sk_receive_queue, skb);
1903 /* When only partial message is copied to the user, increase
1904 * rwnd by that amount. If all the data in the skb is read,
1905 * rwnd is updated when the event is freed.
1907 if (!sctp_ulpevent_is_notification(event))
1908 sctp_assoc_rwnd_increase(event->asoc, copied);
1909 goto out;
1910 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1911 (event->msg_flags & MSG_EOR))
1912 msg->msg_flags |= MSG_EOR;
1913 else
1914 msg->msg_flags &= ~MSG_EOR;
1916 out_free:
1917 if (flags & MSG_PEEK) {
1918 /* Release the skb reference acquired after peeking the skb in
1919 * sctp_skb_recv_datagram().
1921 kfree_skb(skb);
1922 } else {
1923 /* Free the event which includes releasing the reference to
1924 * the owner of the skb, freeing the skb and updating the
1925 * rwnd.
1927 sctp_ulpevent_free(event);
1929 out:
1930 sctp_release_sock(sk);
1931 return err;
1934 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1936 * This option is a on/off flag. If enabled no SCTP message
1937 * fragmentation will be performed. Instead if a message being sent
1938 * exceeds the current PMTU size, the message will NOT be sent and
1939 * instead a error will be indicated to the user.
1941 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1942 char __user *optval, int optlen)
1944 int val;
1946 if (optlen < sizeof(int))
1947 return -EINVAL;
1949 if (get_user(val, (int __user *)optval))
1950 return -EFAULT;
1952 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1954 return 0;
1957 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1958 int optlen)
1960 if (optlen > sizeof(struct sctp_event_subscribe))
1961 return -EINVAL;
1962 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1963 return -EFAULT;
1964 return 0;
1967 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1969 * This socket option is applicable to the UDP-style socket only. When
1970 * set it will cause associations that are idle for more than the
1971 * specified number of seconds to automatically close. An association
1972 * being idle is defined an association that has NOT sent or received
1973 * user data. The special value of '0' indicates that no automatic
1974 * close of any associations should be performed. The option expects an
1975 * integer defining the number of seconds of idle time before an
1976 * association is closed.
1978 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1979 int optlen)
1981 struct sctp_sock *sp = sctp_sk(sk);
1983 /* Applicable to UDP-style socket only */
1984 if (sctp_style(sk, TCP))
1985 return -EOPNOTSUPP;
1986 if (optlen != sizeof(int))
1987 return -EINVAL;
1988 if (copy_from_user(&sp->autoclose, optval, optlen))
1989 return -EFAULT;
1991 return 0;
1994 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
1996 * Applications can enable or disable heartbeats for any peer address of
1997 * an association, modify an address's heartbeat interval, force a
1998 * heartbeat to be sent immediately, and adjust the address's maximum
1999 * number of retransmissions sent before an address is considered
2000 * unreachable. The following structure is used to access and modify an
2001 * address's parameters:
2003 * struct sctp_paddrparams {
2004 * sctp_assoc_t spp_assoc_id;
2005 * struct sockaddr_storage spp_address;
2006 * uint32_t spp_hbinterval;
2007 * uint16_t spp_pathmaxrxt;
2008 * uint32_t spp_pathmtu;
2009 * uint32_t spp_sackdelay;
2010 * uint32_t spp_flags;
2011 * };
2013 * spp_assoc_id - (one-to-many style socket) This is filled in the
2014 * application, and identifies the association for
2015 * this query.
2016 * spp_address - This specifies which address is of interest.
2017 * spp_hbinterval - This contains the value of the heartbeat interval,
2018 * in milliseconds. If a value of zero
2019 * is present in this field then no changes are to
2020 * be made to this parameter.
2021 * spp_pathmaxrxt - This contains the maximum number of
2022 * retransmissions before this address shall be
2023 * considered unreachable. If a value of zero
2024 * is present in this field then no changes are to
2025 * be made to this parameter.
2026 * spp_pathmtu - When Path MTU discovery is disabled the value
2027 * specified here will be the "fixed" path mtu.
2028 * Note that if the spp_address field is empty
2029 * then all associations on this address will
2030 * have this fixed path mtu set upon them.
2032 * spp_sackdelay - When delayed sack is enabled, this value specifies
2033 * the number of milliseconds that sacks will be delayed
2034 * for. This value will apply to all addresses of an
2035 * association if the spp_address field is empty. Note
2036 * also, that if delayed sack is enabled and this
2037 * value is set to 0, no change is made to the last
2038 * recorded delayed sack timer value.
2040 * spp_flags - These flags are used to control various features
2041 * on an association. The flag field may contain
2042 * zero or more of the following options.
2044 * SPP_HB_ENABLE - Enable heartbeats on the
2045 * specified address. Note that if the address
2046 * field is empty all addresses for the association
2047 * have heartbeats enabled upon them.
2049 * SPP_HB_DISABLE - Disable heartbeats on the
2050 * speicifed address. Note that if the address
2051 * field is empty all addresses for the association
2052 * will have their heartbeats disabled. Note also
2053 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2054 * mutually exclusive, only one of these two should
2055 * be specified. Enabling both fields will have
2056 * undetermined results.
2058 * SPP_HB_DEMAND - Request a user initiated heartbeat
2059 * to be made immediately.
2061 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2062 * heartbeat delayis to be set to the value of 0
2063 * milliseconds.
2065 * SPP_PMTUD_ENABLE - This field will enable PMTU
2066 * discovery upon the specified address. Note that
2067 * if the address feild is empty then all addresses
2068 * on the association are effected.
2070 * SPP_PMTUD_DISABLE - This field will disable PMTU
2071 * discovery upon the specified address. Note that
2072 * if the address feild is empty then all addresses
2073 * on the association are effected. Not also that
2074 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2075 * exclusive. Enabling both will have undetermined
2076 * results.
2078 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2079 * on delayed sack. The time specified in spp_sackdelay
2080 * is used to specify the sack delay for this address. Note
2081 * that if spp_address is empty then all addresses will
2082 * enable delayed sack and take on the sack delay
2083 * value specified in spp_sackdelay.
2084 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2085 * off delayed sack. If the spp_address field is blank then
2086 * delayed sack is disabled for the entire association. Note
2087 * also that this field is mutually exclusive to
2088 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2089 * results.
2091 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2092 struct sctp_transport *trans,
2093 struct sctp_association *asoc,
2094 struct sctp_sock *sp,
2095 int hb_change,
2096 int pmtud_change,
2097 int sackdelay_change)
2099 int error;
2101 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2102 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2103 if (error)
2104 return error;
2107 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2108 * this field is ignored. Note also that a value of zero indicates
2109 * the current setting should be left unchanged.
2111 if (params->spp_flags & SPP_HB_ENABLE) {
2113 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2114 * set. This lets us use 0 value when this flag
2115 * is set.
2117 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2118 params->spp_hbinterval = 0;
2120 if (params->spp_hbinterval ||
2121 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2122 if (trans) {
2123 trans->hbinterval =
2124 msecs_to_jiffies(params->spp_hbinterval);
2125 } else if (asoc) {
2126 asoc->hbinterval =
2127 msecs_to_jiffies(params->spp_hbinterval);
2128 } else {
2129 sp->hbinterval = params->spp_hbinterval;
2134 if (hb_change) {
2135 if (trans) {
2136 trans->param_flags =
2137 (trans->param_flags & ~SPP_HB) | hb_change;
2138 } else if (asoc) {
2139 asoc->param_flags =
2140 (asoc->param_flags & ~SPP_HB) | hb_change;
2141 } else {
2142 sp->param_flags =
2143 (sp->param_flags & ~SPP_HB) | hb_change;
2147 /* When Path MTU discovery is disabled the value specified here will
2148 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2149 * include the flag SPP_PMTUD_DISABLE for this field to have any
2150 * effect).
2152 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2153 if (trans) {
2154 trans->pathmtu = params->spp_pathmtu;
2155 sctp_assoc_sync_pmtu(asoc);
2156 } else if (asoc) {
2157 asoc->pathmtu = params->spp_pathmtu;
2158 sctp_frag_point(sp, params->spp_pathmtu);
2159 } else {
2160 sp->pathmtu = params->spp_pathmtu;
2164 if (pmtud_change) {
2165 if (trans) {
2166 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2167 (params->spp_flags & SPP_PMTUD_ENABLE);
2168 trans->param_flags =
2169 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2170 if (update) {
2171 sctp_transport_pmtu(trans);
2172 sctp_assoc_sync_pmtu(asoc);
2174 } else if (asoc) {
2175 asoc->param_flags =
2176 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2177 } else {
2178 sp->param_flags =
2179 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2183 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2184 * value of this field is ignored. Note also that a value of zero
2185 * indicates the current setting should be left unchanged.
2187 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2188 if (trans) {
2189 trans->sackdelay =
2190 msecs_to_jiffies(params->spp_sackdelay);
2191 } else if (asoc) {
2192 asoc->sackdelay =
2193 msecs_to_jiffies(params->spp_sackdelay);
2194 } else {
2195 sp->sackdelay = params->spp_sackdelay;
2199 if (sackdelay_change) {
2200 if (trans) {
2201 trans->param_flags =
2202 (trans->param_flags & ~SPP_SACKDELAY) |
2203 sackdelay_change;
2204 } else if (asoc) {
2205 asoc->param_flags =
2206 (asoc->param_flags & ~SPP_SACKDELAY) |
2207 sackdelay_change;
2208 } else {
2209 sp->param_flags =
2210 (sp->param_flags & ~SPP_SACKDELAY) |
2211 sackdelay_change;
2215 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2216 * of this field is ignored. Note also that a value of zero
2217 * indicates the current setting should be left unchanged.
2219 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2220 if (trans) {
2221 trans->pathmaxrxt = params->spp_pathmaxrxt;
2222 } else if (asoc) {
2223 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2224 } else {
2225 sp->pathmaxrxt = params->spp_pathmaxrxt;
2229 return 0;
2232 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2233 char __user *optval, int optlen)
2235 struct sctp_paddrparams params;
2236 struct sctp_transport *trans = NULL;
2237 struct sctp_association *asoc = NULL;
2238 struct sctp_sock *sp = sctp_sk(sk);
2239 int error;
2240 int hb_change, pmtud_change, sackdelay_change;
2242 if (optlen != sizeof(struct sctp_paddrparams))
2243 return - EINVAL;
2245 if (copy_from_user(&params, optval, optlen))
2246 return -EFAULT;
2248 /* Validate flags and value parameters. */
2249 hb_change = params.spp_flags & SPP_HB;
2250 pmtud_change = params.spp_flags & SPP_PMTUD;
2251 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2253 if (hb_change == SPP_HB ||
2254 pmtud_change == SPP_PMTUD ||
2255 sackdelay_change == SPP_SACKDELAY ||
2256 params.spp_sackdelay > 500 ||
2257 (params.spp_pathmtu
2258 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2259 return -EINVAL;
2261 /* If an address other than INADDR_ANY is specified, and
2262 * no transport is found, then the request is invalid.
2264 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2265 trans = sctp_addr_id2transport(sk, &params.spp_address,
2266 params.spp_assoc_id);
2267 if (!trans)
2268 return -EINVAL;
2271 /* Get association, if assoc_id != 0 and the socket is a one
2272 * to many style socket, and an association was not found, then
2273 * the id was invalid.
2275 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2276 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2277 return -EINVAL;
2279 /* Heartbeat demand can only be sent on a transport or
2280 * association, but not a socket.
2282 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2283 return -EINVAL;
2285 /* Process parameters. */
2286 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2287 hb_change, pmtud_change,
2288 sackdelay_change);
2290 if (error)
2291 return error;
2293 /* If changes are for association, also apply parameters to each
2294 * transport.
2296 if (!trans && asoc) {
2297 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2298 transports) {
2299 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2300 hb_change, pmtud_change,
2301 sackdelay_change);
2305 return 0;
2308 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2310 * This options will get or set the delayed ack timer. The time is set
2311 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2312 * endpoints default delayed ack timer value. If the assoc_id field is
2313 * non-zero, then the set or get effects the specified association.
2315 * struct sctp_assoc_value {
2316 * sctp_assoc_t assoc_id;
2317 * uint32_t assoc_value;
2318 * };
2320 * assoc_id - This parameter, indicates which association the
2321 * user is preforming an action upon. Note that if
2322 * this field's value is zero then the endpoints
2323 * default value is changed (effecting future
2324 * associations only).
2326 * assoc_value - This parameter contains the number of milliseconds
2327 * that the user is requesting the delayed ACK timer
2328 * be set to. Note that this value is defined in
2329 * the standard to be between 200 and 500 milliseconds.
2331 * Note: a value of zero will leave the value alone,
2332 * but disable SACK delay. A non-zero value will also
2333 * enable SACK delay.
2336 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2337 char __user *optval, int optlen)
2339 struct sctp_assoc_value params;
2340 struct sctp_transport *trans = NULL;
2341 struct sctp_association *asoc = NULL;
2342 struct sctp_sock *sp = sctp_sk(sk);
2344 if (optlen != sizeof(struct sctp_assoc_value))
2345 return - EINVAL;
2347 if (copy_from_user(&params, optval, optlen))
2348 return -EFAULT;
2350 /* Validate value parameter. */
2351 if (params.assoc_value > 500)
2352 return -EINVAL;
2354 /* Get association, if assoc_id != 0 and the socket is a one
2355 * to many style socket, and an association was not found, then
2356 * the id was invalid.
2358 asoc = sctp_id2assoc(sk, params.assoc_id);
2359 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2360 return -EINVAL;
2362 if (params.assoc_value) {
2363 if (asoc) {
2364 asoc->sackdelay =
2365 msecs_to_jiffies(params.assoc_value);
2366 asoc->param_flags =
2367 (asoc->param_flags & ~SPP_SACKDELAY) |
2368 SPP_SACKDELAY_ENABLE;
2369 } else {
2370 sp->sackdelay = params.assoc_value;
2371 sp->param_flags =
2372 (sp->param_flags & ~SPP_SACKDELAY) |
2373 SPP_SACKDELAY_ENABLE;
2375 } else {
2376 if (asoc) {
2377 asoc->param_flags =
2378 (asoc->param_flags & ~SPP_SACKDELAY) |
2379 SPP_SACKDELAY_DISABLE;
2380 } else {
2381 sp->param_flags =
2382 (sp->param_flags & ~SPP_SACKDELAY) |
2383 SPP_SACKDELAY_DISABLE;
2387 /* If change is for association, also apply to each transport. */
2388 if (asoc) {
2389 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2390 transports) {
2391 if (params.assoc_value) {
2392 trans->sackdelay =
2393 msecs_to_jiffies(params.assoc_value);
2394 trans->param_flags =
2395 (trans->param_flags & ~SPP_SACKDELAY) |
2396 SPP_SACKDELAY_ENABLE;
2397 } else {
2398 trans->param_flags =
2399 (trans->param_flags & ~SPP_SACKDELAY) |
2400 SPP_SACKDELAY_DISABLE;
2405 return 0;
2408 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2410 * Applications can specify protocol parameters for the default association
2411 * initialization. The option name argument to setsockopt() and getsockopt()
2412 * is SCTP_INITMSG.
2414 * Setting initialization parameters is effective only on an unconnected
2415 * socket (for UDP-style sockets only future associations are effected
2416 * by the change). With TCP-style sockets, this option is inherited by
2417 * sockets derived from a listener socket.
2419 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2421 struct sctp_initmsg sinit;
2422 struct sctp_sock *sp = sctp_sk(sk);
2424 if (optlen != sizeof(struct sctp_initmsg))
2425 return -EINVAL;
2426 if (copy_from_user(&sinit, optval, optlen))
2427 return -EFAULT;
2429 if (sinit.sinit_num_ostreams)
2430 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2431 if (sinit.sinit_max_instreams)
2432 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2433 if (sinit.sinit_max_attempts)
2434 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2435 if (sinit.sinit_max_init_timeo)
2436 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2438 return 0;
2442 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2444 * Applications that wish to use the sendto() system call may wish to
2445 * specify a default set of parameters that would normally be supplied
2446 * through the inclusion of ancillary data. This socket option allows
2447 * such an application to set the default sctp_sndrcvinfo structure.
2448 * The application that wishes to use this socket option simply passes
2449 * in to this call the sctp_sndrcvinfo structure defined in Section
2450 * 5.2.2) The input parameters accepted by this call include
2451 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2452 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2453 * to this call if the caller is using the UDP model.
2455 static int sctp_setsockopt_default_send_param(struct sock *sk,
2456 char __user *optval, int optlen)
2458 struct sctp_sndrcvinfo info;
2459 struct sctp_association *asoc;
2460 struct sctp_sock *sp = sctp_sk(sk);
2462 if (optlen != sizeof(struct sctp_sndrcvinfo))
2463 return -EINVAL;
2464 if (copy_from_user(&info, optval, optlen))
2465 return -EFAULT;
2467 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2468 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2469 return -EINVAL;
2471 if (asoc) {
2472 asoc->default_stream = info.sinfo_stream;
2473 asoc->default_flags = info.sinfo_flags;
2474 asoc->default_ppid = info.sinfo_ppid;
2475 asoc->default_context = info.sinfo_context;
2476 asoc->default_timetolive = info.sinfo_timetolive;
2477 } else {
2478 sp->default_stream = info.sinfo_stream;
2479 sp->default_flags = info.sinfo_flags;
2480 sp->default_ppid = info.sinfo_ppid;
2481 sp->default_context = info.sinfo_context;
2482 sp->default_timetolive = info.sinfo_timetolive;
2485 return 0;
2488 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2490 * Requests that the local SCTP stack use the enclosed peer address as
2491 * the association primary. The enclosed address must be one of the
2492 * association peer's addresses.
2494 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2495 int optlen)
2497 struct sctp_prim prim;
2498 struct sctp_transport *trans;
2500 if (optlen != sizeof(struct sctp_prim))
2501 return -EINVAL;
2503 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2504 return -EFAULT;
2506 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2507 if (!trans)
2508 return -EINVAL;
2510 sctp_assoc_set_primary(trans->asoc, trans);
2512 return 0;
2516 * 7.1.5 SCTP_NODELAY
2518 * Turn on/off any Nagle-like algorithm. This means that packets are
2519 * generally sent as soon as possible and no unnecessary delays are
2520 * introduced, at the cost of more packets in the network. Expects an
2521 * integer boolean flag.
2523 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2524 int optlen)
2526 int val;
2528 if (optlen < sizeof(int))
2529 return -EINVAL;
2530 if (get_user(val, (int __user *)optval))
2531 return -EFAULT;
2533 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2534 return 0;
2539 * 7.1.1 SCTP_RTOINFO
2541 * The protocol parameters used to initialize and bound retransmission
2542 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2543 * and modify these parameters.
2544 * All parameters are time values, in milliseconds. A value of 0, when
2545 * modifying the parameters, indicates that the current value should not
2546 * be changed.
2549 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2550 struct sctp_rtoinfo rtoinfo;
2551 struct sctp_association *asoc;
2553 if (optlen != sizeof (struct sctp_rtoinfo))
2554 return -EINVAL;
2556 if (copy_from_user(&rtoinfo, optval, optlen))
2557 return -EFAULT;
2559 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2561 /* Set the values to the specific association */
2562 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2563 return -EINVAL;
2565 if (asoc) {
2566 if (rtoinfo.srto_initial != 0)
2567 asoc->rto_initial =
2568 msecs_to_jiffies(rtoinfo.srto_initial);
2569 if (rtoinfo.srto_max != 0)
2570 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2571 if (rtoinfo.srto_min != 0)
2572 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2573 } else {
2574 /* If there is no association or the association-id = 0
2575 * set the values to the endpoint.
2577 struct sctp_sock *sp = sctp_sk(sk);
2579 if (rtoinfo.srto_initial != 0)
2580 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2581 if (rtoinfo.srto_max != 0)
2582 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2583 if (rtoinfo.srto_min != 0)
2584 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2587 return 0;
2592 * 7.1.2 SCTP_ASSOCINFO
2594 * This option is used to tune the maximum retransmission attempts
2595 * of the association.
2596 * Returns an error if the new association retransmission value is
2597 * greater than the sum of the retransmission value of the peer.
2598 * See [SCTP] for more information.
2601 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2604 struct sctp_assocparams assocparams;
2605 struct sctp_association *asoc;
2607 if (optlen != sizeof(struct sctp_assocparams))
2608 return -EINVAL;
2609 if (copy_from_user(&assocparams, optval, optlen))
2610 return -EFAULT;
2612 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2614 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2615 return -EINVAL;
2617 /* Set the values to the specific association */
2618 if (asoc) {
2619 if (assocparams.sasoc_asocmaxrxt != 0) {
2620 __u32 path_sum = 0;
2621 int paths = 0;
2622 struct sctp_transport *peer_addr;
2624 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2625 transports) {
2626 path_sum += peer_addr->pathmaxrxt;
2627 paths++;
2630 /* Only validate asocmaxrxt if we have more then
2631 * one path/transport. We do this because path
2632 * retransmissions are only counted when we have more
2633 * then one path.
2635 if (paths > 1 &&
2636 assocparams.sasoc_asocmaxrxt > path_sum)
2637 return -EINVAL;
2639 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2642 if (assocparams.sasoc_cookie_life != 0) {
2643 asoc->cookie_life.tv_sec =
2644 assocparams.sasoc_cookie_life / 1000;
2645 asoc->cookie_life.tv_usec =
2646 (assocparams.sasoc_cookie_life % 1000)
2647 * 1000;
2649 } else {
2650 /* Set the values to the endpoint */
2651 struct sctp_sock *sp = sctp_sk(sk);
2653 if (assocparams.sasoc_asocmaxrxt != 0)
2654 sp->assocparams.sasoc_asocmaxrxt =
2655 assocparams.sasoc_asocmaxrxt;
2656 if (assocparams.sasoc_cookie_life != 0)
2657 sp->assocparams.sasoc_cookie_life =
2658 assocparams.sasoc_cookie_life;
2660 return 0;
2664 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2666 * This socket option is a boolean flag which turns on or off mapped V4
2667 * addresses. If this option is turned on and the socket is type
2668 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2669 * If this option is turned off, then no mapping will be done of V4
2670 * addresses and a user will receive both PF_INET6 and PF_INET type
2671 * addresses on the socket.
2673 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2675 int val;
2676 struct sctp_sock *sp = sctp_sk(sk);
2678 if (optlen < sizeof(int))
2679 return -EINVAL;
2680 if (get_user(val, (int __user *)optval))
2681 return -EFAULT;
2682 if (val)
2683 sp->v4mapped = 1;
2684 else
2685 sp->v4mapped = 0;
2687 return 0;
2691 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2693 * This socket option specifies the maximum size to put in any outgoing
2694 * SCTP chunk. If a message is larger than this size it will be
2695 * fragmented by SCTP into the specified size. Note that the underlying
2696 * SCTP implementation may fragment into smaller sized chunks when the
2697 * PMTU of the underlying association is smaller than the value set by
2698 * the user.
2700 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2702 struct sctp_association *asoc;
2703 struct sctp_sock *sp = sctp_sk(sk);
2704 int val;
2706 if (optlen < sizeof(int))
2707 return -EINVAL;
2708 if (get_user(val, (int __user *)optval))
2709 return -EFAULT;
2710 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2711 return -EINVAL;
2712 sp->user_frag = val;
2714 /* Update the frag_point of the existing associations. */
2715 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
2716 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2719 return 0;
2724 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2726 * Requests that the peer mark the enclosed address as the association
2727 * primary. The enclosed address must be one of the association's
2728 * locally bound addresses. The following structure is used to make a
2729 * set primary request:
2731 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2732 int optlen)
2734 struct sctp_sock *sp;
2735 struct sctp_endpoint *ep;
2736 struct sctp_association *asoc = NULL;
2737 struct sctp_setpeerprim prim;
2738 struct sctp_chunk *chunk;
2739 int err;
2741 sp = sctp_sk(sk);
2742 ep = sp->ep;
2744 if (!sctp_addip_enable)
2745 return -EPERM;
2747 if (optlen != sizeof(struct sctp_setpeerprim))
2748 return -EINVAL;
2750 if (copy_from_user(&prim, optval, optlen))
2751 return -EFAULT;
2753 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2754 if (!asoc)
2755 return -EINVAL;
2757 if (!asoc->peer.asconf_capable)
2758 return -EPERM;
2760 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2761 return -EPERM;
2763 if (!sctp_state(asoc, ESTABLISHED))
2764 return -ENOTCONN;
2766 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2767 return -EADDRNOTAVAIL;
2769 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2770 chunk = sctp_make_asconf_set_prim(asoc,
2771 (union sctp_addr *)&prim.sspp_addr);
2772 if (!chunk)
2773 return -ENOMEM;
2775 err = sctp_send_asconf(asoc, chunk);
2777 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2779 return err;
2782 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2783 int optlen)
2785 struct sctp_setadaptation adaptation;
2787 if (optlen != sizeof(struct sctp_setadaptation))
2788 return -EINVAL;
2789 if (copy_from_user(&adaptation, optval, optlen))
2790 return -EFAULT;
2792 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2794 return 0;
2798 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2800 * The context field in the sctp_sndrcvinfo structure is normally only
2801 * used when a failed message is retrieved holding the value that was
2802 * sent down on the actual send call. This option allows the setting of
2803 * a default context on an association basis that will be received on
2804 * reading messages from the peer. This is especially helpful in the
2805 * one-2-many model for an application to keep some reference to an
2806 * internal state machine that is processing messages on the
2807 * association. Note that the setting of this value only effects
2808 * received messages from the peer and does not effect the value that is
2809 * saved with outbound messages.
2811 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2812 int optlen)
2814 struct sctp_assoc_value params;
2815 struct sctp_sock *sp;
2816 struct sctp_association *asoc;
2818 if (optlen != sizeof(struct sctp_assoc_value))
2819 return -EINVAL;
2820 if (copy_from_user(&params, optval, optlen))
2821 return -EFAULT;
2823 sp = sctp_sk(sk);
2825 if (params.assoc_id != 0) {
2826 asoc = sctp_id2assoc(sk, params.assoc_id);
2827 if (!asoc)
2828 return -EINVAL;
2829 asoc->default_rcv_context = params.assoc_value;
2830 } else {
2831 sp->default_rcv_context = params.assoc_value;
2834 return 0;
2838 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2840 * This options will at a minimum specify if the implementation is doing
2841 * fragmented interleave. Fragmented interleave, for a one to many
2842 * socket, is when subsequent calls to receive a message may return
2843 * parts of messages from different associations. Some implementations
2844 * may allow you to turn this value on or off. If so, when turned off,
2845 * no fragment interleave will occur (which will cause a head of line
2846 * blocking amongst multiple associations sharing the same one to many
2847 * socket). When this option is turned on, then each receive call may
2848 * come from a different association (thus the user must receive data
2849 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2850 * association each receive belongs to.
2852 * This option takes a boolean value. A non-zero value indicates that
2853 * fragmented interleave is on. A value of zero indicates that
2854 * fragmented interleave is off.
2856 * Note that it is important that an implementation that allows this
2857 * option to be turned on, have it off by default. Otherwise an unaware
2858 * application using the one to many model may become confused and act
2859 * incorrectly.
2861 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2862 char __user *optval,
2863 int optlen)
2865 int val;
2867 if (optlen != sizeof(int))
2868 return -EINVAL;
2869 if (get_user(val, (int __user *)optval))
2870 return -EFAULT;
2872 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2874 return 0;
2878 * 7.1.25. Set or Get the sctp partial delivery point
2879 * (SCTP_PARTIAL_DELIVERY_POINT)
2880 * This option will set or get the SCTP partial delivery point. This
2881 * point is the size of a message where the partial delivery API will be
2882 * invoked to help free up rwnd space for the peer. Setting this to a
2883 * lower value will cause partial delivery's to happen more often. The
2884 * calls argument is an integer that sets or gets the partial delivery
2885 * point.
2887 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2888 char __user *optval,
2889 int optlen)
2891 u32 val;
2893 if (optlen != sizeof(u32))
2894 return -EINVAL;
2895 if (get_user(val, (int __user *)optval))
2896 return -EFAULT;
2898 sctp_sk(sk)->pd_point = val;
2900 return 0; /* is this the right error code? */
2904 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2906 * This option will allow a user to change the maximum burst of packets
2907 * that can be emitted by this association. Note that the default value
2908 * is 4, and some implementations may restrict this setting so that it
2909 * can only be lowered.
2911 * NOTE: This text doesn't seem right. Do this on a socket basis with
2912 * future associations inheriting the socket value.
2914 static int sctp_setsockopt_maxburst(struct sock *sk,
2915 char __user *optval,
2916 int optlen)
2918 struct sctp_assoc_value params;
2919 struct sctp_sock *sp;
2920 struct sctp_association *asoc;
2921 int val;
2922 int assoc_id = 0;
2924 if (optlen < sizeof(int))
2925 return -EINVAL;
2927 if (optlen == sizeof(int)) {
2928 printk(KERN_WARNING
2929 "SCTP: Use of int in max_burst socket option deprecated\n");
2930 printk(KERN_WARNING
2931 "SCTP: Use struct sctp_assoc_value instead\n");
2932 if (copy_from_user(&val, optval, optlen))
2933 return -EFAULT;
2934 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2935 if (copy_from_user(&params, optval, optlen))
2936 return -EFAULT;
2937 val = params.assoc_value;
2938 assoc_id = params.assoc_id;
2939 } else
2940 return -EINVAL;
2942 sp = sctp_sk(sk);
2944 if (assoc_id != 0) {
2945 asoc = sctp_id2assoc(sk, assoc_id);
2946 if (!asoc)
2947 return -EINVAL;
2948 asoc->max_burst = val;
2949 } else
2950 sp->max_burst = val;
2952 return 0;
2956 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
2958 * This set option adds a chunk type that the user is requesting to be
2959 * received only in an authenticated way. Changes to the list of chunks
2960 * will only effect future associations on the socket.
2962 static int sctp_setsockopt_auth_chunk(struct sock *sk,
2963 char __user *optval,
2964 int optlen)
2966 struct sctp_authchunk val;
2968 if (optlen != sizeof(struct sctp_authchunk))
2969 return -EINVAL;
2970 if (copy_from_user(&val, optval, optlen))
2971 return -EFAULT;
2973 switch (val.sauth_chunk) {
2974 case SCTP_CID_INIT:
2975 case SCTP_CID_INIT_ACK:
2976 case SCTP_CID_SHUTDOWN_COMPLETE:
2977 case SCTP_CID_AUTH:
2978 return -EINVAL;
2981 /* add this chunk id to the endpoint */
2982 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
2986 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
2988 * This option gets or sets the list of HMAC algorithms that the local
2989 * endpoint requires the peer to use.
2991 static int sctp_setsockopt_hmac_ident(struct sock *sk,
2992 char __user *optval,
2993 int optlen)
2995 struct sctp_hmacalgo *hmacs;
2996 int err;
2998 if (optlen < sizeof(struct sctp_hmacalgo))
2999 return -EINVAL;
3001 hmacs = kmalloc(optlen, GFP_KERNEL);
3002 if (!hmacs)
3003 return -ENOMEM;
3005 if (copy_from_user(hmacs, optval, optlen)) {
3006 err = -EFAULT;
3007 goto out;
3010 if (hmacs->shmac_num_idents == 0 ||
3011 hmacs->shmac_num_idents > SCTP_AUTH_NUM_HMACS) {
3012 err = -EINVAL;
3013 goto out;
3016 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3017 out:
3018 kfree(hmacs);
3019 return err;
3023 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3025 * This option will set a shared secret key which is used to build an
3026 * association shared key.
3028 static int sctp_setsockopt_auth_key(struct sock *sk,
3029 char __user *optval,
3030 int optlen)
3032 struct sctp_authkey *authkey;
3033 struct sctp_association *asoc;
3034 int ret;
3036 if (optlen <= sizeof(struct sctp_authkey))
3037 return -EINVAL;
3039 authkey = kmalloc(optlen, GFP_KERNEL);
3040 if (!authkey)
3041 return -ENOMEM;
3043 if (copy_from_user(authkey, optval, optlen)) {
3044 ret = -EFAULT;
3045 goto out;
3048 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3049 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3050 ret = -EINVAL;
3051 goto out;
3054 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3055 out:
3056 kfree(authkey);
3057 return ret;
3061 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3063 * This option will get or set the active shared key to be used to build
3064 * the association shared key.
3066 static int sctp_setsockopt_active_key(struct sock *sk,
3067 char __user *optval,
3068 int optlen)
3070 struct sctp_authkeyid val;
3071 struct sctp_association *asoc;
3073 if (optlen != sizeof(struct sctp_authkeyid))
3074 return -EINVAL;
3075 if (copy_from_user(&val, optval, optlen))
3076 return -EFAULT;
3078 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3079 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3080 return -EINVAL;
3082 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3083 val.scact_keynumber);
3087 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3089 * This set option will delete a shared secret key from use.
3091 static int sctp_setsockopt_del_key(struct sock *sk,
3092 char __user *optval,
3093 int optlen)
3095 struct sctp_authkeyid val;
3096 struct sctp_association *asoc;
3098 if (optlen != sizeof(struct sctp_authkeyid))
3099 return -EINVAL;
3100 if (copy_from_user(&val, optval, optlen))
3101 return -EFAULT;
3103 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3104 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3105 return -EINVAL;
3107 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3108 val.scact_keynumber);
3113 /* API 6.2 setsockopt(), getsockopt()
3115 * Applications use setsockopt() and getsockopt() to set or retrieve
3116 * socket options. Socket options are used to change the default
3117 * behavior of sockets calls. They are described in Section 7.
3119 * The syntax is:
3121 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3122 * int __user *optlen);
3123 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3124 * int optlen);
3126 * sd - the socket descript.
3127 * level - set to IPPROTO_SCTP for all SCTP options.
3128 * optname - the option name.
3129 * optval - the buffer to store the value of the option.
3130 * optlen - the size of the buffer.
3132 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3133 char __user *optval, int optlen)
3135 int retval = 0;
3137 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3138 sk, optname);
3140 /* I can hardly begin to describe how wrong this is. This is
3141 * so broken as to be worse than useless. The API draft
3142 * REALLY is NOT helpful here... I am not convinced that the
3143 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3144 * are at all well-founded.
3146 if (level != SOL_SCTP) {
3147 struct sctp_af *af = sctp_sk(sk)->pf->af;
3148 retval = af->setsockopt(sk, level, optname, optval, optlen);
3149 goto out_nounlock;
3152 sctp_lock_sock(sk);
3154 switch (optname) {
3155 case SCTP_SOCKOPT_BINDX_ADD:
3156 /* 'optlen' is the size of the addresses buffer. */
3157 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3158 optlen, SCTP_BINDX_ADD_ADDR);
3159 break;
3161 case SCTP_SOCKOPT_BINDX_REM:
3162 /* 'optlen' is the size of the addresses buffer. */
3163 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3164 optlen, SCTP_BINDX_REM_ADDR);
3165 break;
3167 case SCTP_SOCKOPT_CONNECTX:
3168 /* 'optlen' is the size of the addresses buffer. */
3169 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
3170 optlen);
3171 break;
3173 case SCTP_DISABLE_FRAGMENTS:
3174 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3175 break;
3177 case SCTP_EVENTS:
3178 retval = sctp_setsockopt_events(sk, optval, optlen);
3179 break;
3181 case SCTP_AUTOCLOSE:
3182 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3183 break;
3185 case SCTP_PEER_ADDR_PARAMS:
3186 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3187 break;
3189 case SCTP_DELAYED_ACK_TIME:
3190 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
3191 break;
3192 case SCTP_PARTIAL_DELIVERY_POINT:
3193 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3194 break;
3196 case SCTP_INITMSG:
3197 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3198 break;
3199 case SCTP_DEFAULT_SEND_PARAM:
3200 retval = sctp_setsockopt_default_send_param(sk, optval,
3201 optlen);
3202 break;
3203 case SCTP_PRIMARY_ADDR:
3204 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3205 break;
3206 case SCTP_SET_PEER_PRIMARY_ADDR:
3207 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3208 break;
3209 case SCTP_NODELAY:
3210 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3211 break;
3212 case SCTP_RTOINFO:
3213 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3214 break;
3215 case SCTP_ASSOCINFO:
3216 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3217 break;
3218 case SCTP_I_WANT_MAPPED_V4_ADDR:
3219 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3220 break;
3221 case SCTP_MAXSEG:
3222 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3223 break;
3224 case SCTP_ADAPTATION_LAYER:
3225 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3226 break;
3227 case SCTP_CONTEXT:
3228 retval = sctp_setsockopt_context(sk, optval, optlen);
3229 break;
3230 case SCTP_FRAGMENT_INTERLEAVE:
3231 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3232 break;
3233 case SCTP_MAX_BURST:
3234 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3235 break;
3236 case SCTP_AUTH_CHUNK:
3237 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3238 break;
3239 case SCTP_HMAC_IDENT:
3240 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3241 break;
3242 case SCTP_AUTH_KEY:
3243 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3244 break;
3245 case SCTP_AUTH_ACTIVE_KEY:
3246 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3247 break;
3248 case SCTP_AUTH_DELETE_KEY:
3249 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3250 break;
3251 default:
3252 retval = -ENOPROTOOPT;
3253 break;
3256 sctp_release_sock(sk);
3258 out_nounlock:
3259 return retval;
3262 /* API 3.1.6 connect() - UDP Style Syntax
3264 * An application may use the connect() call in the UDP model to initiate an
3265 * association without sending data.
3267 * The syntax is:
3269 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3271 * sd: the socket descriptor to have a new association added to.
3273 * nam: the address structure (either struct sockaddr_in or struct
3274 * sockaddr_in6 defined in RFC2553 [7]).
3276 * len: the size of the address.
3278 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3279 int addr_len)
3281 int err = 0;
3282 struct sctp_af *af;
3284 sctp_lock_sock(sk);
3286 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3287 __func__, sk, addr, addr_len);
3289 /* Validate addr_len before calling common connect/connectx routine. */
3290 af = sctp_get_af_specific(addr->sa_family);
3291 if (!af || addr_len < af->sockaddr_len) {
3292 err = -EINVAL;
3293 } else {
3294 /* Pass correct addr len to common routine (so it knows there
3295 * is only one address being passed.
3297 err = __sctp_connect(sk, addr, af->sockaddr_len);
3300 sctp_release_sock(sk);
3301 return err;
3304 /* FIXME: Write comments. */
3305 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3307 return -EOPNOTSUPP; /* STUB */
3310 /* 4.1.4 accept() - TCP Style Syntax
3312 * Applications use accept() call to remove an established SCTP
3313 * association from the accept queue of the endpoint. A new socket
3314 * descriptor will be returned from accept() to represent the newly
3315 * formed association.
3317 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3319 struct sctp_sock *sp;
3320 struct sctp_endpoint *ep;
3321 struct sock *newsk = NULL;
3322 struct sctp_association *asoc;
3323 long timeo;
3324 int error = 0;
3326 sctp_lock_sock(sk);
3328 sp = sctp_sk(sk);
3329 ep = sp->ep;
3331 if (!sctp_style(sk, TCP)) {
3332 error = -EOPNOTSUPP;
3333 goto out;
3336 if (!sctp_sstate(sk, LISTENING)) {
3337 error = -EINVAL;
3338 goto out;
3341 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3343 error = sctp_wait_for_accept(sk, timeo);
3344 if (error)
3345 goto out;
3347 /* We treat the list of associations on the endpoint as the accept
3348 * queue and pick the first association on the list.
3350 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3352 newsk = sp->pf->create_accept_sk(sk, asoc);
3353 if (!newsk) {
3354 error = -ENOMEM;
3355 goto out;
3358 /* Populate the fields of the newsk from the oldsk and migrate the
3359 * asoc to the newsk.
3361 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3363 out:
3364 sctp_release_sock(sk);
3365 *err = error;
3366 return newsk;
3369 /* The SCTP ioctl handler. */
3370 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3372 return -ENOIOCTLCMD;
3375 /* This is the function which gets called during socket creation to
3376 * initialized the SCTP-specific portion of the sock.
3377 * The sock structure should already be zero-filled memory.
3379 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3381 struct sctp_endpoint *ep;
3382 struct sctp_sock *sp;
3384 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3386 sp = sctp_sk(sk);
3388 /* Initialize the SCTP per socket area. */
3389 switch (sk->sk_type) {
3390 case SOCK_SEQPACKET:
3391 sp->type = SCTP_SOCKET_UDP;
3392 break;
3393 case SOCK_STREAM:
3394 sp->type = SCTP_SOCKET_TCP;
3395 break;
3396 default:
3397 return -ESOCKTNOSUPPORT;
3400 /* Initialize default send parameters. These parameters can be
3401 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3403 sp->default_stream = 0;
3404 sp->default_ppid = 0;
3405 sp->default_flags = 0;
3406 sp->default_context = 0;
3407 sp->default_timetolive = 0;
3409 sp->default_rcv_context = 0;
3410 sp->max_burst = sctp_max_burst;
3412 /* Initialize default setup parameters. These parameters
3413 * can be modified with the SCTP_INITMSG socket option or
3414 * overridden by the SCTP_INIT CMSG.
3416 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3417 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3418 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3419 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3421 /* Initialize default RTO related parameters. These parameters can
3422 * be modified for with the SCTP_RTOINFO socket option.
3424 sp->rtoinfo.srto_initial = sctp_rto_initial;
3425 sp->rtoinfo.srto_max = sctp_rto_max;
3426 sp->rtoinfo.srto_min = sctp_rto_min;
3428 /* Initialize default association related parameters. These parameters
3429 * can be modified with the SCTP_ASSOCINFO socket option.
3431 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3432 sp->assocparams.sasoc_number_peer_destinations = 0;
3433 sp->assocparams.sasoc_peer_rwnd = 0;
3434 sp->assocparams.sasoc_local_rwnd = 0;
3435 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3437 /* Initialize default event subscriptions. By default, all the
3438 * options are off.
3440 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3442 /* Default Peer Address Parameters. These defaults can
3443 * be modified via SCTP_PEER_ADDR_PARAMS
3445 sp->hbinterval = sctp_hb_interval;
3446 sp->pathmaxrxt = sctp_max_retrans_path;
3447 sp->pathmtu = 0; // allow default discovery
3448 sp->sackdelay = sctp_sack_timeout;
3449 sp->param_flags = SPP_HB_ENABLE |
3450 SPP_PMTUD_ENABLE |
3451 SPP_SACKDELAY_ENABLE;
3453 /* If enabled no SCTP message fragmentation will be performed.
3454 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3456 sp->disable_fragments = 0;
3458 /* Enable Nagle algorithm by default. */
3459 sp->nodelay = 0;
3461 /* Enable by default. */
3462 sp->v4mapped = 1;
3464 /* Auto-close idle associations after the configured
3465 * number of seconds. A value of 0 disables this
3466 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3467 * for UDP-style sockets only.
3469 sp->autoclose = 0;
3471 /* User specified fragmentation limit. */
3472 sp->user_frag = 0;
3474 sp->adaptation_ind = 0;
3476 sp->pf = sctp_get_pf_specific(sk->sk_family);
3478 /* Control variables for partial data delivery. */
3479 atomic_set(&sp->pd_mode, 0);
3480 skb_queue_head_init(&sp->pd_lobby);
3481 sp->frag_interleave = 0;
3483 /* Create a per socket endpoint structure. Even if we
3484 * change the data structure relationships, this may still
3485 * be useful for storing pre-connect address information.
3487 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3488 if (!ep)
3489 return -ENOMEM;
3491 sp->ep = ep;
3492 sp->hmac = NULL;
3494 SCTP_DBG_OBJCNT_INC(sock);
3495 atomic_inc(&sctp_sockets_allocated);
3496 return 0;
3499 /* Cleanup any SCTP per socket resources. */
3500 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3502 struct sctp_endpoint *ep;
3504 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3506 /* Release our hold on the endpoint. */
3507 ep = sctp_sk(sk)->ep;
3508 sctp_endpoint_free(ep);
3509 atomic_dec(&sctp_sockets_allocated);
3510 return 0;
3513 /* API 4.1.7 shutdown() - TCP Style Syntax
3514 * int shutdown(int socket, int how);
3516 * sd - the socket descriptor of the association to be closed.
3517 * how - Specifies the type of shutdown. The values are
3518 * as follows:
3519 * SHUT_RD
3520 * Disables further receive operations. No SCTP
3521 * protocol action is taken.
3522 * SHUT_WR
3523 * Disables further send operations, and initiates
3524 * the SCTP shutdown sequence.
3525 * SHUT_RDWR
3526 * Disables further send and receive operations
3527 * and initiates the SCTP shutdown sequence.
3529 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3531 struct sctp_endpoint *ep;
3532 struct sctp_association *asoc;
3534 if (!sctp_style(sk, TCP))
3535 return;
3537 if (how & SEND_SHUTDOWN) {
3538 ep = sctp_sk(sk)->ep;
3539 if (!list_empty(&ep->asocs)) {
3540 asoc = list_entry(ep->asocs.next,
3541 struct sctp_association, asocs);
3542 sctp_primitive_SHUTDOWN(asoc, NULL);
3547 /* 7.2.1 Association Status (SCTP_STATUS)
3549 * Applications can retrieve current status information about an
3550 * association, including association state, peer receiver window size,
3551 * number of unacked data chunks, and number of data chunks pending
3552 * receipt. This information is read-only.
3554 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3555 char __user *optval,
3556 int __user *optlen)
3558 struct sctp_status status;
3559 struct sctp_association *asoc = NULL;
3560 struct sctp_transport *transport;
3561 sctp_assoc_t associd;
3562 int retval = 0;
3564 if (len < sizeof(status)) {
3565 retval = -EINVAL;
3566 goto out;
3569 len = sizeof(status);
3570 if (copy_from_user(&status, optval, len)) {
3571 retval = -EFAULT;
3572 goto out;
3575 associd = status.sstat_assoc_id;
3576 asoc = sctp_id2assoc(sk, associd);
3577 if (!asoc) {
3578 retval = -EINVAL;
3579 goto out;
3582 transport = asoc->peer.primary_path;
3584 status.sstat_assoc_id = sctp_assoc2id(asoc);
3585 status.sstat_state = asoc->state;
3586 status.sstat_rwnd = asoc->peer.rwnd;
3587 status.sstat_unackdata = asoc->unack_data;
3589 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3590 status.sstat_instrms = asoc->c.sinit_max_instreams;
3591 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3592 status.sstat_fragmentation_point = asoc->frag_point;
3593 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3594 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3595 transport->af_specific->sockaddr_len);
3596 /* Map ipv4 address into v4-mapped-on-v6 address. */
3597 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3598 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3599 status.sstat_primary.spinfo_state = transport->state;
3600 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3601 status.sstat_primary.spinfo_srtt = transport->srtt;
3602 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3603 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3605 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3606 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3608 if (put_user(len, optlen)) {
3609 retval = -EFAULT;
3610 goto out;
3613 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3614 len, status.sstat_state, status.sstat_rwnd,
3615 status.sstat_assoc_id);
3617 if (copy_to_user(optval, &status, len)) {
3618 retval = -EFAULT;
3619 goto out;
3622 out:
3623 return (retval);
3627 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3629 * Applications can retrieve information about a specific peer address
3630 * of an association, including its reachability state, congestion
3631 * window, and retransmission timer values. This information is
3632 * read-only.
3634 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3635 char __user *optval,
3636 int __user *optlen)
3638 struct sctp_paddrinfo pinfo;
3639 struct sctp_transport *transport;
3640 int retval = 0;
3642 if (len < sizeof(pinfo)) {
3643 retval = -EINVAL;
3644 goto out;
3647 len = sizeof(pinfo);
3648 if (copy_from_user(&pinfo, optval, len)) {
3649 retval = -EFAULT;
3650 goto out;
3653 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3654 pinfo.spinfo_assoc_id);
3655 if (!transport)
3656 return -EINVAL;
3658 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3659 pinfo.spinfo_state = transport->state;
3660 pinfo.spinfo_cwnd = transport->cwnd;
3661 pinfo.spinfo_srtt = transport->srtt;
3662 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3663 pinfo.spinfo_mtu = transport->pathmtu;
3665 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3666 pinfo.spinfo_state = SCTP_ACTIVE;
3668 if (put_user(len, optlen)) {
3669 retval = -EFAULT;
3670 goto out;
3673 if (copy_to_user(optval, &pinfo, len)) {
3674 retval = -EFAULT;
3675 goto out;
3678 out:
3679 return (retval);
3682 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3684 * This option is a on/off flag. If enabled no SCTP message
3685 * fragmentation will be performed. Instead if a message being sent
3686 * exceeds the current PMTU size, the message will NOT be sent and
3687 * instead a error will be indicated to the user.
3689 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3690 char __user *optval, int __user *optlen)
3692 int val;
3694 if (len < sizeof(int))
3695 return -EINVAL;
3697 len = sizeof(int);
3698 val = (sctp_sk(sk)->disable_fragments == 1);
3699 if (put_user(len, optlen))
3700 return -EFAULT;
3701 if (copy_to_user(optval, &val, len))
3702 return -EFAULT;
3703 return 0;
3706 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3708 * This socket option is used to specify various notifications and
3709 * ancillary data the user wishes to receive.
3711 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3712 int __user *optlen)
3714 if (len < sizeof(struct sctp_event_subscribe))
3715 return -EINVAL;
3716 len = sizeof(struct sctp_event_subscribe);
3717 if (put_user(len, optlen))
3718 return -EFAULT;
3719 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3720 return -EFAULT;
3721 return 0;
3724 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3726 * This socket option is applicable to the UDP-style socket only. When
3727 * set it will cause associations that are idle for more than the
3728 * specified number of seconds to automatically close. An association
3729 * being idle is defined an association that has NOT sent or received
3730 * user data. The special value of '0' indicates that no automatic
3731 * close of any associations should be performed. The option expects an
3732 * integer defining the number of seconds of idle time before an
3733 * association is closed.
3735 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3737 /* Applicable to UDP-style socket only */
3738 if (sctp_style(sk, TCP))
3739 return -EOPNOTSUPP;
3740 if (len < sizeof(int))
3741 return -EINVAL;
3742 len = sizeof(int);
3743 if (put_user(len, optlen))
3744 return -EFAULT;
3745 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3746 return -EFAULT;
3747 return 0;
3750 /* Helper routine to branch off an association to a new socket. */
3751 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3752 struct socket **sockp)
3754 struct sock *sk = asoc->base.sk;
3755 struct socket *sock;
3756 struct inet_sock *inetsk;
3757 struct sctp_af *af;
3758 int err = 0;
3760 /* An association cannot be branched off from an already peeled-off
3761 * socket, nor is this supported for tcp style sockets.
3763 if (!sctp_style(sk, UDP))
3764 return -EINVAL;
3766 /* Create a new socket. */
3767 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3768 if (err < 0)
3769 return err;
3771 /* Populate the fields of the newsk from the oldsk and migrate the
3772 * asoc to the newsk.
3774 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3776 /* Make peeled-off sockets more like 1-1 accepted sockets.
3777 * Set the daddr and initialize id to something more random
3779 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3780 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3781 inetsk = inet_sk(sock->sk);
3782 inetsk->id = asoc->next_tsn ^ jiffies;
3784 *sockp = sock;
3786 return err;
3789 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3791 sctp_peeloff_arg_t peeloff;
3792 struct socket *newsock;
3793 int retval = 0;
3794 struct sctp_association *asoc;
3796 if (len < sizeof(sctp_peeloff_arg_t))
3797 return -EINVAL;
3798 len = sizeof(sctp_peeloff_arg_t);
3799 if (copy_from_user(&peeloff, optval, len))
3800 return -EFAULT;
3802 asoc = sctp_id2assoc(sk, peeloff.associd);
3803 if (!asoc) {
3804 retval = -EINVAL;
3805 goto out;
3808 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
3810 retval = sctp_do_peeloff(asoc, &newsock);
3811 if (retval < 0)
3812 goto out;
3814 /* Map the socket to an unused fd that can be returned to the user. */
3815 retval = sock_map_fd(newsock);
3816 if (retval < 0) {
3817 sock_release(newsock);
3818 goto out;
3821 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3822 __func__, sk, asoc, newsock->sk, retval);
3824 /* Return the fd mapped to the new socket. */
3825 peeloff.sd = retval;
3826 if (put_user(len, optlen))
3827 return -EFAULT;
3828 if (copy_to_user(optval, &peeloff, len))
3829 retval = -EFAULT;
3831 out:
3832 return retval;
3835 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3837 * Applications can enable or disable heartbeats for any peer address of
3838 * an association, modify an address's heartbeat interval, force a
3839 * heartbeat to be sent immediately, and adjust the address's maximum
3840 * number of retransmissions sent before an address is considered
3841 * unreachable. The following structure is used to access and modify an
3842 * address's parameters:
3844 * struct sctp_paddrparams {
3845 * sctp_assoc_t spp_assoc_id;
3846 * struct sockaddr_storage spp_address;
3847 * uint32_t spp_hbinterval;
3848 * uint16_t spp_pathmaxrxt;
3849 * uint32_t spp_pathmtu;
3850 * uint32_t spp_sackdelay;
3851 * uint32_t spp_flags;
3852 * };
3854 * spp_assoc_id - (one-to-many style socket) This is filled in the
3855 * application, and identifies the association for
3856 * this query.
3857 * spp_address - This specifies which address is of interest.
3858 * spp_hbinterval - This contains the value of the heartbeat interval,
3859 * in milliseconds. If a value of zero
3860 * is present in this field then no changes are to
3861 * be made to this parameter.
3862 * spp_pathmaxrxt - This contains the maximum number of
3863 * retransmissions before this address shall be
3864 * considered unreachable. If a value of zero
3865 * is present in this field then no changes are to
3866 * be made to this parameter.
3867 * spp_pathmtu - When Path MTU discovery is disabled the value
3868 * specified here will be the "fixed" path mtu.
3869 * Note that if the spp_address field is empty
3870 * then all associations on this address will
3871 * have this fixed path mtu set upon them.
3873 * spp_sackdelay - When delayed sack is enabled, this value specifies
3874 * the number of milliseconds that sacks will be delayed
3875 * for. This value will apply to all addresses of an
3876 * association if the spp_address field is empty. Note
3877 * also, that if delayed sack is enabled and this
3878 * value is set to 0, no change is made to the last
3879 * recorded delayed sack timer value.
3881 * spp_flags - These flags are used to control various features
3882 * on an association. The flag field may contain
3883 * zero or more of the following options.
3885 * SPP_HB_ENABLE - Enable heartbeats on the
3886 * specified address. Note that if the address
3887 * field is empty all addresses for the association
3888 * have heartbeats enabled upon them.
3890 * SPP_HB_DISABLE - Disable heartbeats on the
3891 * speicifed address. Note that if the address
3892 * field is empty all addresses for the association
3893 * will have their heartbeats disabled. Note also
3894 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3895 * mutually exclusive, only one of these two should
3896 * be specified. Enabling both fields will have
3897 * undetermined results.
3899 * SPP_HB_DEMAND - Request a user initiated heartbeat
3900 * to be made immediately.
3902 * SPP_PMTUD_ENABLE - This field will enable PMTU
3903 * discovery upon the specified address. Note that
3904 * if the address feild is empty then all addresses
3905 * on the association are effected.
3907 * SPP_PMTUD_DISABLE - This field will disable PMTU
3908 * discovery upon the specified address. Note that
3909 * if the address feild is empty then all addresses
3910 * on the association are effected. Not also that
3911 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3912 * exclusive. Enabling both will have undetermined
3913 * results.
3915 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3916 * on delayed sack. The time specified in spp_sackdelay
3917 * is used to specify the sack delay for this address. Note
3918 * that if spp_address is empty then all addresses will
3919 * enable delayed sack and take on the sack delay
3920 * value specified in spp_sackdelay.
3921 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3922 * off delayed sack. If the spp_address field is blank then
3923 * delayed sack is disabled for the entire association. Note
3924 * also that this field is mutually exclusive to
3925 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3926 * results.
3928 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3929 char __user *optval, int __user *optlen)
3931 struct sctp_paddrparams params;
3932 struct sctp_transport *trans = NULL;
3933 struct sctp_association *asoc = NULL;
3934 struct sctp_sock *sp = sctp_sk(sk);
3936 if (len < sizeof(struct sctp_paddrparams))
3937 return -EINVAL;
3938 len = sizeof(struct sctp_paddrparams);
3939 if (copy_from_user(&params, optval, len))
3940 return -EFAULT;
3942 /* If an address other than INADDR_ANY is specified, and
3943 * no transport is found, then the request is invalid.
3945 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3946 trans = sctp_addr_id2transport(sk, &params.spp_address,
3947 params.spp_assoc_id);
3948 if (!trans) {
3949 SCTP_DEBUG_PRINTK("Failed no transport\n");
3950 return -EINVAL;
3954 /* Get association, if assoc_id != 0 and the socket is a one
3955 * to many style socket, and an association was not found, then
3956 * the id was invalid.
3958 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3959 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3960 SCTP_DEBUG_PRINTK("Failed no association\n");
3961 return -EINVAL;
3964 if (trans) {
3965 /* Fetch transport values. */
3966 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3967 params.spp_pathmtu = trans->pathmtu;
3968 params.spp_pathmaxrxt = trans->pathmaxrxt;
3969 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3971 /*draft-11 doesn't say what to return in spp_flags*/
3972 params.spp_flags = trans->param_flags;
3973 } else if (asoc) {
3974 /* Fetch association values. */
3975 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3976 params.spp_pathmtu = asoc->pathmtu;
3977 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3978 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3980 /*draft-11 doesn't say what to return in spp_flags*/
3981 params.spp_flags = asoc->param_flags;
3982 } else {
3983 /* Fetch socket values. */
3984 params.spp_hbinterval = sp->hbinterval;
3985 params.spp_pathmtu = sp->pathmtu;
3986 params.spp_sackdelay = sp->sackdelay;
3987 params.spp_pathmaxrxt = sp->pathmaxrxt;
3989 /*draft-11 doesn't say what to return in spp_flags*/
3990 params.spp_flags = sp->param_flags;
3993 if (copy_to_user(optval, &params, len))
3994 return -EFAULT;
3996 if (put_user(len, optlen))
3997 return -EFAULT;
3999 return 0;
4002 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
4004 * This options will get or set the delayed ack timer. The time is set
4005 * in milliseconds. If the assoc_id is 0, then this sets or gets the
4006 * endpoints default delayed ack timer value. If the assoc_id field is
4007 * non-zero, then the set or get effects the specified association.
4009 * struct sctp_assoc_value {
4010 * sctp_assoc_t assoc_id;
4011 * uint32_t assoc_value;
4012 * };
4014 * assoc_id - This parameter, indicates which association the
4015 * user is preforming an action upon. Note that if
4016 * this field's value is zero then the endpoints
4017 * default value is changed (effecting future
4018 * associations only).
4020 * assoc_value - This parameter contains the number of milliseconds
4021 * that the user is requesting the delayed ACK timer
4022 * be set to. Note that this value is defined in
4023 * the standard to be between 200 and 500 milliseconds.
4025 * Note: a value of zero will leave the value alone,
4026 * but disable SACK delay. A non-zero value will also
4027 * enable SACK delay.
4029 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
4030 char __user *optval,
4031 int __user *optlen)
4033 struct sctp_assoc_value params;
4034 struct sctp_association *asoc = NULL;
4035 struct sctp_sock *sp = sctp_sk(sk);
4037 if (len < sizeof(struct sctp_assoc_value))
4038 return - EINVAL;
4040 len = sizeof(struct sctp_assoc_value);
4042 if (copy_from_user(&params, optval, len))
4043 return -EFAULT;
4045 /* Get association, if assoc_id != 0 and the socket is a one
4046 * to many style socket, and an association was not found, then
4047 * the id was invalid.
4049 asoc = sctp_id2assoc(sk, params.assoc_id);
4050 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4051 return -EINVAL;
4053 if (asoc) {
4054 /* Fetch association values. */
4055 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
4056 params.assoc_value = jiffies_to_msecs(
4057 asoc->sackdelay);
4058 else
4059 params.assoc_value = 0;
4060 } else {
4061 /* Fetch socket values. */
4062 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
4063 params.assoc_value = sp->sackdelay;
4064 else
4065 params.assoc_value = 0;
4068 if (copy_to_user(optval, &params, len))
4069 return -EFAULT;
4071 if (put_user(len, optlen))
4072 return -EFAULT;
4074 return 0;
4077 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4079 * Applications can specify protocol parameters for the default association
4080 * initialization. The option name argument to setsockopt() and getsockopt()
4081 * is SCTP_INITMSG.
4083 * Setting initialization parameters is effective only on an unconnected
4084 * socket (for UDP-style sockets only future associations are effected
4085 * by the change). With TCP-style sockets, this option is inherited by
4086 * sockets derived from a listener socket.
4088 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4090 if (len < sizeof(struct sctp_initmsg))
4091 return -EINVAL;
4092 len = sizeof(struct sctp_initmsg);
4093 if (put_user(len, optlen))
4094 return -EFAULT;
4095 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4096 return -EFAULT;
4097 return 0;
4100 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4101 char __user *optval,
4102 int __user *optlen)
4104 sctp_assoc_t id;
4105 struct sctp_association *asoc;
4106 struct list_head *pos;
4107 int cnt = 0;
4109 if (len < sizeof(sctp_assoc_t))
4110 return -EINVAL;
4112 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4113 return -EFAULT;
4115 /* For UDP-style sockets, id specifies the association to query. */
4116 asoc = sctp_id2assoc(sk, id);
4117 if (!asoc)
4118 return -EINVAL;
4120 list_for_each(pos, &asoc->peer.transport_addr_list) {
4121 cnt ++;
4124 return cnt;
4128 * Old API for getting list of peer addresses. Does not work for 32-bit
4129 * programs running on a 64-bit kernel
4131 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4132 char __user *optval,
4133 int __user *optlen)
4135 struct sctp_association *asoc;
4136 int cnt = 0;
4137 struct sctp_getaddrs_old getaddrs;
4138 struct sctp_transport *from;
4139 void __user *to;
4140 union sctp_addr temp;
4141 struct sctp_sock *sp = sctp_sk(sk);
4142 int addrlen;
4144 if (len < sizeof(struct sctp_getaddrs_old))
4145 return -EINVAL;
4147 len = sizeof(struct sctp_getaddrs_old);
4149 if (copy_from_user(&getaddrs, optval, len))
4150 return -EFAULT;
4152 if (getaddrs.addr_num <= 0) return -EINVAL;
4154 /* For UDP-style sockets, id specifies the association to query. */
4155 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4156 if (!asoc)
4157 return -EINVAL;
4159 to = (void __user *)getaddrs.addrs;
4160 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4161 transports) {
4162 memcpy(&temp, &from->ipaddr, sizeof(temp));
4163 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4164 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4165 if (copy_to_user(to, &temp, addrlen))
4166 return -EFAULT;
4167 to += addrlen ;
4168 cnt ++;
4169 if (cnt >= getaddrs.addr_num) break;
4171 getaddrs.addr_num = cnt;
4172 if (put_user(len, optlen))
4173 return -EFAULT;
4174 if (copy_to_user(optval, &getaddrs, len))
4175 return -EFAULT;
4177 return 0;
4180 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4181 char __user *optval, int __user *optlen)
4183 struct sctp_association *asoc;
4184 int cnt = 0;
4185 struct sctp_getaddrs getaddrs;
4186 struct sctp_transport *from;
4187 void __user *to;
4188 union sctp_addr temp;
4189 struct sctp_sock *sp = sctp_sk(sk);
4190 int addrlen;
4191 size_t space_left;
4192 int bytes_copied;
4194 if (len < sizeof(struct sctp_getaddrs))
4195 return -EINVAL;
4197 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4198 return -EFAULT;
4200 /* For UDP-style sockets, id specifies the association to query. */
4201 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4202 if (!asoc)
4203 return -EINVAL;
4205 to = optval + offsetof(struct sctp_getaddrs,addrs);
4206 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4208 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4209 transports) {
4210 memcpy(&temp, &from->ipaddr, sizeof(temp));
4211 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4212 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4213 if (space_left < addrlen)
4214 return -ENOMEM;
4215 if (copy_to_user(to, &temp, addrlen))
4216 return -EFAULT;
4217 to += addrlen;
4218 cnt++;
4219 space_left -= addrlen;
4222 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4223 return -EFAULT;
4224 bytes_copied = ((char __user *)to) - optval;
4225 if (put_user(bytes_copied, optlen))
4226 return -EFAULT;
4228 return 0;
4231 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4232 char __user *optval,
4233 int __user *optlen)
4235 sctp_assoc_t id;
4236 struct sctp_bind_addr *bp;
4237 struct sctp_association *asoc;
4238 struct sctp_sockaddr_entry *addr;
4239 int cnt = 0;
4241 if (len < sizeof(sctp_assoc_t))
4242 return -EINVAL;
4244 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4245 return -EFAULT;
4248 * For UDP-style sockets, id specifies the association to query.
4249 * If the id field is set to the value '0' then the locally bound
4250 * addresses are returned without regard to any particular
4251 * association.
4253 if (0 == id) {
4254 bp = &sctp_sk(sk)->ep->base.bind_addr;
4255 } else {
4256 asoc = sctp_id2assoc(sk, id);
4257 if (!asoc)
4258 return -EINVAL;
4259 bp = &asoc->base.bind_addr;
4262 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4263 * addresses from the global local address list.
4265 if (sctp_list_single_entry(&bp->address_list)) {
4266 addr = list_entry(bp->address_list.next,
4267 struct sctp_sockaddr_entry, list);
4268 if (sctp_is_any(&addr->a)) {
4269 rcu_read_lock();
4270 list_for_each_entry_rcu(addr,
4271 &sctp_local_addr_list, list) {
4272 if (!addr->valid)
4273 continue;
4275 if ((PF_INET == sk->sk_family) &&
4276 (AF_INET6 == addr->a.sa.sa_family))
4277 continue;
4279 cnt++;
4281 rcu_read_unlock();
4282 } else {
4283 cnt = 1;
4285 goto done;
4288 /* Protection on the bound address list is not needed,
4289 * since in the socket option context we hold the socket lock,
4290 * so there is no way that the bound address list can change.
4292 list_for_each_entry(addr, &bp->address_list, list) {
4293 cnt ++;
4295 done:
4296 return cnt;
4299 /* Helper function that copies local addresses to user and returns the number
4300 * of addresses copied.
4302 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4303 int max_addrs, void *to,
4304 int *bytes_copied)
4306 struct sctp_sockaddr_entry *addr;
4307 union sctp_addr temp;
4308 int cnt = 0;
4309 int addrlen;
4311 rcu_read_lock();
4312 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4313 if (!addr->valid)
4314 continue;
4316 if ((PF_INET == sk->sk_family) &&
4317 (AF_INET6 == addr->a.sa.sa_family))
4318 continue;
4319 memcpy(&temp, &addr->a, sizeof(temp));
4320 if (!temp.v4.sin_port)
4321 temp.v4.sin_port = htons(port);
4323 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4324 &temp);
4325 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4326 memcpy(to, &temp, addrlen);
4328 to += addrlen;
4329 *bytes_copied += addrlen;
4330 cnt ++;
4331 if (cnt >= max_addrs) break;
4333 rcu_read_unlock();
4335 return cnt;
4338 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4339 size_t space_left, int *bytes_copied)
4341 struct sctp_sockaddr_entry *addr;
4342 union sctp_addr temp;
4343 int cnt = 0;
4344 int addrlen;
4346 rcu_read_lock();
4347 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4348 if (!addr->valid)
4349 continue;
4351 if ((PF_INET == sk->sk_family) &&
4352 (AF_INET6 == addr->a.sa.sa_family))
4353 continue;
4354 memcpy(&temp, &addr->a, sizeof(temp));
4355 if (!temp.v4.sin_port)
4356 temp.v4.sin_port = htons(port);
4358 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4359 &temp);
4360 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4361 if (space_left < addrlen) {
4362 cnt = -ENOMEM;
4363 break;
4365 memcpy(to, &temp, addrlen);
4367 to += addrlen;
4368 cnt ++;
4369 space_left -= addrlen;
4370 *bytes_copied += addrlen;
4372 rcu_read_unlock();
4374 return cnt;
4377 /* Old API for getting list of local addresses. Does not work for 32-bit
4378 * programs running on a 64-bit kernel
4380 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4381 char __user *optval, int __user *optlen)
4383 struct sctp_bind_addr *bp;
4384 struct sctp_association *asoc;
4385 int cnt = 0;
4386 struct sctp_getaddrs_old getaddrs;
4387 struct sctp_sockaddr_entry *addr;
4388 void __user *to;
4389 union sctp_addr temp;
4390 struct sctp_sock *sp = sctp_sk(sk);
4391 int addrlen;
4392 int err = 0;
4393 void *addrs;
4394 void *buf;
4395 int bytes_copied = 0;
4397 if (len < sizeof(struct sctp_getaddrs_old))
4398 return -EINVAL;
4400 len = sizeof(struct sctp_getaddrs_old);
4401 if (copy_from_user(&getaddrs, optval, len))
4402 return -EFAULT;
4404 if (getaddrs.addr_num <= 0 ||
4405 getaddrs.addr_num >= (INT_MAX / sizeof(union sctp_addr)))
4406 return -EINVAL;
4408 * For UDP-style sockets, id specifies the association to query.
4409 * If the id field is set to the value '0' then the locally bound
4410 * addresses are returned without regard to any particular
4411 * association.
4413 if (0 == getaddrs.assoc_id) {
4414 bp = &sctp_sk(sk)->ep->base.bind_addr;
4415 } else {
4416 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4417 if (!asoc)
4418 return -EINVAL;
4419 bp = &asoc->base.bind_addr;
4422 to = getaddrs.addrs;
4424 /* Allocate space for a local instance of packed array to hold all
4425 * the data. We store addresses here first and then put write them
4426 * to the user in one shot.
4428 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4429 GFP_KERNEL);
4430 if (!addrs)
4431 return -ENOMEM;
4433 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4434 * addresses from the global local address list.
4436 if (sctp_list_single_entry(&bp->address_list)) {
4437 addr = list_entry(bp->address_list.next,
4438 struct sctp_sockaddr_entry, list);
4439 if (sctp_is_any(&addr->a)) {
4440 cnt = sctp_copy_laddrs_old(sk, bp->port,
4441 getaddrs.addr_num,
4442 addrs, &bytes_copied);
4443 goto copy_getaddrs;
4447 buf = addrs;
4448 /* Protection on the bound address list is not needed since
4449 * in the socket option context we hold a socket lock and
4450 * thus the bound address list can't change.
4452 list_for_each_entry(addr, &bp->address_list, list) {
4453 memcpy(&temp, &addr->a, sizeof(temp));
4454 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4455 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4456 memcpy(buf, &temp, addrlen);
4457 buf += addrlen;
4458 bytes_copied += addrlen;
4459 cnt ++;
4460 if (cnt >= getaddrs.addr_num) break;
4463 copy_getaddrs:
4464 /* copy the entire address list into the user provided space */
4465 if (copy_to_user(to, addrs, bytes_copied)) {
4466 err = -EFAULT;
4467 goto error;
4470 /* copy the leading structure back to user */
4471 getaddrs.addr_num = cnt;
4472 if (copy_to_user(optval, &getaddrs, len))
4473 err = -EFAULT;
4475 error:
4476 kfree(addrs);
4477 return err;
4480 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4481 char __user *optval, int __user *optlen)
4483 struct sctp_bind_addr *bp;
4484 struct sctp_association *asoc;
4485 int cnt = 0;
4486 struct sctp_getaddrs getaddrs;
4487 struct sctp_sockaddr_entry *addr;
4488 void __user *to;
4489 union sctp_addr temp;
4490 struct sctp_sock *sp = sctp_sk(sk);
4491 int addrlen;
4492 int err = 0;
4493 size_t space_left;
4494 int bytes_copied = 0;
4495 void *addrs;
4496 void *buf;
4498 if (len < sizeof(struct sctp_getaddrs))
4499 return -EINVAL;
4501 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4502 return -EFAULT;
4505 * For UDP-style sockets, id specifies the association to query.
4506 * If the id field is set to the value '0' then the locally bound
4507 * addresses are returned without regard to any particular
4508 * association.
4510 if (0 == getaddrs.assoc_id) {
4511 bp = &sctp_sk(sk)->ep->base.bind_addr;
4512 } else {
4513 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4514 if (!asoc)
4515 return -EINVAL;
4516 bp = &asoc->base.bind_addr;
4519 to = optval + offsetof(struct sctp_getaddrs,addrs);
4520 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4522 addrs = kmalloc(space_left, GFP_KERNEL);
4523 if (!addrs)
4524 return -ENOMEM;
4526 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4527 * addresses from the global local address list.
4529 if (sctp_list_single_entry(&bp->address_list)) {
4530 addr = list_entry(bp->address_list.next,
4531 struct sctp_sockaddr_entry, list);
4532 if (sctp_is_any(&addr->a)) {
4533 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4534 space_left, &bytes_copied);
4535 if (cnt < 0) {
4536 err = cnt;
4537 goto out;
4539 goto copy_getaddrs;
4543 buf = addrs;
4544 /* Protection on the bound address list is not needed since
4545 * in the socket option context we hold a socket lock and
4546 * thus the bound address list can't change.
4548 list_for_each_entry(addr, &bp->address_list, list) {
4549 memcpy(&temp, &addr->a, sizeof(temp));
4550 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4551 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4552 if (space_left < addrlen) {
4553 err = -ENOMEM; /*fixme: right error?*/
4554 goto out;
4556 memcpy(buf, &temp, addrlen);
4557 buf += addrlen;
4558 bytes_copied += addrlen;
4559 cnt ++;
4560 space_left -= addrlen;
4563 copy_getaddrs:
4564 if (copy_to_user(to, addrs, bytes_copied)) {
4565 err = -EFAULT;
4566 goto out;
4568 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4569 err = -EFAULT;
4570 goto out;
4572 if (put_user(bytes_copied, optlen))
4573 err = -EFAULT;
4574 out:
4575 kfree(addrs);
4576 return err;
4579 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4581 * Requests that the local SCTP stack use the enclosed peer address as
4582 * the association primary. The enclosed address must be one of the
4583 * association peer's addresses.
4585 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4586 char __user *optval, int __user *optlen)
4588 struct sctp_prim prim;
4589 struct sctp_association *asoc;
4590 struct sctp_sock *sp = sctp_sk(sk);
4592 if (len < sizeof(struct sctp_prim))
4593 return -EINVAL;
4595 len = sizeof(struct sctp_prim);
4597 if (copy_from_user(&prim, optval, len))
4598 return -EFAULT;
4600 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4601 if (!asoc)
4602 return -EINVAL;
4604 if (!asoc->peer.primary_path)
4605 return -ENOTCONN;
4607 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4608 asoc->peer.primary_path->af_specific->sockaddr_len);
4610 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4611 (union sctp_addr *)&prim.ssp_addr);
4613 if (put_user(len, optlen))
4614 return -EFAULT;
4615 if (copy_to_user(optval, &prim, len))
4616 return -EFAULT;
4618 return 0;
4622 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4624 * Requests that the local endpoint set the specified Adaptation Layer
4625 * Indication parameter for all future INIT and INIT-ACK exchanges.
4627 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4628 char __user *optval, int __user *optlen)
4630 struct sctp_setadaptation adaptation;
4632 if (len < sizeof(struct sctp_setadaptation))
4633 return -EINVAL;
4635 len = sizeof(struct sctp_setadaptation);
4637 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4639 if (put_user(len, optlen))
4640 return -EFAULT;
4641 if (copy_to_user(optval, &adaptation, len))
4642 return -EFAULT;
4644 return 0;
4649 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4651 * Applications that wish to use the sendto() system call may wish to
4652 * specify a default set of parameters that would normally be supplied
4653 * through the inclusion of ancillary data. This socket option allows
4654 * such an application to set the default sctp_sndrcvinfo structure.
4657 * The application that wishes to use this socket option simply passes
4658 * in to this call the sctp_sndrcvinfo structure defined in Section
4659 * 5.2.2) The input parameters accepted by this call include
4660 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4661 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4662 * to this call if the caller is using the UDP model.
4664 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4666 static int sctp_getsockopt_default_send_param(struct sock *sk,
4667 int len, char __user *optval,
4668 int __user *optlen)
4670 struct sctp_sndrcvinfo info;
4671 struct sctp_association *asoc;
4672 struct sctp_sock *sp = sctp_sk(sk);
4674 if (len < sizeof(struct sctp_sndrcvinfo))
4675 return -EINVAL;
4677 len = sizeof(struct sctp_sndrcvinfo);
4679 if (copy_from_user(&info, optval, len))
4680 return -EFAULT;
4682 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4683 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4684 return -EINVAL;
4686 if (asoc) {
4687 info.sinfo_stream = asoc->default_stream;
4688 info.sinfo_flags = asoc->default_flags;
4689 info.sinfo_ppid = asoc->default_ppid;
4690 info.sinfo_context = asoc->default_context;
4691 info.sinfo_timetolive = asoc->default_timetolive;
4692 } else {
4693 info.sinfo_stream = sp->default_stream;
4694 info.sinfo_flags = sp->default_flags;
4695 info.sinfo_ppid = sp->default_ppid;
4696 info.sinfo_context = sp->default_context;
4697 info.sinfo_timetolive = sp->default_timetolive;
4700 if (put_user(len, optlen))
4701 return -EFAULT;
4702 if (copy_to_user(optval, &info, len))
4703 return -EFAULT;
4705 return 0;
4710 * 7.1.5 SCTP_NODELAY
4712 * Turn on/off any Nagle-like algorithm. This means that packets are
4713 * generally sent as soon as possible and no unnecessary delays are
4714 * introduced, at the cost of more packets in the network. Expects an
4715 * integer boolean flag.
4718 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4719 char __user *optval, int __user *optlen)
4721 int val;
4723 if (len < sizeof(int))
4724 return -EINVAL;
4726 len = sizeof(int);
4727 val = (sctp_sk(sk)->nodelay == 1);
4728 if (put_user(len, optlen))
4729 return -EFAULT;
4730 if (copy_to_user(optval, &val, len))
4731 return -EFAULT;
4732 return 0;
4737 * 7.1.1 SCTP_RTOINFO
4739 * The protocol parameters used to initialize and bound retransmission
4740 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4741 * and modify these parameters.
4742 * All parameters are time values, in milliseconds. A value of 0, when
4743 * modifying the parameters, indicates that the current value should not
4744 * be changed.
4747 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4748 char __user *optval,
4749 int __user *optlen) {
4750 struct sctp_rtoinfo rtoinfo;
4751 struct sctp_association *asoc;
4753 if (len < sizeof (struct sctp_rtoinfo))
4754 return -EINVAL;
4756 len = sizeof(struct sctp_rtoinfo);
4758 if (copy_from_user(&rtoinfo, optval, len))
4759 return -EFAULT;
4761 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4763 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4764 return -EINVAL;
4766 /* Values corresponding to the specific association. */
4767 if (asoc) {
4768 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4769 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4770 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4771 } else {
4772 /* Values corresponding to the endpoint. */
4773 struct sctp_sock *sp = sctp_sk(sk);
4775 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4776 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4777 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4780 if (put_user(len, optlen))
4781 return -EFAULT;
4783 if (copy_to_user(optval, &rtoinfo, len))
4784 return -EFAULT;
4786 return 0;
4791 * 7.1.2 SCTP_ASSOCINFO
4793 * This option is used to tune the maximum retransmission attempts
4794 * of the association.
4795 * Returns an error if the new association retransmission value is
4796 * greater than the sum of the retransmission value of the peer.
4797 * See [SCTP] for more information.
4800 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4801 char __user *optval,
4802 int __user *optlen)
4805 struct sctp_assocparams assocparams;
4806 struct sctp_association *asoc;
4807 struct list_head *pos;
4808 int cnt = 0;
4810 if (len < sizeof (struct sctp_assocparams))
4811 return -EINVAL;
4813 len = sizeof(struct sctp_assocparams);
4815 if (copy_from_user(&assocparams, optval, len))
4816 return -EFAULT;
4818 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4820 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4821 return -EINVAL;
4823 /* Values correspoinding to the specific association */
4824 if (asoc) {
4825 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4826 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4827 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4828 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4829 * 1000) +
4830 (asoc->cookie_life.tv_usec
4831 / 1000);
4833 list_for_each(pos, &asoc->peer.transport_addr_list) {
4834 cnt ++;
4837 assocparams.sasoc_number_peer_destinations = cnt;
4838 } else {
4839 /* Values corresponding to the endpoint */
4840 struct sctp_sock *sp = sctp_sk(sk);
4842 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4843 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4844 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4845 assocparams.sasoc_cookie_life =
4846 sp->assocparams.sasoc_cookie_life;
4847 assocparams.sasoc_number_peer_destinations =
4848 sp->assocparams.
4849 sasoc_number_peer_destinations;
4852 if (put_user(len, optlen))
4853 return -EFAULT;
4855 if (copy_to_user(optval, &assocparams, len))
4856 return -EFAULT;
4858 return 0;
4862 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4864 * This socket option is a boolean flag which turns on or off mapped V4
4865 * addresses. If this option is turned on and the socket is type
4866 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4867 * If this option is turned off, then no mapping will be done of V4
4868 * addresses and a user will receive both PF_INET6 and PF_INET type
4869 * addresses on the socket.
4871 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4872 char __user *optval, int __user *optlen)
4874 int val;
4875 struct sctp_sock *sp = sctp_sk(sk);
4877 if (len < sizeof(int))
4878 return -EINVAL;
4880 len = sizeof(int);
4881 val = sp->v4mapped;
4882 if (put_user(len, optlen))
4883 return -EFAULT;
4884 if (copy_to_user(optval, &val, len))
4885 return -EFAULT;
4887 return 0;
4891 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4892 * (chapter and verse is quoted at sctp_setsockopt_context())
4894 static int sctp_getsockopt_context(struct sock *sk, int len,
4895 char __user *optval, int __user *optlen)
4897 struct sctp_assoc_value params;
4898 struct sctp_sock *sp;
4899 struct sctp_association *asoc;
4901 if (len < sizeof(struct sctp_assoc_value))
4902 return -EINVAL;
4904 len = sizeof(struct sctp_assoc_value);
4906 if (copy_from_user(&params, optval, len))
4907 return -EFAULT;
4909 sp = sctp_sk(sk);
4911 if (params.assoc_id != 0) {
4912 asoc = sctp_id2assoc(sk, params.assoc_id);
4913 if (!asoc)
4914 return -EINVAL;
4915 params.assoc_value = asoc->default_rcv_context;
4916 } else {
4917 params.assoc_value = sp->default_rcv_context;
4920 if (put_user(len, optlen))
4921 return -EFAULT;
4922 if (copy_to_user(optval, &params, len))
4923 return -EFAULT;
4925 return 0;
4929 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4931 * This socket option specifies the maximum size to put in any outgoing
4932 * SCTP chunk. If a message is larger than this size it will be
4933 * fragmented by SCTP into the specified size. Note that the underlying
4934 * SCTP implementation may fragment into smaller sized chunks when the
4935 * PMTU of the underlying association is smaller than the value set by
4936 * the user.
4938 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4939 char __user *optval, int __user *optlen)
4941 int val;
4943 if (len < sizeof(int))
4944 return -EINVAL;
4946 len = sizeof(int);
4948 val = sctp_sk(sk)->user_frag;
4949 if (put_user(len, optlen))
4950 return -EFAULT;
4951 if (copy_to_user(optval, &val, len))
4952 return -EFAULT;
4954 return 0;
4958 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4959 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4961 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4962 char __user *optval, int __user *optlen)
4964 int val;
4966 if (len < sizeof(int))
4967 return -EINVAL;
4969 len = sizeof(int);
4971 val = sctp_sk(sk)->frag_interleave;
4972 if (put_user(len, optlen))
4973 return -EFAULT;
4974 if (copy_to_user(optval, &val, len))
4975 return -EFAULT;
4977 return 0;
4981 * 7.1.25. Set or Get the sctp partial delivery point
4982 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
4984 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
4985 char __user *optval,
4986 int __user *optlen)
4988 u32 val;
4990 if (len < sizeof(u32))
4991 return -EINVAL;
4993 len = sizeof(u32);
4995 val = sctp_sk(sk)->pd_point;
4996 if (put_user(len, optlen))
4997 return -EFAULT;
4998 if (copy_to_user(optval, &val, len))
4999 return -EFAULT;
5001 return -ENOTSUPP;
5005 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5006 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5008 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5009 char __user *optval,
5010 int __user *optlen)
5012 struct sctp_assoc_value params;
5013 struct sctp_sock *sp;
5014 struct sctp_association *asoc;
5016 if (len < sizeof(int))
5017 return -EINVAL;
5019 if (len == sizeof(int)) {
5020 printk(KERN_WARNING
5021 "SCTP: Use of int in max_burst socket option deprecated\n");
5022 printk(KERN_WARNING
5023 "SCTP: Use struct sctp_assoc_value instead\n");
5024 params.assoc_id = 0;
5025 } else if (len == sizeof (struct sctp_assoc_value)) {
5026 if (copy_from_user(&params, optval, len))
5027 return -EFAULT;
5028 } else
5029 return -EINVAL;
5031 sp = sctp_sk(sk);
5033 if (params.assoc_id != 0) {
5034 asoc = sctp_id2assoc(sk, params.assoc_id);
5035 if (!asoc)
5036 return -EINVAL;
5037 params.assoc_value = asoc->max_burst;
5038 } else
5039 params.assoc_value = sp->max_burst;
5041 if (len == sizeof(int)) {
5042 if (copy_to_user(optval, &params.assoc_value, len))
5043 return -EFAULT;
5044 } else {
5045 if (copy_to_user(optval, &params, len))
5046 return -EFAULT;
5049 return 0;
5053 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5054 char __user *optval, int __user *optlen)
5056 struct sctp_hmac_algo_param *hmacs;
5057 __u16 param_len;
5059 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5060 param_len = ntohs(hmacs->param_hdr.length);
5062 if (len < param_len)
5063 return -EINVAL;
5064 if (put_user(len, optlen))
5065 return -EFAULT;
5066 if (copy_to_user(optval, hmacs->hmac_ids, len))
5067 return -EFAULT;
5069 return 0;
5072 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5073 char __user *optval, int __user *optlen)
5075 struct sctp_authkeyid val;
5076 struct sctp_association *asoc;
5078 if (len < sizeof(struct sctp_authkeyid))
5079 return -EINVAL;
5080 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5081 return -EFAULT;
5083 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5084 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5085 return -EINVAL;
5087 if (asoc)
5088 val.scact_keynumber = asoc->active_key_id;
5089 else
5090 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5092 return 0;
5095 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5096 char __user *optval, int __user *optlen)
5098 struct sctp_authchunks __user *p = (void __user *)optval;
5099 struct sctp_authchunks val;
5100 struct sctp_association *asoc;
5101 struct sctp_chunks_param *ch;
5102 u32 num_chunks;
5103 char __user *to;
5105 if (len <= sizeof(struct sctp_authchunks))
5106 return -EINVAL;
5108 if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5109 return -EFAULT;
5111 to = p->gauth_chunks;
5112 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5113 if (!asoc)
5114 return -EINVAL;
5116 ch = asoc->peer.peer_chunks;
5118 /* See if the user provided enough room for all the data */
5119 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5120 if (len < num_chunks)
5121 return -EINVAL;
5123 len = num_chunks;
5124 if (put_user(len, optlen))
5125 return -EFAULT;
5126 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5127 return -EFAULT;
5128 if (copy_to_user(to, ch->chunks, len))
5129 return -EFAULT;
5131 return 0;
5134 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5135 char __user *optval, int __user *optlen)
5137 struct sctp_authchunks __user *p = (void __user *)optval;
5138 struct sctp_authchunks val;
5139 struct sctp_association *asoc;
5140 struct sctp_chunks_param *ch;
5141 u32 num_chunks;
5142 char __user *to;
5144 if (len <= sizeof(struct sctp_authchunks))
5145 return -EINVAL;
5147 if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5148 return -EFAULT;
5150 to = p->gauth_chunks;
5151 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5152 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5153 return -EINVAL;
5155 if (asoc)
5156 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5157 else
5158 ch = sctp_sk(sk)->ep->auth_chunk_list;
5160 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5161 if (len < num_chunks)
5162 return -EINVAL;
5164 len = num_chunks;
5165 if (put_user(len, optlen))
5166 return -EFAULT;
5167 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5168 return -EFAULT;
5169 if (copy_to_user(to, ch->chunks, len))
5170 return -EFAULT;
5172 return 0;
5175 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5176 char __user *optval, int __user *optlen)
5178 int retval = 0;
5179 int len;
5181 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5182 sk, optname);
5184 /* I can hardly begin to describe how wrong this is. This is
5185 * so broken as to be worse than useless. The API draft
5186 * REALLY is NOT helpful here... I am not convinced that the
5187 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5188 * are at all well-founded.
5190 if (level != SOL_SCTP) {
5191 struct sctp_af *af = sctp_sk(sk)->pf->af;
5193 retval = af->getsockopt(sk, level, optname, optval, optlen);
5194 return retval;
5197 if (get_user(len, optlen))
5198 return -EFAULT;
5200 sctp_lock_sock(sk);
5202 switch (optname) {
5203 case SCTP_STATUS:
5204 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5205 break;
5206 case SCTP_DISABLE_FRAGMENTS:
5207 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5208 optlen);
5209 break;
5210 case SCTP_EVENTS:
5211 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5212 break;
5213 case SCTP_AUTOCLOSE:
5214 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5215 break;
5216 case SCTP_SOCKOPT_PEELOFF:
5217 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5218 break;
5219 case SCTP_PEER_ADDR_PARAMS:
5220 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5221 optlen);
5222 break;
5223 case SCTP_DELAYED_ACK_TIME:
5224 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
5225 optlen);
5226 break;
5227 case SCTP_INITMSG:
5228 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5229 break;
5230 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5231 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5232 optlen);
5233 break;
5234 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5235 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5236 optlen);
5237 break;
5238 case SCTP_GET_PEER_ADDRS_OLD:
5239 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5240 optlen);
5241 break;
5242 case SCTP_GET_LOCAL_ADDRS_OLD:
5243 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5244 optlen);
5245 break;
5246 case SCTP_GET_PEER_ADDRS:
5247 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5248 optlen);
5249 break;
5250 case SCTP_GET_LOCAL_ADDRS:
5251 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5252 optlen);
5253 break;
5254 case SCTP_DEFAULT_SEND_PARAM:
5255 retval = sctp_getsockopt_default_send_param(sk, len,
5256 optval, optlen);
5257 break;
5258 case SCTP_PRIMARY_ADDR:
5259 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5260 break;
5261 case SCTP_NODELAY:
5262 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5263 break;
5264 case SCTP_RTOINFO:
5265 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5266 break;
5267 case SCTP_ASSOCINFO:
5268 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5269 break;
5270 case SCTP_I_WANT_MAPPED_V4_ADDR:
5271 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5272 break;
5273 case SCTP_MAXSEG:
5274 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5275 break;
5276 case SCTP_GET_PEER_ADDR_INFO:
5277 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5278 optlen);
5279 break;
5280 case SCTP_ADAPTATION_LAYER:
5281 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5282 optlen);
5283 break;
5284 case SCTP_CONTEXT:
5285 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5286 break;
5287 case SCTP_FRAGMENT_INTERLEAVE:
5288 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5289 optlen);
5290 break;
5291 case SCTP_PARTIAL_DELIVERY_POINT:
5292 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5293 optlen);
5294 break;
5295 case SCTP_MAX_BURST:
5296 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5297 break;
5298 case SCTP_AUTH_KEY:
5299 case SCTP_AUTH_CHUNK:
5300 case SCTP_AUTH_DELETE_KEY:
5301 retval = -EOPNOTSUPP;
5302 break;
5303 case SCTP_HMAC_IDENT:
5304 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5305 break;
5306 case SCTP_AUTH_ACTIVE_KEY:
5307 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5308 break;
5309 case SCTP_PEER_AUTH_CHUNKS:
5310 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5311 optlen);
5312 break;
5313 case SCTP_LOCAL_AUTH_CHUNKS:
5314 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5315 optlen);
5316 break;
5317 default:
5318 retval = -ENOPROTOOPT;
5319 break;
5322 sctp_release_sock(sk);
5323 return retval;
5326 static void sctp_hash(struct sock *sk)
5328 /* STUB */
5331 static void sctp_unhash(struct sock *sk)
5333 /* STUB */
5336 /* Check if port is acceptable. Possibly find first available port.
5338 * The port hash table (contained in the 'global' SCTP protocol storage
5339 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5340 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5341 * list (the list number is the port number hashed out, so as you
5342 * would expect from a hash function, all the ports in a given list have
5343 * such a number that hashes out to the same list number; you were
5344 * expecting that, right?); so each list has a set of ports, with a
5345 * link to the socket (struct sock) that uses it, the port number and
5346 * a fastreuse flag (FIXME: NPI ipg).
5348 static struct sctp_bind_bucket *sctp_bucket_create(
5349 struct sctp_bind_hashbucket *head, unsigned short snum);
5351 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5353 struct sctp_bind_hashbucket *head; /* hash list */
5354 struct sctp_bind_bucket *pp; /* hash list port iterator */
5355 struct hlist_node *node;
5356 unsigned short snum;
5357 int ret;
5359 snum = ntohs(addr->v4.sin_port);
5361 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5362 sctp_local_bh_disable();
5364 if (snum == 0) {
5365 /* Search for an available port. */
5366 int low, high, remaining, index;
5367 unsigned int rover;
5369 inet_get_local_port_range(&low, &high);
5370 remaining = (high - low) + 1;
5371 rover = net_random() % remaining + low;
5373 do {
5374 rover++;
5375 if ((rover < low) || (rover > high))
5376 rover = low;
5377 index = sctp_phashfn(rover);
5378 head = &sctp_port_hashtable[index];
5379 sctp_spin_lock(&head->lock);
5380 sctp_for_each_hentry(pp, node, &head->chain)
5381 if (pp->port == rover)
5382 goto next;
5383 break;
5384 next:
5385 sctp_spin_unlock(&head->lock);
5386 } while (--remaining > 0);
5388 /* Exhausted local port range during search? */
5389 ret = 1;
5390 if (remaining <= 0)
5391 goto fail;
5393 /* OK, here is the one we will use. HEAD (the port
5394 * hash table list entry) is non-NULL and we hold it's
5395 * mutex.
5397 snum = rover;
5398 } else {
5399 /* We are given an specific port number; we verify
5400 * that it is not being used. If it is used, we will
5401 * exahust the search in the hash list corresponding
5402 * to the port number (snum) - we detect that with the
5403 * port iterator, pp being NULL.
5405 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5406 sctp_spin_lock(&head->lock);
5407 sctp_for_each_hentry(pp, node, &head->chain) {
5408 if (pp->port == snum)
5409 goto pp_found;
5412 pp = NULL;
5413 goto pp_not_found;
5414 pp_found:
5415 if (!hlist_empty(&pp->owner)) {
5416 /* We had a port hash table hit - there is an
5417 * available port (pp != NULL) and it is being
5418 * used by other socket (pp->owner not empty); that other
5419 * socket is going to be sk2.
5421 int reuse = sk->sk_reuse;
5422 struct sock *sk2;
5423 struct hlist_node *node;
5425 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5426 if (pp->fastreuse && sk->sk_reuse &&
5427 sk->sk_state != SCTP_SS_LISTENING)
5428 goto success;
5430 /* Run through the list of sockets bound to the port
5431 * (pp->port) [via the pointers bind_next and
5432 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5433 * we get the endpoint they describe and run through
5434 * the endpoint's list of IP (v4 or v6) addresses,
5435 * comparing each of the addresses with the address of
5436 * the socket sk. If we find a match, then that means
5437 * that this port/socket (sk) combination are already
5438 * in an endpoint.
5440 sk_for_each_bound(sk2, node, &pp->owner) {
5441 struct sctp_endpoint *ep2;
5442 ep2 = sctp_sk(sk2)->ep;
5444 if (reuse && sk2->sk_reuse &&
5445 sk2->sk_state != SCTP_SS_LISTENING)
5446 continue;
5448 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5449 sctp_sk(sk))) {
5450 ret = (long)sk2;
5451 goto fail_unlock;
5454 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5456 pp_not_found:
5457 /* If there was a hash table miss, create a new port. */
5458 ret = 1;
5459 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5460 goto fail_unlock;
5462 /* In either case (hit or miss), make sure fastreuse is 1 only
5463 * if sk->sk_reuse is too (that is, if the caller requested
5464 * SO_REUSEADDR on this socket -sk-).
5466 if (hlist_empty(&pp->owner)) {
5467 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5468 pp->fastreuse = 1;
5469 else
5470 pp->fastreuse = 0;
5471 } else if (pp->fastreuse &&
5472 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5473 pp->fastreuse = 0;
5475 /* We are set, so fill up all the data in the hash table
5476 * entry, tie the socket list information with the rest of the
5477 * sockets FIXME: Blurry, NPI (ipg).
5479 success:
5480 if (!sctp_sk(sk)->bind_hash) {
5481 inet_sk(sk)->num = snum;
5482 sk_add_bind_node(sk, &pp->owner);
5483 sctp_sk(sk)->bind_hash = pp;
5485 ret = 0;
5487 fail_unlock:
5488 sctp_spin_unlock(&head->lock);
5490 fail:
5491 sctp_local_bh_enable();
5492 return ret;
5495 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5496 * port is requested.
5498 static int sctp_get_port(struct sock *sk, unsigned short snum)
5500 long ret;
5501 union sctp_addr addr;
5502 struct sctp_af *af = sctp_sk(sk)->pf->af;
5504 /* Set up a dummy address struct from the sk. */
5505 af->from_sk(&addr, sk);
5506 addr.v4.sin_port = htons(snum);
5508 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5509 ret = sctp_get_port_local(sk, &addr);
5511 return (ret ? 1 : 0);
5515 * 3.1.3 listen() - UDP Style Syntax
5517 * By default, new associations are not accepted for UDP style sockets.
5518 * An application uses listen() to mark a socket as being able to
5519 * accept new associations.
5521 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5523 struct sctp_sock *sp = sctp_sk(sk);
5524 struct sctp_endpoint *ep = sp->ep;
5526 /* Only UDP style sockets that are not peeled off are allowed to
5527 * listen().
5529 if (!sctp_style(sk, UDP))
5530 return -EINVAL;
5532 /* If backlog is zero, disable listening. */
5533 if (!backlog) {
5534 if (sctp_sstate(sk, CLOSED))
5535 return 0;
5537 sctp_unhash_endpoint(ep);
5538 sk->sk_state = SCTP_SS_CLOSED;
5539 return 0;
5542 /* Return if we are already listening. */
5543 if (sctp_sstate(sk, LISTENING))
5544 return 0;
5547 * If a bind() or sctp_bindx() is not called prior to a listen()
5548 * call that allows new associations to be accepted, the system
5549 * picks an ephemeral port and will choose an address set equivalent
5550 * to binding with a wildcard address.
5552 * This is not currently spelled out in the SCTP sockets
5553 * extensions draft, but follows the practice as seen in TCP
5554 * sockets.
5556 * Additionally, turn off fastreuse flag since we are not listening
5558 sk->sk_state = SCTP_SS_LISTENING;
5559 if (!ep->base.bind_addr.port) {
5560 if (sctp_autobind(sk))
5561 return -EAGAIN;
5562 } else
5563 sctp_sk(sk)->bind_hash->fastreuse = 0;
5565 sctp_hash_endpoint(ep);
5566 return 0;
5570 * 4.1.3 listen() - TCP Style Syntax
5572 * Applications uses listen() to ready the SCTP endpoint for accepting
5573 * inbound associations.
5575 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5577 struct sctp_sock *sp = sctp_sk(sk);
5578 struct sctp_endpoint *ep = sp->ep;
5580 /* If backlog is zero, disable listening. */
5581 if (!backlog) {
5582 if (sctp_sstate(sk, CLOSED))
5583 return 0;
5585 sctp_unhash_endpoint(ep);
5586 sk->sk_state = SCTP_SS_CLOSED;
5587 return 0;
5590 if (sctp_sstate(sk, LISTENING))
5591 return 0;
5594 * If a bind() or sctp_bindx() is not called prior to a listen()
5595 * call that allows new associations to be accepted, the system
5596 * picks an ephemeral port and will choose an address set equivalent
5597 * to binding with a wildcard address.
5599 * This is not currently spelled out in the SCTP sockets
5600 * extensions draft, but follows the practice as seen in TCP
5601 * sockets.
5603 sk->sk_state = SCTP_SS_LISTENING;
5604 if (!ep->base.bind_addr.port) {
5605 if (sctp_autobind(sk))
5606 return -EAGAIN;
5607 } else
5608 sctp_sk(sk)->bind_hash->fastreuse = 0;
5610 sk->sk_max_ack_backlog = backlog;
5611 sctp_hash_endpoint(ep);
5612 return 0;
5616 * Move a socket to LISTENING state.
5618 int sctp_inet_listen(struct socket *sock, int backlog)
5620 struct sock *sk = sock->sk;
5621 struct crypto_hash *tfm = NULL;
5622 int err = -EINVAL;
5624 if (unlikely(backlog < 0))
5625 goto out;
5627 sctp_lock_sock(sk);
5629 if (sock->state != SS_UNCONNECTED)
5630 goto out;
5632 /* Allocate HMAC for generating cookie. */
5633 if (sctp_hmac_alg) {
5634 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5635 if (IS_ERR(tfm)) {
5636 if (net_ratelimit()) {
5637 printk(KERN_INFO
5638 "SCTP: failed to load transform for %s: %ld\n",
5639 sctp_hmac_alg, PTR_ERR(tfm));
5641 err = -ENOSYS;
5642 goto out;
5646 switch (sock->type) {
5647 case SOCK_SEQPACKET:
5648 err = sctp_seqpacket_listen(sk, backlog);
5649 break;
5650 case SOCK_STREAM:
5651 err = sctp_stream_listen(sk, backlog);
5652 break;
5653 default:
5654 break;
5657 if (err)
5658 goto cleanup;
5660 /* Store away the transform reference. */
5661 sctp_sk(sk)->hmac = tfm;
5662 out:
5663 sctp_release_sock(sk);
5664 return err;
5665 cleanup:
5666 crypto_free_hash(tfm);
5667 goto out;
5671 * This function is done by modeling the current datagram_poll() and the
5672 * tcp_poll(). Note that, based on these implementations, we don't
5673 * lock the socket in this function, even though it seems that,
5674 * ideally, locking or some other mechanisms can be used to ensure
5675 * the integrity of the counters (sndbuf and wmem_alloc) used
5676 * in this place. We assume that we don't need locks either until proven
5677 * otherwise.
5679 * Another thing to note is that we include the Async I/O support
5680 * here, again, by modeling the current TCP/UDP code. We don't have
5681 * a good way to test with it yet.
5683 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5685 struct sock *sk = sock->sk;
5686 struct sctp_sock *sp = sctp_sk(sk);
5687 unsigned int mask;
5689 poll_wait(file, sk->sk_sleep, wait);
5691 /* A TCP-style listening socket becomes readable when the accept queue
5692 * is not empty.
5694 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5695 return (!list_empty(&sp->ep->asocs)) ?
5696 (POLLIN | POLLRDNORM) : 0;
5698 mask = 0;
5700 /* Is there any exceptional events? */
5701 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5702 mask |= POLLERR;
5703 if (sk->sk_shutdown & RCV_SHUTDOWN)
5704 mask |= POLLRDHUP;
5705 if (sk->sk_shutdown == SHUTDOWN_MASK)
5706 mask |= POLLHUP;
5708 /* Is it readable? Reconsider this code with TCP-style support. */
5709 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5710 (sk->sk_shutdown & RCV_SHUTDOWN))
5711 mask |= POLLIN | POLLRDNORM;
5713 /* The association is either gone or not ready. */
5714 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5715 return mask;
5717 /* Is it writable? */
5718 if (sctp_writeable(sk)) {
5719 mask |= POLLOUT | POLLWRNORM;
5720 } else {
5721 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5723 * Since the socket is not locked, the buffer
5724 * might be made available after the writeable check and
5725 * before the bit is set. This could cause a lost I/O
5726 * signal. tcp_poll() has a race breaker for this race
5727 * condition. Based on their implementation, we put
5728 * in the following code to cover it as well.
5730 if (sctp_writeable(sk))
5731 mask |= POLLOUT | POLLWRNORM;
5733 return mask;
5736 /********************************************************************
5737 * 2nd Level Abstractions
5738 ********************************************************************/
5740 static struct sctp_bind_bucket *sctp_bucket_create(
5741 struct sctp_bind_hashbucket *head, unsigned short snum)
5743 struct sctp_bind_bucket *pp;
5745 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5746 if (pp) {
5747 SCTP_DBG_OBJCNT_INC(bind_bucket);
5748 pp->port = snum;
5749 pp->fastreuse = 0;
5750 INIT_HLIST_HEAD(&pp->owner);
5751 hlist_add_head(&pp->node, &head->chain);
5753 return pp;
5756 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5757 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5759 if (pp && hlist_empty(&pp->owner)) {
5760 __hlist_del(&pp->node);
5761 kmem_cache_free(sctp_bucket_cachep, pp);
5762 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5766 /* Release this socket's reference to a local port. */
5767 static inline void __sctp_put_port(struct sock *sk)
5769 struct sctp_bind_hashbucket *head =
5770 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5771 struct sctp_bind_bucket *pp;
5773 sctp_spin_lock(&head->lock);
5774 pp = sctp_sk(sk)->bind_hash;
5775 __sk_del_bind_node(sk);
5776 sctp_sk(sk)->bind_hash = NULL;
5777 inet_sk(sk)->num = 0;
5778 sctp_bucket_destroy(pp);
5779 sctp_spin_unlock(&head->lock);
5782 void sctp_put_port(struct sock *sk)
5784 sctp_local_bh_disable();
5785 __sctp_put_port(sk);
5786 sctp_local_bh_enable();
5790 * The system picks an ephemeral port and choose an address set equivalent
5791 * to binding with a wildcard address.
5792 * One of those addresses will be the primary address for the association.
5793 * This automatically enables the multihoming capability of SCTP.
5795 static int sctp_autobind(struct sock *sk)
5797 union sctp_addr autoaddr;
5798 struct sctp_af *af;
5799 __be16 port;
5801 /* Initialize a local sockaddr structure to INADDR_ANY. */
5802 af = sctp_sk(sk)->pf->af;
5804 port = htons(inet_sk(sk)->num);
5805 af->inaddr_any(&autoaddr, port);
5807 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5810 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5812 * From RFC 2292
5813 * 4.2 The cmsghdr Structure *
5815 * When ancillary data is sent or received, any number of ancillary data
5816 * objects can be specified by the msg_control and msg_controllen members of
5817 * the msghdr structure, because each object is preceded by
5818 * a cmsghdr structure defining the object's length (the cmsg_len member).
5819 * Historically Berkeley-derived implementations have passed only one object
5820 * at a time, but this API allows multiple objects to be
5821 * passed in a single call to sendmsg() or recvmsg(). The following example
5822 * shows two ancillary data objects in a control buffer.
5824 * |<--------------------------- msg_controllen -------------------------->|
5825 * | |
5827 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5829 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5830 * | | |
5832 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5834 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5835 * | | | | |
5837 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5838 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5840 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5842 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5846 * msg_control
5847 * points here
5849 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5850 sctp_cmsgs_t *cmsgs)
5852 struct cmsghdr *cmsg;
5853 struct msghdr *my_msg = (struct msghdr *)msg;
5855 for (cmsg = CMSG_FIRSTHDR(msg);
5856 cmsg != NULL;
5857 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5858 if (!CMSG_OK(my_msg, cmsg))
5859 return -EINVAL;
5861 /* Should we parse this header or ignore? */
5862 if (cmsg->cmsg_level != IPPROTO_SCTP)
5863 continue;
5865 /* Strictly check lengths following example in SCM code. */
5866 switch (cmsg->cmsg_type) {
5867 case SCTP_INIT:
5868 /* SCTP Socket API Extension
5869 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5871 * This cmsghdr structure provides information for
5872 * initializing new SCTP associations with sendmsg().
5873 * The SCTP_INITMSG socket option uses this same data
5874 * structure. This structure is not used for
5875 * recvmsg().
5877 * cmsg_level cmsg_type cmsg_data[]
5878 * ------------ ------------ ----------------------
5879 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5881 if (cmsg->cmsg_len !=
5882 CMSG_LEN(sizeof(struct sctp_initmsg)))
5883 return -EINVAL;
5884 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5885 break;
5887 case SCTP_SNDRCV:
5888 /* SCTP Socket API Extension
5889 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5891 * This cmsghdr structure specifies SCTP options for
5892 * sendmsg() and describes SCTP header information
5893 * about a received message through recvmsg().
5895 * cmsg_level cmsg_type cmsg_data[]
5896 * ------------ ------------ ----------------------
5897 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5899 if (cmsg->cmsg_len !=
5900 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5901 return -EINVAL;
5903 cmsgs->info =
5904 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5906 /* Minimally, validate the sinfo_flags. */
5907 if (cmsgs->info->sinfo_flags &
5908 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5909 SCTP_ABORT | SCTP_EOF))
5910 return -EINVAL;
5911 break;
5913 default:
5914 return -EINVAL;
5917 return 0;
5921 * Wait for a packet..
5922 * Note: This function is the same function as in core/datagram.c
5923 * with a few modifications to make lksctp work.
5925 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5927 int error;
5928 DEFINE_WAIT(wait);
5930 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5932 /* Socket errors? */
5933 error = sock_error(sk);
5934 if (error)
5935 goto out;
5937 if (!skb_queue_empty(&sk->sk_receive_queue))
5938 goto ready;
5940 /* Socket shut down? */
5941 if (sk->sk_shutdown & RCV_SHUTDOWN)
5942 goto out;
5944 /* Sequenced packets can come disconnected. If so we report the
5945 * problem.
5947 error = -ENOTCONN;
5949 /* Is there a good reason to think that we may receive some data? */
5950 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5951 goto out;
5953 /* Handle signals. */
5954 if (signal_pending(current))
5955 goto interrupted;
5957 /* Let another process have a go. Since we are going to sleep
5958 * anyway. Note: This may cause odd behaviors if the message
5959 * does not fit in the user's buffer, but this seems to be the
5960 * only way to honor MSG_DONTWAIT realistically.
5962 sctp_release_sock(sk);
5963 *timeo_p = schedule_timeout(*timeo_p);
5964 sctp_lock_sock(sk);
5966 ready:
5967 finish_wait(sk->sk_sleep, &wait);
5968 return 0;
5970 interrupted:
5971 error = sock_intr_errno(*timeo_p);
5973 out:
5974 finish_wait(sk->sk_sleep, &wait);
5975 *err = error;
5976 return error;
5979 /* Receive a datagram.
5980 * Note: This is pretty much the same routine as in core/datagram.c
5981 * with a few changes to make lksctp work.
5983 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
5984 int noblock, int *err)
5986 int error;
5987 struct sk_buff *skb;
5988 long timeo;
5990 timeo = sock_rcvtimeo(sk, noblock);
5992 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
5993 timeo, MAX_SCHEDULE_TIMEOUT);
5995 do {
5996 /* Again only user level code calls this function,
5997 * so nothing interrupt level
5998 * will suddenly eat the receive_queue.
6000 * Look at current nfs client by the way...
6001 * However, this function was corrent in any case. 8)
6003 if (flags & MSG_PEEK) {
6004 spin_lock_bh(&sk->sk_receive_queue.lock);
6005 skb = skb_peek(&sk->sk_receive_queue);
6006 if (skb)
6007 atomic_inc(&skb->users);
6008 spin_unlock_bh(&sk->sk_receive_queue.lock);
6009 } else {
6010 skb = skb_dequeue(&sk->sk_receive_queue);
6013 if (skb)
6014 return skb;
6016 /* Caller is allowed not to check sk->sk_err before calling. */
6017 error = sock_error(sk);
6018 if (error)
6019 goto no_packet;
6021 if (sk->sk_shutdown & RCV_SHUTDOWN)
6022 break;
6024 /* User doesn't want to wait. */
6025 error = -EAGAIN;
6026 if (!timeo)
6027 goto no_packet;
6028 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6030 return NULL;
6032 no_packet:
6033 *err = error;
6034 return NULL;
6037 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6038 static void __sctp_write_space(struct sctp_association *asoc)
6040 struct sock *sk = asoc->base.sk;
6041 struct socket *sock = sk->sk_socket;
6043 if ((sctp_wspace(asoc) > 0) && sock) {
6044 if (waitqueue_active(&asoc->wait))
6045 wake_up_interruptible(&asoc->wait);
6047 if (sctp_writeable(sk)) {
6048 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6049 wake_up_interruptible(sk->sk_sleep);
6051 /* Note that we try to include the Async I/O support
6052 * here by modeling from the current TCP/UDP code.
6053 * We have not tested with it yet.
6055 if (sock->fasync_list &&
6056 !(sk->sk_shutdown & SEND_SHUTDOWN))
6057 sock_wake_async(sock,
6058 SOCK_WAKE_SPACE, POLL_OUT);
6063 /* Do accounting for the sndbuf space.
6064 * Decrement the used sndbuf space of the corresponding association by the
6065 * data size which was just transmitted(freed).
6067 static void sctp_wfree(struct sk_buff *skb)
6069 struct sctp_association *asoc;
6070 struct sctp_chunk *chunk;
6071 struct sock *sk;
6073 /* Get the saved chunk pointer. */
6074 chunk = *((struct sctp_chunk **)(skb->cb));
6075 asoc = chunk->asoc;
6076 sk = asoc->base.sk;
6077 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6078 sizeof(struct sk_buff) +
6079 sizeof(struct sctp_chunk);
6081 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6084 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6086 sk->sk_wmem_queued -= skb->truesize;
6087 sk_mem_uncharge(sk, skb->truesize);
6089 sock_wfree(skb);
6090 __sctp_write_space(asoc);
6092 sctp_association_put(asoc);
6095 /* Do accounting for the receive space on the socket.
6096 * Accounting for the association is done in ulpevent.c
6097 * We set this as a destructor for the cloned data skbs so that
6098 * accounting is done at the correct time.
6100 void sctp_sock_rfree(struct sk_buff *skb)
6102 struct sock *sk = skb->sk;
6103 struct sctp_ulpevent *event = sctp_skb2event(skb);
6105 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6108 * Mimic the behavior of sock_rfree
6110 sk_mem_uncharge(sk, event->rmem_len);
6114 /* Helper function to wait for space in the sndbuf. */
6115 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6116 size_t msg_len)
6118 struct sock *sk = asoc->base.sk;
6119 int err = 0;
6120 long current_timeo = *timeo_p;
6121 DEFINE_WAIT(wait);
6123 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6124 asoc, (long)(*timeo_p), msg_len);
6126 /* Increment the association's refcnt. */
6127 sctp_association_hold(asoc);
6129 /* Wait on the association specific sndbuf space. */
6130 for (;;) {
6131 prepare_to_wait_exclusive(&asoc->wait, &wait,
6132 TASK_INTERRUPTIBLE);
6133 if (!*timeo_p)
6134 goto do_nonblock;
6135 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6136 asoc->base.dead)
6137 goto do_error;
6138 if (signal_pending(current))
6139 goto do_interrupted;
6140 if (msg_len <= sctp_wspace(asoc))
6141 break;
6143 /* Let another process have a go. Since we are going
6144 * to sleep anyway.
6146 sctp_release_sock(sk);
6147 current_timeo = schedule_timeout(current_timeo);
6148 BUG_ON(sk != asoc->base.sk);
6149 sctp_lock_sock(sk);
6151 *timeo_p = current_timeo;
6154 out:
6155 finish_wait(&asoc->wait, &wait);
6157 /* Release the association's refcnt. */
6158 sctp_association_put(asoc);
6160 return err;
6162 do_error:
6163 err = -EPIPE;
6164 goto out;
6166 do_interrupted:
6167 err = sock_intr_errno(*timeo_p);
6168 goto out;
6170 do_nonblock:
6171 err = -EAGAIN;
6172 goto out;
6175 /* If socket sndbuf has changed, wake up all per association waiters. */
6176 void sctp_write_space(struct sock *sk)
6178 struct sctp_association *asoc;
6180 /* Wake up the tasks in each wait queue. */
6181 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6182 __sctp_write_space(asoc);
6186 /* Is there any sndbuf space available on the socket?
6188 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6189 * associations on the same socket. For a UDP-style socket with
6190 * multiple associations, it is possible for it to be "unwriteable"
6191 * prematurely. I assume that this is acceptable because
6192 * a premature "unwriteable" is better than an accidental "writeable" which
6193 * would cause an unwanted block under certain circumstances. For the 1-1
6194 * UDP-style sockets or TCP-style sockets, this code should work.
6195 * - Daisy
6197 static int sctp_writeable(struct sock *sk)
6199 int amt = 0;
6201 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
6202 if (amt < 0)
6203 amt = 0;
6204 return amt;
6207 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6208 * returns immediately with EINPROGRESS.
6210 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6212 struct sock *sk = asoc->base.sk;
6213 int err = 0;
6214 long current_timeo = *timeo_p;
6215 DEFINE_WAIT(wait);
6217 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6218 (long)(*timeo_p));
6220 /* Increment the association's refcnt. */
6221 sctp_association_hold(asoc);
6223 for (;;) {
6224 prepare_to_wait_exclusive(&asoc->wait, &wait,
6225 TASK_INTERRUPTIBLE);
6226 if (!*timeo_p)
6227 goto do_nonblock;
6228 if (sk->sk_shutdown & RCV_SHUTDOWN)
6229 break;
6230 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6231 asoc->base.dead)
6232 goto do_error;
6233 if (signal_pending(current))
6234 goto do_interrupted;
6236 if (sctp_state(asoc, ESTABLISHED))
6237 break;
6239 /* Let another process have a go. Since we are going
6240 * to sleep anyway.
6242 sctp_release_sock(sk);
6243 current_timeo = schedule_timeout(current_timeo);
6244 sctp_lock_sock(sk);
6246 *timeo_p = current_timeo;
6249 out:
6250 finish_wait(&asoc->wait, &wait);
6252 /* Release the association's refcnt. */
6253 sctp_association_put(asoc);
6255 return err;
6257 do_error:
6258 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6259 err = -ETIMEDOUT;
6260 else
6261 err = -ECONNREFUSED;
6262 goto out;
6264 do_interrupted:
6265 err = sock_intr_errno(*timeo_p);
6266 goto out;
6268 do_nonblock:
6269 err = -EINPROGRESS;
6270 goto out;
6273 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6275 struct sctp_endpoint *ep;
6276 int err = 0;
6277 DEFINE_WAIT(wait);
6279 ep = sctp_sk(sk)->ep;
6282 for (;;) {
6283 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6284 TASK_INTERRUPTIBLE);
6286 if (list_empty(&ep->asocs)) {
6287 sctp_release_sock(sk);
6288 timeo = schedule_timeout(timeo);
6289 sctp_lock_sock(sk);
6292 err = -EINVAL;
6293 if (!sctp_sstate(sk, LISTENING))
6294 break;
6296 err = 0;
6297 if (!list_empty(&ep->asocs))
6298 break;
6300 err = sock_intr_errno(timeo);
6301 if (signal_pending(current))
6302 break;
6304 err = -EAGAIN;
6305 if (!timeo)
6306 break;
6309 finish_wait(sk->sk_sleep, &wait);
6311 return err;
6314 static void sctp_wait_for_close(struct sock *sk, long timeout)
6316 DEFINE_WAIT(wait);
6318 do {
6319 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6320 if (list_empty(&sctp_sk(sk)->ep->asocs))
6321 break;
6322 sctp_release_sock(sk);
6323 timeout = schedule_timeout(timeout);
6324 sctp_lock_sock(sk);
6325 } while (!signal_pending(current) && timeout);
6327 finish_wait(sk->sk_sleep, &wait);
6330 static void sctp_sock_rfree_frag(struct sk_buff *skb)
6332 struct sk_buff *frag;
6334 if (!skb->data_len)
6335 goto done;
6337 /* Don't forget the fragments. */
6338 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6339 sctp_sock_rfree_frag(frag);
6341 done:
6342 sctp_sock_rfree(skb);
6345 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6347 struct sk_buff *frag;
6349 if (!skb->data_len)
6350 goto done;
6352 /* Don't forget the fragments. */
6353 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6354 sctp_skb_set_owner_r_frag(frag, sk);
6356 done:
6357 sctp_skb_set_owner_r(skb, sk);
6360 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6361 * and its messages to the newsk.
6363 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6364 struct sctp_association *assoc,
6365 sctp_socket_type_t type)
6367 struct sctp_sock *oldsp = sctp_sk(oldsk);
6368 struct sctp_sock *newsp = sctp_sk(newsk);
6369 struct sctp_bind_bucket *pp; /* hash list port iterator */
6370 struct sctp_endpoint *newep = newsp->ep;
6371 struct sk_buff *skb, *tmp;
6372 struct sctp_ulpevent *event;
6373 struct sctp_bind_hashbucket *head;
6375 /* Migrate socket buffer sizes and all the socket level options to the
6376 * new socket.
6378 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6379 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6380 /* Brute force copy old sctp opt. */
6381 inet_sk_copy_descendant(newsk, oldsk);
6383 /* Restore the ep value that was overwritten with the above structure
6384 * copy.
6386 newsp->ep = newep;
6387 newsp->hmac = NULL;
6389 /* Hook this new socket in to the bind_hash list. */
6390 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6391 sctp_local_bh_disable();
6392 sctp_spin_lock(&head->lock);
6393 pp = sctp_sk(oldsk)->bind_hash;
6394 sk_add_bind_node(newsk, &pp->owner);
6395 sctp_sk(newsk)->bind_hash = pp;
6396 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6397 sctp_spin_unlock(&head->lock);
6398 sctp_local_bh_enable();
6400 /* Copy the bind_addr list from the original endpoint to the new
6401 * endpoint so that we can handle restarts properly
6403 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6404 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6406 /* Move any messages in the old socket's receive queue that are for the
6407 * peeled off association to the new socket's receive queue.
6409 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6410 event = sctp_skb2event(skb);
6411 if (event->asoc == assoc) {
6412 sctp_sock_rfree_frag(skb);
6413 __skb_unlink(skb, &oldsk->sk_receive_queue);
6414 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6415 sctp_skb_set_owner_r_frag(skb, newsk);
6419 /* Clean up any messages pending delivery due to partial
6420 * delivery. Three cases:
6421 * 1) No partial deliver; no work.
6422 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6423 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6425 skb_queue_head_init(&newsp->pd_lobby);
6426 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6428 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6429 struct sk_buff_head *queue;
6431 /* Decide which queue to move pd_lobby skbs to. */
6432 if (assoc->ulpq.pd_mode) {
6433 queue = &newsp->pd_lobby;
6434 } else
6435 queue = &newsk->sk_receive_queue;
6437 /* Walk through the pd_lobby, looking for skbs that
6438 * need moved to the new socket.
6440 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6441 event = sctp_skb2event(skb);
6442 if (event->asoc == assoc) {
6443 sctp_sock_rfree_frag(skb);
6444 __skb_unlink(skb, &oldsp->pd_lobby);
6445 __skb_queue_tail(queue, skb);
6446 sctp_skb_set_owner_r_frag(skb, newsk);
6450 /* Clear up any skbs waiting for the partial
6451 * delivery to finish.
6453 if (assoc->ulpq.pd_mode)
6454 sctp_clear_pd(oldsk, NULL);
6458 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6459 sctp_sock_rfree_frag(skb);
6460 sctp_skb_set_owner_r_frag(skb, newsk);
6463 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6464 sctp_sock_rfree_frag(skb);
6465 sctp_skb_set_owner_r_frag(skb, newsk);
6468 /* Set the type of socket to indicate that it is peeled off from the
6469 * original UDP-style socket or created with the accept() call on a
6470 * TCP-style socket..
6472 newsp->type = type;
6474 /* Mark the new socket "in-use" by the user so that any packets
6475 * that may arrive on the association after we've moved it are
6476 * queued to the backlog. This prevents a potential race between
6477 * backlog processing on the old socket and new-packet processing
6478 * on the new socket.
6480 * The caller has just allocated newsk so we can guarantee that other
6481 * paths won't try to lock it and then oldsk.
6483 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6484 sctp_assoc_migrate(assoc, newsk);
6486 /* If the association on the newsk is already closed before accept()
6487 * is called, set RCV_SHUTDOWN flag.
6489 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6490 newsk->sk_shutdown |= RCV_SHUTDOWN;
6492 newsk->sk_state = SCTP_SS_ESTABLISHED;
6493 sctp_release_sock(newsk);
6497 /* This proto struct describes the ULP interface for SCTP. */
6498 struct proto sctp_prot = {
6499 .name = "SCTP",
6500 .owner = THIS_MODULE,
6501 .close = sctp_close,
6502 .connect = sctp_connect,
6503 .disconnect = sctp_disconnect,
6504 .accept = sctp_accept,
6505 .ioctl = sctp_ioctl,
6506 .init = sctp_init_sock,
6507 .destroy = sctp_destroy_sock,
6508 .shutdown = sctp_shutdown,
6509 .setsockopt = sctp_setsockopt,
6510 .getsockopt = sctp_getsockopt,
6511 .sendmsg = sctp_sendmsg,
6512 .recvmsg = sctp_recvmsg,
6513 .bind = sctp_bind,
6514 .backlog_rcv = sctp_backlog_rcv,
6515 .hash = sctp_hash,
6516 .unhash = sctp_unhash,
6517 .get_port = sctp_get_port,
6518 .obj_size = sizeof(struct sctp_sock),
6519 .sysctl_mem = sysctl_sctp_mem,
6520 .sysctl_rmem = sysctl_sctp_rmem,
6521 .sysctl_wmem = sysctl_sctp_wmem,
6522 .memory_pressure = &sctp_memory_pressure,
6523 .enter_memory_pressure = sctp_enter_memory_pressure,
6524 .memory_allocated = &sctp_memory_allocated,
6525 .sockets_allocated = &sctp_sockets_allocated,
6528 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6530 struct proto sctpv6_prot = {
6531 .name = "SCTPv6",
6532 .owner = THIS_MODULE,
6533 .close = sctp_close,
6534 .connect = sctp_connect,
6535 .disconnect = sctp_disconnect,
6536 .accept = sctp_accept,
6537 .ioctl = sctp_ioctl,
6538 .init = sctp_init_sock,
6539 .destroy = sctp_destroy_sock,
6540 .shutdown = sctp_shutdown,
6541 .setsockopt = sctp_setsockopt,
6542 .getsockopt = sctp_getsockopt,
6543 .sendmsg = sctp_sendmsg,
6544 .recvmsg = sctp_recvmsg,
6545 .bind = sctp_bind,
6546 .backlog_rcv = sctp_backlog_rcv,
6547 .hash = sctp_hash,
6548 .unhash = sctp_unhash,
6549 .get_port = sctp_get_port,
6550 .obj_size = sizeof(struct sctp6_sock),
6551 .sysctl_mem = sysctl_sctp_mem,
6552 .sysctl_rmem = sysctl_sctp_rmem,
6553 .sysctl_wmem = sysctl_sctp_wmem,
6554 .memory_pressure = &sctp_memory_pressure,
6555 .enter_memory_pressure = sctp_enter_memory_pressure,
6556 .memory_allocated = &sctp_memory_allocated,
6557 .sockets_allocated = &sctp_sockets_allocated,
6559 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */