USB: fix bug in sg initialization in usbtest
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / sctp / socket.c
blobd994d822900de2b9662fe70a846c35c44719ed0d
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 *pos;
517 struct list_head *p;
518 int i;
519 int retval = 0;
521 if (!sctp_addip_enable)
522 return retval;
524 sp = sctp_sk(sk);
525 ep = sp->ep;
527 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
528 __FUNCTION__, sk, addrs, addrcnt);
530 list_for_each(pos, &ep->asocs) {
531 asoc = list_entry(pos, struct sctp_association, asocs);
533 if (!asoc->peer.asconf_capable)
534 continue;
536 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
537 continue;
539 if (!sctp_state(asoc, ESTABLISHED))
540 continue;
542 /* Check if any address in the packed array of addresses is
543 * in the bind address list of the association. If so,
544 * do not send the asconf chunk to its peer, but continue with
545 * other associations.
547 addr_buf = addrs;
548 for (i = 0; i < addrcnt; i++) {
549 addr = (union sctp_addr *)addr_buf;
550 af = sctp_get_af_specific(addr->v4.sin_family);
551 if (!af) {
552 retval = -EINVAL;
553 goto out;
556 if (sctp_assoc_lookup_laddr(asoc, addr))
557 break;
559 addr_buf += af->sockaddr_len;
561 if (i < addrcnt)
562 continue;
564 /* Use the first valid address in bind addr list of
565 * association as Address Parameter of ASCONF CHUNK.
567 bp = &asoc->base.bind_addr;
568 p = bp->address_list.next;
569 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
570 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
571 addrcnt, SCTP_PARAM_ADD_IP);
572 if (!chunk) {
573 retval = -ENOMEM;
574 goto out;
577 retval = sctp_send_asconf(asoc, chunk);
578 if (retval)
579 goto out;
581 /* Add the new addresses to the bind address list with
582 * use_as_src set to 0.
584 addr_buf = addrs;
585 for (i = 0; i < addrcnt; i++) {
586 addr = (union sctp_addr *)addr_buf;
587 af = sctp_get_af_specific(addr->v4.sin_family);
588 memcpy(&saveaddr, addr, af->sockaddr_len);
589 retval = sctp_add_bind_addr(bp, &saveaddr,
590 SCTP_ADDR_NEW, GFP_ATOMIC);
591 addr_buf += af->sockaddr_len;
595 out:
596 return retval;
599 /* Remove a list of addresses from bind addresses list. Do not remove the
600 * last address.
602 * Basically run through each address specified in the addrs/addrcnt
603 * array/length pair, determine if it is IPv6 or IPv4 and call
604 * sctp_del_bind() on it.
606 * If any of them fails, then the operation will be reversed and the
607 * ones that were removed will be added back.
609 * At least one address has to be left; if only one address is
610 * available, the operation will return -EBUSY.
612 * Only sctp_setsockopt_bindx() is supposed to call this function.
614 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
616 struct sctp_sock *sp = sctp_sk(sk);
617 struct sctp_endpoint *ep = sp->ep;
618 int cnt;
619 struct sctp_bind_addr *bp = &ep->base.bind_addr;
620 int retval = 0;
621 void *addr_buf;
622 union sctp_addr *sa_addr;
623 struct sctp_af *af;
625 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
626 sk, addrs, addrcnt);
628 addr_buf = addrs;
629 for (cnt = 0; cnt < addrcnt; cnt++) {
630 /* If the bind address list is empty or if there is only one
631 * bind address, there is nothing more to be removed (we need
632 * at least one address here).
634 if (list_empty(&bp->address_list) ||
635 (sctp_list_single_entry(&bp->address_list))) {
636 retval = -EBUSY;
637 goto err_bindx_rem;
640 sa_addr = (union sctp_addr *)addr_buf;
641 af = sctp_get_af_specific(sa_addr->sa.sa_family);
642 if (!af) {
643 retval = -EINVAL;
644 goto err_bindx_rem;
647 if (!af->addr_valid(sa_addr, sp, NULL)) {
648 retval = -EADDRNOTAVAIL;
649 goto err_bindx_rem;
652 if (sa_addr->v4.sin_port != htons(bp->port)) {
653 retval = -EINVAL;
654 goto err_bindx_rem;
657 /* FIXME - There is probably a need to check if sk->sk_saddr and
658 * sk->sk_rcv_addr are currently set to one of the addresses to
659 * be removed. This is something which needs to be looked into
660 * when we are fixing the outstanding issues with multi-homing
661 * socket routing and failover schemes. Refer to comments in
662 * sctp_do_bind(). -daisy
664 retval = sctp_del_bind_addr(bp, sa_addr);
666 addr_buf += af->sockaddr_len;
667 err_bindx_rem:
668 if (retval < 0) {
669 /* Failed. Add the ones that has been removed back */
670 if (cnt > 0)
671 sctp_bindx_add(sk, addrs, cnt);
672 return retval;
676 return retval;
679 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
680 * the associations that are part of the endpoint indicating that a list of
681 * local addresses are removed from the endpoint.
683 * If any of the addresses is already in the bind address list of the
684 * association, we do not send the chunk for that association. But it will not
685 * affect other associations.
687 * Only sctp_setsockopt_bindx() is supposed to call this function.
689 static int sctp_send_asconf_del_ip(struct sock *sk,
690 struct sockaddr *addrs,
691 int addrcnt)
693 struct sctp_sock *sp;
694 struct sctp_endpoint *ep;
695 struct sctp_association *asoc;
696 struct sctp_transport *transport;
697 struct sctp_bind_addr *bp;
698 struct sctp_chunk *chunk;
699 union sctp_addr *laddr;
700 void *addr_buf;
701 struct sctp_af *af;
702 struct list_head *pos, *pos1;
703 struct sctp_sockaddr_entry *saddr;
704 int i;
705 int retval = 0;
707 if (!sctp_addip_enable)
708 return retval;
710 sp = sctp_sk(sk);
711 ep = sp->ep;
713 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
714 __FUNCTION__, sk, addrs, addrcnt);
716 list_for_each(pos, &ep->asocs) {
717 asoc = list_entry(pos, struct sctp_association, asocs);
719 if (!asoc->peer.asconf_capable)
720 continue;
722 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
723 continue;
725 if (!sctp_state(asoc, ESTABLISHED))
726 continue;
728 /* Check if any address in the packed array of addresses is
729 * not present in the bind address list of the association.
730 * If so, do not send the asconf chunk to its peer, but
731 * continue with other associations.
733 addr_buf = addrs;
734 for (i = 0; i < addrcnt; i++) {
735 laddr = (union sctp_addr *)addr_buf;
736 af = sctp_get_af_specific(laddr->v4.sin_family);
737 if (!af) {
738 retval = -EINVAL;
739 goto out;
742 if (!sctp_assoc_lookup_laddr(asoc, laddr))
743 break;
745 addr_buf += af->sockaddr_len;
747 if (i < addrcnt)
748 continue;
750 /* Find one address in the association's bind address list
751 * that is not in the packed array of addresses. This is to
752 * make sure that we do not delete all the addresses in the
753 * association.
755 bp = &asoc->base.bind_addr;
756 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
757 addrcnt, sp);
758 if (!laddr)
759 continue;
761 /* We do not need RCU protection throughout this loop
762 * because this is done under a socket lock from the
763 * setsockopt call.
765 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
766 SCTP_PARAM_DEL_IP);
767 if (!chunk) {
768 retval = -ENOMEM;
769 goto out;
772 /* Reset use_as_src flag for the addresses in the bind address
773 * list that are to be deleted.
775 addr_buf = addrs;
776 for (i = 0; i < addrcnt; i++) {
777 laddr = (union sctp_addr *)addr_buf;
778 af = sctp_get_af_specific(laddr->v4.sin_family);
779 list_for_each_entry(saddr, &bp->address_list, list) {
780 if (sctp_cmp_addr_exact(&saddr->a, laddr))
781 saddr->state = SCTP_ADDR_DEL;
783 addr_buf += af->sockaddr_len;
786 /* Update the route and saddr entries for all the transports
787 * as some of the addresses in the bind address list are
788 * about to be deleted and cannot be used as source addresses.
790 list_for_each(pos1, &asoc->peer.transport_addr_list) {
791 transport = list_entry(pos1, struct sctp_transport,
792 transports);
793 dst_release(transport->dst);
794 sctp_transport_route(transport, NULL,
795 sctp_sk(asoc->base.sk));
798 retval = sctp_send_asconf(asoc, chunk);
800 out:
801 return retval;
804 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
806 * API 8.1
807 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
808 * int flags);
810 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
811 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
812 * or IPv6 addresses.
814 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
815 * Section 3.1.2 for this usage.
817 * addrs is a pointer to an array of one or more socket addresses. Each
818 * address is contained in its appropriate structure (i.e. struct
819 * sockaddr_in or struct sockaddr_in6) the family of the address type
820 * must be used to distinguish the address length (note that this
821 * representation is termed a "packed array" of addresses). The caller
822 * specifies the number of addresses in the array with addrcnt.
824 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
825 * -1, and sets errno to the appropriate error code.
827 * For SCTP, the port given in each socket address must be the same, or
828 * sctp_bindx() will fail, setting errno to EINVAL.
830 * The flags parameter is formed from the bitwise OR of zero or more of
831 * the following currently defined flags:
833 * SCTP_BINDX_ADD_ADDR
835 * SCTP_BINDX_REM_ADDR
837 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
838 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
839 * addresses from the association. The two flags are mutually exclusive;
840 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
841 * not remove all addresses from an association; sctp_bindx() will
842 * reject such an attempt with EINVAL.
844 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
845 * additional addresses with an endpoint after calling bind(). Or use
846 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
847 * socket is associated with so that no new association accepted will be
848 * associated with those addresses. If the endpoint supports dynamic
849 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
850 * endpoint to send the appropriate message to the peer to change the
851 * peers address lists.
853 * Adding and removing addresses from a connected association is
854 * optional functionality. Implementations that do not support this
855 * functionality should return EOPNOTSUPP.
857 * Basically do nothing but copying the addresses from user to kernel
858 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
859 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
860 * from userspace.
862 * We don't use copy_from_user() for optimization: we first do the
863 * sanity checks (buffer size -fast- and access check-healthy
864 * pointer); if all of those succeed, then we can alloc the memory
865 * (expensive operation) needed to copy the data to kernel. Then we do
866 * the copying without checking the user space area
867 * (__copy_from_user()).
869 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
870 * it.
872 * sk The sk of the socket
873 * addrs The pointer to the addresses in user land
874 * addrssize Size of the addrs buffer
875 * op Operation to perform (add or remove, see the flags of
876 * sctp_bindx)
878 * Returns 0 if ok, <0 errno code on error.
880 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
881 struct sockaddr __user *addrs,
882 int addrs_size, int op)
884 struct sockaddr *kaddrs;
885 int err;
886 int addrcnt = 0;
887 int walk_size = 0;
888 struct sockaddr *sa_addr;
889 void *addr_buf;
890 struct sctp_af *af;
892 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
893 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
895 if (unlikely(addrs_size <= 0))
896 return -EINVAL;
898 /* Check the user passed a healthy pointer. */
899 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
900 return -EFAULT;
902 /* Alloc space for the address array in kernel memory. */
903 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
904 if (unlikely(!kaddrs))
905 return -ENOMEM;
907 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
908 kfree(kaddrs);
909 return -EFAULT;
912 /* Walk through the addrs buffer and count the number of addresses. */
913 addr_buf = kaddrs;
914 while (walk_size < addrs_size) {
915 sa_addr = (struct sockaddr *)addr_buf;
916 af = sctp_get_af_specific(sa_addr->sa_family);
918 /* If the address family is not supported or if this address
919 * causes the address buffer to overflow return EINVAL.
921 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
922 kfree(kaddrs);
923 return -EINVAL;
925 addrcnt++;
926 addr_buf += af->sockaddr_len;
927 walk_size += af->sockaddr_len;
930 /* Do the work. */
931 switch (op) {
932 case SCTP_BINDX_ADD_ADDR:
933 err = sctp_bindx_add(sk, kaddrs, addrcnt);
934 if (err)
935 goto out;
936 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
937 break;
939 case SCTP_BINDX_REM_ADDR:
940 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
941 if (err)
942 goto out;
943 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
944 break;
946 default:
947 err = -EINVAL;
948 break;
951 out:
952 kfree(kaddrs);
954 return err;
957 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
959 * Common routine for handling connect() and sctp_connectx().
960 * Connect will come in with just a single address.
962 static int __sctp_connect(struct sock* sk,
963 struct sockaddr *kaddrs,
964 int addrs_size)
966 struct sctp_sock *sp;
967 struct sctp_endpoint *ep;
968 struct sctp_association *asoc = NULL;
969 struct sctp_association *asoc2;
970 struct sctp_transport *transport;
971 union sctp_addr to;
972 struct sctp_af *af;
973 sctp_scope_t scope;
974 long timeo;
975 int err = 0;
976 int addrcnt = 0;
977 int walk_size = 0;
978 union sctp_addr *sa_addr = NULL;
979 void *addr_buf;
980 unsigned short port;
981 unsigned int f_flags = 0;
983 sp = sctp_sk(sk);
984 ep = sp->ep;
986 /* connect() cannot be done on a socket that is already in ESTABLISHED
987 * state - UDP-style peeled off socket or a TCP-style socket that
988 * is already connected.
989 * It cannot be done even on a TCP-style listening socket.
991 if (sctp_sstate(sk, ESTABLISHED) ||
992 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
993 err = -EISCONN;
994 goto out_free;
997 /* Walk through the addrs buffer and count the number of addresses. */
998 addr_buf = kaddrs;
999 while (walk_size < addrs_size) {
1000 sa_addr = (union sctp_addr *)addr_buf;
1001 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1002 port = ntohs(sa_addr->v4.sin_port);
1004 /* If the address family is not supported or if this address
1005 * causes the address buffer to overflow return EINVAL.
1007 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1008 err = -EINVAL;
1009 goto out_free;
1012 /* Save current address so we can work with it */
1013 memcpy(&to, sa_addr, af->sockaddr_len);
1015 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1016 if (err)
1017 goto out_free;
1019 /* Make sure the destination port is correctly set
1020 * in all addresses.
1022 if (asoc && asoc->peer.port && asoc->peer.port != port)
1023 goto out_free;
1026 /* Check if there already is a matching association on the
1027 * endpoint (other than the one created here).
1029 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1030 if (asoc2 && asoc2 != asoc) {
1031 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1032 err = -EISCONN;
1033 else
1034 err = -EALREADY;
1035 goto out_free;
1038 /* If we could not find a matching association on the endpoint,
1039 * make sure that there is no peeled-off association matching
1040 * the peer address even on another socket.
1042 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1043 err = -EADDRNOTAVAIL;
1044 goto out_free;
1047 if (!asoc) {
1048 /* If a bind() or sctp_bindx() is not called prior to
1049 * an sctp_connectx() call, the system picks an
1050 * ephemeral port and will choose an address set
1051 * equivalent to binding with a wildcard address.
1053 if (!ep->base.bind_addr.port) {
1054 if (sctp_autobind(sk)) {
1055 err = -EAGAIN;
1056 goto out_free;
1058 } else {
1060 * If an unprivileged user inherits a 1-many
1061 * style socket with open associations on a
1062 * privileged port, it MAY be permitted to
1063 * accept new associations, but it SHOULD NOT
1064 * be permitted to open new associations.
1066 if (ep->base.bind_addr.port < PROT_SOCK &&
1067 !capable(CAP_NET_BIND_SERVICE)) {
1068 err = -EACCES;
1069 goto out_free;
1073 scope = sctp_scope(&to);
1074 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1075 if (!asoc) {
1076 err = -ENOMEM;
1077 goto out_free;
1081 /* Prime the peer's transport structures. */
1082 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1083 SCTP_UNKNOWN);
1084 if (!transport) {
1085 err = -ENOMEM;
1086 goto out_free;
1089 addrcnt++;
1090 addr_buf += af->sockaddr_len;
1091 walk_size += af->sockaddr_len;
1094 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1095 if (err < 0) {
1096 goto out_free;
1099 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1100 if (err < 0) {
1101 goto out_free;
1104 /* Initialize sk's dport and daddr for getpeername() */
1105 inet_sk(sk)->dport = htons(asoc->peer.port);
1106 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1107 af->to_sk_daddr(sa_addr, sk);
1108 sk->sk_err = 0;
1110 /* in-kernel sockets don't generally have a file allocated to them
1111 * if all they do is call sock_create_kern().
1113 if (sk->sk_socket->file)
1114 f_flags = sk->sk_socket->file->f_flags;
1116 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1118 err = sctp_wait_for_connect(asoc, &timeo);
1120 /* Don't free association on exit. */
1121 asoc = NULL;
1123 out_free:
1125 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1126 " kaddrs: %p err: %d\n",
1127 asoc, kaddrs, err);
1128 if (asoc)
1129 sctp_association_free(asoc);
1130 return err;
1133 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1135 * API 8.9
1136 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt);
1138 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1139 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1140 * or IPv6 addresses.
1142 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1143 * Section 3.1.2 for this usage.
1145 * addrs is a pointer to an array of one or more socket addresses. Each
1146 * address is contained in its appropriate structure (i.e. struct
1147 * sockaddr_in or struct sockaddr_in6) the family of the address type
1148 * must be used to distengish the address length (note that this
1149 * representation is termed a "packed array" of addresses). The caller
1150 * specifies the number of addresses in the array with addrcnt.
1152 * On success, sctp_connectx() returns 0. On failure, sctp_connectx() returns
1153 * -1, and sets errno to the appropriate error code.
1155 * For SCTP, the port given in each socket address must be the same, or
1156 * sctp_connectx() will fail, setting errno to EINVAL.
1158 * An application can use sctp_connectx to initiate an association with
1159 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1160 * allows a caller to specify multiple addresses at which a peer can be
1161 * reached. The way the SCTP stack uses the list of addresses to set up
1162 * the association is implementation dependant. This function only
1163 * specifies that the stack will try to make use of all the addresses in
1164 * the list when needed.
1166 * Note that the list of addresses passed in is only used for setting up
1167 * the association. It does not necessarily equal the set of addresses
1168 * the peer uses for the resulting association. If the caller wants to
1169 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1170 * retrieve them after the association has been set up.
1172 * Basically do nothing but copying the addresses from user to kernel
1173 * land and invoking either sctp_connectx(). This is used for tunneling
1174 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1176 * We don't use copy_from_user() for optimization: we first do the
1177 * sanity checks (buffer size -fast- and access check-healthy
1178 * pointer); if all of those succeed, then we can alloc the memory
1179 * (expensive operation) needed to copy the data to kernel. Then we do
1180 * the copying without checking the user space area
1181 * (__copy_from_user()).
1183 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1184 * it.
1186 * sk The sk of the socket
1187 * addrs The pointer to the addresses in user land
1188 * addrssize Size of the addrs buffer
1190 * Returns 0 if ok, <0 errno code on error.
1192 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1193 struct sockaddr __user *addrs,
1194 int addrs_size)
1196 int err = 0;
1197 struct sockaddr *kaddrs;
1199 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1200 __FUNCTION__, sk, addrs, addrs_size);
1202 if (unlikely(addrs_size <= 0))
1203 return -EINVAL;
1205 /* Check the user passed a healthy pointer. */
1206 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1207 return -EFAULT;
1209 /* Alloc space for the address array in kernel memory. */
1210 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1211 if (unlikely(!kaddrs))
1212 return -ENOMEM;
1214 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1215 err = -EFAULT;
1216 } else {
1217 err = __sctp_connect(sk, kaddrs, addrs_size);
1220 kfree(kaddrs);
1221 return err;
1224 /* API 3.1.4 close() - UDP Style Syntax
1225 * Applications use close() to perform graceful shutdown (as described in
1226 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1227 * by a UDP-style socket.
1229 * The syntax is
1231 * ret = close(int sd);
1233 * sd - the socket descriptor of the associations to be closed.
1235 * To gracefully shutdown a specific association represented by the
1236 * UDP-style socket, an application should use the sendmsg() call,
1237 * passing no user data, but including the appropriate flag in the
1238 * ancillary data (see Section xxxx).
1240 * If sd in the close() call is a branched-off socket representing only
1241 * one association, the shutdown is performed on that association only.
1243 * 4.1.6 close() - TCP Style Syntax
1245 * Applications use close() to gracefully close down an association.
1247 * The syntax is:
1249 * int close(int sd);
1251 * sd - the socket descriptor of the association to be closed.
1253 * After an application calls close() on a socket descriptor, no further
1254 * socket operations will succeed on that descriptor.
1256 * API 7.1.4 SO_LINGER
1258 * An application using the TCP-style socket can use this option to
1259 * perform the SCTP ABORT primitive. The linger option structure is:
1261 * struct linger {
1262 * int l_onoff; // option on/off
1263 * int l_linger; // linger time
1264 * };
1266 * To enable the option, set l_onoff to 1. If the l_linger value is set
1267 * to 0, calling close() is the same as the ABORT primitive. If the
1268 * value is set to a negative value, the setsockopt() call will return
1269 * an error. If the value is set to a positive value linger_time, the
1270 * close() can be blocked for at most linger_time ms. If the graceful
1271 * shutdown phase does not finish during this period, close() will
1272 * return but the graceful shutdown phase continues in the system.
1274 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1276 struct sctp_endpoint *ep;
1277 struct sctp_association *asoc;
1278 struct list_head *pos, *temp;
1280 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1282 sctp_lock_sock(sk);
1283 sk->sk_shutdown = SHUTDOWN_MASK;
1285 ep = sctp_sk(sk)->ep;
1287 /* Walk all associations on an endpoint. */
1288 list_for_each_safe(pos, temp, &ep->asocs) {
1289 asoc = list_entry(pos, struct sctp_association, asocs);
1291 if (sctp_style(sk, TCP)) {
1292 /* A closed association can still be in the list if
1293 * it belongs to a TCP-style listening socket that is
1294 * not yet accepted. If so, free it. If not, send an
1295 * ABORT or SHUTDOWN based on the linger options.
1297 if (sctp_state(asoc, CLOSED)) {
1298 sctp_unhash_established(asoc);
1299 sctp_association_free(asoc);
1300 continue;
1304 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1305 struct sctp_chunk *chunk;
1307 chunk = sctp_make_abort_user(asoc, NULL, 0);
1308 if (chunk)
1309 sctp_primitive_ABORT(asoc, chunk);
1310 } else
1311 sctp_primitive_SHUTDOWN(asoc, NULL);
1314 /* Clean up any skbs sitting on the receive queue. */
1315 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1316 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1318 /* On a TCP-style socket, block for at most linger_time if set. */
1319 if (sctp_style(sk, TCP) && timeout)
1320 sctp_wait_for_close(sk, timeout);
1322 /* This will run the backlog queue. */
1323 sctp_release_sock(sk);
1325 /* Supposedly, no process has access to the socket, but
1326 * the net layers still may.
1328 sctp_local_bh_disable();
1329 sctp_bh_lock_sock(sk);
1331 /* Hold the sock, since sk_common_release() will put sock_put()
1332 * and we have just a little more cleanup.
1334 sock_hold(sk);
1335 sk_common_release(sk);
1337 sctp_bh_unlock_sock(sk);
1338 sctp_local_bh_enable();
1340 sock_put(sk);
1342 SCTP_DBG_OBJCNT_DEC(sock);
1345 /* Handle EPIPE error. */
1346 static int sctp_error(struct sock *sk, int flags, int err)
1348 if (err == -EPIPE)
1349 err = sock_error(sk) ? : -EPIPE;
1350 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1351 send_sig(SIGPIPE, current, 0);
1352 return err;
1355 /* API 3.1.3 sendmsg() - UDP Style Syntax
1357 * An application uses sendmsg() and recvmsg() calls to transmit data to
1358 * and receive data from its peer.
1360 * ssize_t sendmsg(int socket, const struct msghdr *message,
1361 * int flags);
1363 * socket - the socket descriptor of the endpoint.
1364 * message - pointer to the msghdr structure which contains a single
1365 * user message and possibly some ancillary data.
1367 * See Section 5 for complete description of the data
1368 * structures.
1370 * flags - flags sent or received with the user message, see Section
1371 * 5 for complete description of the flags.
1373 * Note: This function could use a rewrite especially when explicit
1374 * connect support comes in.
1376 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1378 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1380 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1381 struct msghdr *msg, size_t msg_len)
1383 struct sctp_sock *sp;
1384 struct sctp_endpoint *ep;
1385 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1386 struct sctp_transport *transport, *chunk_tp;
1387 struct sctp_chunk *chunk;
1388 union sctp_addr to;
1389 struct sockaddr *msg_name = NULL;
1390 struct sctp_sndrcvinfo default_sinfo = { 0 };
1391 struct sctp_sndrcvinfo *sinfo;
1392 struct sctp_initmsg *sinit;
1393 sctp_assoc_t associd = 0;
1394 sctp_cmsgs_t cmsgs = { NULL };
1395 int err;
1396 sctp_scope_t scope;
1397 long timeo;
1398 __u16 sinfo_flags = 0;
1399 struct sctp_datamsg *datamsg;
1400 struct list_head *pos;
1401 int msg_flags = msg->msg_flags;
1403 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1404 sk, msg, msg_len);
1406 err = 0;
1407 sp = sctp_sk(sk);
1408 ep = sp->ep;
1410 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1412 /* We cannot send a message over a TCP-style listening socket. */
1413 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1414 err = -EPIPE;
1415 goto out_nounlock;
1418 /* Parse out the SCTP CMSGs. */
1419 err = sctp_msghdr_parse(msg, &cmsgs);
1421 if (err) {
1422 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1423 goto out_nounlock;
1426 /* Fetch the destination address for this packet. This
1427 * address only selects the association--it is not necessarily
1428 * the address we will send to.
1429 * For a peeled-off socket, msg_name is ignored.
1431 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1432 int msg_namelen = msg->msg_namelen;
1434 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1435 msg_namelen);
1436 if (err)
1437 return err;
1439 if (msg_namelen > sizeof(to))
1440 msg_namelen = sizeof(to);
1441 memcpy(&to, msg->msg_name, msg_namelen);
1442 msg_name = msg->msg_name;
1445 sinfo = cmsgs.info;
1446 sinit = cmsgs.init;
1448 /* Did the user specify SNDRCVINFO? */
1449 if (sinfo) {
1450 sinfo_flags = sinfo->sinfo_flags;
1451 associd = sinfo->sinfo_assoc_id;
1454 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1455 msg_len, sinfo_flags);
1457 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1458 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1459 err = -EINVAL;
1460 goto out_nounlock;
1463 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1464 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1465 * If SCTP_ABORT is set, the message length could be non zero with
1466 * the msg_iov set to the user abort reason.
1468 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1469 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1470 err = -EINVAL;
1471 goto out_nounlock;
1474 /* If SCTP_ADDR_OVER is set, there must be an address
1475 * specified in msg_name.
1477 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1478 err = -EINVAL;
1479 goto out_nounlock;
1482 transport = NULL;
1484 SCTP_DEBUG_PRINTK("About to look up association.\n");
1486 sctp_lock_sock(sk);
1488 /* If a msg_name has been specified, assume this is to be used. */
1489 if (msg_name) {
1490 /* Look for a matching association on the endpoint. */
1491 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1492 if (!asoc) {
1493 /* If we could not find a matching association on the
1494 * endpoint, make sure that it is not a TCP-style
1495 * socket that already has an association or there is
1496 * no peeled-off association on another socket.
1498 if ((sctp_style(sk, TCP) &&
1499 sctp_sstate(sk, ESTABLISHED)) ||
1500 sctp_endpoint_is_peeled_off(ep, &to)) {
1501 err = -EADDRNOTAVAIL;
1502 goto out_unlock;
1505 } else {
1506 asoc = sctp_id2assoc(sk, associd);
1507 if (!asoc) {
1508 err = -EPIPE;
1509 goto out_unlock;
1513 if (asoc) {
1514 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1516 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1517 * socket that has an association in CLOSED state. This can
1518 * happen when an accepted socket has an association that is
1519 * already CLOSED.
1521 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1522 err = -EPIPE;
1523 goto out_unlock;
1526 if (sinfo_flags & SCTP_EOF) {
1527 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1528 asoc);
1529 sctp_primitive_SHUTDOWN(asoc, NULL);
1530 err = 0;
1531 goto out_unlock;
1533 if (sinfo_flags & SCTP_ABORT) {
1535 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1536 if (!chunk) {
1537 err = -ENOMEM;
1538 goto out_unlock;
1541 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1542 sctp_primitive_ABORT(asoc, chunk);
1543 err = 0;
1544 goto out_unlock;
1548 /* Do we need to create the association? */
1549 if (!asoc) {
1550 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1552 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1553 err = -EINVAL;
1554 goto out_unlock;
1557 /* Check for invalid stream against the stream counts,
1558 * either the default or the user specified stream counts.
1560 if (sinfo) {
1561 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1562 /* Check against the defaults. */
1563 if (sinfo->sinfo_stream >=
1564 sp->initmsg.sinit_num_ostreams) {
1565 err = -EINVAL;
1566 goto out_unlock;
1568 } else {
1569 /* Check against the requested. */
1570 if (sinfo->sinfo_stream >=
1571 sinit->sinit_num_ostreams) {
1572 err = -EINVAL;
1573 goto out_unlock;
1579 * API 3.1.2 bind() - UDP Style Syntax
1580 * If a bind() or sctp_bindx() is not called prior to a
1581 * sendmsg() call that initiates a new association, the
1582 * system picks an ephemeral port and will choose an address
1583 * set equivalent to binding with a wildcard address.
1585 if (!ep->base.bind_addr.port) {
1586 if (sctp_autobind(sk)) {
1587 err = -EAGAIN;
1588 goto out_unlock;
1590 } else {
1592 * If an unprivileged user inherits a one-to-many
1593 * style socket with open associations on a privileged
1594 * port, it MAY be permitted to accept new associations,
1595 * but it SHOULD NOT be permitted to open new
1596 * associations.
1598 if (ep->base.bind_addr.port < PROT_SOCK &&
1599 !capable(CAP_NET_BIND_SERVICE)) {
1600 err = -EACCES;
1601 goto out_unlock;
1605 scope = sctp_scope(&to);
1606 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1607 if (!new_asoc) {
1608 err = -ENOMEM;
1609 goto out_unlock;
1611 asoc = new_asoc;
1613 /* If the SCTP_INIT ancillary data is specified, set all
1614 * the association init values accordingly.
1616 if (sinit) {
1617 if (sinit->sinit_num_ostreams) {
1618 asoc->c.sinit_num_ostreams =
1619 sinit->sinit_num_ostreams;
1621 if (sinit->sinit_max_instreams) {
1622 asoc->c.sinit_max_instreams =
1623 sinit->sinit_max_instreams;
1625 if (sinit->sinit_max_attempts) {
1626 asoc->max_init_attempts
1627 = sinit->sinit_max_attempts;
1629 if (sinit->sinit_max_init_timeo) {
1630 asoc->max_init_timeo =
1631 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1635 /* Prime the peer's transport structures. */
1636 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1637 if (!transport) {
1638 err = -ENOMEM;
1639 goto out_free;
1641 err = sctp_assoc_set_bind_addr_from_ep(asoc, GFP_KERNEL);
1642 if (err < 0) {
1643 err = -ENOMEM;
1644 goto out_free;
1648 /* ASSERT: we have a valid association at this point. */
1649 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1651 if (!sinfo) {
1652 /* If the user didn't specify SNDRCVINFO, make up one with
1653 * some defaults.
1655 default_sinfo.sinfo_stream = asoc->default_stream;
1656 default_sinfo.sinfo_flags = asoc->default_flags;
1657 default_sinfo.sinfo_ppid = asoc->default_ppid;
1658 default_sinfo.sinfo_context = asoc->default_context;
1659 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1660 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1661 sinfo = &default_sinfo;
1664 /* API 7.1.7, the sndbuf size per association bounds the
1665 * maximum size of data that can be sent in a single send call.
1667 if (msg_len > sk->sk_sndbuf) {
1668 err = -EMSGSIZE;
1669 goto out_free;
1672 if (asoc->pmtu_pending)
1673 sctp_assoc_pending_pmtu(asoc);
1675 /* If fragmentation is disabled and the message length exceeds the
1676 * association fragmentation point, return EMSGSIZE. The I-D
1677 * does not specify what this error is, but this looks like
1678 * a great fit.
1680 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1681 err = -EMSGSIZE;
1682 goto out_free;
1685 if (sinfo) {
1686 /* Check for invalid stream. */
1687 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1688 err = -EINVAL;
1689 goto out_free;
1693 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1694 if (!sctp_wspace(asoc)) {
1695 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1696 if (err)
1697 goto out_free;
1700 /* If an address is passed with the sendto/sendmsg call, it is used
1701 * to override the primary destination address in the TCP model, or
1702 * when SCTP_ADDR_OVER flag is set in the UDP model.
1704 if ((sctp_style(sk, TCP) && msg_name) ||
1705 (sinfo_flags & SCTP_ADDR_OVER)) {
1706 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1707 if (!chunk_tp) {
1708 err = -EINVAL;
1709 goto out_free;
1711 } else
1712 chunk_tp = NULL;
1714 /* Auto-connect, if we aren't connected already. */
1715 if (sctp_state(asoc, CLOSED)) {
1716 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1717 if (err < 0)
1718 goto out_free;
1719 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1722 /* Break the message into multiple chunks of maximum size. */
1723 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1724 if (!datamsg) {
1725 err = -ENOMEM;
1726 goto out_free;
1729 /* Now send the (possibly) fragmented message. */
1730 list_for_each(pos, &datamsg->chunks) {
1731 chunk = list_entry(pos, struct sctp_chunk, frag_list);
1732 sctp_datamsg_track(chunk);
1734 /* Do accounting for the write space. */
1735 sctp_set_owner_w(chunk);
1737 chunk->transport = chunk_tp;
1739 /* Send it to the lower layers. Note: all chunks
1740 * must either fail or succeed. The lower layer
1741 * works that way today. Keep it that way or this
1742 * breaks.
1744 err = sctp_primitive_SEND(asoc, chunk);
1745 /* Did the lower layer accept the chunk? */
1746 if (err)
1747 sctp_chunk_free(chunk);
1748 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1751 sctp_datamsg_free(datamsg);
1752 if (err)
1753 goto out_free;
1754 else
1755 err = msg_len;
1757 /* If we are already past ASSOCIATE, the lower
1758 * layers are responsible for association cleanup.
1760 goto out_unlock;
1762 out_free:
1763 if (new_asoc)
1764 sctp_association_free(asoc);
1765 out_unlock:
1766 sctp_release_sock(sk);
1768 out_nounlock:
1769 return sctp_error(sk, msg_flags, err);
1771 #if 0
1772 do_sock_err:
1773 if (msg_len)
1774 err = msg_len;
1775 else
1776 err = sock_error(sk);
1777 goto out;
1779 do_interrupted:
1780 if (msg_len)
1781 err = msg_len;
1782 goto out;
1783 #endif /* 0 */
1786 /* This is an extended version of skb_pull() that removes the data from the
1787 * start of a skb even when data is spread across the list of skb's in the
1788 * frag_list. len specifies the total amount of data that needs to be removed.
1789 * when 'len' bytes could be removed from the skb, it returns 0.
1790 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1791 * could not be removed.
1793 static int sctp_skb_pull(struct sk_buff *skb, int len)
1795 struct sk_buff *list;
1796 int skb_len = skb_headlen(skb);
1797 int rlen;
1799 if (len <= skb_len) {
1800 __skb_pull(skb, len);
1801 return 0;
1803 len -= skb_len;
1804 __skb_pull(skb, skb_len);
1806 for (list = skb_shinfo(skb)->frag_list; list; list = list->next) {
1807 rlen = sctp_skb_pull(list, len);
1808 skb->len -= (len-rlen);
1809 skb->data_len -= (len-rlen);
1811 if (!rlen)
1812 return 0;
1814 len = rlen;
1817 return len;
1820 /* API 3.1.3 recvmsg() - UDP Style Syntax
1822 * ssize_t recvmsg(int socket, struct msghdr *message,
1823 * int flags);
1825 * socket - the socket descriptor of the endpoint.
1826 * message - pointer to the msghdr structure which contains a single
1827 * user message and possibly some ancillary data.
1829 * See Section 5 for complete description of the data
1830 * structures.
1832 * flags - flags sent or received with the user message, see Section
1833 * 5 for complete description of the flags.
1835 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1837 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1838 struct msghdr *msg, size_t len, int noblock,
1839 int flags, int *addr_len)
1841 struct sctp_ulpevent *event = NULL;
1842 struct sctp_sock *sp = sctp_sk(sk);
1843 struct sk_buff *skb;
1844 int copied;
1845 int err = 0;
1846 int skb_len;
1848 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1849 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1850 "len", len, "knoblauch", noblock,
1851 "flags", flags, "addr_len", addr_len);
1853 sctp_lock_sock(sk);
1855 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1856 err = -ENOTCONN;
1857 goto out;
1860 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1861 if (!skb)
1862 goto out;
1864 /* Get the total length of the skb including any skb's in the
1865 * frag_list.
1867 skb_len = skb->len;
1869 copied = skb_len;
1870 if (copied > len)
1871 copied = len;
1873 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1875 event = sctp_skb2event(skb);
1877 if (err)
1878 goto out_free;
1880 sock_recv_timestamp(msg, sk, skb);
1881 if (sctp_ulpevent_is_notification(event)) {
1882 msg->msg_flags |= MSG_NOTIFICATION;
1883 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1884 } else {
1885 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1888 /* Check if we allow SCTP_SNDRCVINFO. */
1889 if (sp->subscribe.sctp_data_io_event)
1890 sctp_ulpevent_read_sndrcvinfo(event, msg);
1891 #if 0
1892 /* FIXME: we should be calling IP/IPv6 layers. */
1893 if (sk->sk_protinfo.af_inet.cmsg_flags)
1894 ip_cmsg_recv(msg, skb);
1895 #endif
1897 err = copied;
1899 /* If skb's length exceeds the user's buffer, update the skb and
1900 * push it back to the receive_queue so that the next call to
1901 * recvmsg() will return the remaining data. Don't set MSG_EOR.
1903 if (skb_len > copied) {
1904 msg->msg_flags &= ~MSG_EOR;
1905 if (flags & MSG_PEEK)
1906 goto out_free;
1907 sctp_skb_pull(skb, copied);
1908 skb_queue_head(&sk->sk_receive_queue, skb);
1910 /* When only partial message is copied to the user, increase
1911 * rwnd by that amount. If all the data in the skb is read,
1912 * rwnd is updated when the event is freed.
1914 if (!sctp_ulpevent_is_notification(event))
1915 sctp_assoc_rwnd_increase(event->asoc, copied);
1916 goto out;
1917 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
1918 (event->msg_flags & MSG_EOR))
1919 msg->msg_flags |= MSG_EOR;
1920 else
1921 msg->msg_flags &= ~MSG_EOR;
1923 out_free:
1924 if (flags & MSG_PEEK) {
1925 /* Release the skb reference acquired after peeking the skb in
1926 * sctp_skb_recv_datagram().
1928 kfree_skb(skb);
1929 } else {
1930 /* Free the event which includes releasing the reference to
1931 * the owner of the skb, freeing the skb and updating the
1932 * rwnd.
1934 sctp_ulpevent_free(event);
1936 out:
1937 sctp_release_sock(sk);
1938 return err;
1941 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
1943 * This option is a on/off flag. If enabled no SCTP message
1944 * fragmentation will be performed. Instead if a message being sent
1945 * exceeds the current PMTU size, the message will NOT be sent and
1946 * instead a error will be indicated to the user.
1948 static int sctp_setsockopt_disable_fragments(struct sock *sk,
1949 char __user *optval, int optlen)
1951 int val;
1953 if (optlen < sizeof(int))
1954 return -EINVAL;
1956 if (get_user(val, (int __user *)optval))
1957 return -EFAULT;
1959 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
1961 return 0;
1964 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
1965 int optlen)
1967 if (optlen > sizeof(struct sctp_event_subscribe))
1968 return -EINVAL;
1969 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
1970 return -EFAULT;
1971 return 0;
1974 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
1976 * This socket option is applicable to the UDP-style socket only. When
1977 * set it will cause associations that are idle for more than the
1978 * specified number of seconds to automatically close. An association
1979 * being idle is defined an association that has NOT sent or received
1980 * user data. The special value of '0' indicates that no automatic
1981 * close of any associations should be performed. The option expects an
1982 * integer defining the number of seconds of idle time before an
1983 * association is closed.
1985 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
1986 int optlen)
1988 struct sctp_sock *sp = sctp_sk(sk);
1990 /* Applicable to UDP-style socket only */
1991 if (sctp_style(sk, TCP))
1992 return -EOPNOTSUPP;
1993 if (optlen != sizeof(int))
1994 return -EINVAL;
1995 if (copy_from_user(&sp->autoclose, optval, optlen))
1996 return -EFAULT;
1998 return 0;
2001 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2003 * Applications can enable or disable heartbeats for any peer address of
2004 * an association, modify an address's heartbeat interval, force a
2005 * heartbeat to be sent immediately, and adjust the address's maximum
2006 * number of retransmissions sent before an address is considered
2007 * unreachable. The following structure is used to access and modify an
2008 * address's parameters:
2010 * struct sctp_paddrparams {
2011 * sctp_assoc_t spp_assoc_id;
2012 * struct sockaddr_storage spp_address;
2013 * uint32_t spp_hbinterval;
2014 * uint16_t spp_pathmaxrxt;
2015 * uint32_t spp_pathmtu;
2016 * uint32_t spp_sackdelay;
2017 * uint32_t spp_flags;
2018 * };
2020 * spp_assoc_id - (one-to-many style socket) This is filled in the
2021 * application, and identifies the association for
2022 * this query.
2023 * spp_address - This specifies which address is of interest.
2024 * spp_hbinterval - This contains the value of the heartbeat interval,
2025 * in milliseconds. If a value of zero
2026 * is present in this field then no changes are to
2027 * be made to this parameter.
2028 * spp_pathmaxrxt - This contains the maximum number of
2029 * retransmissions before this address shall be
2030 * considered unreachable. If a value of zero
2031 * is present in this field then no changes are to
2032 * be made to this parameter.
2033 * spp_pathmtu - When Path MTU discovery is disabled the value
2034 * specified here will be the "fixed" path mtu.
2035 * Note that if the spp_address field is empty
2036 * then all associations on this address will
2037 * have this fixed path mtu set upon them.
2039 * spp_sackdelay - When delayed sack is enabled, this value specifies
2040 * the number of milliseconds that sacks will be delayed
2041 * for. This value will apply to all addresses of an
2042 * association if the spp_address field is empty. Note
2043 * also, that if delayed sack is enabled and this
2044 * value is set to 0, no change is made to the last
2045 * recorded delayed sack timer value.
2047 * spp_flags - These flags are used to control various features
2048 * on an association. The flag field may contain
2049 * zero or more of the following options.
2051 * SPP_HB_ENABLE - Enable heartbeats on the
2052 * specified address. Note that if the address
2053 * field is empty all addresses for the association
2054 * have heartbeats enabled upon them.
2056 * SPP_HB_DISABLE - Disable heartbeats on the
2057 * speicifed address. Note that if the address
2058 * field is empty all addresses for the association
2059 * will have their heartbeats disabled. Note also
2060 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2061 * mutually exclusive, only one of these two should
2062 * be specified. Enabling both fields will have
2063 * undetermined results.
2065 * SPP_HB_DEMAND - Request a user initiated heartbeat
2066 * to be made immediately.
2068 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2069 * heartbeat delayis to be set to the value of 0
2070 * milliseconds.
2072 * SPP_PMTUD_ENABLE - This field will enable PMTU
2073 * discovery upon the specified address. Note that
2074 * if the address feild is empty then all addresses
2075 * on the association are effected.
2077 * SPP_PMTUD_DISABLE - This field will disable PMTU
2078 * discovery upon the specified address. Note that
2079 * if the address feild is empty then all addresses
2080 * on the association are effected. Not also that
2081 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2082 * exclusive. Enabling both will have undetermined
2083 * results.
2085 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2086 * on delayed sack. The time specified in spp_sackdelay
2087 * is used to specify the sack delay for this address. Note
2088 * that if spp_address is empty then all addresses will
2089 * enable delayed sack and take on the sack delay
2090 * value specified in spp_sackdelay.
2091 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2092 * off delayed sack. If the spp_address field is blank then
2093 * delayed sack is disabled for the entire association. Note
2094 * also that this field is mutually exclusive to
2095 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2096 * results.
2098 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2099 struct sctp_transport *trans,
2100 struct sctp_association *asoc,
2101 struct sctp_sock *sp,
2102 int hb_change,
2103 int pmtud_change,
2104 int sackdelay_change)
2106 int error;
2108 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2109 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2110 if (error)
2111 return error;
2114 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2115 * this field is ignored. Note also that a value of zero indicates
2116 * the current setting should be left unchanged.
2118 if (params->spp_flags & SPP_HB_ENABLE) {
2120 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2121 * set. This lets us use 0 value when this flag
2122 * is set.
2124 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2125 params->spp_hbinterval = 0;
2127 if (params->spp_hbinterval ||
2128 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2129 if (trans) {
2130 trans->hbinterval =
2131 msecs_to_jiffies(params->spp_hbinterval);
2132 } else if (asoc) {
2133 asoc->hbinterval =
2134 msecs_to_jiffies(params->spp_hbinterval);
2135 } else {
2136 sp->hbinterval = params->spp_hbinterval;
2141 if (hb_change) {
2142 if (trans) {
2143 trans->param_flags =
2144 (trans->param_flags & ~SPP_HB) | hb_change;
2145 } else if (asoc) {
2146 asoc->param_flags =
2147 (asoc->param_flags & ~SPP_HB) | hb_change;
2148 } else {
2149 sp->param_flags =
2150 (sp->param_flags & ~SPP_HB) | hb_change;
2154 /* When Path MTU discovery is disabled the value specified here will
2155 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2156 * include the flag SPP_PMTUD_DISABLE for this field to have any
2157 * effect).
2159 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2160 if (trans) {
2161 trans->pathmtu = params->spp_pathmtu;
2162 sctp_assoc_sync_pmtu(asoc);
2163 } else if (asoc) {
2164 asoc->pathmtu = params->spp_pathmtu;
2165 sctp_frag_point(sp, params->spp_pathmtu);
2166 } else {
2167 sp->pathmtu = params->spp_pathmtu;
2171 if (pmtud_change) {
2172 if (trans) {
2173 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2174 (params->spp_flags & SPP_PMTUD_ENABLE);
2175 trans->param_flags =
2176 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2177 if (update) {
2178 sctp_transport_pmtu(trans);
2179 sctp_assoc_sync_pmtu(asoc);
2181 } else if (asoc) {
2182 asoc->param_flags =
2183 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2184 } else {
2185 sp->param_flags =
2186 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2190 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2191 * value of this field is ignored. Note also that a value of zero
2192 * indicates the current setting should be left unchanged.
2194 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2195 if (trans) {
2196 trans->sackdelay =
2197 msecs_to_jiffies(params->spp_sackdelay);
2198 } else if (asoc) {
2199 asoc->sackdelay =
2200 msecs_to_jiffies(params->spp_sackdelay);
2201 } else {
2202 sp->sackdelay = params->spp_sackdelay;
2206 if (sackdelay_change) {
2207 if (trans) {
2208 trans->param_flags =
2209 (trans->param_flags & ~SPP_SACKDELAY) |
2210 sackdelay_change;
2211 } else if (asoc) {
2212 asoc->param_flags =
2213 (asoc->param_flags & ~SPP_SACKDELAY) |
2214 sackdelay_change;
2215 } else {
2216 sp->param_flags =
2217 (sp->param_flags & ~SPP_SACKDELAY) |
2218 sackdelay_change;
2222 /* Note that unless the spp_flag is set to SPP_PMTUD_ENABLE the value
2223 * of this field is ignored. Note also that a value of zero
2224 * indicates the current setting should be left unchanged.
2226 if ((params->spp_flags & SPP_PMTUD_ENABLE) && params->spp_pathmaxrxt) {
2227 if (trans) {
2228 trans->pathmaxrxt = params->spp_pathmaxrxt;
2229 } else if (asoc) {
2230 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2231 } else {
2232 sp->pathmaxrxt = params->spp_pathmaxrxt;
2236 return 0;
2239 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2240 char __user *optval, int optlen)
2242 struct sctp_paddrparams params;
2243 struct sctp_transport *trans = NULL;
2244 struct sctp_association *asoc = NULL;
2245 struct sctp_sock *sp = sctp_sk(sk);
2246 int error;
2247 int hb_change, pmtud_change, sackdelay_change;
2249 if (optlen != sizeof(struct sctp_paddrparams))
2250 return - EINVAL;
2252 if (copy_from_user(&params, optval, optlen))
2253 return -EFAULT;
2255 /* Validate flags and value parameters. */
2256 hb_change = params.spp_flags & SPP_HB;
2257 pmtud_change = params.spp_flags & SPP_PMTUD;
2258 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2260 if (hb_change == SPP_HB ||
2261 pmtud_change == SPP_PMTUD ||
2262 sackdelay_change == SPP_SACKDELAY ||
2263 params.spp_sackdelay > 500 ||
2264 (params.spp_pathmtu
2265 && params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2266 return -EINVAL;
2268 /* If an address other than INADDR_ANY is specified, and
2269 * no transport is found, then the request is invalid.
2271 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
2272 trans = sctp_addr_id2transport(sk, &params.spp_address,
2273 params.spp_assoc_id);
2274 if (!trans)
2275 return -EINVAL;
2278 /* Get association, if assoc_id != 0 and the socket is a one
2279 * to many style socket, and an association was not found, then
2280 * the id was invalid.
2282 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2283 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2284 return -EINVAL;
2286 /* Heartbeat demand can only be sent on a transport or
2287 * association, but not a socket.
2289 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2290 return -EINVAL;
2292 /* Process parameters. */
2293 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2294 hb_change, pmtud_change,
2295 sackdelay_change);
2297 if (error)
2298 return error;
2300 /* If changes are for association, also apply parameters to each
2301 * transport.
2303 if (!trans && asoc) {
2304 struct list_head *pos;
2306 list_for_each(pos, &asoc->peer.transport_addr_list) {
2307 trans = list_entry(pos, struct sctp_transport,
2308 transports);
2309 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2310 hb_change, pmtud_change,
2311 sackdelay_change);
2315 return 0;
2318 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
2320 * This options will get or set the delayed ack timer. The time is set
2321 * in milliseconds. If the assoc_id is 0, then this sets or gets the
2322 * endpoints default delayed ack timer value. If the assoc_id field is
2323 * non-zero, then the set or get effects the specified association.
2325 * struct sctp_assoc_value {
2326 * sctp_assoc_t assoc_id;
2327 * uint32_t assoc_value;
2328 * };
2330 * assoc_id - This parameter, indicates which association the
2331 * user is preforming an action upon. Note that if
2332 * this field's value is zero then the endpoints
2333 * default value is changed (effecting future
2334 * associations only).
2336 * assoc_value - This parameter contains the number of milliseconds
2337 * that the user is requesting the delayed ACK timer
2338 * be set to. Note that this value is defined in
2339 * the standard to be between 200 and 500 milliseconds.
2341 * Note: a value of zero will leave the value alone,
2342 * but disable SACK delay. A non-zero value will also
2343 * enable SACK delay.
2346 static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
2347 char __user *optval, int optlen)
2349 struct sctp_assoc_value params;
2350 struct sctp_transport *trans = NULL;
2351 struct sctp_association *asoc = NULL;
2352 struct sctp_sock *sp = sctp_sk(sk);
2354 if (optlen != sizeof(struct sctp_assoc_value))
2355 return - EINVAL;
2357 if (copy_from_user(&params, optval, optlen))
2358 return -EFAULT;
2360 /* Validate value parameter. */
2361 if (params.assoc_value > 500)
2362 return -EINVAL;
2364 /* Get association, if assoc_id != 0 and the socket is a one
2365 * to many style socket, and an association was not found, then
2366 * the id was invalid.
2368 asoc = sctp_id2assoc(sk, params.assoc_id);
2369 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2370 return -EINVAL;
2372 if (params.assoc_value) {
2373 if (asoc) {
2374 asoc->sackdelay =
2375 msecs_to_jiffies(params.assoc_value);
2376 asoc->param_flags =
2377 (asoc->param_flags & ~SPP_SACKDELAY) |
2378 SPP_SACKDELAY_ENABLE;
2379 } else {
2380 sp->sackdelay = params.assoc_value;
2381 sp->param_flags =
2382 (sp->param_flags & ~SPP_SACKDELAY) |
2383 SPP_SACKDELAY_ENABLE;
2385 } else {
2386 if (asoc) {
2387 asoc->param_flags =
2388 (asoc->param_flags & ~SPP_SACKDELAY) |
2389 SPP_SACKDELAY_DISABLE;
2390 } else {
2391 sp->param_flags =
2392 (sp->param_flags & ~SPP_SACKDELAY) |
2393 SPP_SACKDELAY_DISABLE;
2397 /* If change is for association, also apply to each transport. */
2398 if (asoc) {
2399 struct list_head *pos;
2401 list_for_each(pos, &asoc->peer.transport_addr_list) {
2402 trans = list_entry(pos, struct sctp_transport,
2403 transports);
2404 if (params.assoc_value) {
2405 trans->sackdelay =
2406 msecs_to_jiffies(params.assoc_value);
2407 trans->param_flags =
2408 (trans->param_flags & ~SPP_SACKDELAY) |
2409 SPP_SACKDELAY_ENABLE;
2410 } else {
2411 trans->param_flags =
2412 (trans->param_flags & ~SPP_SACKDELAY) |
2413 SPP_SACKDELAY_DISABLE;
2418 return 0;
2421 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2423 * Applications can specify protocol parameters for the default association
2424 * initialization. The option name argument to setsockopt() and getsockopt()
2425 * is SCTP_INITMSG.
2427 * Setting initialization parameters is effective only on an unconnected
2428 * socket (for UDP-style sockets only future associations are effected
2429 * by the change). With TCP-style sockets, this option is inherited by
2430 * sockets derived from a listener socket.
2432 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, int optlen)
2434 struct sctp_initmsg sinit;
2435 struct sctp_sock *sp = sctp_sk(sk);
2437 if (optlen != sizeof(struct sctp_initmsg))
2438 return -EINVAL;
2439 if (copy_from_user(&sinit, optval, optlen))
2440 return -EFAULT;
2442 if (sinit.sinit_num_ostreams)
2443 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2444 if (sinit.sinit_max_instreams)
2445 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2446 if (sinit.sinit_max_attempts)
2447 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2448 if (sinit.sinit_max_init_timeo)
2449 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2451 return 0;
2455 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2457 * Applications that wish to use the sendto() system call may wish to
2458 * specify a default set of parameters that would normally be supplied
2459 * through the inclusion of ancillary data. This socket option allows
2460 * such an application to set the default sctp_sndrcvinfo structure.
2461 * The application that wishes to use this socket option simply passes
2462 * in to this call the sctp_sndrcvinfo structure defined in Section
2463 * 5.2.2) The input parameters accepted by this call include
2464 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2465 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2466 * to this call if the caller is using the UDP model.
2468 static int sctp_setsockopt_default_send_param(struct sock *sk,
2469 char __user *optval, int optlen)
2471 struct sctp_sndrcvinfo info;
2472 struct sctp_association *asoc;
2473 struct sctp_sock *sp = sctp_sk(sk);
2475 if (optlen != sizeof(struct sctp_sndrcvinfo))
2476 return -EINVAL;
2477 if (copy_from_user(&info, optval, optlen))
2478 return -EFAULT;
2480 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2481 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2482 return -EINVAL;
2484 if (asoc) {
2485 asoc->default_stream = info.sinfo_stream;
2486 asoc->default_flags = info.sinfo_flags;
2487 asoc->default_ppid = info.sinfo_ppid;
2488 asoc->default_context = info.sinfo_context;
2489 asoc->default_timetolive = info.sinfo_timetolive;
2490 } else {
2491 sp->default_stream = info.sinfo_stream;
2492 sp->default_flags = info.sinfo_flags;
2493 sp->default_ppid = info.sinfo_ppid;
2494 sp->default_context = info.sinfo_context;
2495 sp->default_timetolive = info.sinfo_timetolive;
2498 return 0;
2501 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2503 * Requests that the local SCTP stack use the enclosed peer address as
2504 * the association primary. The enclosed address must be one of the
2505 * association peer's addresses.
2507 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2508 int optlen)
2510 struct sctp_prim prim;
2511 struct sctp_transport *trans;
2513 if (optlen != sizeof(struct sctp_prim))
2514 return -EINVAL;
2516 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2517 return -EFAULT;
2519 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2520 if (!trans)
2521 return -EINVAL;
2523 sctp_assoc_set_primary(trans->asoc, trans);
2525 return 0;
2529 * 7.1.5 SCTP_NODELAY
2531 * Turn on/off any Nagle-like algorithm. This means that packets are
2532 * generally sent as soon as possible and no unnecessary delays are
2533 * introduced, at the cost of more packets in the network. Expects an
2534 * integer boolean flag.
2536 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2537 int optlen)
2539 int val;
2541 if (optlen < sizeof(int))
2542 return -EINVAL;
2543 if (get_user(val, (int __user *)optval))
2544 return -EFAULT;
2546 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2547 return 0;
2552 * 7.1.1 SCTP_RTOINFO
2554 * The protocol parameters used to initialize and bound retransmission
2555 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2556 * and modify these parameters.
2557 * All parameters are time values, in milliseconds. A value of 0, when
2558 * modifying the parameters, indicates that the current value should not
2559 * be changed.
2562 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, int optlen) {
2563 struct sctp_rtoinfo rtoinfo;
2564 struct sctp_association *asoc;
2566 if (optlen != sizeof (struct sctp_rtoinfo))
2567 return -EINVAL;
2569 if (copy_from_user(&rtoinfo, optval, optlen))
2570 return -EFAULT;
2572 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2574 /* Set the values to the specific association */
2575 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2576 return -EINVAL;
2578 if (asoc) {
2579 if (rtoinfo.srto_initial != 0)
2580 asoc->rto_initial =
2581 msecs_to_jiffies(rtoinfo.srto_initial);
2582 if (rtoinfo.srto_max != 0)
2583 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2584 if (rtoinfo.srto_min != 0)
2585 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2586 } else {
2587 /* If there is no association or the association-id = 0
2588 * set the values to the endpoint.
2590 struct sctp_sock *sp = sctp_sk(sk);
2592 if (rtoinfo.srto_initial != 0)
2593 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2594 if (rtoinfo.srto_max != 0)
2595 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2596 if (rtoinfo.srto_min != 0)
2597 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2600 return 0;
2605 * 7.1.2 SCTP_ASSOCINFO
2607 * This option is used to tune the maximum retransmission attempts
2608 * of the association.
2609 * Returns an error if the new association retransmission value is
2610 * greater than the sum of the retransmission value of the peer.
2611 * See [SCTP] for more information.
2614 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, int optlen)
2617 struct sctp_assocparams assocparams;
2618 struct sctp_association *asoc;
2620 if (optlen != sizeof(struct sctp_assocparams))
2621 return -EINVAL;
2622 if (copy_from_user(&assocparams, optval, optlen))
2623 return -EFAULT;
2625 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2627 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2628 return -EINVAL;
2630 /* Set the values to the specific association */
2631 if (asoc) {
2632 if (assocparams.sasoc_asocmaxrxt != 0) {
2633 __u32 path_sum = 0;
2634 int paths = 0;
2635 struct list_head *pos;
2636 struct sctp_transport *peer_addr;
2638 list_for_each(pos, &asoc->peer.transport_addr_list) {
2639 peer_addr = list_entry(pos,
2640 struct sctp_transport,
2641 transports);
2642 path_sum += peer_addr->pathmaxrxt;
2643 paths++;
2646 /* Only validate asocmaxrxt if we have more then
2647 * one path/transport. We do this because path
2648 * retransmissions are only counted when we have more
2649 * then one path.
2651 if (paths > 1 &&
2652 assocparams.sasoc_asocmaxrxt > path_sum)
2653 return -EINVAL;
2655 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2658 if (assocparams.sasoc_cookie_life != 0) {
2659 asoc->cookie_life.tv_sec =
2660 assocparams.sasoc_cookie_life / 1000;
2661 asoc->cookie_life.tv_usec =
2662 (assocparams.sasoc_cookie_life % 1000)
2663 * 1000;
2665 } else {
2666 /* Set the values to the endpoint */
2667 struct sctp_sock *sp = sctp_sk(sk);
2669 if (assocparams.sasoc_asocmaxrxt != 0)
2670 sp->assocparams.sasoc_asocmaxrxt =
2671 assocparams.sasoc_asocmaxrxt;
2672 if (assocparams.sasoc_cookie_life != 0)
2673 sp->assocparams.sasoc_cookie_life =
2674 assocparams.sasoc_cookie_life;
2676 return 0;
2680 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2682 * This socket option is a boolean flag which turns on or off mapped V4
2683 * addresses. If this option is turned on and the socket is type
2684 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2685 * If this option is turned off, then no mapping will be done of V4
2686 * addresses and a user will receive both PF_INET6 and PF_INET type
2687 * addresses on the socket.
2689 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, int optlen)
2691 int val;
2692 struct sctp_sock *sp = sctp_sk(sk);
2694 if (optlen < sizeof(int))
2695 return -EINVAL;
2696 if (get_user(val, (int __user *)optval))
2697 return -EFAULT;
2698 if (val)
2699 sp->v4mapped = 1;
2700 else
2701 sp->v4mapped = 0;
2703 return 0;
2707 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
2709 * This socket option specifies the maximum size to put in any outgoing
2710 * SCTP chunk. If a message is larger than this size it will be
2711 * fragmented by SCTP into the specified size. Note that the underlying
2712 * SCTP implementation may fragment into smaller sized chunks when the
2713 * PMTU of the underlying association is smaller than the value set by
2714 * the user.
2716 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, int optlen)
2718 struct sctp_association *asoc;
2719 struct list_head *pos;
2720 struct sctp_sock *sp = sctp_sk(sk);
2721 int val;
2723 if (optlen < sizeof(int))
2724 return -EINVAL;
2725 if (get_user(val, (int __user *)optval))
2726 return -EFAULT;
2727 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2728 return -EINVAL;
2729 sp->user_frag = val;
2731 /* Update the frag_point of the existing associations. */
2732 list_for_each(pos, &(sp->ep->asocs)) {
2733 asoc = list_entry(pos, struct sctp_association, asocs);
2734 asoc->frag_point = sctp_frag_point(sp, asoc->pathmtu);
2737 return 0;
2742 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2744 * Requests that the peer mark the enclosed address as the association
2745 * primary. The enclosed address must be one of the association's
2746 * locally bound addresses. The following structure is used to make a
2747 * set primary request:
2749 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2750 int optlen)
2752 struct sctp_sock *sp;
2753 struct sctp_endpoint *ep;
2754 struct sctp_association *asoc = NULL;
2755 struct sctp_setpeerprim prim;
2756 struct sctp_chunk *chunk;
2757 int err;
2759 sp = sctp_sk(sk);
2760 ep = sp->ep;
2762 if (!sctp_addip_enable)
2763 return -EPERM;
2765 if (optlen != sizeof(struct sctp_setpeerprim))
2766 return -EINVAL;
2768 if (copy_from_user(&prim, optval, optlen))
2769 return -EFAULT;
2771 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2772 if (!asoc)
2773 return -EINVAL;
2775 if (!asoc->peer.asconf_capable)
2776 return -EPERM;
2778 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2779 return -EPERM;
2781 if (!sctp_state(asoc, ESTABLISHED))
2782 return -ENOTCONN;
2784 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2785 return -EADDRNOTAVAIL;
2787 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2788 chunk = sctp_make_asconf_set_prim(asoc,
2789 (union sctp_addr *)&prim.sspp_addr);
2790 if (!chunk)
2791 return -ENOMEM;
2793 err = sctp_send_asconf(asoc, chunk);
2795 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2797 return err;
2800 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2801 int optlen)
2803 struct sctp_setadaptation adaptation;
2805 if (optlen != sizeof(struct sctp_setadaptation))
2806 return -EINVAL;
2807 if (copy_from_user(&adaptation, optval, optlen))
2808 return -EFAULT;
2810 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2812 return 0;
2816 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
2818 * The context field in the sctp_sndrcvinfo structure is normally only
2819 * used when a failed message is retrieved holding the value that was
2820 * sent down on the actual send call. This option allows the setting of
2821 * a default context on an association basis that will be received on
2822 * reading messages from the peer. This is especially helpful in the
2823 * one-2-many model for an application to keep some reference to an
2824 * internal state machine that is processing messages on the
2825 * association. Note that the setting of this value only effects
2826 * received messages from the peer and does not effect the value that is
2827 * saved with outbound messages.
2829 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
2830 int optlen)
2832 struct sctp_assoc_value params;
2833 struct sctp_sock *sp;
2834 struct sctp_association *asoc;
2836 if (optlen != sizeof(struct sctp_assoc_value))
2837 return -EINVAL;
2838 if (copy_from_user(&params, optval, optlen))
2839 return -EFAULT;
2841 sp = sctp_sk(sk);
2843 if (params.assoc_id != 0) {
2844 asoc = sctp_id2assoc(sk, params.assoc_id);
2845 if (!asoc)
2846 return -EINVAL;
2847 asoc->default_rcv_context = params.assoc_value;
2848 } else {
2849 sp->default_rcv_context = params.assoc_value;
2852 return 0;
2856 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
2858 * This options will at a minimum specify if the implementation is doing
2859 * fragmented interleave. Fragmented interleave, for a one to many
2860 * socket, is when subsequent calls to receive a message may return
2861 * parts of messages from different associations. Some implementations
2862 * may allow you to turn this value on or off. If so, when turned off,
2863 * no fragment interleave will occur (which will cause a head of line
2864 * blocking amongst multiple associations sharing the same one to many
2865 * socket). When this option is turned on, then each receive call may
2866 * come from a different association (thus the user must receive data
2867 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
2868 * association each receive belongs to.
2870 * This option takes a boolean value. A non-zero value indicates that
2871 * fragmented interleave is on. A value of zero indicates that
2872 * fragmented interleave is off.
2874 * Note that it is important that an implementation that allows this
2875 * option to be turned on, have it off by default. Otherwise an unaware
2876 * application using the one to many model may become confused and act
2877 * incorrectly.
2879 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
2880 char __user *optval,
2881 int optlen)
2883 int val;
2885 if (optlen != sizeof(int))
2886 return -EINVAL;
2887 if (get_user(val, (int __user *)optval))
2888 return -EFAULT;
2890 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
2892 return 0;
2896 * 7.1.25. Set or Get the sctp partial delivery point
2897 * (SCTP_PARTIAL_DELIVERY_POINT)
2898 * This option will set or get the SCTP partial delivery point. This
2899 * point is the size of a message where the partial delivery API will be
2900 * invoked to help free up rwnd space for the peer. Setting this to a
2901 * lower value will cause partial delivery's to happen more often. The
2902 * calls argument is an integer that sets or gets the partial delivery
2903 * point.
2905 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
2906 char __user *optval,
2907 int optlen)
2909 u32 val;
2911 if (optlen != sizeof(u32))
2912 return -EINVAL;
2913 if (get_user(val, (int __user *)optval))
2914 return -EFAULT;
2916 sctp_sk(sk)->pd_point = val;
2918 return 0; /* is this the right error code? */
2922 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
2924 * This option will allow a user to change the maximum burst of packets
2925 * that can be emitted by this association. Note that the default value
2926 * is 4, and some implementations may restrict this setting so that it
2927 * can only be lowered.
2929 * NOTE: This text doesn't seem right. Do this on a socket basis with
2930 * future associations inheriting the socket value.
2932 static int sctp_setsockopt_maxburst(struct sock *sk,
2933 char __user *optval,
2934 int optlen)
2936 struct sctp_assoc_value params;
2937 struct sctp_sock *sp;
2938 struct sctp_association *asoc;
2939 int val;
2940 int assoc_id = 0;
2942 if (optlen < sizeof(int))
2943 return -EINVAL;
2945 if (optlen == sizeof(int)) {
2946 printk(KERN_WARNING
2947 "SCTP: Use of int in max_burst socket option deprecated\n");
2948 printk(KERN_WARNING
2949 "SCTP: Use struct sctp_assoc_value instead\n");
2950 if (copy_from_user(&val, optval, optlen))
2951 return -EFAULT;
2952 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2953 if (copy_from_user(&params, optval, optlen))
2954 return -EFAULT;
2955 val = params.assoc_value;
2956 assoc_id = params.assoc_id;
2957 } else
2958 return -EINVAL;
2960 sp = sctp_sk(sk);
2962 if (assoc_id != 0) {
2963 asoc = sctp_id2assoc(sk, assoc_id);
2964 if (!asoc)
2965 return -EINVAL;
2966 asoc->max_burst = val;
2967 } else
2968 sp->max_burst = val;
2970 return 0;
2974 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
2976 * This set option adds a chunk type that the user is requesting to be
2977 * received only in an authenticated way. Changes to the list of chunks
2978 * will only effect future associations on the socket.
2980 static int sctp_setsockopt_auth_chunk(struct sock *sk,
2981 char __user *optval,
2982 int optlen)
2984 struct sctp_authchunk val;
2986 if (optlen != sizeof(struct sctp_authchunk))
2987 return -EINVAL;
2988 if (copy_from_user(&val, optval, optlen))
2989 return -EFAULT;
2991 switch (val.sauth_chunk) {
2992 case SCTP_CID_INIT:
2993 case SCTP_CID_INIT_ACK:
2994 case SCTP_CID_SHUTDOWN_COMPLETE:
2995 case SCTP_CID_AUTH:
2996 return -EINVAL;
2999 /* add this chunk id to the endpoint */
3000 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3004 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3006 * This option gets or sets the list of HMAC algorithms that the local
3007 * endpoint requires the peer to use.
3009 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3010 char __user *optval,
3011 int optlen)
3013 struct sctp_hmacalgo *hmacs;
3014 int err;
3016 if (optlen < sizeof(struct sctp_hmacalgo))
3017 return -EINVAL;
3019 hmacs = kmalloc(optlen, GFP_KERNEL);
3020 if (!hmacs)
3021 return -ENOMEM;
3023 if (copy_from_user(hmacs, optval, optlen)) {
3024 err = -EFAULT;
3025 goto out;
3028 if (hmacs->shmac_num_idents == 0 ||
3029 hmacs->shmac_num_idents > SCTP_AUTH_NUM_HMACS) {
3030 err = -EINVAL;
3031 goto out;
3034 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3035 out:
3036 kfree(hmacs);
3037 return err;
3041 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3043 * This option will set a shared secret key which is used to build an
3044 * association shared key.
3046 static int sctp_setsockopt_auth_key(struct sock *sk,
3047 char __user *optval,
3048 int optlen)
3050 struct sctp_authkey *authkey;
3051 struct sctp_association *asoc;
3052 int ret;
3054 if (optlen <= sizeof(struct sctp_authkey))
3055 return -EINVAL;
3057 authkey = kmalloc(optlen, GFP_KERNEL);
3058 if (!authkey)
3059 return -ENOMEM;
3061 if (copy_from_user(authkey, optval, optlen)) {
3062 ret = -EFAULT;
3063 goto out;
3066 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3067 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3068 ret = -EINVAL;
3069 goto out;
3072 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3073 out:
3074 kfree(authkey);
3075 return ret;
3079 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3081 * This option will get or set the active shared key to be used to build
3082 * the association shared key.
3084 static int sctp_setsockopt_active_key(struct sock *sk,
3085 char __user *optval,
3086 int optlen)
3088 struct sctp_authkeyid val;
3089 struct sctp_association *asoc;
3091 if (optlen != sizeof(struct sctp_authkeyid))
3092 return -EINVAL;
3093 if (copy_from_user(&val, optval, optlen))
3094 return -EFAULT;
3096 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3097 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3098 return -EINVAL;
3100 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3101 val.scact_keynumber);
3105 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3107 * This set option will delete a shared secret key from use.
3109 static int sctp_setsockopt_del_key(struct sock *sk,
3110 char __user *optval,
3111 int optlen)
3113 struct sctp_authkeyid val;
3114 struct sctp_association *asoc;
3116 if (optlen != sizeof(struct sctp_authkeyid))
3117 return -EINVAL;
3118 if (copy_from_user(&val, optval, optlen))
3119 return -EFAULT;
3121 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3122 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3123 return -EINVAL;
3125 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3126 val.scact_keynumber);
3131 /* API 6.2 setsockopt(), getsockopt()
3133 * Applications use setsockopt() and getsockopt() to set or retrieve
3134 * socket options. Socket options are used to change the default
3135 * behavior of sockets calls. They are described in Section 7.
3137 * The syntax is:
3139 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3140 * int __user *optlen);
3141 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3142 * int optlen);
3144 * sd - the socket descript.
3145 * level - set to IPPROTO_SCTP for all SCTP options.
3146 * optname - the option name.
3147 * optval - the buffer to store the value of the option.
3148 * optlen - the size of the buffer.
3150 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3151 char __user *optval, int optlen)
3153 int retval = 0;
3155 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3156 sk, optname);
3158 /* I can hardly begin to describe how wrong this is. This is
3159 * so broken as to be worse than useless. The API draft
3160 * REALLY is NOT helpful here... I am not convinced that the
3161 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3162 * are at all well-founded.
3164 if (level != SOL_SCTP) {
3165 struct sctp_af *af = sctp_sk(sk)->pf->af;
3166 retval = af->setsockopt(sk, level, optname, optval, optlen);
3167 goto out_nounlock;
3170 sctp_lock_sock(sk);
3172 switch (optname) {
3173 case SCTP_SOCKOPT_BINDX_ADD:
3174 /* 'optlen' is the size of the addresses buffer. */
3175 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3176 optlen, SCTP_BINDX_ADD_ADDR);
3177 break;
3179 case SCTP_SOCKOPT_BINDX_REM:
3180 /* 'optlen' is the size of the addresses buffer. */
3181 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3182 optlen, SCTP_BINDX_REM_ADDR);
3183 break;
3185 case SCTP_SOCKOPT_CONNECTX:
3186 /* 'optlen' is the size of the addresses buffer. */
3187 retval = sctp_setsockopt_connectx(sk, (struct sockaddr __user *)optval,
3188 optlen);
3189 break;
3191 case SCTP_DISABLE_FRAGMENTS:
3192 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3193 break;
3195 case SCTP_EVENTS:
3196 retval = sctp_setsockopt_events(sk, optval, optlen);
3197 break;
3199 case SCTP_AUTOCLOSE:
3200 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3201 break;
3203 case SCTP_PEER_ADDR_PARAMS:
3204 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3205 break;
3207 case SCTP_DELAYED_ACK_TIME:
3208 retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
3209 break;
3210 case SCTP_PARTIAL_DELIVERY_POINT:
3211 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3212 break;
3214 case SCTP_INITMSG:
3215 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3216 break;
3217 case SCTP_DEFAULT_SEND_PARAM:
3218 retval = sctp_setsockopt_default_send_param(sk, optval,
3219 optlen);
3220 break;
3221 case SCTP_PRIMARY_ADDR:
3222 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3223 break;
3224 case SCTP_SET_PEER_PRIMARY_ADDR:
3225 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3226 break;
3227 case SCTP_NODELAY:
3228 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3229 break;
3230 case SCTP_RTOINFO:
3231 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3232 break;
3233 case SCTP_ASSOCINFO:
3234 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3235 break;
3236 case SCTP_I_WANT_MAPPED_V4_ADDR:
3237 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3238 break;
3239 case SCTP_MAXSEG:
3240 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3241 break;
3242 case SCTP_ADAPTATION_LAYER:
3243 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3244 break;
3245 case SCTP_CONTEXT:
3246 retval = sctp_setsockopt_context(sk, optval, optlen);
3247 break;
3248 case SCTP_FRAGMENT_INTERLEAVE:
3249 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3250 break;
3251 case SCTP_MAX_BURST:
3252 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3253 break;
3254 case SCTP_AUTH_CHUNK:
3255 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3256 break;
3257 case SCTP_HMAC_IDENT:
3258 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3259 break;
3260 case SCTP_AUTH_KEY:
3261 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3262 break;
3263 case SCTP_AUTH_ACTIVE_KEY:
3264 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3265 break;
3266 case SCTP_AUTH_DELETE_KEY:
3267 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3268 break;
3269 default:
3270 retval = -ENOPROTOOPT;
3271 break;
3274 sctp_release_sock(sk);
3276 out_nounlock:
3277 return retval;
3280 /* API 3.1.6 connect() - UDP Style Syntax
3282 * An application may use the connect() call in the UDP model to initiate an
3283 * association without sending data.
3285 * The syntax is:
3287 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3289 * sd: the socket descriptor to have a new association added to.
3291 * nam: the address structure (either struct sockaddr_in or struct
3292 * sockaddr_in6 defined in RFC2553 [7]).
3294 * len: the size of the address.
3296 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3297 int addr_len)
3299 int err = 0;
3300 struct sctp_af *af;
3302 sctp_lock_sock(sk);
3304 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3305 __FUNCTION__, sk, addr, addr_len);
3307 /* Validate addr_len before calling common connect/connectx routine. */
3308 af = sctp_get_af_specific(addr->sa_family);
3309 if (!af || addr_len < af->sockaddr_len) {
3310 err = -EINVAL;
3311 } else {
3312 /* Pass correct addr len to common routine (so it knows there
3313 * is only one address being passed.
3315 err = __sctp_connect(sk, addr, af->sockaddr_len);
3318 sctp_release_sock(sk);
3319 return err;
3322 /* FIXME: Write comments. */
3323 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3325 return -EOPNOTSUPP; /* STUB */
3328 /* 4.1.4 accept() - TCP Style Syntax
3330 * Applications use accept() call to remove an established SCTP
3331 * association from the accept queue of the endpoint. A new socket
3332 * descriptor will be returned from accept() to represent the newly
3333 * formed association.
3335 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3337 struct sctp_sock *sp;
3338 struct sctp_endpoint *ep;
3339 struct sock *newsk = NULL;
3340 struct sctp_association *asoc;
3341 long timeo;
3342 int error = 0;
3344 sctp_lock_sock(sk);
3346 sp = sctp_sk(sk);
3347 ep = sp->ep;
3349 if (!sctp_style(sk, TCP)) {
3350 error = -EOPNOTSUPP;
3351 goto out;
3354 if (!sctp_sstate(sk, LISTENING)) {
3355 error = -EINVAL;
3356 goto out;
3359 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3361 error = sctp_wait_for_accept(sk, timeo);
3362 if (error)
3363 goto out;
3365 /* We treat the list of associations on the endpoint as the accept
3366 * queue and pick the first association on the list.
3368 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3370 newsk = sp->pf->create_accept_sk(sk, asoc);
3371 if (!newsk) {
3372 error = -ENOMEM;
3373 goto out;
3376 /* Populate the fields of the newsk from the oldsk and migrate the
3377 * asoc to the newsk.
3379 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3381 out:
3382 sctp_release_sock(sk);
3383 *err = error;
3384 return newsk;
3387 /* The SCTP ioctl handler. */
3388 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3390 return -ENOIOCTLCMD;
3393 /* This is the function which gets called during socket creation to
3394 * initialized the SCTP-specific portion of the sock.
3395 * The sock structure should already be zero-filled memory.
3397 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3399 struct sctp_endpoint *ep;
3400 struct sctp_sock *sp;
3402 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3404 sp = sctp_sk(sk);
3406 /* Initialize the SCTP per socket area. */
3407 switch (sk->sk_type) {
3408 case SOCK_SEQPACKET:
3409 sp->type = SCTP_SOCKET_UDP;
3410 break;
3411 case SOCK_STREAM:
3412 sp->type = SCTP_SOCKET_TCP;
3413 break;
3414 default:
3415 return -ESOCKTNOSUPPORT;
3418 /* Initialize default send parameters. These parameters can be
3419 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3421 sp->default_stream = 0;
3422 sp->default_ppid = 0;
3423 sp->default_flags = 0;
3424 sp->default_context = 0;
3425 sp->default_timetolive = 0;
3427 sp->default_rcv_context = 0;
3428 sp->max_burst = sctp_max_burst;
3430 /* Initialize default setup parameters. These parameters
3431 * can be modified with the SCTP_INITMSG socket option or
3432 * overridden by the SCTP_INIT CMSG.
3434 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3435 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3436 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3437 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3439 /* Initialize default RTO related parameters. These parameters can
3440 * be modified for with the SCTP_RTOINFO socket option.
3442 sp->rtoinfo.srto_initial = sctp_rto_initial;
3443 sp->rtoinfo.srto_max = sctp_rto_max;
3444 sp->rtoinfo.srto_min = sctp_rto_min;
3446 /* Initialize default association related parameters. These parameters
3447 * can be modified with the SCTP_ASSOCINFO socket option.
3449 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3450 sp->assocparams.sasoc_number_peer_destinations = 0;
3451 sp->assocparams.sasoc_peer_rwnd = 0;
3452 sp->assocparams.sasoc_local_rwnd = 0;
3453 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3455 /* Initialize default event subscriptions. By default, all the
3456 * options are off.
3458 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3460 /* Default Peer Address Parameters. These defaults can
3461 * be modified via SCTP_PEER_ADDR_PARAMS
3463 sp->hbinterval = sctp_hb_interval;
3464 sp->pathmaxrxt = sctp_max_retrans_path;
3465 sp->pathmtu = 0; // allow default discovery
3466 sp->sackdelay = sctp_sack_timeout;
3467 sp->param_flags = SPP_HB_ENABLE |
3468 SPP_PMTUD_ENABLE |
3469 SPP_SACKDELAY_ENABLE;
3471 /* If enabled no SCTP message fragmentation will be performed.
3472 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3474 sp->disable_fragments = 0;
3476 /* Enable Nagle algorithm by default. */
3477 sp->nodelay = 0;
3479 /* Enable by default. */
3480 sp->v4mapped = 1;
3482 /* Auto-close idle associations after the configured
3483 * number of seconds. A value of 0 disables this
3484 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3485 * for UDP-style sockets only.
3487 sp->autoclose = 0;
3489 /* User specified fragmentation limit. */
3490 sp->user_frag = 0;
3492 sp->adaptation_ind = 0;
3494 sp->pf = sctp_get_pf_specific(sk->sk_family);
3496 /* Control variables for partial data delivery. */
3497 atomic_set(&sp->pd_mode, 0);
3498 skb_queue_head_init(&sp->pd_lobby);
3499 sp->frag_interleave = 0;
3501 /* Create a per socket endpoint structure. Even if we
3502 * change the data structure relationships, this may still
3503 * be useful for storing pre-connect address information.
3505 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3506 if (!ep)
3507 return -ENOMEM;
3509 sp->ep = ep;
3510 sp->hmac = NULL;
3512 SCTP_DBG_OBJCNT_INC(sock);
3513 atomic_inc(&sctp_sockets_allocated);
3514 return 0;
3517 /* Cleanup any SCTP per socket resources. */
3518 SCTP_STATIC int sctp_destroy_sock(struct sock *sk)
3520 struct sctp_endpoint *ep;
3522 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3524 /* Release our hold on the endpoint. */
3525 ep = sctp_sk(sk)->ep;
3526 sctp_endpoint_free(ep);
3527 atomic_dec(&sctp_sockets_allocated);
3528 return 0;
3531 /* API 4.1.7 shutdown() - TCP Style Syntax
3532 * int shutdown(int socket, int how);
3534 * sd - the socket descriptor of the association to be closed.
3535 * how - Specifies the type of shutdown. The values are
3536 * as follows:
3537 * SHUT_RD
3538 * Disables further receive operations. No SCTP
3539 * protocol action is taken.
3540 * SHUT_WR
3541 * Disables further send operations, and initiates
3542 * the SCTP shutdown sequence.
3543 * SHUT_RDWR
3544 * Disables further send and receive operations
3545 * and initiates the SCTP shutdown sequence.
3547 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3549 struct sctp_endpoint *ep;
3550 struct sctp_association *asoc;
3552 if (!sctp_style(sk, TCP))
3553 return;
3555 if (how & SEND_SHUTDOWN) {
3556 ep = sctp_sk(sk)->ep;
3557 if (!list_empty(&ep->asocs)) {
3558 asoc = list_entry(ep->asocs.next,
3559 struct sctp_association, asocs);
3560 sctp_primitive_SHUTDOWN(asoc, NULL);
3565 /* 7.2.1 Association Status (SCTP_STATUS)
3567 * Applications can retrieve current status information about an
3568 * association, including association state, peer receiver window size,
3569 * number of unacked data chunks, and number of data chunks pending
3570 * receipt. This information is read-only.
3572 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3573 char __user *optval,
3574 int __user *optlen)
3576 struct sctp_status status;
3577 struct sctp_association *asoc = NULL;
3578 struct sctp_transport *transport;
3579 sctp_assoc_t associd;
3580 int retval = 0;
3582 if (len < sizeof(status)) {
3583 retval = -EINVAL;
3584 goto out;
3587 len = sizeof(status);
3588 if (copy_from_user(&status, optval, len)) {
3589 retval = -EFAULT;
3590 goto out;
3593 associd = status.sstat_assoc_id;
3594 asoc = sctp_id2assoc(sk, associd);
3595 if (!asoc) {
3596 retval = -EINVAL;
3597 goto out;
3600 transport = asoc->peer.primary_path;
3602 status.sstat_assoc_id = sctp_assoc2id(asoc);
3603 status.sstat_state = asoc->state;
3604 status.sstat_rwnd = asoc->peer.rwnd;
3605 status.sstat_unackdata = asoc->unack_data;
3607 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3608 status.sstat_instrms = asoc->c.sinit_max_instreams;
3609 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3610 status.sstat_fragmentation_point = asoc->frag_point;
3611 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3612 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3613 transport->af_specific->sockaddr_len);
3614 /* Map ipv4 address into v4-mapped-on-v6 address. */
3615 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3616 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3617 status.sstat_primary.spinfo_state = transport->state;
3618 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3619 status.sstat_primary.spinfo_srtt = transport->srtt;
3620 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3621 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3623 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3624 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3626 if (put_user(len, optlen)) {
3627 retval = -EFAULT;
3628 goto out;
3631 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3632 len, status.sstat_state, status.sstat_rwnd,
3633 status.sstat_assoc_id);
3635 if (copy_to_user(optval, &status, len)) {
3636 retval = -EFAULT;
3637 goto out;
3640 out:
3641 return (retval);
3645 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3647 * Applications can retrieve information about a specific peer address
3648 * of an association, including its reachability state, congestion
3649 * window, and retransmission timer values. This information is
3650 * read-only.
3652 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3653 char __user *optval,
3654 int __user *optlen)
3656 struct sctp_paddrinfo pinfo;
3657 struct sctp_transport *transport;
3658 int retval = 0;
3660 if (len < sizeof(pinfo)) {
3661 retval = -EINVAL;
3662 goto out;
3665 len = sizeof(pinfo);
3666 if (copy_from_user(&pinfo, optval, len)) {
3667 retval = -EFAULT;
3668 goto out;
3671 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3672 pinfo.spinfo_assoc_id);
3673 if (!transport)
3674 return -EINVAL;
3676 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3677 pinfo.spinfo_state = transport->state;
3678 pinfo.spinfo_cwnd = transport->cwnd;
3679 pinfo.spinfo_srtt = transport->srtt;
3680 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3681 pinfo.spinfo_mtu = transport->pathmtu;
3683 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3684 pinfo.spinfo_state = SCTP_ACTIVE;
3686 if (put_user(len, optlen)) {
3687 retval = -EFAULT;
3688 goto out;
3691 if (copy_to_user(optval, &pinfo, len)) {
3692 retval = -EFAULT;
3693 goto out;
3696 out:
3697 return (retval);
3700 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3702 * This option is a on/off flag. If enabled no SCTP message
3703 * fragmentation will be performed. Instead if a message being sent
3704 * exceeds the current PMTU size, the message will NOT be sent and
3705 * instead a error will be indicated to the user.
3707 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3708 char __user *optval, int __user *optlen)
3710 int val;
3712 if (len < sizeof(int))
3713 return -EINVAL;
3715 len = sizeof(int);
3716 val = (sctp_sk(sk)->disable_fragments == 1);
3717 if (put_user(len, optlen))
3718 return -EFAULT;
3719 if (copy_to_user(optval, &val, len))
3720 return -EFAULT;
3721 return 0;
3724 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3726 * This socket option is used to specify various notifications and
3727 * ancillary data the user wishes to receive.
3729 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3730 int __user *optlen)
3732 if (len < sizeof(struct sctp_event_subscribe))
3733 return -EINVAL;
3734 len = sizeof(struct sctp_event_subscribe);
3735 if (put_user(len, optlen))
3736 return -EFAULT;
3737 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3738 return -EFAULT;
3739 return 0;
3742 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
3744 * This socket option is applicable to the UDP-style socket only. When
3745 * set it will cause associations that are idle for more than the
3746 * specified number of seconds to automatically close. An association
3747 * being idle is defined an association that has NOT sent or received
3748 * user data. The special value of '0' indicates that no automatic
3749 * close of any associations should be performed. The option expects an
3750 * integer defining the number of seconds of idle time before an
3751 * association is closed.
3753 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
3755 /* Applicable to UDP-style socket only */
3756 if (sctp_style(sk, TCP))
3757 return -EOPNOTSUPP;
3758 if (len < sizeof(int))
3759 return -EINVAL;
3760 len = sizeof(int);
3761 if (put_user(len, optlen))
3762 return -EFAULT;
3763 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
3764 return -EFAULT;
3765 return 0;
3768 /* Helper routine to branch off an association to a new socket. */
3769 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
3770 struct socket **sockp)
3772 struct sock *sk = asoc->base.sk;
3773 struct socket *sock;
3774 struct inet_sock *inetsk;
3775 struct sctp_af *af;
3776 int err = 0;
3778 /* An association cannot be branched off from an already peeled-off
3779 * socket, nor is this supported for tcp style sockets.
3781 if (!sctp_style(sk, UDP))
3782 return -EINVAL;
3784 /* Create a new socket. */
3785 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
3786 if (err < 0)
3787 return err;
3789 /* Populate the fields of the newsk from the oldsk and migrate the
3790 * asoc to the newsk.
3792 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
3794 /* Make peeled-off sockets more like 1-1 accepted sockets.
3795 * Set the daddr and initialize id to something more random
3797 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
3798 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
3799 inetsk = inet_sk(sock->sk);
3800 inetsk->id = asoc->next_tsn ^ jiffies;
3802 *sockp = sock;
3804 return err;
3807 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
3809 sctp_peeloff_arg_t peeloff;
3810 struct socket *newsock;
3811 int retval = 0;
3812 struct sctp_association *asoc;
3814 if (len < sizeof(sctp_peeloff_arg_t))
3815 return -EINVAL;
3816 len = sizeof(sctp_peeloff_arg_t);
3817 if (copy_from_user(&peeloff, optval, len))
3818 return -EFAULT;
3820 asoc = sctp_id2assoc(sk, peeloff.associd);
3821 if (!asoc) {
3822 retval = -EINVAL;
3823 goto out;
3826 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __FUNCTION__, sk, asoc);
3828 retval = sctp_do_peeloff(asoc, &newsock);
3829 if (retval < 0)
3830 goto out;
3832 /* Map the socket to an unused fd that can be returned to the user. */
3833 retval = sock_map_fd(newsock);
3834 if (retval < 0) {
3835 sock_release(newsock);
3836 goto out;
3839 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
3840 __FUNCTION__, sk, asoc, newsock->sk, retval);
3842 /* Return the fd mapped to the new socket. */
3843 peeloff.sd = retval;
3844 if (put_user(len, optlen))
3845 return -EFAULT;
3846 if (copy_to_user(optval, &peeloff, len))
3847 retval = -EFAULT;
3849 out:
3850 return retval;
3853 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
3855 * Applications can enable or disable heartbeats for any peer address of
3856 * an association, modify an address's heartbeat interval, force a
3857 * heartbeat to be sent immediately, and adjust the address's maximum
3858 * number of retransmissions sent before an address is considered
3859 * unreachable. The following structure is used to access and modify an
3860 * address's parameters:
3862 * struct sctp_paddrparams {
3863 * sctp_assoc_t spp_assoc_id;
3864 * struct sockaddr_storage spp_address;
3865 * uint32_t spp_hbinterval;
3866 * uint16_t spp_pathmaxrxt;
3867 * uint32_t spp_pathmtu;
3868 * uint32_t spp_sackdelay;
3869 * uint32_t spp_flags;
3870 * };
3872 * spp_assoc_id - (one-to-many style socket) This is filled in the
3873 * application, and identifies the association for
3874 * this query.
3875 * spp_address - This specifies which address is of interest.
3876 * spp_hbinterval - This contains the value of the heartbeat interval,
3877 * in milliseconds. If a value of zero
3878 * is present in this field then no changes are to
3879 * be made to this parameter.
3880 * spp_pathmaxrxt - This contains the maximum number of
3881 * retransmissions before this address shall be
3882 * considered unreachable. If a value of zero
3883 * is present in this field then no changes are to
3884 * be made to this parameter.
3885 * spp_pathmtu - When Path MTU discovery is disabled the value
3886 * specified here will be the "fixed" path mtu.
3887 * Note that if the spp_address field is empty
3888 * then all associations on this address will
3889 * have this fixed path mtu set upon them.
3891 * spp_sackdelay - When delayed sack is enabled, this value specifies
3892 * the number of milliseconds that sacks will be delayed
3893 * for. This value will apply to all addresses of an
3894 * association if the spp_address field is empty. Note
3895 * also, that if delayed sack is enabled and this
3896 * value is set to 0, no change is made to the last
3897 * recorded delayed sack timer value.
3899 * spp_flags - These flags are used to control various features
3900 * on an association. The flag field may contain
3901 * zero or more of the following options.
3903 * SPP_HB_ENABLE - Enable heartbeats on the
3904 * specified address. Note that if the address
3905 * field is empty all addresses for the association
3906 * have heartbeats enabled upon them.
3908 * SPP_HB_DISABLE - Disable heartbeats on the
3909 * speicifed address. Note that if the address
3910 * field is empty all addresses for the association
3911 * will have their heartbeats disabled. Note also
3912 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
3913 * mutually exclusive, only one of these two should
3914 * be specified. Enabling both fields will have
3915 * undetermined results.
3917 * SPP_HB_DEMAND - Request a user initiated heartbeat
3918 * to be made immediately.
3920 * SPP_PMTUD_ENABLE - This field will enable PMTU
3921 * discovery upon the specified address. Note that
3922 * if the address feild is empty then all addresses
3923 * on the association are effected.
3925 * SPP_PMTUD_DISABLE - This field will disable PMTU
3926 * discovery upon the specified address. Note that
3927 * if the address feild is empty then all addresses
3928 * on the association are effected. Not also that
3929 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
3930 * exclusive. Enabling both will have undetermined
3931 * results.
3933 * SPP_SACKDELAY_ENABLE - Setting this flag turns
3934 * on delayed sack. The time specified in spp_sackdelay
3935 * is used to specify the sack delay for this address. Note
3936 * that if spp_address is empty then all addresses will
3937 * enable delayed sack and take on the sack delay
3938 * value specified in spp_sackdelay.
3939 * SPP_SACKDELAY_DISABLE - Setting this flag turns
3940 * off delayed sack. If the spp_address field is blank then
3941 * delayed sack is disabled for the entire association. Note
3942 * also that this field is mutually exclusive to
3943 * SPP_SACKDELAY_ENABLE, setting both will have undefined
3944 * results.
3946 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
3947 char __user *optval, int __user *optlen)
3949 struct sctp_paddrparams params;
3950 struct sctp_transport *trans = NULL;
3951 struct sctp_association *asoc = NULL;
3952 struct sctp_sock *sp = sctp_sk(sk);
3954 if (len < sizeof(struct sctp_paddrparams))
3955 return -EINVAL;
3956 len = sizeof(struct sctp_paddrparams);
3957 if (copy_from_user(&params, optval, len))
3958 return -EFAULT;
3960 /* If an address other than INADDR_ANY is specified, and
3961 * no transport is found, then the request is invalid.
3963 if (!sctp_is_any(( union sctp_addr *)&params.spp_address)) {
3964 trans = sctp_addr_id2transport(sk, &params.spp_address,
3965 params.spp_assoc_id);
3966 if (!trans) {
3967 SCTP_DEBUG_PRINTK("Failed no transport\n");
3968 return -EINVAL;
3972 /* Get association, if assoc_id != 0 and the socket is a one
3973 * to many style socket, and an association was not found, then
3974 * the id was invalid.
3976 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
3977 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
3978 SCTP_DEBUG_PRINTK("Failed no association\n");
3979 return -EINVAL;
3982 if (trans) {
3983 /* Fetch transport values. */
3984 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
3985 params.spp_pathmtu = trans->pathmtu;
3986 params.spp_pathmaxrxt = trans->pathmaxrxt;
3987 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
3989 /*draft-11 doesn't say what to return in spp_flags*/
3990 params.spp_flags = trans->param_flags;
3991 } else if (asoc) {
3992 /* Fetch association values. */
3993 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
3994 params.spp_pathmtu = asoc->pathmtu;
3995 params.spp_pathmaxrxt = asoc->pathmaxrxt;
3996 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
3998 /*draft-11 doesn't say what to return in spp_flags*/
3999 params.spp_flags = asoc->param_flags;
4000 } else {
4001 /* Fetch socket values. */
4002 params.spp_hbinterval = sp->hbinterval;
4003 params.spp_pathmtu = sp->pathmtu;
4004 params.spp_sackdelay = sp->sackdelay;
4005 params.spp_pathmaxrxt = sp->pathmaxrxt;
4007 /*draft-11 doesn't say what to return in spp_flags*/
4008 params.spp_flags = sp->param_flags;
4011 if (copy_to_user(optval, &params, len))
4012 return -EFAULT;
4014 if (put_user(len, optlen))
4015 return -EFAULT;
4017 return 0;
4020 /* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
4022 * This options will get or set the delayed ack timer. The time is set
4023 * in milliseconds. If the assoc_id is 0, then this sets or gets the
4024 * endpoints default delayed ack timer value. If the assoc_id field is
4025 * non-zero, then the set or get effects the specified association.
4027 * struct sctp_assoc_value {
4028 * sctp_assoc_t assoc_id;
4029 * uint32_t assoc_value;
4030 * };
4032 * assoc_id - This parameter, indicates which association the
4033 * user is preforming an action upon. Note that if
4034 * this field's value is zero then the endpoints
4035 * default value is changed (effecting future
4036 * associations only).
4038 * assoc_value - This parameter contains the number of milliseconds
4039 * that the user is requesting the delayed ACK timer
4040 * be set to. Note that this value is defined in
4041 * the standard to be between 200 and 500 milliseconds.
4043 * Note: a value of zero will leave the value alone,
4044 * but disable SACK delay. A non-zero value will also
4045 * enable SACK delay.
4047 static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
4048 char __user *optval,
4049 int __user *optlen)
4051 struct sctp_assoc_value params;
4052 struct sctp_association *asoc = NULL;
4053 struct sctp_sock *sp = sctp_sk(sk);
4055 if (len < sizeof(struct sctp_assoc_value))
4056 return - EINVAL;
4058 len = sizeof(struct sctp_assoc_value);
4060 if (copy_from_user(&params, optval, len))
4061 return -EFAULT;
4063 /* Get association, if assoc_id != 0 and the socket is a one
4064 * to many style socket, and an association was not found, then
4065 * the id was invalid.
4067 asoc = sctp_id2assoc(sk, params.assoc_id);
4068 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4069 return -EINVAL;
4071 if (asoc) {
4072 /* Fetch association values. */
4073 if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
4074 params.assoc_value = jiffies_to_msecs(
4075 asoc->sackdelay);
4076 else
4077 params.assoc_value = 0;
4078 } else {
4079 /* Fetch socket values. */
4080 if (sp->param_flags & SPP_SACKDELAY_ENABLE)
4081 params.assoc_value = sp->sackdelay;
4082 else
4083 params.assoc_value = 0;
4086 if (copy_to_user(optval, &params, len))
4087 return -EFAULT;
4089 if (put_user(len, optlen))
4090 return -EFAULT;
4092 return 0;
4095 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4097 * Applications can specify protocol parameters for the default association
4098 * initialization. The option name argument to setsockopt() and getsockopt()
4099 * is SCTP_INITMSG.
4101 * Setting initialization parameters is effective only on an unconnected
4102 * socket (for UDP-style sockets only future associations are effected
4103 * by the change). With TCP-style sockets, this option is inherited by
4104 * sockets derived from a listener socket.
4106 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4108 if (len < sizeof(struct sctp_initmsg))
4109 return -EINVAL;
4110 len = sizeof(struct sctp_initmsg);
4111 if (put_user(len, optlen))
4112 return -EFAULT;
4113 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4114 return -EFAULT;
4115 return 0;
4118 static int sctp_getsockopt_peer_addrs_num_old(struct sock *sk, int len,
4119 char __user *optval,
4120 int __user *optlen)
4122 sctp_assoc_t id;
4123 struct sctp_association *asoc;
4124 struct list_head *pos;
4125 int cnt = 0;
4127 if (len < sizeof(sctp_assoc_t))
4128 return -EINVAL;
4130 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4131 return -EFAULT;
4133 /* For UDP-style sockets, id specifies the association to query. */
4134 asoc = sctp_id2assoc(sk, id);
4135 if (!asoc)
4136 return -EINVAL;
4138 list_for_each(pos, &asoc->peer.transport_addr_list) {
4139 cnt ++;
4142 return cnt;
4146 * Old API for getting list of peer addresses. Does not work for 32-bit
4147 * programs running on a 64-bit kernel
4149 static int sctp_getsockopt_peer_addrs_old(struct sock *sk, int len,
4150 char __user *optval,
4151 int __user *optlen)
4153 struct sctp_association *asoc;
4154 struct list_head *pos;
4155 int cnt = 0;
4156 struct sctp_getaddrs_old getaddrs;
4157 struct sctp_transport *from;
4158 void __user *to;
4159 union sctp_addr temp;
4160 struct sctp_sock *sp = sctp_sk(sk);
4161 int addrlen;
4163 if (len < sizeof(struct sctp_getaddrs_old))
4164 return -EINVAL;
4166 len = sizeof(struct sctp_getaddrs_old);
4168 if (copy_from_user(&getaddrs, optval, len))
4169 return -EFAULT;
4171 if (getaddrs.addr_num <= 0) return -EINVAL;
4173 /* For UDP-style sockets, id specifies the association to query. */
4174 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4175 if (!asoc)
4176 return -EINVAL;
4178 to = (void __user *)getaddrs.addrs;
4179 list_for_each(pos, &asoc->peer.transport_addr_list) {
4180 from = list_entry(pos, struct sctp_transport, transports);
4181 memcpy(&temp, &from->ipaddr, sizeof(temp));
4182 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4183 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4184 if (copy_to_user(to, &temp, addrlen))
4185 return -EFAULT;
4186 to += addrlen ;
4187 cnt ++;
4188 if (cnt >= getaddrs.addr_num) break;
4190 getaddrs.addr_num = cnt;
4191 if (put_user(len, optlen))
4192 return -EFAULT;
4193 if (copy_to_user(optval, &getaddrs, len))
4194 return -EFAULT;
4196 return 0;
4199 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4200 char __user *optval, int __user *optlen)
4202 struct sctp_association *asoc;
4203 struct list_head *pos;
4204 int cnt = 0;
4205 struct sctp_getaddrs getaddrs;
4206 struct sctp_transport *from;
4207 void __user *to;
4208 union sctp_addr temp;
4209 struct sctp_sock *sp = sctp_sk(sk);
4210 int addrlen;
4211 size_t space_left;
4212 int bytes_copied;
4214 if (len < sizeof(struct sctp_getaddrs))
4215 return -EINVAL;
4217 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4218 return -EFAULT;
4220 /* For UDP-style sockets, id specifies the association to query. */
4221 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4222 if (!asoc)
4223 return -EINVAL;
4225 to = optval + offsetof(struct sctp_getaddrs,addrs);
4226 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4228 list_for_each(pos, &asoc->peer.transport_addr_list) {
4229 from = list_entry(pos, struct sctp_transport, transports);
4230 memcpy(&temp, &from->ipaddr, sizeof(temp));
4231 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4232 addrlen = sctp_get_af_specific(sk->sk_family)->sockaddr_len;
4233 if (space_left < addrlen)
4234 return -ENOMEM;
4235 if (copy_to_user(to, &temp, addrlen))
4236 return -EFAULT;
4237 to += addrlen;
4238 cnt++;
4239 space_left -= addrlen;
4242 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4243 return -EFAULT;
4244 bytes_copied = ((char __user *)to) - optval;
4245 if (put_user(bytes_copied, optlen))
4246 return -EFAULT;
4248 return 0;
4251 static int sctp_getsockopt_local_addrs_num_old(struct sock *sk, int len,
4252 char __user *optval,
4253 int __user *optlen)
4255 sctp_assoc_t id;
4256 struct sctp_bind_addr *bp;
4257 struct sctp_association *asoc;
4258 struct sctp_sockaddr_entry *addr;
4259 int cnt = 0;
4261 if (len < sizeof(sctp_assoc_t))
4262 return -EINVAL;
4264 if (copy_from_user(&id, optval, sizeof(sctp_assoc_t)))
4265 return -EFAULT;
4268 * For UDP-style sockets, id specifies the association to query.
4269 * If the id field is set to the value '0' then the locally bound
4270 * addresses are returned without regard to any particular
4271 * association.
4273 if (0 == id) {
4274 bp = &sctp_sk(sk)->ep->base.bind_addr;
4275 } else {
4276 asoc = sctp_id2assoc(sk, id);
4277 if (!asoc)
4278 return -EINVAL;
4279 bp = &asoc->base.bind_addr;
4282 /* If the endpoint is bound to 0.0.0.0 or ::0, count the valid
4283 * addresses from the global local address list.
4285 if (sctp_list_single_entry(&bp->address_list)) {
4286 addr = list_entry(bp->address_list.next,
4287 struct sctp_sockaddr_entry, list);
4288 if (sctp_is_any(&addr->a)) {
4289 rcu_read_lock();
4290 list_for_each_entry_rcu(addr,
4291 &sctp_local_addr_list, list) {
4292 if (!addr->valid)
4293 continue;
4295 if ((PF_INET == sk->sk_family) &&
4296 (AF_INET6 == addr->a.sa.sa_family))
4297 continue;
4299 cnt++;
4301 rcu_read_unlock();
4302 } else {
4303 cnt = 1;
4305 goto done;
4308 /* Protection on the bound address list is not needed,
4309 * since in the socket option context we hold the socket lock,
4310 * so there is no way that the bound address list can change.
4312 list_for_each_entry(addr, &bp->address_list, list) {
4313 cnt ++;
4315 done:
4316 return cnt;
4319 /* Helper function that copies local addresses to user and returns the number
4320 * of addresses copied.
4322 static int sctp_copy_laddrs_old(struct sock *sk, __u16 port,
4323 int max_addrs, void *to,
4324 int *bytes_copied)
4326 struct sctp_sockaddr_entry *addr;
4327 union sctp_addr temp;
4328 int cnt = 0;
4329 int addrlen;
4331 rcu_read_lock();
4332 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4333 if (!addr->valid)
4334 continue;
4336 if ((PF_INET == sk->sk_family) &&
4337 (AF_INET6 == addr->a.sa.sa_family))
4338 continue;
4339 memcpy(&temp, &addr->a, sizeof(temp));
4340 if (!temp.v4.sin_port)
4341 temp.v4.sin_port = htons(port);
4343 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4344 &temp);
4345 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4346 memcpy(to, &temp, addrlen);
4348 to += addrlen;
4349 *bytes_copied += addrlen;
4350 cnt ++;
4351 if (cnt >= max_addrs) break;
4353 rcu_read_unlock();
4355 return cnt;
4358 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4359 size_t space_left, int *bytes_copied)
4361 struct sctp_sockaddr_entry *addr;
4362 union sctp_addr temp;
4363 int cnt = 0;
4364 int addrlen;
4366 rcu_read_lock();
4367 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4368 if (!addr->valid)
4369 continue;
4371 if ((PF_INET == sk->sk_family) &&
4372 (AF_INET6 == addr->a.sa.sa_family))
4373 continue;
4374 memcpy(&temp, &addr->a, sizeof(temp));
4375 if (!temp.v4.sin_port)
4376 temp.v4.sin_port = htons(port);
4378 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4379 &temp);
4380 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4381 if (space_left < addrlen) {
4382 cnt = -ENOMEM;
4383 break;
4385 memcpy(to, &temp, addrlen);
4387 to += addrlen;
4388 cnt ++;
4389 space_left -= addrlen;
4390 *bytes_copied += addrlen;
4392 rcu_read_unlock();
4394 return cnt;
4397 /* Old API for getting list of local addresses. Does not work for 32-bit
4398 * programs running on a 64-bit kernel
4400 static int sctp_getsockopt_local_addrs_old(struct sock *sk, int len,
4401 char __user *optval, int __user *optlen)
4403 struct sctp_bind_addr *bp;
4404 struct sctp_association *asoc;
4405 int cnt = 0;
4406 struct sctp_getaddrs_old getaddrs;
4407 struct sctp_sockaddr_entry *addr;
4408 void __user *to;
4409 union sctp_addr temp;
4410 struct sctp_sock *sp = sctp_sk(sk);
4411 int addrlen;
4412 int err = 0;
4413 void *addrs;
4414 void *buf;
4415 int bytes_copied = 0;
4417 if (len < sizeof(struct sctp_getaddrs_old))
4418 return -EINVAL;
4420 len = sizeof(struct sctp_getaddrs_old);
4421 if (copy_from_user(&getaddrs, optval, len))
4422 return -EFAULT;
4424 if (getaddrs.addr_num <= 0) return -EINVAL;
4426 * For UDP-style sockets, id specifies the association to query.
4427 * If the id field is set to the value '0' then the locally bound
4428 * addresses are returned without regard to any particular
4429 * association.
4431 if (0 == getaddrs.assoc_id) {
4432 bp = &sctp_sk(sk)->ep->base.bind_addr;
4433 } else {
4434 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4435 if (!asoc)
4436 return -EINVAL;
4437 bp = &asoc->base.bind_addr;
4440 to = getaddrs.addrs;
4442 /* Allocate space for a local instance of packed array to hold all
4443 * the data. We store addresses here first and then put write them
4444 * to the user in one shot.
4446 addrs = kmalloc(sizeof(union sctp_addr) * getaddrs.addr_num,
4447 GFP_KERNEL);
4448 if (!addrs)
4449 return -ENOMEM;
4451 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4452 * addresses from the global local address list.
4454 if (sctp_list_single_entry(&bp->address_list)) {
4455 addr = list_entry(bp->address_list.next,
4456 struct sctp_sockaddr_entry, list);
4457 if (sctp_is_any(&addr->a)) {
4458 cnt = sctp_copy_laddrs_old(sk, bp->port,
4459 getaddrs.addr_num,
4460 addrs, &bytes_copied);
4461 goto copy_getaddrs;
4465 buf = addrs;
4466 /* Protection on the bound address list is not needed since
4467 * in the socket option context we hold a socket lock and
4468 * thus the bound address list can't change.
4470 list_for_each_entry(addr, &bp->address_list, list) {
4471 memcpy(&temp, &addr->a, sizeof(temp));
4472 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4473 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4474 memcpy(buf, &temp, addrlen);
4475 buf += addrlen;
4476 bytes_copied += addrlen;
4477 cnt ++;
4478 if (cnt >= getaddrs.addr_num) break;
4481 copy_getaddrs:
4482 /* copy the entire address list into the user provided space */
4483 if (copy_to_user(to, addrs, bytes_copied)) {
4484 err = -EFAULT;
4485 goto error;
4488 /* copy the leading structure back to user */
4489 getaddrs.addr_num = cnt;
4490 if (copy_to_user(optval, &getaddrs, len))
4491 err = -EFAULT;
4493 error:
4494 kfree(addrs);
4495 return err;
4498 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4499 char __user *optval, int __user *optlen)
4501 struct sctp_bind_addr *bp;
4502 struct sctp_association *asoc;
4503 int cnt = 0;
4504 struct sctp_getaddrs getaddrs;
4505 struct sctp_sockaddr_entry *addr;
4506 void __user *to;
4507 union sctp_addr temp;
4508 struct sctp_sock *sp = sctp_sk(sk);
4509 int addrlen;
4510 int err = 0;
4511 size_t space_left;
4512 int bytes_copied = 0;
4513 void *addrs;
4514 void *buf;
4516 if (len < sizeof(struct sctp_getaddrs))
4517 return -EINVAL;
4519 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4520 return -EFAULT;
4523 * For UDP-style sockets, id specifies the association to query.
4524 * If the id field is set to the value '0' then the locally bound
4525 * addresses are returned without regard to any particular
4526 * association.
4528 if (0 == getaddrs.assoc_id) {
4529 bp = &sctp_sk(sk)->ep->base.bind_addr;
4530 } else {
4531 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4532 if (!asoc)
4533 return -EINVAL;
4534 bp = &asoc->base.bind_addr;
4537 to = optval + offsetof(struct sctp_getaddrs,addrs);
4538 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4540 addrs = kmalloc(space_left, GFP_KERNEL);
4541 if (!addrs)
4542 return -ENOMEM;
4544 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4545 * addresses from the global local address list.
4547 if (sctp_list_single_entry(&bp->address_list)) {
4548 addr = list_entry(bp->address_list.next,
4549 struct sctp_sockaddr_entry, list);
4550 if (sctp_is_any(&addr->a)) {
4551 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4552 space_left, &bytes_copied);
4553 if (cnt < 0) {
4554 err = cnt;
4555 goto out;
4557 goto copy_getaddrs;
4561 buf = addrs;
4562 /* Protection on the bound address list is not needed since
4563 * in the socket option context we hold a socket lock and
4564 * thus the bound address list can't change.
4566 list_for_each_entry(addr, &bp->address_list, list) {
4567 memcpy(&temp, &addr->a, sizeof(temp));
4568 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4569 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4570 if (space_left < addrlen) {
4571 err = -ENOMEM; /*fixme: right error?*/
4572 goto out;
4574 memcpy(buf, &temp, addrlen);
4575 buf += addrlen;
4576 bytes_copied += addrlen;
4577 cnt ++;
4578 space_left -= addrlen;
4581 copy_getaddrs:
4582 if (copy_to_user(to, addrs, bytes_copied)) {
4583 err = -EFAULT;
4584 goto out;
4586 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4587 err = -EFAULT;
4588 goto out;
4590 if (put_user(bytes_copied, optlen))
4591 err = -EFAULT;
4592 out:
4593 kfree(addrs);
4594 return err;
4597 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4599 * Requests that the local SCTP stack use the enclosed peer address as
4600 * the association primary. The enclosed address must be one of the
4601 * association peer's addresses.
4603 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4604 char __user *optval, int __user *optlen)
4606 struct sctp_prim prim;
4607 struct sctp_association *asoc;
4608 struct sctp_sock *sp = sctp_sk(sk);
4610 if (len < sizeof(struct sctp_prim))
4611 return -EINVAL;
4613 len = sizeof(struct sctp_prim);
4615 if (copy_from_user(&prim, optval, len))
4616 return -EFAULT;
4618 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4619 if (!asoc)
4620 return -EINVAL;
4622 if (!asoc->peer.primary_path)
4623 return -ENOTCONN;
4625 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4626 asoc->peer.primary_path->af_specific->sockaddr_len);
4628 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4629 (union sctp_addr *)&prim.ssp_addr);
4631 if (put_user(len, optlen))
4632 return -EFAULT;
4633 if (copy_to_user(optval, &prim, len))
4634 return -EFAULT;
4636 return 0;
4640 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4642 * Requests that the local endpoint set the specified Adaptation Layer
4643 * Indication parameter for all future INIT and INIT-ACK exchanges.
4645 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4646 char __user *optval, int __user *optlen)
4648 struct sctp_setadaptation adaptation;
4650 if (len < sizeof(struct sctp_setadaptation))
4651 return -EINVAL;
4653 len = sizeof(struct sctp_setadaptation);
4655 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4657 if (put_user(len, optlen))
4658 return -EFAULT;
4659 if (copy_to_user(optval, &adaptation, len))
4660 return -EFAULT;
4662 return 0;
4667 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4669 * Applications that wish to use the sendto() system call may wish to
4670 * specify a default set of parameters that would normally be supplied
4671 * through the inclusion of ancillary data. This socket option allows
4672 * such an application to set the default sctp_sndrcvinfo structure.
4675 * The application that wishes to use this socket option simply passes
4676 * in to this call the sctp_sndrcvinfo structure defined in Section
4677 * 5.2.2) The input parameters accepted by this call include
4678 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4679 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4680 * to this call if the caller is using the UDP model.
4682 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4684 static int sctp_getsockopt_default_send_param(struct sock *sk,
4685 int len, char __user *optval,
4686 int __user *optlen)
4688 struct sctp_sndrcvinfo info;
4689 struct sctp_association *asoc;
4690 struct sctp_sock *sp = sctp_sk(sk);
4692 if (len < sizeof(struct sctp_sndrcvinfo))
4693 return -EINVAL;
4695 len = sizeof(struct sctp_sndrcvinfo);
4697 if (copy_from_user(&info, optval, len))
4698 return -EFAULT;
4700 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4701 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4702 return -EINVAL;
4704 if (asoc) {
4705 info.sinfo_stream = asoc->default_stream;
4706 info.sinfo_flags = asoc->default_flags;
4707 info.sinfo_ppid = asoc->default_ppid;
4708 info.sinfo_context = asoc->default_context;
4709 info.sinfo_timetolive = asoc->default_timetolive;
4710 } else {
4711 info.sinfo_stream = sp->default_stream;
4712 info.sinfo_flags = sp->default_flags;
4713 info.sinfo_ppid = sp->default_ppid;
4714 info.sinfo_context = sp->default_context;
4715 info.sinfo_timetolive = sp->default_timetolive;
4718 if (put_user(len, optlen))
4719 return -EFAULT;
4720 if (copy_to_user(optval, &info, len))
4721 return -EFAULT;
4723 return 0;
4728 * 7.1.5 SCTP_NODELAY
4730 * Turn on/off any Nagle-like algorithm. This means that packets are
4731 * generally sent as soon as possible and no unnecessary delays are
4732 * introduced, at the cost of more packets in the network. Expects an
4733 * integer boolean flag.
4736 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4737 char __user *optval, int __user *optlen)
4739 int val;
4741 if (len < sizeof(int))
4742 return -EINVAL;
4744 len = sizeof(int);
4745 val = (sctp_sk(sk)->nodelay == 1);
4746 if (put_user(len, optlen))
4747 return -EFAULT;
4748 if (copy_to_user(optval, &val, len))
4749 return -EFAULT;
4750 return 0;
4755 * 7.1.1 SCTP_RTOINFO
4757 * The protocol parameters used to initialize and bound retransmission
4758 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4759 * and modify these parameters.
4760 * All parameters are time values, in milliseconds. A value of 0, when
4761 * modifying the parameters, indicates that the current value should not
4762 * be changed.
4765 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4766 char __user *optval,
4767 int __user *optlen) {
4768 struct sctp_rtoinfo rtoinfo;
4769 struct sctp_association *asoc;
4771 if (len < sizeof (struct sctp_rtoinfo))
4772 return -EINVAL;
4774 len = sizeof(struct sctp_rtoinfo);
4776 if (copy_from_user(&rtoinfo, optval, len))
4777 return -EFAULT;
4779 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4781 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4782 return -EINVAL;
4784 /* Values corresponding to the specific association. */
4785 if (asoc) {
4786 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4787 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4788 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4789 } else {
4790 /* Values corresponding to the endpoint. */
4791 struct sctp_sock *sp = sctp_sk(sk);
4793 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4794 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4795 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4798 if (put_user(len, optlen))
4799 return -EFAULT;
4801 if (copy_to_user(optval, &rtoinfo, len))
4802 return -EFAULT;
4804 return 0;
4809 * 7.1.2 SCTP_ASSOCINFO
4811 * This option is used to tune the maximum retransmission attempts
4812 * of the association.
4813 * Returns an error if the new association retransmission value is
4814 * greater than the sum of the retransmission value of the peer.
4815 * See [SCTP] for more information.
4818 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4819 char __user *optval,
4820 int __user *optlen)
4823 struct sctp_assocparams assocparams;
4824 struct sctp_association *asoc;
4825 struct list_head *pos;
4826 int cnt = 0;
4828 if (len < sizeof (struct sctp_assocparams))
4829 return -EINVAL;
4831 len = sizeof(struct sctp_assocparams);
4833 if (copy_from_user(&assocparams, optval, len))
4834 return -EFAULT;
4836 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4838 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4839 return -EINVAL;
4841 /* Values correspoinding to the specific association */
4842 if (asoc) {
4843 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4844 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4845 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4846 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4847 * 1000) +
4848 (asoc->cookie_life.tv_usec
4849 / 1000);
4851 list_for_each(pos, &asoc->peer.transport_addr_list) {
4852 cnt ++;
4855 assocparams.sasoc_number_peer_destinations = cnt;
4856 } else {
4857 /* Values corresponding to the endpoint */
4858 struct sctp_sock *sp = sctp_sk(sk);
4860 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4861 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4862 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4863 assocparams.sasoc_cookie_life =
4864 sp->assocparams.sasoc_cookie_life;
4865 assocparams.sasoc_number_peer_destinations =
4866 sp->assocparams.
4867 sasoc_number_peer_destinations;
4870 if (put_user(len, optlen))
4871 return -EFAULT;
4873 if (copy_to_user(optval, &assocparams, len))
4874 return -EFAULT;
4876 return 0;
4880 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4882 * This socket option is a boolean flag which turns on or off mapped V4
4883 * addresses. If this option is turned on and the socket is type
4884 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4885 * If this option is turned off, then no mapping will be done of V4
4886 * addresses and a user will receive both PF_INET6 and PF_INET type
4887 * addresses on the socket.
4889 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4890 char __user *optval, int __user *optlen)
4892 int val;
4893 struct sctp_sock *sp = sctp_sk(sk);
4895 if (len < sizeof(int))
4896 return -EINVAL;
4898 len = sizeof(int);
4899 val = sp->v4mapped;
4900 if (put_user(len, optlen))
4901 return -EFAULT;
4902 if (copy_to_user(optval, &val, len))
4903 return -EFAULT;
4905 return 0;
4909 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4910 * (chapter and verse is quoted at sctp_setsockopt_context())
4912 static int sctp_getsockopt_context(struct sock *sk, int len,
4913 char __user *optval, int __user *optlen)
4915 struct sctp_assoc_value params;
4916 struct sctp_sock *sp;
4917 struct sctp_association *asoc;
4919 if (len < sizeof(struct sctp_assoc_value))
4920 return -EINVAL;
4922 len = sizeof(struct sctp_assoc_value);
4924 if (copy_from_user(&params, optval, len))
4925 return -EFAULT;
4927 sp = sctp_sk(sk);
4929 if (params.assoc_id != 0) {
4930 asoc = sctp_id2assoc(sk, params.assoc_id);
4931 if (!asoc)
4932 return -EINVAL;
4933 params.assoc_value = asoc->default_rcv_context;
4934 } else {
4935 params.assoc_value = sp->default_rcv_context;
4938 if (put_user(len, optlen))
4939 return -EFAULT;
4940 if (copy_to_user(optval, &params, len))
4941 return -EFAULT;
4943 return 0;
4947 * 7.1.17 Set the maximum fragrmentation size (SCTP_MAXSEG)
4949 * This socket option specifies the maximum size to put in any outgoing
4950 * SCTP chunk. If a message is larger than this size it will be
4951 * fragmented by SCTP into the specified size. Note that the underlying
4952 * SCTP implementation may fragment into smaller sized chunks when the
4953 * PMTU of the underlying association is smaller than the value set by
4954 * the user.
4956 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4957 char __user *optval, int __user *optlen)
4959 int val;
4961 if (len < sizeof(int))
4962 return -EINVAL;
4964 len = sizeof(int);
4966 val = sctp_sk(sk)->user_frag;
4967 if (put_user(len, optlen))
4968 return -EFAULT;
4969 if (copy_to_user(optval, &val, len))
4970 return -EFAULT;
4972 return 0;
4976 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
4977 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
4979 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
4980 char __user *optval, int __user *optlen)
4982 int val;
4984 if (len < sizeof(int))
4985 return -EINVAL;
4987 len = sizeof(int);
4989 val = sctp_sk(sk)->frag_interleave;
4990 if (put_user(len, optlen))
4991 return -EFAULT;
4992 if (copy_to_user(optval, &val, len))
4993 return -EFAULT;
4995 return 0;
4999 * 7.1.25. Set or Get the sctp partial delivery point
5000 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5002 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5003 char __user *optval,
5004 int __user *optlen)
5006 u32 val;
5008 if (len < sizeof(u32))
5009 return -EINVAL;
5011 len = sizeof(u32);
5013 val = sctp_sk(sk)->pd_point;
5014 if (put_user(len, optlen))
5015 return -EFAULT;
5016 if (copy_to_user(optval, &val, len))
5017 return -EFAULT;
5019 return -ENOTSUPP;
5023 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5024 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5026 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5027 char __user *optval,
5028 int __user *optlen)
5030 struct sctp_assoc_value params;
5031 struct sctp_sock *sp;
5032 struct sctp_association *asoc;
5034 if (len < sizeof(int))
5035 return -EINVAL;
5037 if (len == sizeof(int)) {
5038 printk(KERN_WARNING
5039 "SCTP: Use of int in max_burst socket option deprecated\n");
5040 printk(KERN_WARNING
5041 "SCTP: Use struct sctp_assoc_value instead\n");
5042 params.assoc_id = 0;
5043 } else if (len == sizeof (struct sctp_assoc_value)) {
5044 if (copy_from_user(&params, optval, len))
5045 return -EFAULT;
5046 } else
5047 return -EINVAL;
5049 sp = sctp_sk(sk);
5051 if (params.assoc_id != 0) {
5052 asoc = sctp_id2assoc(sk, params.assoc_id);
5053 if (!asoc)
5054 return -EINVAL;
5055 params.assoc_value = asoc->max_burst;
5056 } else
5057 params.assoc_value = sp->max_burst;
5059 if (len == sizeof(int)) {
5060 if (copy_to_user(optval, &params.assoc_value, len))
5061 return -EFAULT;
5062 } else {
5063 if (copy_to_user(optval, &params, len))
5064 return -EFAULT;
5067 return 0;
5071 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5072 char __user *optval, int __user *optlen)
5074 struct sctp_hmac_algo_param *hmacs;
5075 __u16 param_len;
5077 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5078 param_len = ntohs(hmacs->param_hdr.length);
5080 if (len < param_len)
5081 return -EINVAL;
5082 if (put_user(len, optlen))
5083 return -EFAULT;
5084 if (copy_to_user(optval, hmacs->hmac_ids, len))
5085 return -EFAULT;
5087 return 0;
5090 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5091 char __user *optval, int __user *optlen)
5093 struct sctp_authkeyid val;
5094 struct sctp_association *asoc;
5096 if (len < sizeof(struct sctp_authkeyid))
5097 return -EINVAL;
5098 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5099 return -EFAULT;
5101 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5102 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5103 return -EINVAL;
5105 if (asoc)
5106 val.scact_keynumber = asoc->active_key_id;
5107 else
5108 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5110 return 0;
5113 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5114 char __user *optval, int __user *optlen)
5116 struct sctp_authchunks __user *p = (void __user *)optval;
5117 struct sctp_authchunks val;
5118 struct sctp_association *asoc;
5119 struct sctp_chunks_param *ch;
5120 u32 num_chunks;
5121 char __user *to;
5123 if (len <= sizeof(struct sctp_authchunks))
5124 return -EINVAL;
5126 if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5127 return -EFAULT;
5129 to = p->gauth_chunks;
5130 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5131 if (!asoc)
5132 return -EINVAL;
5134 ch = asoc->peer.peer_chunks;
5136 /* See if the user provided enough room for all the data */
5137 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5138 if (len < num_chunks)
5139 return -EINVAL;
5141 len = num_chunks;
5142 if (put_user(len, optlen))
5143 return -EFAULT;
5144 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5145 return -EFAULT;
5146 if (copy_to_user(to, ch->chunks, len))
5147 return -EFAULT;
5149 return 0;
5152 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5153 char __user *optval, int __user *optlen)
5155 struct sctp_authchunks __user *p = (void __user *)optval;
5156 struct sctp_authchunks val;
5157 struct sctp_association *asoc;
5158 struct sctp_chunks_param *ch;
5159 u32 num_chunks;
5160 char __user *to;
5162 if (len <= sizeof(struct sctp_authchunks))
5163 return -EINVAL;
5165 if (copy_from_user(&val, p, sizeof(struct sctp_authchunks)))
5166 return -EFAULT;
5168 to = p->gauth_chunks;
5169 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5170 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5171 return -EINVAL;
5173 if (asoc)
5174 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5175 else
5176 ch = sctp_sk(sk)->ep->auth_chunk_list;
5178 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5179 if (len < num_chunks)
5180 return -EINVAL;
5182 len = num_chunks;
5183 if (put_user(len, optlen))
5184 return -EFAULT;
5185 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5186 return -EFAULT;
5187 if (copy_to_user(to, ch->chunks, len))
5188 return -EFAULT;
5190 return 0;
5193 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5194 char __user *optval, int __user *optlen)
5196 int retval = 0;
5197 int len;
5199 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5200 sk, optname);
5202 /* I can hardly begin to describe how wrong this is. This is
5203 * so broken as to be worse than useless. The API draft
5204 * REALLY is NOT helpful here... I am not convinced that the
5205 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5206 * are at all well-founded.
5208 if (level != SOL_SCTP) {
5209 struct sctp_af *af = sctp_sk(sk)->pf->af;
5211 retval = af->getsockopt(sk, level, optname, optval, optlen);
5212 return retval;
5215 if (get_user(len, optlen))
5216 return -EFAULT;
5218 sctp_lock_sock(sk);
5220 switch (optname) {
5221 case SCTP_STATUS:
5222 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5223 break;
5224 case SCTP_DISABLE_FRAGMENTS:
5225 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5226 optlen);
5227 break;
5228 case SCTP_EVENTS:
5229 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5230 break;
5231 case SCTP_AUTOCLOSE:
5232 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5233 break;
5234 case SCTP_SOCKOPT_PEELOFF:
5235 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5236 break;
5237 case SCTP_PEER_ADDR_PARAMS:
5238 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5239 optlen);
5240 break;
5241 case SCTP_DELAYED_ACK_TIME:
5242 retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
5243 optlen);
5244 break;
5245 case SCTP_INITMSG:
5246 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5247 break;
5248 case SCTP_GET_PEER_ADDRS_NUM_OLD:
5249 retval = sctp_getsockopt_peer_addrs_num_old(sk, len, optval,
5250 optlen);
5251 break;
5252 case SCTP_GET_LOCAL_ADDRS_NUM_OLD:
5253 retval = sctp_getsockopt_local_addrs_num_old(sk, len, optval,
5254 optlen);
5255 break;
5256 case SCTP_GET_PEER_ADDRS_OLD:
5257 retval = sctp_getsockopt_peer_addrs_old(sk, len, optval,
5258 optlen);
5259 break;
5260 case SCTP_GET_LOCAL_ADDRS_OLD:
5261 retval = sctp_getsockopt_local_addrs_old(sk, len, optval,
5262 optlen);
5263 break;
5264 case SCTP_GET_PEER_ADDRS:
5265 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5266 optlen);
5267 break;
5268 case SCTP_GET_LOCAL_ADDRS:
5269 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5270 optlen);
5271 break;
5272 case SCTP_DEFAULT_SEND_PARAM:
5273 retval = sctp_getsockopt_default_send_param(sk, len,
5274 optval, optlen);
5275 break;
5276 case SCTP_PRIMARY_ADDR:
5277 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5278 break;
5279 case SCTP_NODELAY:
5280 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5281 break;
5282 case SCTP_RTOINFO:
5283 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5284 break;
5285 case SCTP_ASSOCINFO:
5286 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5287 break;
5288 case SCTP_I_WANT_MAPPED_V4_ADDR:
5289 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5290 break;
5291 case SCTP_MAXSEG:
5292 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5293 break;
5294 case SCTP_GET_PEER_ADDR_INFO:
5295 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5296 optlen);
5297 break;
5298 case SCTP_ADAPTATION_LAYER:
5299 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5300 optlen);
5301 break;
5302 case SCTP_CONTEXT:
5303 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5304 break;
5305 case SCTP_FRAGMENT_INTERLEAVE:
5306 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5307 optlen);
5308 break;
5309 case SCTP_PARTIAL_DELIVERY_POINT:
5310 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5311 optlen);
5312 break;
5313 case SCTP_MAX_BURST:
5314 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5315 break;
5316 case SCTP_AUTH_KEY:
5317 case SCTP_AUTH_CHUNK:
5318 case SCTP_AUTH_DELETE_KEY:
5319 retval = -EOPNOTSUPP;
5320 break;
5321 case SCTP_HMAC_IDENT:
5322 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5323 break;
5324 case SCTP_AUTH_ACTIVE_KEY:
5325 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5326 break;
5327 case SCTP_PEER_AUTH_CHUNKS:
5328 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5329 optlen);
5330 break;
5331 case SCTP_LOCAL_AUTH_CHUNKS:
5332 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5333 optlen);
5334 break;
5335 default:
5336 retval = -ENOPROTOOPT;
5337 break;
5340 sctp_release_sock(sk);
5341 return retval;
5344 static void sctp_hash(struct sock *sk)
5346 /* STUB */
5349 static void sctp_unhash(struct sock *sk)
5351 /* STUB */
5354 /* Check if port is acceptable. Possibly find first available port.
5356 * The port hash table (contained in the 'global' SCTP protocol storage
5357 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5358 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5359 * list (the list number is the port number hashed out, so as you
5360 * would expect from a hash function, all the ports in a given list have
5361 * such a number that hashes out to the same list number; you were
5362 * expecting that, right?); so each list has a set of ports, with a
5363 * link to the socket (struct sock) that uses it, the port number and
5364 * a fastreuse flag (FIXME: NPI ipg).
5366 static struct sctp_bind_bucket *sctp_bucket_create(
5367 struct sctp_bind_hashbucket *head, unsigned short snum);
5369 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5371 struct sctp_bind_hashbucket *head; /* hash list */
5372 struct sctp_bind_bucket *pp; /* hash list port iterator */
5373 struct hlist_node *node;
5374 unsigned short snum;
5375 int ret;
5377 snum = ntohs(addr->v4.sin_port);
5379 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5380 sctp_local_bh_disable();
5382 if (snum == 0) {
5383 /* Search for an available port. */
5384 int low, high, remaining, index;
5385 unsigned int rover;
5387 inet_get_local_port_range(&low, &high);
5388 remaining = (high - low) + 1;
5389 rover = net_random() % remaining + low;
5391 do {
5392 rover++;
5393 if ((rover < low) || (rover > high))
5394 rover = low;
5395 index = sctp_phashfn(rover);
5396 head = &sctp_port_hashtable[index];
5397 sctp_spin_lock(&head->lock);
5398 sctp_for_each_hentry(pp, node, &head->chain)
5399 if (pp->port == rover)
5400 goto next;
5401 break;
5402 next:
5403 sctp_spin_unlock(&head->lock);
5404 } while (--remaining > 0);
5406 /* Exhausted local port range during search? */
5407 ret = 1;
5408 if (remaining <= 0)
5409 goto fail;
5411 /* OK, here is the one we will use. HEAD (the port
5412 * hash table list entry) is non-NULL and we hold it's
5413 * mutex.
5415 snum = rover;
5416 } else {
5417 /* We are given an specific port number; we verify
5418 * that it is not being used. If it is used, we will
5419 * exahust the search in the hash list corresponding
5420 * to the port number (snum) - we detect that with the
5421 * port iterator, pp being NULL.
5423 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5424 sctp_spin_lock(&head->lock);
5425 sctp_for_each_hentry(pp, node, &head->chain) {
5426 if (pp->port == snum)
5427 goto pp_found;
5430 pp = NULL;
5431 goto pp_not_found;
5432 pp_found:
5433 if (!hlist_empty(&pp->owner)) {
5434 /* We had a port hash table hit - there is an
5435 * available port (pp != NULL) and it is being
5436 * used by other socket (pp->owner not empty); that other
5437 * socket is going to be sk2.
5439 int reuse = sk->sk_reuse;
5440 struct sock *sk2;
5441 struct hlist_node *node;
5443 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5444 if (pp->fastreuse && sk->sk_reuse &&
5445 sk->sk_state != SCTP_SS_LISTENING)
5446 goto success;
5448 /* Run through the list of sockets bound to the port
5449 * (pp->port) [via the pointers bind_next and
5450 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5451 * we get the endpoint they describe and run through
5452 * the endpoint's list of IP (v4 or v6) addresses,
5453 * comparing each of the addresses with the address of
5454 * the socket sk. If we find a match, then that means
5455 * that this port/socket (sk) combination are already
5456 * in an endpoint.
5458 sk_for_each_bound(sk2, node, &pp->owner) {
5459 struct sctp_endpoint *ep2;
5460 ep2 = sctp_sk(sk2)->ep;
5462 if (reuse && sk2->sk_reuse &&
5463 sk2->sk_state != SCTP_SS_LISTENING)
5464 continue;
5466 if (sctp_bind_addr_match(&ep2->base.bind_addr, addr,
5467 sctp_sk(sk))) {
5468 ret = (long)sk2;
5469 goto fail_unlock;
5472 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5474 pp_not_found:
5475 /* If there was a hash table miss, create a new port. */
5476 ret = 1;
5477 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5478 goto fail_unlock;
5480 /* In either case (hit or miss), make sure fastreuse is 1 only
5481 * if sk->sk_reuse is too (that is, if the caller requested
5482 * SO_REUSEADDR on this socket -sk-).
5484 if (hlist_empty(&pp->owner)) {
5485 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5486 pp->fastreuse = 1;
5487 else
5488 pp->fastreuse = 0;
5489 } else if (pp->fastreuse &&
5490 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5491 pp->fastreuse = 0;
5493 /* We are set, so fill up all the data in the hash table
5494 * entry, tie the socket list information with the rest of the
5495 * sockets FIXME: Blurry, NPI (ipg).
5497 success:
5498 if (!sctp_sk(sk)->bind_hash) {
5499 inet_sk(sk)->num = snum;
5500 sk_add_bind_node(sk, &pp->owner);
5501 sctp_sk(sk)->bind_hash = pp;
5503 ret = 0;
5505 fail_unlock:
5506 sctp_spin_unlock(&head->lock);
5508 fail:
5509 sctp_local_bh_enable();
5510 return ret;
5513 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5514 * port is requested.
5516 static int sctp_get_port(struct sock *sk, unsigned short snum)
5518 long ret;
5519 union sctp_addr addr;
5520 struct sctp_af *af = sctp_sk(sk)->pf->af;
5522 /* Set up a dummy address struct from the sk. */
5523 af->from_sk(&addr, sk);
5524 addr.v4.sin_port = htons(snum);
5526 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5527 ret = sctp_get_port_local(sk, &addr);
5529 return (ret ? 1 : 0);
5533 * 3.1.3 listen() - UDP Style Syntax
5535 * By default, new associations are not accepted for UDP style sockets.
5536 * An application uses listen() to mark a socket as being able to
5537 * accept new associations.
5539 SCTP_STATIC int sctp_seqpacket_listen(struct sock *sk, int backlog)
5541 struct sctp_sock *sp = sctp_sk(sk);
5542 struct sctp_endpoint *ep = sp->ep;
5544 /* Only UDP style sockets that are not peeled off are allowed to
5545 * listen().
5547 if (!sctp_style(sk, UDP))
5548 return -EINVAL;
5550 /* If backlog is zero, disable listening. */
5551 if (!backlog) {
5552 if (sctp_sstate(sk, CLOSED))
5553 return 0;
5555 sctp_unhash_endpoint(ep);
5556 sk->sk_state = SCTP_SS_CLOSED;
5557 return 0;
5560 /* Return if we are already listening. */
5561 if (sctp_sstate(sk, LISTENING))
5562 return 0;
5565 * If a bind() or sctp_bindx() is not called prior to a listen()
5566 * call that allows new associations to be accepted, the system
5567 * picks an ephemeral port and will choose an address set equivalent
5568 * to binding with a wildcard address.
5570 * This is not currently spelled out in the SCTP sockets
5571 * extensions draft, but follows the practice as seen in TCP
5572 * sockets.
5574 * Additionally, turn off fastreuse flag since we are not listening
5576 sk->sk_state = SCTP_SS_LISTENING;
5577 if (!ep->base.bind_addr.port) {
5578 if (sctp_autobind(sk))
5579 return -EAGAIN;
5580 } else
5581 sctp_sk(sk)->bind_hash->fastreuse = 0;
5583 sctp_hash_endpoint(ep);
5584 return 0;
5588 * 4.1.3 listen() - TCP Style Syntax
5590 * Applications uses listen() to ready the SCTP endpoint for accepting
5591 * inbound associations.
5593 SCTP_STATIC int sctp_stream_listen(struct sock *sk, int backlog)
5595 struct sctp_sock *sp = sctp_sk(sk);
5596 struct sctp_endpoint *ep = sp->ep;
5598 /* If backlog is zero, disable listening. */
5599 if (!backlog) {
5600 if (sctp_sstate(sk, CLOSED))
5601 return 0;
5603 sctp_unhash_endpoint(ep);
5604 sk->sk_state = SCTP_SS_CLOSED;
5605 return 0;
5608 if (sctp_sstate(sk, LISTENING))
5609 return 0;
5612 * If a bind() or sctp_bindx() is not called prior to a listen()
5613 * call that allows new associations to be accepted, the system
5614 * picks an ephemeral port and will choose an address set equivalent
5615 * to binding with a wildcard address.
5617 * This is not currently spelled out in the SCTP sockets
5618 * extensions draft, but follows the practice as seen in TCP
5619 * sockets.
5621 sk->sk_state = SCTP_SS_LISTENING;
5622 if (!ep->base.bind_addr.port) {
5623 if (sctp_autobind(sk))
5624 return -EAGAIN;
5625 } else
5626 sctp_sk(sk)->bind_hash->fastreuse = 0;
5628 sk->sk_max_ack_backlog = backlog;
5629 sctp_hash_endpoint(ep);
5630 return 0;
5634 * Move a socket to LISTENING state.
5636 int sctp_inet_listen(struct socket *sock, int backlog)
5638 struct sock *sk = sock->sk;
5639 struct crypto_hash *tfm = NULL;
5640 int err = -EINVAL;
5642 if (unlikely(backlog < 0))
5643 goto out;
5645 sctp_lock_sock(sk);
5647 if (sock->state != SS_UNCONNECTED)
5648 goto out;
5650 /* Allocate HMAC for generating cookie. */
5651 if (sctp_hmac_alg) {
5652 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5653 if (IS_ERR(tfm)) {
5654 if (net_ratelimit()) {
5655 printk(KERN_INFO
5656 "SCTP: failed to load transform for %s: %ld\n",
5657 sctp_hmac_alg, PTR_ERR(tfm));
5659 err = -ENOSYS;
5660 goto out;
5664 switch (sock->type) {
5665 case SOCK_SEQPACKET:
5666 err = sctp_seqpacket_listen(sk, backlog);
5667 break;
5668 case SOCK_STREAM:
5669 err = sctp_stream_listen(sk, backlog);
5670 break;
5671 default:
5672 break;
5675 if (err)
5676 goto cleanup;
5678 /* Store away the transform reference. */
5679 sctp_sk(sk)->hmac = tfm;
5680 out:
5681 sctp_release_sock(sk);
5682 return err;
5683 cleanup:
5684 crypto_free_hash(tfm);
5685 goto out;
5689 * This function is done by modeling the current datagram_poll() and the
5690 * tcp_poll(). Note that, based on these implementations, we don't
5691 * lock the socket in this function, even though it seems that,
5692 * ideally, locking or some other mechanisms can be used to ensure
5693 * the integrity of the counters (sndbuf and wmem_alloc) used
5694 * in this place. We assume that we don't need locks either until proven
5695 * otherwise.
5697 * Another thing to note is that we include the Async I/O support
5698 * here, again, by modeling the current TCP/UDP code. We don't have
5699 * a good way to test with it yet.
5701 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5703 struct sock *sk = sock->sk;
5704 struct sctp_sock *sp = sctp_sk(sk);
5705 unsigned int mask;
5707 poll_wait(file, sk->sk_sleep, wait);
5709 /* A TCP-style listening socket becomes readable when the accept queue
5710 * is not empty.
5712 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5713 return (!list_empty(&sp->ep->asocs)) ?
5714 (POLLIN | POLLRDNORM) : 0;
5716 mask = 0;
5718 /* Is there any exceptional events? */
5719 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5720 mask |= POLLERR;
5721 if (sk->sk_shutdown & RCV_SHUTDOWN)
5722 mask |= POLLRDHUP;
5723 if (sk->sk_shutdown == SHUTDOWN_MASK)
5724 mask |= POLLHUP;
5726 /* Is it readable? Reconsider this code with TCP-style support. */
5727 if (!skb_queue_empty(&sk->sk_receive_queue) ||
5728 (sk->sk_shutdown & RCV_SHUTDOWN))
5729 mask |= POLLIN | POLLRDNORM;
5731 /* The association is either gone or not ready. */
5732 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5733 return mask;
5735 /* Is it writable? */
5736 if (sctp_writeable(sk)) {
5737 mask |= POLLOUT | POLLWRNORM;
5738 } else {
5739 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5741 * Since the socket is not locked, the buffer
5742 * might be made available after the writeable check and
5743 * before the bit is set. This could cause a lost I/O
5744 * signal. tcp_poll() has a race breaker for this race
5745 * condition. Based on their implementation, we put
5746 * in the following code to cover it as well.
5748 if (sctp_writeable(sk))
5749 mask |= POLLOUT | POLLWRNORM;
5751 return mask;
5754 /********************************************************************
5755 * 2nd Level Abstractions
5756 ********************************************************************/
5758 static struct sctp_bind_bucket *sctp_bucket_create(
5759 struct sctp_bind_hashbucket *head, unsigned short snum)
5761 struct sctp_bind_bucket *pp;
5763 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5764 SCTP_DBG_OBJCNT_INC(bind_bucket);
5765 if (pp) {
5766 pp->port = snum;
5767 pp->fastreuse = 0;
5768 INIT_HLIST_HEAD(&pp->owner);
5769 hlist_add_head(&pp->node, &head->chain);
5771 return pp;
5774 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5775 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5777 if (pp && hlist_empty(&pp->owner)) {
5778 __hlist_del(&pp->node);
5779 kmem_cache_free(sctp_bucket_cachep, pp);
5780 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5784 /* Release this socket's reference to a local port. */
5785 static inline void __sctp_put_port(struct sock *sk)
5787 struct sctp_bind_hashbucket *head =
5788 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->num)];
5789 struct sctp_bind_bucket *pp;
5791 sctp_spin_lock(&head->lock);
5792 pp = sctp_sk(sk)->bind_hash;
5793 __sk_del_bind_node(sk);
5794 sctp_sk(sk)->bind_hash = NULL;
5795 inet_sk(sk)->num = 0;
5796 sctp_bucket_destroy(pp);
5797 sctp_spin_unlock(&head->lock);
5800 void sctp_put_port(struct sock *sk)
5802 sctp_local_bh_disable();
5803 __sctp_put_port(sk);
5804 sctp_local_bh_enable();
5808 * The system picks an ephemeral port and choose an address set equivalent
5809 * to binding with a wildcard address.
5810 * One of those addresses will be the primary address for the association.
5811 * This automatically enables the multihoming capability of SCTP.
5813 static int sctp_autobind(struct sock *sk)
5815 union sctp_addr autoaddr;
5816 struct sctp_af *af;
5817 __be16 port;
5819 /* Initialize a local sockaddr structure to INADDR_ANY. */
5820 af = sctp_sk(sk)->pf->af;
5822 port = htons(inet_sk(sk)->num);
5823 af->inaddr_any(&autoaddr, port);
5825 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5828 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5830 * From RFC 2292
5831 * 4.2 The cmsghdr Structure *
5833 * When ancillary data is sent or received, any number of ancillary data
5834 * objects can be specified by the msg_control and msg_controllen members of
5835 * the msghdr structure, because each object is preceded by
5836 * a cmsghdr structure defining the object's length (the cmsg_len member).
5837 * Historically Berkeley-derived implementations have passed only one object
5838 * at a time, but this API allows multiple objects to be
5839 * passed in a single call to sendmsg() or recvmsg(). The following example
5840 * shows two ancillary data objects in a control buffer.
5842 * |<--------------------------- msg_controllen -------------------------->|
5843 * | |
5845 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5847 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5848 * | | |
5850 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5852 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5853 * | | | | |
5855 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5856 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5858 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5860 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5864 * msg_control
5865 * points here
5867 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5868 sctp_cmsgs_t *cmsgs)
5870 struct cmsghdr *cmsg;
5872 for (cmsg = CMSG_FIRSTHDR(msg);
5873 cmsg != NULL;
5874 cmsg = CMSG_NXTHDR((struct msghdr*)msg, cmsg)) {
5875 if (!CMSG_OK(msg, cmsg))
5876 return -EINVAL;
5878 /* Should we parse this header or ignore? */
5879 if (cmsg->cmsg_level != IPPROTO_SCTP)
5880 continue;
5882 /* Strictly check lengths following example in SCM code. */
5883 switch (cmsg->cmsg_type) {
5884 case SCTP_INIT:
5885 /* SCTP Socket API Extension
5886 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5888 * This cmsghdr structure provides information for
5889 * initializing new SCTP associations with sendmsg().
5890 * The SCTP_INITMSG socket option uses this same data
5891 * structure. This structure is not used for
5892 * recvmsg().
5894 * cmsg_level cmsg_type cmsg_data[]
5895 * ------------ ------------ ----------------------
5896 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5898 if (cmsg->cmsg_len !=
5899 CMSG_LEN(sizeof(struct sctp_initmsg)))
5900 return -EINVAL;
5901 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5902 break;
5904 case SCTP_SNDRCV:
5905 /* SCTP Socket API Extension
5906 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5908 * This cmsghdr structure specifies SCTP options for
5909 * sendmsg() and describes SCTP header information
5910 * about a received message through recvmsg().
5912 * cmsg_level cmsg_type cmsg_data[]
5913 * ------------ ------------ ----------------------
5914 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5916 if (cmsg->cmsg_len !=
5917 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5918 return -EINVAL;
5920 cmsgs->info =
5921 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5923 /* Minimally, validate the sinfo_flags. */
5924 if (cmsgs->info->sinfo_flags &
5925 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5926 SCTP_ABORT | SCTP_EOF))
5927 return -EINVAL;
5928 break;
5930 default:
5931 return -EINVAL;
5934 return 0;
5938 * Wait for a packet..
5939 * Note: This function is the same function as in core/datagram.c
5940 * with a few modifications to make lksctp work.
5942 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5944 int error;
5945 DEFINE_WAIT(wait);
5947 prepare_to_wait_exclusive(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
5949 /* Socket errors? */
5950 error = sock_error(sk);
5951 if (error)
5952 goto out;
5954 if (!skb_queue_empty(&sk->sk_receive_queue))
5955 goto ready;
5957 /* Socket shut down? */
5958 if (sk->sk_shutdown & RCV_SHUTDOWN)
5959 goto out;
5961 /* Sequenced packets can come disconnected. If so we report the
5962 * problem.
5964 error = -ENOTCONN;
5966 /* Is there a good reason to think that we may receive some data? */
5967 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
5968 goto out;
5970 /* Handle signals. */
5971 if (signal_pending(current))
5972 goto interrupted;
5974 /* Let another process have a go. Since we are going to sleep
5975 * anyway. Note: This may cause odd behaviors if the message
5976 * does not fit in the user's buffer, but this seems to be the
5977 * only way to honor MSG_DONTWAIT realistically.
5979 sctp_release_sock(sk);
5980 *timeo_p = schedule_timeout(*timeo_p);
5981 sctp_lock_sock(sk);
5983 ready:
5984 finish_wait(sk->sk_sleep, &wait);
5985 return 0;
5987 interrupted:
5988 error = sock_intr_errno(*timeo_p);
5990 out:
5991 finish_wait(sk->sk_sleep, &wait);
5992 *err = error;
5993 return error;
5996 /* Receive a datagram.
5997 * Note: This is pretty much the same routine as in core/datagram.c
5998 * with a few changes to make lksctp work.
6000 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6001 int noblock, int *err)
6003 int error;
6004 struct sk_buff *skb;
6005 long timeo;
6007 timeo = sock_rcvtimeo(sk, noblock);
6009 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6010 timeo, MAX_SCHEDULE_TIMEOUT);
6012 do {
6013 /* Again only user level code calls this function,
6014 * so nothing interrupt level
6015 * will suddenly eat the receive_queue.
6017 * Look at current nfs client by the way...
6018 * However, this function was corrent in any case. 8)
6020 if (flags & MSG_PEEK) {
6021 spin_lock_bh(&sk->sk_receive_queue.lock);
6022 skb = skb_peek(&sk->sk_receive_queue);
6023 if (skb)
6024 atomic_inc(&skb->users);
6025 spin_unlock_bh(&sk->sk_receive_queue.lock);
6026 } else {
6027 skb = skb_dequeue(&sk->sk_receive_queue);
6030 if (skb)
6031 return skb;
6033 /* Caller is allowed not to check sk->sk_err before calling. */
6034 error = sock_error(sk);
6035 if (error)
6036 goto no_packet;
6038 if (sk->sk_shutdown & RCV_SHUTDOWN)
6039 break;
6041 /* User doesn't want to wait. */
6042 error = -EAGAIN;
6043 if (!timeo)
6044 goto no_packet;
6045 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6047 return NULL;
6049 no_packet:
6050 *err = error;
6051 return NULL;
6054 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6055 static void __sctp_write_space(struct sctp_association *asoc)
6057 struct sock *sk = asoc->base.sk;
6058 struct socket *sock = sk->sk_socket;
6060 if ((sctp_wspace(asoc) > 0) && sock) {
6061 if (waitqueue_active(&asoc->wait))
6062 wake_up_interruptible(&asoc->wait);
6064 if (sctp_writeable(sk)) {
6065 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
6066 wake_up_interruptible(sk->sk_sleep);
6068 /* Note that we try to include the Async I/O support
6069 * here by modeling from the current TCP/UDP code.
6070 * We have not tested with it yet.
6072 if (sock->fasync_list &&
6073 !(sk->sk_shutdown & SEND_SHUTDOWN))
6074 sock_wake_async(sock,
6075 SOCK_WAKE_SPACE, POLL_OUT);
6080 /* Do accounting for the sndbuf space.
6081 * Decrement the used sndbuf space of the corresponding association by the
6082 * data size which was just transmitted(freed).
6084 static void sctp_wfree(struct sk_buff *skb)
6086 struct sctp_association *asoc;
6087 struct sctp_chunk *chunk;
6088 struct sock *sk;
6090 /* Get the saved chunk pointer. */
6091 chunk = *((struct sctp_chunk **)(skb->cb));
6092 asoc = chunk->asoc;
6093 sk = asoc->base.sk;
6094 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6095 sizeof(struct sk_buff) +
6096 sizeof(struct sctp_chunk);
6098 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6101 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6103 sk->sk_wmem_queued -= skb->truesize;
6104 sk_mem_uncharge(sk, skb->truesize);
6106 sock_wfree(skb);
6107 __sctp_write_space(asoc);
6109 sctp_association_put(asoc);
6112 /* Do accounting for the receive space on the socket.
6113 * Accounting for the association is done in ulpevent.c
6114 * We set this as a destructor for the cloned data skbs so that
6115 * accounting is done at the correct time.
6117 void sctp_sock_rfree(struct sk_buff *skb)
6119 struct sock *sk = skb->sk;
6120 struct sctp_ulpevent *event = sctp_skb2event(skb);
6122 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6125 * Mimic the behavior of sock_rfree
6127 sk_mem_uncharge(sk, event->rmem_len);
6131 /* Helper function to wait for space in the sndbuf. */
6132 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6133 size_t msg_len)
6135 struct sock *sk = asoc->base.sk;
6136 int err = 0;
6137 long current_timeo = *timeo_p;
6138 DEFINE_WAIT(wait);
6140 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6141 asoc, (long)(*timeo_p), msg_len);
6143 /* Increment the association's refcnt. */
6144 sctp_association_hold(asoc);
6146 /* Wait on the association specific sndbuf space. */
6147 for (;;) {
6148 prepare_to_wait_exclusive(&asoc->wait, &wait,
6149 TASK_INTERRUPTIBLE);
6150 if (!*timeo_p)
6151 goto do_nonblock;
6152 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6153 asoc->base.dead)
6154 goto do_error;
6155 if (signal_pending(current))
6156 goto do_interrupted;
6157 if (msg_len <= sctp_wspace(asoc))
6158 break;
6160 /* Let another process have a go. Since we are going
6161 * to sleep anyway.
6163 sctp_release_sock(sk);
6164 current_timeo = schedule_timeout(current_timeo);
6165 BUG_ON(sk != asoc->base.sk);
6166 sctp_lock_sock(sk);
6168 *timeo_p = current_timeo;
6171 out:
6172 finish_wait(&asoc->wait, &wait);
6174 /* Release the association's refcnt. */
6175 sctp_association_put(asoc);
6177 return err;
6179 do_error:
6180 err = -EPIPE;
6181 goto out;
6183 do_interrupted:
6184 err = sock_intr_errno(*timeo_p);
6185 goto out;
6187 do_nonblock:
6188 err = -EAGAIN;
6189 goto out;
6192 /* If socket sndbuf has changed, wake up all per association waiters. */
6193 void sctp_write_space(struct sock *sk)
6195 struct sctp_association *asoc;
6196 struct list_head *pos;
6198 /* Wake up the tasks in each wait queue. */
6199 list_for_each(pos, &((sctp_sk(sk))->ep->asocs)) {
6200 asoc = list_entry(pos, struct sctp_association, asocs);
6201 __sctp_write_space(asoc);
6205 /* Is there any sndbuf space available on the socket?
6207 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6208 * associations on the same socket. For a UDP-style socket with
6209 * multiple associations, it is possible for it to be "unwriteable"
6210 * prematurely. I assume that this is acceptable because
6211 * a premature "unwriteable" is better than an accidental "writeable" which
6212 * would cause an unwanted block under certain circumstances. For the 1-1
6213 * UDP-style sockets or TCP-style sockets, this code should work.
6214 * - Daisy
6216 static int sctp_writeable(struct sock *sk)
6218 int amt = 0;
6220 amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
6221 if (amt < 0)
6222 amt = 0;
6223 return amt;
6226 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6227 * returns immediately with EINPROGRESS.
6229 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6231 struct sock *sk = asoc->base.sk;
6232 int err = 0;
6233 long current_timeo = *timeo_p;
6234 DEFINE_WAIT(wait);
6236 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __FUNCTION__, asoc,
6237 (long)(*timeo_p));
6239 /* Increment the association's refcnt. */
6240 sctp_association_hold(asoc);
6242 for (;;) {
6243 prepare_to_wait_exclusive(&asoc->wait, &wait,
6244 TASK_INTERRUPTIBLE);
6245 if (!*timeo_p)
6246 goto do_nonblock;
6247 if (sk->sk_shutdown & RCV_SHUTDOWN)
6248 break;
6249 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6250 asoc->base.dead)
6251 goto do_error;
6252 if (signal_pending(current))
6253 goto do_interrupted;
6255 if (sctp_state(asoc, ESTABLISHED))
6256 break;
6258 /* Let another process have a go. Since we are going
6259 * to sleep anyway.
6261 sctp_release_sock(sk);
6262 current_timeo = schedule_timeout(current_timeo);
6263 sctp_lock_sock(sk);
6265 *timeo_p = current_timeo;
6268 out:
6269 finish_wait(&asoc->wait, &wait);
6271 /* Release the association's refcnt. */
6272 sctp_association_put(asoc);
6274 return err;
6276 do_error:
6277 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6278 err = -ETIMEDOUT;
6279 else
6280 err = -ECONNREFUSED;
6281 goto out;
6283 do_interrupted:
6284 err = sock_intr_errno(*timeo_p);
6285 goto out;
6287 do_nonblock:
6288 err = -EINPROGRESS;
6289 goto out;
6292 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6294 struct sctp_endpoint *ep;
6295 int err = 0;
6296 DEFINE_WAIT(wait);
6298 ep = sctp_sk(sk)->ep;
6301 for (;;) {
6302 prepare_to_wait_exclusive(sk->sk_sleep, &wait,
6303 TASK_INTERRUPTIBLE);
6305 if (list_empty(&ep->asocs)) {
6306 sctp_release_sock(sk);
6307 timeo = schedule_timeout(timeo);
6308 sctp_lock_sock(sk);
6311 err = -EINVAL;
6312 if (!sctp_sstate(sk, LISTENING))
6313 break;
6315 err = 0;
6316 if (!list_empty(&ep->asocs))
6317 break;
6319 err = sock_intr_errno(timeo);
6320 if (signal_pending(current))
6321 break;
6323 err = -EAGAIN;
6324 if (!timeo)
6325 break;
6328 finish_wait(sk->sk_sleep, &wait);
6330 return err;
6333 static void sctp_wait_for_close(struct sock *sk, long timeout)
6335 DEFINE_WAIT(wait);
6337 do {
6338 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
6339 if (list_empty(&sctp_sk(sk)->ep->asocs))
6340 break;
6341 sctp_release_sock(sk);
6342 timeout = schedule_timeout(timeout);
6343 sctp_lock_sock(sk);
6344 } while (!signal_pending(current) && timeout);
6346 finish_wait(sk->sk_sleep, &wait);
6349 static void sctp_sock_rfree_frag(struct sk_buff *skb)
6351 struct sk_buff *frag;
6353 if (!skb->data_len)
6354 goto done;
6356 /* Don't forget the fragments. */
6357 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6358 sctp_sock_rfree_frag(frag);
6360 done:
6361 sctp_sock_rfree(skb);
6364 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6366 struct sk_buff *frag;
6368 if (!skb->data_len)
6369 goto done;
6371 /* Don't forget the fragments. */
6372 for (frag = skb_shinfo(skb)->frag_list; frag; frag = frag->next)
6373 sctp_skb_set_owner_r_frag(frag, sk);
6375 done:
6376 sctp_skb_set_owner_r(skb, sk);
6379 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6380 * and its messages to the newsk.
6382 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6383 struct sctp_association *assoc,
6384 sctp_socket_type_t type)
6386 struct sctp_sock *oldsp = sctp_sk(oldsk);
6387 struct sctp_sock *newsp = sctp_sk(newsk);
6388 struct sctp_bind_bucket *pp; /* hash list port iterator */
6389 struct sctp_endpoint *newep = newsp->ep;
6390 struct sk_buff *skb, *tmp;
6391 struct sctp_ulpevent *event;
6392 struct sctp_bind_hashbucket *head;
6394 /* Migrate socket buffer sizes and all the socket level options to the
6395 * new socket.
6397 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6398 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6399 /* Brute force copy old sctp opt. */
6400 inet_sk_copy_descendant(newsk, oldsk);
6402 /* Restore the ep value that was overwritten with the above structure
6403 * copy.
6405 newsp->ep = newep;
6406 newsp->hmac = NULL;
6408 /* Hook this new socket in to the bind_hash list. */
6409 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->num)];
6410 sctp_local_bh_disable();
6411 sctp_spin_lock(&head->lock);
6412 pp = sctp_sk(oldsk)->bind_hash;
6413 sk_add_bind_node(newsk, &pp->owner);
6414 sctp_sk(newsk)->bind_hash = pp;
6415 inet_sk(newsk)->num = inet_sk(oldsk)->num;
6416 sctp_spin_unlock(&head->lock);
6417 sctp_local_bh_enable();
6419 /* Copy the bind_addr list from the original endpoint to the new
6420 * endpoint so that we can handle restarts properly
6422 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6423 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6425 /* Move any messages in the old socket's receive queue that are for the
6426 * peeled off association to the new socket's receive queue.
6428 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6429 event = sctp_skb2event(skb);
6430 if (event->asoc == assoc) {
6431 sctp_sock_rfree_frag(skb);
6432 __skb_unlink(skb, &oldsk->sk_receive_queue);
6433 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6434 sctp_skb_set_owner_r_frag(skb, newsk);
6438 /* Clean up any messages pending delivery due to partial
6439 * delivery. Three cases:
6440 * 1) No partial deliver; no work.
6441 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6442 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6444 skb_queue_head_init(&newsp->pd_lobby);
6445 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6447 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6448 struct sk_buff_head *queue;
6450 /* Decide which queue to move pd_lobby skbs to. */
6451 if (assoc->ulpq.pd_mode) {
6452 queue = &newsp->pd_lobby;
6453 } else
6454 queue = &newsk->sk_receive_queue;
6456 /* Walk through the pd_lobby, looking for skbs that
6457 * need moved to the new socket.
6459 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6460 event = sctp_skb2event(skb);
6461 if (event->asoc == assoc) {
6462 sctp_sock_rfree_frag(skb);
6463 __skb_unlink(skb, &oldsp->pd_lobby);
6464 __skb_queue_tail(queue, skb);
6465 sctp_skb_set_owner_r_frag(skb, newsk);
6469 /* Clear up any skbs waiting for the partial
6470 * delivery to finish.
6472 if (assoc->ulpq.pd_mode)
6473 sctp_clear_pd(oldsk, NULL);
6477 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp) {
6478 sctp_sock_rfree_frag(skb);
6479 sctp_skb_set_owner_r_frag(skb, newsk);
6482 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp) {
6483 sctp_sock_rfree_frag(skb);
6484 sctp_skb_set_owner_r_frag(skb, newsk);
6487 /* Set the type of socket to indicate that it is peeled off from the
6488 * original UDP-style socket or created with the accept() call on a
6489 * TCP-style socket..
6491 newsp->type = type;
6493 /* Mark the new socket "in-use" by the user so that any packets
6494 * that may arrive on the association after we've moved it are
6495 * queued to the backlog. This prevents a potential race between
6496 * backlog processing on the old socket and new-packet processing
6497 * on the new socket.
6499 * The caller has just allocated newsk so we can guarantee that other
6500 * paths won't try to lock it and then oldsk.
6502 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6503 sctp_assoc_migrate(assoc, newsk);
6505 /* If the association on the newsk is already closed before accept()
6506 * is called, set RCV_SHUTDOWN flag.
6508 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6509 newsk->sk_shutdown |= RCV_SHUTDOWN;
6511 newsk->sk_state = SCTP_SS_ESTABLISHED;
6512 sctp_release_sock(newsk);
6516 DEFINE_PROTO_INUSE(sctp)
6518 /* This proto struct describes the ULP interface for SCTP. */
6519 struct proto sctp_prot = {
6520 .name = "SCTP",
6521 .owner = THIS_MODULE,
6522 .close = sctp_close,
6523 .connect = sctp_connect,
6524 .disconnect = sctp_disconnect,
6525 .accept = sctp_accept,
6526 .ioctl = sctp_ioctl,
6527 .init = sctp_init_sock,
6528 .destroy = sctp_destroy_sock,
6529 .shutdown = sctp_shutdown,
6530 .setsockopt = sctp_setsockopt,
6531 .getsockopt = sctp_getsockopt,
6532 .sendmsg = sctp_sendmsg,
6533 .recvmsg = sctp_recvmsg,
6534 .bind = sctp_bind,
6535 .backlog_rcv = sctp_backlog_rcv,
6536 .hash = sctp_hash,
6537 .unhash = sctp_unhash,
6538 .get_port = sctp_get_port,
6539 .obj_size = sizeof(struct sctp_sock),
6540 .sysctl_mem = sysctl_sctp_mem,
6541 .sysctl_rmem = sysctl_sctp_rmem,
6542 .sysctl_wmem = sysctl_sctp_wmem,
6543 .memory_pressure = &sctp_memory_pressure,
6544 .enter_memory_pressure = sctp_enter_memory_pressure,
6545 .memory_allocated = &sctp_memory_allocated,
6546 .sockets_allocated = &sctp_sockets_allocated,
6547 REF_PROTO_INUSE(sctp)
6550 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6551 DEFINE_PROTO_INUSE(sctpv6)
6553 struct proto sctpv6_prot = {
6554 .name = "SCTPv6",
6555 .owner = THIS_MODULE,
6556 .close = sctp_close,
6557 .connect = sctp_connect,
6558 .disconnect = sctp_disconnect,
6559 .accept = sctp_accept,
6560 .ioctl = sctp_ioctl,
6561 .init = sctp_init_sock,
6562 .destroy = sctp_destroy_sock,
6563 .shutdown = sctp_shutdown,
6564 .setsockopt = sctp_setsockopt,
6565 .getsockopt = sctp_getsockopt,
6566 .sendmsg = sctp_sendmsg,
6567 .recvmsg = sctp_recvmsg,
6568 .bind = sctp_bind,
6569 .backlog_rcv = sctp_backlog_rcv,
6570 .hash = sctp_hash,
6571 .unhash = sctp_unhash,
6572 .get_port = sctp_get_port,
6573 .obj_size = sizeof(struct sctp6_sock),
6574 .sysctl_mem = sysctl_sctp_mem,
6575 .sysctl_rmem = sysctl_sctp_rmem,
6576 .sysctl_wmem = sysctl_sctp_wmem,
6577 .memory_pressure = &sctp_memory_pressure,
6578 .enter_memory_pressure = sctp_enter_memory_pressure,
6579 .memory_allocated = &sctp_memory_allocated,
6580 .sockets_allocated = &sctp_sockets_allocated,
6581 REF_PROTO_INUSE(sctpv6)
6583 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */