Add support for allocating irqs for bootbus devices
[pohmelfs.git] / net / sctp / socket.c
blob6766913a53e626279a7a2f4eb543e1f9fa7b5263
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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
62 #include <linux/types.h>
63 #include <linux/kernel.h>
64 #include <linux/wait.h>
65 #include <linux/time.h>
66 #include <linux/ip.h>
67 #include <linux/capability.h>
68 #include <linux/fcntl.h>
69 #include <linux/poll.h>
70 #include <linux/init.h>
71 #include <linux/crypto.h>
72 #include <linux/slab.h>
74 #include <net/ip.h>
75 #include <net/icmp.h>
76 #include <net/route.h>
77 #include <net/ipv6.h>
78 #include <net/inet_common.h>
80 #include <linux/socket.h> /* for sa_family_t */
81 #include <net/sock.h>
82 #include <net/sctp/sctp.h>
83 #include <net/sctp/sm.h>
85 /* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
90 /* Forward declarations for internal helper functions. */
91 static int sctp_writeable(struct sock *sk);
92 static void sctp_wfree(struct sk_buff *skb);
93 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
94 size_t msg_len);
95 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
96 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
97 static int sctp_wait_for_accept(struct sock *sk, long timeo);
98 static void sctp_wait_for_close(struct sock *sk, long timeo);
99 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
100 union sctp_addr *addr, int len);
101 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
102 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
103 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
104 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
105 static int sctp_send_asconf(struct sctp_association *asoc,
106 struct sctp_chunk *chunk);
107 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
108 static int sctp_autobind(struct sock *sk);
109 static void sctp_sock_migrate(struct sock *, struct sock *,
110 struct sctp_association *, sctp_socket_type_t);
111 static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
113 extern struct kmem_cache *sctp_bucket_cachep;
114 extern long sysctl_sctp_mem[3];
115 extern int sysctl_sctp_rmem[3];
116 extern int sysctl_sctp_wmem[3];
118 static int sctp_memory_pressure;
119 static atomic_long_t sctp_memory_allocated;
120 struct percpu_counter sctp_sockets_allocated;
122 static void sctp_enter_memory_pressure(struct sock *sk)
124 sctp_memory_pressure = 1;
128 /* Get the sndbuf space available at the time on the association. */
129 static inline int sctp_wspace(struct sctp_association *asoc)
131 int amt;
133 if (asoc->ep->sndbuf_policy)
134 amt = asoc->sndbuf_used;
135 else
136 amt = sk_wmem_alloc_get(asoc->base.sk);
138 if (amt >= asoc->base.sk->sk_sndbuf) {
139 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
140 amt = 0;
141 else {
142 amt = sk_stream_wspace(asoc->base.sk);
143 if (amt < 0)
144 amt = 0;
146 } else {
147 amt = asoc->base.sk->sk_sndbuf - amt;
149 return amt;
152 /* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
159 * tracking.
161 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
163 struct sctp_association *asoc = chunk->asoc;
164 struct sock *sk = asoc->base.sk;
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc);
169 skb_set_owner_w(chunk->skb, sk);
171 chunk->skb->destructor = sctp_wfree;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
175 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
176 sizeof(struct sk_buff) +
177 sizeof(struct sctp_chunk);
179 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
180 sk->sk_wmem_queued += chunk->skb->truesize;
181 sk_mem_charge(sk, chunk->skb->truesize);
184 /* Verify that this is a valid address. */
185 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
186 int len)
188 struct sctp_af *af;
190 /* Verify basic sockaddr. */
191 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 if (!af)
193 return -EINVAL;
195 /* Is this a valid SCTP address? */
196 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
197 return -EINVAL;
199 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
200 return -EINVAL;
202 return 0;
205 /* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
208 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
210 struct sctp_association *asoc = NULL;
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk, UDP)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
218 if (!sctp_sstate(sk, ESTABLISHED))
219 return NULL;
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk)->ep->asocs))
223 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
224 struct sctp_association, asocs);
225 return asoc;
228 /* Otherwise this is a UDP-style socket. */
229 if (!id || (id == (sctp_assoc_t)-1))
230 return NULL;
232 spin_lock_bh(&sctp_assocs_id_lock);
233 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
234 spin_unlock_bh(&sctp_assocs_id_lock);
236 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
237 return NULL;
239 return asoc;
242 /* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
244 * the same.
246 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
247 struct sockaddr_storage *addr,
248 sctp_assoc_t id)
250 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
251 struct sctp_transport *transport;
252 union sctp_addr *laddr = (union sctp_addr *)addr;
254 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
255 laddr,
256 &transport);
258 if (!addr_asoc)
259 return NULL;
261 id_asoc = sctp_id2assoc(sk, id);
262 if (id_asoc && (id_asoc != addr_asoc))
263 return NULL;
265 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
266 (union sctp_addr *)addr);
268 return transport;
271 /* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
281 SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
283 int retval = 0;
285 sctp_lock_sock(sk);
287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
288 sk, addr, addr_len);
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk)->ep->base.bind_addr.port)
292 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
293 addr_len);
294 else
295 retval = -EINVAL;
297 sctp_release_sock(sk);
299 return retval;
302 static long sctp_get_port_local(struct sock *, union sctp_addr *);
304 /* Verify this is a valid sockaddr. */
305 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
306 union sctp_addr *addr, int len)
308 struct sctp_af *af;
310 /* Check minimum size. */
311 if (len < sizeof (struct sockaddr))
312 return NULL;
314 /* V4 mapped address are really of AF_INET family */
315 if (addr->sa.sa_family == AF_INET6 &&
316 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
317 if (!opt->pf->af_supported(AF_INET, opt))
318 return NULL;
319 } else {
320 /* Does this PF support this AF? */
321 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 return NULL;
325 /* If we get this far, af is valid. */
326 af = sctp_get_af_specific(addr->sa.sa_family);
328 if (len < af->sockaddr_len)
329 return NULL;
331 return af;
334 /* Bind a local address either to an endpoint or to an association. */
335 SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
337 struct sctp_sock *sp = sctp_sk(sk);
338 struct sctp_endpoint *ep = sp->ep;
339 struct sctp_bind_addr *bp = &ep->base.bind_addr;
340 struct sctp_af *af;
341 unsigned short snum;
342 int ret = 0;
344 /* Common sockaddr verification. */
345 af = sctp_sockaddr_af(sp, addr, len);
346 if (!af) {
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
348 sk, addr, len);
349 return -EINVAL;
352 snum = ntohs(addr->v4.sin_port);
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
357 addr,
358 bp->port, snum,
359 len);
361 /* PF specific bind() address verification. */
362 if (!sp->pf->bind_verify(sp, addr))
363 return -EADDRNOTAVAIL;
365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
369 if (bp->port) {
370 if (!snum)
371 snum = bp->port;
372 else if (snum != bp->port) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
374 " New port %d does not match existing port "
375 "%d.\n", snum, bp->port);
376 return -EINVAL;
380 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
381 return -EACCES;
383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
386 if (sctp_bind_addr_match(bp, addr, sp))
387 return -EINVAL;
389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
391 * detection.
393 addr->v4.sin_port = htons(snum);
394 if ((ret = sctp_get_port_local(sk, addr))) {
395 return -EADDRINUSE;
398 /* Refresh ephemeral port. */
399 if (!bp->port)
400 bp->port = inet_sk(sk)->inet_num;
402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
405 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
407 /* Copy back into socket for getsockname() use. */
408 if (!ret) {
409 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
410 af->to_sk_saddr(addr, sk);
413 return ret;
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
419 * at any one time. If a sender, after sending an ASCONF chunk, decides
420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
424 * from each endpoint).
426 static int sctp_send_asconf(struct sctp_association *asoc,
427 struct sctp_chunk *chunk)
429 int retval = 0;
431 /* If there is an outstanding ASCONF chunk, queue it for later
432 * transmission.
434 if (asoc->addip_last_asconf) {
435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
436 goto out;
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(asoc, chunk);
442 if (retval)
443 sctp_chunk_free(chunk);
444 else
445 asoc->addip_last_asconf = chunk;
447 out:
448 return retval;
451 /* Add a list of addresses as bind addresses to local endpoint or
452 * association.
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
463 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
465 int cnt;
466 int retval = 0;
467 void *addr_buf;
468 struct sockaddr *sa_addr;
469 struct sctp_af *af;
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
472 sk, addrs, addrcnt);
474 addr_buf = addrs;
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
479 sa_addr = (struct sockaddr *)addr_buf;
480 af = sctp_get_af_specific(sa_addr->sa_family);
481 if (!af) {
482 retval = -EINVAL;
483 goto err_bindx_add;
486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
487 af->sockaddr_len);
489 addr_buf += af->sockaddr_len;
491 err_bindx_add:
492 if (retval < 0) {
493 /* Failed. Cleanup the ones that have been added */
494 if (cnt > 0)
495 sctp_bindx_rem(sk, addrs, cnt);
496 return retval;
500 return retval;
503 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
507 * If any of the addresses is already in the bind address list of the
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
513 static int sctp_send_asconf_add_ip(struct sock *sk,
514 struct sockaddr *addrs,
515 int addrcnt)
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
524 union sctp_addr saveaddr;
525 void *addr_buf;
526 struct sctp_af *af;
527 struct list_head *p;
528 int i;
529 int retval = 0;
531 if (!sctp_addip_enable)
532 return retval;
534 sp = sctp_sk(sk);
535 ep = sp->ep;
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
538 __func__, sk, addrs, addrcnt);
540 list_for_each_entry(asoc, &ep->asocs, asocs) {
542 if (!asoc->peer.asconf_capable)
543 continue;
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
546 continue;
548 if (!sctp_state(asoc, ESTABLISHED))
549 continue;
551 /* Check if any address in the packed array of addresses is
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
554 * other associations.
556 addr_buf = addrs;
557 for (i = 0; i < addrcnt; i++) {
558 addr = (union sctp_addr *)addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
560 if (!af) {
561 retval = -EINVAL;
562 goto out;
565 if (sctp_assoc_lookup_laddr(asoc, addr))
566 break;
568 addr_buf += af->sockaddr_len;
570 if (i < addrcnt)
571 continue;
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
580 addrcnt, SCTP_PARAM_ADD_IP);
581 if (!chunk) {
582 retval = -ENOMEM;
583 goto out;
586 retval = sctp_send_asconf(asoc, chunk);
587 if (retval)
588 goto out;
590 /* Add the new addresses to the bind address list with
591 * use_as_src set to 0.
593 addr_buf = addrs;
594 for (i = 0; i < addrcnt; i++) {
595 addr = (union sctp_addr *)addr_buf;
596 af = sctp_get_af_specific(addr->v4.sin_family);
597 memcpy(&saveaddr, addr, af->sockaddr_len);
598 retval = sctp_add_bind_addr(bp, &saveaddr,
599 SCTP_ADDR_NEW, GFP_ATOMIC);
600 addr_buf += af->sockaddr_len;
604 out:
605 return retval;
608 /* Remove a list of addresses from bind addresses list. Do not remove the
609 * last address.
611 * Basically run through each address specified in the addrs/addrcnt
612 * array/length pair, determine if it is IPv6 or IPv4 and call
613 * sctp_del_bind() on it.
615 * If any of them fails, then the operation will be reversed and the
616 * ones that were removed will be added back.
618 * At least one address has to be left; if only one address is
619 * available, the operation will return -EBUSY.
621 * Only sctp_setsockopt_bindx() is supposed to call this function.
623 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
625 struct sctp_sock *sp = sctp_sk(sk);
626 struct sctp_endpoint *ep = sp->ep;
627 int cnt;
628 struct sctp_bind_addr *bp = &ep->base.bind_addr;
629 int retval = 0;
630 void *addr_buf;
631 union sctp_addr *sa_addr;
632 struct sctp_af *af;
634 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
635 sk, addrs, addrcnt);
637 addr_buf = addrs;
638 for (cnt = 0; cnt < addrcnt; cnt++) {
639 /* If the bind address list is empty or if there is only one
640 * bind address, there is nothing more to be removed (we need
641 * at least one address here).
643 if (list_empty(&bp->address_list) ||
644 (sctp_list_single_entry(&bp->address_list))) {
645 retval = -EBUSY;
646 goto err_bindx_rem;
649 sa_addr = (union sctp_addr *)addr_buf;
650 af = sctp_get_af_specific(sa_addr->sa.sa_family);
651 if (!af) {
652 retval = -EINVAL;
653 goto err_bindx_rem;
656 if (!af->addr_valid(sa_addr, sp, NULL)) {
657 retval = -EADDRNOTAVAIL;
658 goto err_bindx_rem;
661 if (sa_addr->v4.sin_port &&
662 sa_addr->v4.sin_port != htons(bp->port)) {
663 retval = -EINVAL;
664 goto err_bindx_rem;
667 if (!sa_addr->v4.sin_port)
668 sa_addr->v4.sin_port = htons(bp->port);
670 /* FIXME - There is probably a need to check if sk->sk_saddr and
671 * sk->sk_rcv_addr are currently set to one of the addresses to
672 * be removed. This is something which needs to be looked into
673 * when we are fixing the outstanding issues with multi-homing
674 * socket routing and failover schemes. Refer to comments in
675 * sctp_do_bind(). -daisy
677 retval = sctp_del_bind_addr(bp, sa_addr);
679 addr_buf += af->sockaddr_len;
680 err_bindx_rem:
681 if (retval < 0) {
682 /* Failed. Add the ones that has been removed back */
683 if (cnt > 0)
684 sctp_bindx_add(sk, addrs, cnt);
685 return retval;
689 return retval;
692 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
693 * the associations that are part of the endpoint indicating that a list of
694 * local addresses are removed from the endpoint.
696 * If any of the addresses is already in the bind address list of the
697 * association, we do not send the chunk for that association. But it will not
698 * affect other associations.
700 * Only sctp_setsockopt_bindx() is supposed to call this function.
702 static int sctp_send_asconf_del_ip(struct sock *sk,
703 struct sockaddr *addrs,
704 int addrcnt)
706 struct sctp_sock *sp;
707 struct sctp_endpoint *ep;
708 struct sctp_association *asoc;
709 struct sctp_transport *transport;
710 struct sctp_bind_addr *bp;
711 struct sctp_chunk *chunk;
712 union sctp_addr *laddr;
713 void *addr_buf;
714 struct sctp_af *af;
715 struct sctp_sockaddr_entry *saddr;
716 int i;
717 int retval = 0;
719 if (!sctp_addip_enable)
720 return retval;
722 sp = sctp_sk(sk);
723 ep = sp->ep;
725 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
726 __func__, sk, addrs, addrcnt);
728 list_for_each_entry(asoc, &ep->asocs, asocs) {
730 if (!asoc->peer.asconf_capable)
731 continue;
733 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
734 continue;
736 if (!sctp_state(asoc, ESTABLISHED))
737 continue;
739 /* Check if any address in the packed array of addresses is
740 * not present in the bind address list of the association.
741 * If so, do not send the asconf chunk to its peer, but
742 * continue with other associations.
744 addr_buf = addrs;
745 for (i = 0; i < addrcnt; i++) {
746 laddr = (union sctp_addr *)addr_buf;
747 af = sctp_get_af_specific(laddr->v4.sin_family);
748 if (!af) {
749 retval = -EINVAL;
750 goto out;
753 if (!sctp_assoc_lookup_laddr(asoc, laddr))
754 break;
756 addr_buf += af->sockaddr_len;
758 if (i < addrcnt)
759 continue;
761 /* Find one address in the association's bind address list
762 * that is not in the packed array of addresses. This is to
763 * make sure that we do not delete all the addresses in the
764 * association.
766 bp = &asoc->base.bind_addr;
767 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
768 addrcnt, sp);
769 if (!laddr)
770 continue;
772 /* We do not need RCU protection throughout this loop
773 * because this is done under a socket lock from the
774 * setsockopt call.
776 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
777 SCTP_PARAM_DEL_IP);
778 if (!chunk) {
779 retval = -ENOMEM;
780 goto out;
783 /* Reset use_as_src flag for the addresses in the bind address
784 * list that are to be deleted.
786 addr_buf = addrs;
787 for (i = 0; i < addrcnt; i++) {
788 laddr = (union sctp_addr *)addr_buf;
789 af = sctp_get_af_specific(laddr->v4.sin_family);
790 list_for_each_entry(saddr, &bp->address_list, list) {
791 if (sctp_cmp_addr_exact(&saddr->a, laddr))
792 saddr->state = SCTP_ADDR_DEL;
794 addr_buf += af->sockaddr_len;
797 /* Update the route and saddr entries for all the transports
798 * as some of the addresses in the bind address list are
799 * about to be deleted and cannot be used as source addresses.
801 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
802 transports) {
803 dst_release(transport->dst);
804 sctp_transport_route(transport, NULL,
805 sctp_sk(asoc->base.sk));
808 retval = sctp_send_asconf(asoc, chunk);
810 out:
811 return retval;
814 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
816 * API 8.1
817 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
818 * int flags);
820 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
821 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
822 * or IPv6 addresses.
824 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
825 * Section 3.1.2 for this usage.
827 * addrs is a pointer to an array of one or more socket addresses. Each
828 * address is contained in its appropriate structure (i.e. struct
829 * sockaddr_in or struct sockaddr_in6) the family of the address type
830 * must be used to distinguish the address length (note that this
831 * representation is termed a "packed array" of addresses). The caller
832 * specifies the number of addresses in the array with addrcnt.
834 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
835 * -1, and sets errno to the appropriate error code.
837 * For SCTP, the port given in each socket address must be the same, or
838 * sctp_bindx() will fail, setting errno to EINVAL.
840 * The flags parameter is formed from the bitwise OR of zero or more of
841 * the following currently defined flags:
843 * SCTP_BINDX_ADD_ADDR
845 * SCTP_BINDX_REM_ADDR
847 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
848 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
849 * addresses from the association. The two flags are mutually exclusive;
850 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
851 * not remove all addresses from an association; sctp_bindx() will
852 * reject such an attempt with EINVAL.
854 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
855 * additional addresses with an endpoint after calling bind(). Or use
856 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
857 * socket is associated with so that no new association accepted will be
858 * associated with those addresses. If the endpoint supports dynamic
859 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
860 * endpoint to send the appropriate message to the peer to change the
861 * peers address lists.
863 * Adding and removing addresses from a connected association is
864 * optional functionality. Implementations that do not support this
865 * functionality should return EOPNOTSUPP.
867 * Basically do nothing but copying the addresses from user to kernel
868 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
869 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
870 * from userspace.
872 * We don't use copy_from_user() for optimization: we first do the
873 * sanity checks (buffer size -fast- and access check-healthy
874 * pointer); if all of those succeed, then we can alloc the memory
875 * (expensive operation) needed to copy the data to kernel. Then we do
876 * the copying without checking the user space area
877 * (__copy_from_user()).
879 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
880 * it.
882 * sk The sk of the socket
883 * addrs The pointer to the addresses in user land
884 * addrssize Size of the addrs buffer
885 * op Operation to perform (add or remove, see the flags of
886 * sctp_bindx)
888 * Returns 0 if ok, <0 errno code on error.
890 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
891 struct sockaddr __user *addrs,
892 int addrs_size, int op)
894 struct sockaddr *kaddrs;
895 int err;
896 int addrcnt = 0;
897 int walk_size = 0;
898 struct sockaddr *sa_addr;
899 void *addr_buf;
900 struct sctp_af *af;
902 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
903 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
905 if (unlikely(addrs_size <= 0))
906 return -EINVAL;
908 /* Check the user passed a healthy pointer. */
909 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
910 return -EFAULT;
912 /* Alloc space for the address array in kernel memory. */
913 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
914 if (unlikely(!kaddrs))
915 return -ENOMEM;
917 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
918 kfree(kaddrs);
919 return -EFAULT;
922 /* Walk through the addrs buffer and count the number of addresses. */
923 addr_buf = kaddrs;
924 while (walk_size < addrs_size) {
925 if (walk_size + sizeof(sa_family_t) > addrs_size) {
926 kfree(kaddrs);
927 return -EINVAL;
930 sa_addr = (struct sockaddr *)addr_buf;
931 af = sctp_get_af_specific(sa_addr->sa_family);
933 /* If the address family is not supported or if this address
934 * causes the address buffer to overflow return EINVAL.
936 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
937 kfree(kaddrs);
938 return -EINVAL;
940 addrcnt++;
941 addr_buf += af->sockaddr_len;
942 walk_size += af->sockaddr_len;
945 /* Do the work. */
946 switch (op) {
947 case SCTP_BINDX_ADD_ADDR:
948 err = sctp_bindx_add(sk, kaddrs, addrcnt);
949 if (err)
950 goto out;
951 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
952 break;
954 case SCTP_BINDX_REM_ADDR:
955 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
956 if (err)
957 goto out;
958 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
959 break;
961 default:
962 err = -EINVAL;
963 break;
966 out:
967 kfree(kaddrs);
969 return err;
972 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
974 * Common routine for handling connect() and sctp_connectx().
975 * Connect will come in with just a single address.
977 static int __sctp_connect(struct sock* sk,
978 struct sockaddr *kaddrs,
979 int addrs_size,
980 sctp_assoc_t *assoc_id)
982 struct sctp_sock *sp;
983 struct sctp_endpoint *ep;
984 struct sctp_association *asoc = NULL;
985 struct sctp_association *asoc2;
986 struct sctp_transport *transport;
987 union sctp_addr to;
988 struct sctp_af *af;
989 sctp_scope_t scope;
990 long timeo;
991 int err = 0;
992 int addrcnt = 0;
993 int walk_size = 0;
994 union sctp_addr *sa_addr = NULL;
995 void *addr_buf;
996 unsigned short port;
997 unsigned int f_flags = 0;
999 sp = sctp_sk(sk);
1000 ep = sp->ep;
1002 /* connect() cannot be done on a socket that is already in ESTABLISHED
1003 * state - UDP-style peeled off socket or a TCP-style socket that
1004 * is already connected.
1005 * It cannot be done even on a TCP-style listening socket.
1007 if (sctp_sstate(sk, ESTABLISHED) ||
1008 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1009 err = -EISCONN;
1010 goto out_free;
1013 /* Walk through the addrs buffer and count the number of addresses. */
1014 addr_buf = kaddrs;
1015 while (walk_size < addrs_size) {
1016 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1017 err = -EINVAL;
1018 goto out_free;
1021 sa_addr = (union sctp_addr *)addr_buf;
1022 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1024 /* If the address family is not supported or if this address
1025 * causes the address buffer to overflow return EINVAL.
1027 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1028 err = -EINVAL;
1029 goto out_free;
1032 port = ntohs(sa_addr->v4.sin_port);
1034 /* Save current address so we can work with it */
1035 memcpy(&to, sa_addr, af->sockaddr_len);
1037 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1038 if (err)
1039 goto out_free;
1041 /* Make sure the destination port is correctly set
1042 * in all addresses.
1044 if (asoc && asoc->peer.port && asoc->peer.port != port)
1045 goto out_free;
1048 /* Check if there already is a matching association on the
1049 * endpoint (other than the one created here).
1051 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1052 if (asoc2 && asoc2 != asoc) {
1053 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1054 err = -EISCONN;
1055 else
1056 err = -EALREADY;
1057 goto out_free;
1060 /* If we could not find a matching association on the endpoint,
1061 * make sure that there is no peeled-off association matching
1062 * the peer address even on another socket.
1064 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1065 err = -EADDRNOTAVAIL;
1066 goto out_free;
1069 if (!asoc) {
1070 /* If a bind() or sctp_bindx() is not called prior to
1071 * an sctp_connectx() call, the system picks an
1072 * ephemeral port and will choose an address set
1073 * equivalent to binding with a wildcard address.
1075 if (!ep->base.bind_addr.port) {
1076 if (sctp_autobind(sk)) {
1077 err = -EAGAIN;
1078 goto out_free;
1080 } else {
1082 * If an unprivileged user inherits a 1-many
1083 * style socket with open associations on a
1084 * privileged port, it MAY be permitted to
1085 * accept new associations, but it SHOULD NOT
1086 * be permitted to open new associations.
1088 if (ep->base.bind_addr.port < PROT_SOCK &&
1089 !capable(CAP_NET_BIND_SERVICE)) {
1090 err = -EACCES;
1091 goto out_free;
1095 scope = sctp_scope(&to);
1096 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1097 if (!asoc) {
1098 err = -ENOMEM;
1099 goto out_free;
1102 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1103 GFP_KERNEL);
1104 if (err < 0) {
1105 goto out_free;
1110 /* Prime the peer's transport structures. */
1111 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1112 SCTP_UNKNOWN);
1113 if (!transport) {
1114 err = -ENOMEM;
1115 goto out_free;
1118 addrcnt++;
1119 addr_buf += af->sockaddr_len;
1120 walk_size += af->sockaddr_len;
1123 /* In case the user of sctp_connectx() wants an association
1124 * id back, assign one now.
1126 if (assoc_id) {
1127 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1128 if (err < 0)
1129 goto out_free;
1132 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1133 if (err < 0) {
1134 goto out_free;
1137 /* Initialize sk's dport and daddr for getpeername() */
1138 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1139 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1140 af->to_sk_daddr(sa_addr, sk);
1141 sk->sk_err = 0;
1143 /* in-kernel sockets don't generally have a file allocated to them
1144 * if all they do is call sock_create_kern().
1146 if (sk->sk_socket->file)
1147 f_flags = sk->sk_socket->file->f_flags;
1149 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1151 err = sctp_wait_for_connect(asoc, &timeo);
1152 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1153 *assoc_id = asoc->assoc_id;
1155 /* Don't free association on exit. */
1156 asoc = NULL;
1158 out_free:
1160 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1161 " kaddrs: %p err: %d\n",
1162 asoc, kaddrs, err);
1163 if (asoc)
1164 sctp_association_free(asoc);
1165 return err;
1168 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1170 * API 8.9
1171 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1172 * sctp_assoc_t *asoc);
1174 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1175 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1176 * or IPv6 addresses.
1178 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1179 * Section 3.1.2 for this usage.
1181 * addrs is a pointer to an array of one or more socket addresses. Each
1182 * address is contained in its appropriate structure (i.e. struct
1183 * sockaddr_in or struct sockaddr_in6) the family of the address type
1184 * must be used to distengish the address length (note that this
1185 * representation is termed a "packed array" of addresses). The caller
1186 * specifies the number of addresses in the array with addrcnt.
1188 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1189 * the association id of the new association. On failure, sctp_connectx()
1190 * returns -1, and sets errno to the appropriate error code. The assoc_id
1191 * is not touched by the kernel.
1193 * For SCTP, the port given in each socket address must be the same, or
1194 * sctp_connectx() will fail, setting errno to EINVAL.
1196 * An application can use sctp_connectx to initiate an association with
1197 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1198 * allows a caller to specify multiple addresses at which a peer can be
1199 * reached. The way the SCTP stack uses the list of addresses to set up
1200 * the association is implementation dependent. This function only
1201 * specifies that the stack will try to make use of all the addresses in
1202 * the list when needed.
1204 * Note that the list of addresses passed in is only used for setting up
1205 * the association. It does not necessarily equal the set of addresses
1206 * the peer uses for the resulting association. If the caller wants to
1207 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1208 * retrieve them after the association has been set up.
1210 * Basically do nothing but copying the addresses from user to kernel
1211 * land and invoking either sctp_connectx(). This is used for tunneling
1212 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1214 * We don't use copy_from_user() for optimization: we first do the
1215 * sanity checks (buffer size -fast- and access check-healthy
1216 * pointer); if all of those succeed, then we can alloc the memory
1217 * (expensive operation) needed to copy the data to kernel. Then we do
1218 * the copying without checking the user space area
1219 * (__copy_from_user()).
1221 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1222 * it.
1224 * sk The sk of the socket
1225 * addrs The pointer to the addresses in user land
1226 * addrssize Size of the addrs buffer
1228 * Returns >=0 if ok, <0 errno code on error.
1230 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1231 struct sockaddr __user *addrs,
1232 int addrs_size,
1233 sctp_assoc_t *assoc_id)
1235 int err = 0;
1236 struct sockaddr *kaddrs;
1238 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1239 __func__, sk, addrs, addrs_size);
1241 if (unlikely(addrs_size <= 0))
1242 return -EINVAL;
1244 /* Check the user passed a healthy pointer. */
1245 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1246 return -EFAULT;
1248 /* Alloc space for the address array in kernel memory. */
1249 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1250 if (unlikely(!kaddrs))
1251 return -ENOMEM;
1253 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1254 err = -EFAULT;
1255 } else {
1256 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1259 kfree(kaddrs);
1261 return err;
1265 * This is an older interface. It's kept for backward compatibility
1266 * to the option that doesn't provide association id.
1268 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1269 struct sockaddr __user *addrs,
1270 int addrs_size)
1272 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1276 * New interface for the API. The since the API is done with a socket
1277 * option, to make it simple we feed back the association id is as a return
1278 * indication to the call. Error is always negative and association id is
1279 * always positive.
1281 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1282 struct sockaddr __user *addrs,
1283 int addrs_size)
1285 sctp_assoc_t assoc_id = 0;
1286 int err = 0;
1288 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1290 if (err)
1291 return err;
1292 else
1293 return assoc_id;
1297 * New (hopefully final) interface for the API.
1298 * We use the sctp_getaddrs_old structure so that use-space library
1299 * can avoid any unnecessary allocations. The only defferent part
1300 * is that we store the actual length of the address buffer into the
1301 * addrs_num structure member. That way we can re-use the existing
1302 * code.
1304 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1305 char __user *optval,
1306 int __user *optlen)
1308 struct sctp_getaddrs_old param;
1309 sctp_assoc_t assoc_id = 0;
1310 int err = 0;
1312 if (len < sizeof(param))
1313 return -EINVAL;
1315 if (copy_from_user(&param, optval, sizeof(param)))
1316 return -EFAULT;
1318 err = __sctp_setsockopt_connectx(sk,
1319 (struct sockaddr __user *)param.addrs,
1320 param.addr_num, &assoc_id);
1322 if (err == 0 || err == -EINPROGRESS) {
1323 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1324 return -EFAULT;
1325 if (put_user(sizeof(assoc_id), optlen))
1326 return -EFAULT;
1329 return err;
1332 /* API 3.1.4 close() - UDP Style Syntax
1333 * Applications use close() to perform graceful shutdown (as described in
1334 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1335 * by a UDP-style socket.
1337 * The syntax is
1339 * ret = close(int sd);
1341 * sd - the socket descriptor of the associations to be closed.
1343 * To gracefully shutdown a specific association represented by the
1344 * UDP-style socket, an application should use the sendmsg() call,
1345 * passing no user data, but including the appropriate flag in the
1346 * ancillary data (see Section xxxx).
1348 * If sd in the close() call is a branched-off socket representing only
1349 * one association, the shutdown is performed on that association only.
1351 * 4.1.6 close() - TCP Style Syntax
1353 * Applications use close() to gracefully close down an association.
1355 * The syntax is:
1357 * int close(int sd);
1359 * sd - the socket descriptor of the association to be closed.
1361 * After an application calls close() on a socket descriptor, no further
1362 * socket operations will succeed on that descriptor.
1364 * API 7.1.4 SO_LINGER
1366 * An application using the TCP-style socket can use this option to
1367 * perform the SCTP ABORT primitive. The linger option structure is:
1369 * struct linger {
1370 * int l_onoff; // option on/off
1371 * int l_linger; // linger time
1372 * };
1374 * To enable the option, set l_onoff to 1. If the l_linger value is set
1375 * to 0, calling close() is the same as the ABORT primitive. If the
1376 * value is set to a negative value, the setsockopt() call will return
1377 * an error. If the value is set to a positive value linger_time, the
1378 * close() can be blocked for at most linger_time ms. If the graceful
1379 * shutdown phase does not finish during this period, close() will
1380 * return but the graceful shutdown phase continues in the system.
1382 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1384 struct sctp_endpoint *ep;
1385 struct sctp_association *asoc;
1386 struct list_head *pos, *temp;
1388 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1390 sctp_lock_sock(sk);
1391 sk->sk_shutdown = SHUTDOWN_MASK;
1392 sk->sk_state = SCTP_SS_CLOSING;
1394 ep = sctp_sk(sk)->ep;
1396 /* Walk all associations on an endpoint. */
1397 list_for_each_safe(pos, temp, &ep->asocs) {
1398 asoc = list_entry(pos, struct sctp_association, asocs);
1400 if (sctp_style(sk, TCP)) {
1401 /* A closed association can still be in the list if
1402 * it belongs to a TCP-style listening socket that is
1403 * not yet accepted. If so, free it. If not, send an
1404 * ABORT or SHUTDOWN based on the linger options.
1406 if (sctp_state(asoc, CLOSED)) {
1407 sctp_unhash_established(asoc);
1408 sctp_association_free(asoc);
1409 continue;
1413 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1414 struct sctp_chunk *chunk;
1416 chunk = sctp_make_abort_user(asoc, NULL, 0);
1417 if (chunk)
1418 sctp_primitive_ABORT(asoc, chunk);
1419 } else
1420 sctp_primitive_SHUTDOWN(asoc, NULL);
1423 /* Clean up any skbs sitting on the receive queue. */
1424 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1425 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1427 /* On a TCP-style socket, block for at most linger_time if set. */
1428 if (sctp_style(sk, TCP) && timeout)
1429 sctp_wait_for_close(sk, timeout);
1431 /* This will run the backlog queue. */
1432 sctp_release_sock(sk);
1434 /* Supposedly, no process has access to the socket, but
1435 * the net layers still may.
1437 sctp_local_bh_disable();
1438 sctp_bh_lock_sock(sk);
1440 /* Hold the sock, since sk_common_release() will put sock_put()
1441 * and we have just a little more cleanup.
1443 sock_hold(sk);
1444 sk_common_release(sk);
1446 sctp_bh_unlock_sock(sk);
1447 sctp_local_bh_enable();
1449 sock_put(sk);
1451 SCTP_DBG_OBJCNT_DEC(sock);
1454 /* Handle EPIPE error. */
1455 static int sctp_error(struct sock *sk, int flags, int err)
1457 if (err == -EPIPE)
1458 err = sock_error(sk) ? : -EPIPE;
1459 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1460 send_sig(SIGPIPE, current, 0);
1461 return err;
1464 /* API 3.1.3 sendmsg() - UDP Style Syntax
1466 * An application uses sendmsg() and recvmsg() calls to transmit data to
1467 * and receive data from its peer.
1469 * ssize_t sendmsg(int socket, const struct msghdr *message,
1470 * int flags);
1472 * socket - the socket descriptor of the endpoint.
1473 * message - pointer to the msghdr structure which contains a single
1474 * user message and possibly some ancillary data.
1476 * See Section 5 for complete description of the data
1477 * structures.
1479 * flags - flags sent or received with the user message, see Section
1480 * 5 for complete description of the flags.
1482 * Note: This function could use a rewrite especially when explicit
1483 * connect support comes in.
1485 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1487 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1489 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1490 struct msghdr *msg, size_t msg_len)
1492 struct sctp_sock *sp;
1493 struct sctp_endpoint *ep;
1494 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1495 struct sctp_transport *transport, *chunk_tp;
1496 struct sctp_chunk *chunk;
1497 union sctp_addr to;
1498 struct sockaddr *msg_name = NULL;
1499 struct sctp_sndrcvinfo default_sinfo;
1500 struct sctp_sndrcvinfo *sinfo;
1501 struct sctp_initmsg *sinit;
1502 sctp_assoc_t associd = 0;
1503 sctp_cmsgs_t cmsgs = { NULL };
1504 int err;
1505 sctp_scope_t scope;
1506 long timeo;
1507 __u16 sinfo_flags = 0;
1508 struct sctp_datamsg *datamsg;
1509 int msg_flags = msg->msg_flags;
1511 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1512 sk, msg, msg_len);
1514 err = 0;
1515 sp = sctp_sk(sk);
1516 ep = sp->ep;
1518 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1520 /* We cannot send a message over a TCP-style listening socket. */
1521 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1522 err = -EPIPE;
1523 goto out_nounlock;
1526 /* Parse out the SCTP CMSGs. */
1527 err = sctp_msghdr_parse(msg, &cmsgs);
1529 if (err) {
1530 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1531 goto out_nounlock;
1534 /* Fetch the destination address for this packet. This
1535 * address only selects the association--it is not necessarily
1536 * the address we will send to.
1537 * For a peeled-off socket, msg_name is ignored.
1539 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1540 int msg_namelen = msg->msg_namelen;
1542 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1543 msg_namelen);
1544 if (err)
1545 return err;
1547 if (msg_namelen > sizeof(to))
1548 msg_namelen = sizeof(to);
1549 memcpy(&to, msg->msg_name, msg_namelen);
1550 msg_name = msg->msg_name;
1553 sinfo = cmsgs.info;
1554 sinit = cmsgs.init;
1556 /* Did the user specify SNDRCVINFO? */
1557 if (sinfo) {
1558 sinfo_flags = sinfo->sinfo_flags;
1559 associd = sinfo->sinfo_assoc_id;
1562 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1563 msg_len, sinfo_flags);
1565 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1566 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1567 err = -EINVAL;
1568 goto out_nounlock;
1571 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1572 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1573 * If SCTP_ABORT is set, the message length could be non zero with
1574 * the msg_iov set to the user abort reason.
1576 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1577 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1578 err = -EINVAL;
1579 goto out_nounlock;
1582 /* If SCTP_ADDR_OVER is set, there must be an address
1583 * specified in msg_name.
1585 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1586 err = -EINVAL;
1587 goto out_nounlock;
1590 transport = NULL;
1592 SCTP_DEBUG_PRINTK("About to look up association.\n");
1594 sctp_lock_sock(sk);
1596 /* If a msg_name has been specified, assume this is to be used. */
1597 if (msg_name) {
1598 /* Look for a matching association on the endpoint. */
1599 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1600 if (!asoc) {
1601 /* If we could not find a matching association on the
1602 * endpoint, make sure that it is not a TCP-style
1603 * socket that already has an association or there is
1604 * no peeled-off association on another socket.
1606 if ((sctp_style(sk, TCP) &&
1607 sctp_sstate(sk, ESTABLISHED)) ||
1608 sctp_endpoint_is_peeled_off(ep, &to)) {
1609 err = -EADDRNOTAVAIL;
1610 goto out_unlock;
1613 } else {
1614 asoc = sctp_id2assoc(sk, associd);
1615 if (!asoc) {
1616 err = -EPIPE;
1617 goto out_unlock;
1621 if (asoc) {
1622 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1624 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1625 * socket that has an association in CLOSED state. This can
1626 * happen when an accepted socket has an association that is
1627 * already CLOSED.
1629 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1630 err = -EPIPE;
1631 goto out_unlock;
1634 if (sinfo_flags & SCTP_EOF) {
1635 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1636 asoc);
1637 sctp_primitive_SHUTDOWN(asoc, NULL);
1638 err = 0;
1639 goto out_unlock;
1641 if (sinfo_flags & SCTP_ABORT) {
1643 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1644 if (!chunk) {
1645 err = -ENOMEM;
1646 goto out_unlock;
1649 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1650 sctp_primitive_ABORT(asoc, chunk);
1651 err = 0;
1652 goto out_unlock;
1656 /* Do we need to create the association? */
1657 if (!asoc) {
1658 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1660 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1661 err = -EINVAL;
1662 goto out_unlock;
1665 /* Check for invalid stream against the stream counts,
1666 * either the default or the user specified stream counts.
1668 if (sinfo) {
1669 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1670 /* Check against the defaults. */
1671 if (sinfo->sinfo_stream >=
1672 sp->initmsg.sinit_num_ostreams) {
1673 err = -EINVAL;
1674 goto out_unlock;
1676 } else {
1677 /* Check against the requested. */
1678 if (sinfo->sinfo_stream >=
1679 sinit->sinit_num_ostreams) {
1680 err = -EINVAL;
1681 goto out_unlock;
1687 * API 3.1.2 bind() - UDP Style Syntax
1688 * If a bind() or sctp_bindx() is not called prior to a
1689 * sendmsg() call that initiates a new association, the
1690 * system picks an ephemeral port and will choose an address
1691 * set equivalent to binding with a wildcard address.
1693 if (!ep->base.bind_addr.port) {
1694 if (sctp_autobind(sk)) {
1695 err = -EAGAIN;
1696 goto out_unlock;
1698 } else {
1700 * If an unprivileged user inherits a one-to-many
1701 * style socket with open associations on a privileged
1702 * port, it MAY be permitted to accept new associations,
1703 * but it SHOULD NOT be permitted to open new
1704 * associations.
1706 if (ep->base.bind_addr.port < PROT_SOCK &&
1707 !capable(CAP_NET_BIND_SERVICE)) {
1708 err = -EACCES;
1709 goto out_unlock;
1713 scope = sctp_scope(&to);
1714 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1715 if (!new_asoc) {
1716 err = -ENOMEM;
1717 goto out_unlock;
1719 asoc = new_asoc;
1720 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1721 if (err < 0) {
1722 err = -ENOMEM;
1723 goto out_free;
1726 /* If the SCTP_INIT ancillary data is specified, set all
1727 * the association init values accordingly.
1729 if (sinit) {
1730 if (sinit->sinit_num_ostreams) {
1731 asoc->c.sinit_num_ostreams =
1732 sinit->sinit_num_ostreams;
1734 if (sinit->sinit_max_instreams) {
1735 asoc->c.sinit_max_instreams =
1736 sinit->sinit_max_instreams;
1738 if (sinit->sinit_max_attempts) {
1739 asoc->max_init_attempts
1740 = sinit->sinit_max_attempts;
1742 if (sinit->sinit_max_init_timeo) {
1743 asoc->max_init_timeo =
1744 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1748 /* Prime the peer's transport structures. */
1749 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1750 if (!transport) {
1751 err = -ENOMEM;
1752 goto out_free;
1756 /* ASSERT: we have a valid association at this point. */
1757 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1759 if (!sinfo) {
1760 /* If the user didn't specify SNDRCVINFO, make up one with
1761 * some defaults.
1763 memset(&default_sinfo, 0, sizeof(default_sinfo));
1764 default_sinfo.sinfo_stream = asoc->default_stream;
1765 default_sinfo.sinfo_flags = asoc->default_flags;
1766 default_sinfo.sinfo_ppid = asoc->default_ppid;
1767 default_sinfo.sinfo_context = asoc->default_context;
1768 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1769 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1770 sinfo = &default_sinfo;
1773 /* API 7.1.7, the sndbuf size per association bounds the
1774 * maximum size of data that can be sent in a single send call.
1776 if (msg_len > sk->sk_sndbuf) {
1777 err = -EMSGSIZE;
1778 goto out_free;
1781 if (asoc->pmtu_pending)
1782 sctp_assoc_pending_pmtu(asoc);
1784 /* If fragmentation is disabled and the message length exceeds the
1785 * association fragmentation point, return EMSGSIZE. The I-D
1786 * does not specify what this error is, but this looks like
1787 * a great fit.
1789 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1790 err = -EMSGSIZE;
1791 goto out_free;
1794 /* Check for invalid stream. */
1795 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1796 err = -EINVAL;
1797 goto out_free;
1800 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1801 if (!sctp_wspace(asoc)) {
1802 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1803 if (err)
1804 goto out_free;
1807 /* If an address is passed with the sendto/sendmsg call, it is used
1808 * to override the primary destination address in the TCP model, or
1809 * when SCTP_ADDR_OVER flag is set in the UDP model.
1811 if ((sctp_style(sk, TCP) && msg_name) ||
1812 (sinfo_flags & SCTP_ADDR_OVER)) {
1813 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1814 if (!chunk_tp) {
1815 err = -EINVAL;
1816 goto out_free;
1818 } else
1819 chunk_tp = NULL;
1821 /* Auto-connect, if we aren't connected already. */
1822 if (sctp_state(asoc, CLOSED)) {
1823 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1824 if (err < 0)
1825 goto out_free;
1826 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1829 /* Break the message into multiple chunks of maximum size. */
1830 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1831 if (!datamsg) {
1832 err = -ENOMEM;
1833 goto out_free;
1836 /* Now send the (possibly) fragmented message. */
1837 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1838 sctp_chunk_hold(chunk);
1840 /* Do accounting for the write space. */
1841 sctp_set_owner_w(chunk);
1843 chunk->transport = chunk_tp;
1846 /* Send it to the lower layers. Note: all chunks
1847 * must either fail or succeed. The lower layer
1848 * works that way today. Keep it that way or this
1849 * breaks.
1851 err = sctp_primitive_SEND(asoc, datamsg);
1852 /* Did the lower layer accept the chunk? */
1853 if (err)
1854 sctp_datamsg_free(datamsg);
1855 else
1856 sctp_datamsg_put(datamsg);
1858 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1860 if (err)
1861 goto out_free;
1862 else
1863 err = msg_len;
1865 /* If we are already past ASSOCIATE, the lower
1866 * layers are responsible for association cleanup.
1868 goto out_unlock;
1870 out_free:
1871 if (new_asoc)
1872 sctp_association_free(asoc);
1873 out_unlock:
1874 sctp_release_sock(sk);
1876 out_nounlock:
1877 return sctp_error(sk, msg_flags, err);
1879 #if 0
1880 do_sock_err:
1881 if (msg_len)
1882 err = msg_len;
1883 else
1884 err = sock_error(sk);
1885 goto out;
1887 do_interrupted:
1888 if (msg_len)
1889 err = msg_len;
1890 goto out;
1891 #endif /* 0 */
1894 /* This is an extended version of skb_pull() that removes the data from the
1895 * start of a skb even when data is spread across the list of skb's in the
1896 * frag_list. len specifies the total amount of data that needs to be removed.
1897 * when 'len' bytes could be removed from the skb, it returns 0.
1898 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1899 * could not be removed.
1901 static int sctp_skb_pull(struct sk_buff *skb, int len)
1903 struct sk_buff *list;
1904 int skb_len = skb_headlen(skb);
1905 int rlen;
1907 if (len <= skb_len) {
1908 __skb_pull(skb, len);
1909 return 0;
1911 len -= skb_len;
1912 __skb_pull(skb, skb_len);
1914 skb_walk_frags(skb, list) {
1915 rlen = sctp_skb_pull(list, len);
1916 skb->len -= (len-rlen);
1917 skb->data_len -= (len-rlen);
1919 if (!rlen)
1920 return 0;
1922 len = rlen;
1925 return len;
1928 /* API 3.1.3 recvmsg() - UDP Style Syntax
1930 * ssize_t recvmsg(int socket, struct msghdr *message,
1931 * int flags);
1933 * socket - the socket descriptor of the endpoint.
1934 * message - pointer to the msghdr structure which contains a single
1935 * user message and possibly some ancillary data.
1937 * See Section 5 for complete description of the data
1938 * structures.
1940 * flags - flags sent or received with the user message, see Section
1941 * 5 for complete description of the flags.
1943 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1945 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1946 struct msghdr *msg, size_t len, int noblock,
1947 int flags, int *addr_len)
1949 struct sctp_ulpevent *event = NULL;
1950 struct sctp_sock *sp = sctp_sk(sk);
1951 struct sk_buff *skb;
1952 int copied;
1953 int err = 0;
1954 int skb_len;
1956 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1957 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1958 "len", len, "knoblauch", noblock,
1959 "flags", flags, "addr_len", addr_len);
1961 sctp_lock_sock(sk);
1963 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1964 err = -ENOTCONN;
1965 goto out;
1968 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1969 if (!skb)
1970 goto out;
1972 /* Get the total length of the skb including any skb's in the
1973 * frag_list.
1975 skb_len = skb->len;
1977 copied = skb_len;
1978 if (copied > len)
1979 copied = len;
1981 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1983 event = sctp_skb2event(skb);
1985 if (err)
1986 goto out_free;
1988 sock_recv_ts_and_drops(msg, sk, skb);
1989 if (sctp_ulpevent_is_notification(event)) {
1990 msg->msg_flags |= MSG_NOTIFICATION;
1991 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1992 } else {
1993 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1996 /* Check if we allow SCTP_SNDRCVINFO. */
1997 if (sp->subscribe.sctp_data_io_event)
1998 sctp_ulpevent_read_sndrcvinfo(event, msg);
1999 #if 0
2000 /* FIXME: we should be calling IP/IPv6 layers. */
2001 if (sk->sk_protinfo.af_inet.cmsg_flags)
2002 ip_cmsg_recv(msg, skb);
2003 #endif
2005 err = copied;
2007 /* If skb's length exceeds the user's buffer, update the skb and
2008 * push it back to the receive_queue so that the next call to
2009 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2011 if (skb_len > copied) {
2012 msg->msg_flags &= ~MSG_EOR;
2013 if (flags & MSG_PEEK)
2014 goto out_free;
2015 sctp_skb_pull(skb, copied);
2016 skb_queue_head(&sk->sk_receive_queue, skb);
2018 /* When only partial message is copied to the user, increase
2019 * rwnd by that amount. If all the data in the skb is read,
2020 * rwnd is updated when the event is freed.
2022 if (!sctp_ulpevent_is_notification(event))
2023 sctp_assoc_rwnd_increase(event->asoc, copied);
2024 goto out;
2025 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2026 (event->msg_flags & MSG_EOR))
2027 msg->msg_flags |= MSG_EOR;
2028 else
2029 msg->msg_flags &= ~MSG_EOR;
2031 out_free:
2032 if (flags & MSG_PEEK) {
2033 /* Release the skb reference acquired after peeking the skb in
2034 * sctp_skb_recv_datagram().
2036 kfree_skb(skb);
2037 } else {
2038 /* Free the event which includes releasing the reference to
2039 * the owner of the skb, freeing the skb and updating the
2040 * rwnd.
2042 sctp_ulpevent_free(event);
2044 out:
2045 sctp_release_sock(sk);
2046 return err;
2049 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2051 * This option is a on/off flag. If enabled no SCTP message
2052 * fragmentation will be performed. Instead if a message being sent
2053 * exceeds the current PMTU size, the message will NOT be sent and
2054 * instead a error will be indicated to the user.
2056 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2057 char __user *optval,
2058 unsigned int optlen)
2060 int val;
2062 if (optlen < sizeof(int))
2063 return -EINVAL;
2065 if (get_user(val, (int __user *)optval))
2066 return -EFAULT;
2068 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2070 return 0;
2073 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2074 unsigned int optlen)
2076 if (optlen > sizeof(struct sctp_event_subscribe))
2077 return -EINVAL;
2078 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2079 return -EFAULT;
2080 return 0;
2083 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2085 * This socket option is applicable to the UDP-style socket only. When
2086 * set it will cause associations that are idle for more than the
2087 * specified number of seconds to automatically close. An association
2088 * being idle is defined an association that has NOT sent or received
2089 * user data. The special value of '0' indicates that no automatic
2090 * close of any associations should be performed. The option expects an
2091 * integer defining the number of seconds of idle time before an
2092 * association is closed.
2094 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2095 unsigned int optlen)
2097 struct sctp_sock *sp = sctp_sk(sk);
2099 /* Applicable to UDP-style socket only */
2100 if (sctp_style(sk, TCP))
2101 return -EOPNOTSUPP;
2102 if (optlen != sizeof(int))
2103 return -EINVAL;
2104 if (copy_from_user(&sp->autoclose, optval, optlen))
2105 return -EFAULT;
2106 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2107 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2109 return 0;
2112 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2114 * Applications can enable or disable heartbeats for any peer address of
2115 * an association, modify an address's heartbeat interval, force a
2116 * heartbeat to be sent immediately, and adjust the address's maximum
2117 * number of retransmissions sent before an address is considered
2118 * unreachable. The following structure is used to access and modify an
2119 * address's parameters:
2121 * struct sctp_paddrparams {
2122 * sctp_assoc_t spp_assoc_id;
2123 * struct sockaddr_storage spp_address;
2124 * uint32_t spp_hbinterval;
2125 * uint16_t spp_pathmaxrxt;
2126 * uint32_t spp_pathmtu;
2127 * uint32_t spp_sackdelay;
2128 * uint32_t spp_flags;
2129 * };
2131 * spp_assoc_id - (one-to-many style socket) This is filled in the
2132 * application, and identifies the association for
2133 * this query.
2134 * spp_address - This specifies which address is of interest.
2135 * spp_hbinterval - This contains the value of the heartbeat interval,
2136 * in milliseconds. If a value of zero
2137 * is present in this field then no changes are to
2138 * be made to this parameter.
2139 * spp_pathmaxrxt - This contains the maximum number of
2140 * retransmissions before this address shall be
2141 * considered unreachable. If a value of zero
2142 * is present in this field then no changes are to
2143 * be made to this parameter.
2144 * spp_pathmtu - When Path MTU discovery is disabled the value
2145 * specified here will be the "fixed" path mtu.
2146 * Note that if the spp_address field is empty
2147 * then all associations on this address will
2148 * have this fixed path mtu set upon them.
2150 * spp_sackdelay - When delayed sack is enabled, this value specifies
2151 * the number of milliseconds that sacks will be delayed
2152 * for. This value will apply to all addresses of an
2153 * association if the spp_address field is empty. Note
2154 * also, that if delayed sack is enabled and this
2155 * value is set to 0, no change is made to the last
2156 * recorded delayed sack timer value.
2158 * spp_flags - These flags are used to control various features
2159 * on an association. The flag field may contain
2160 * zero or more of the following options.
2162 * SPP_HB_ENABLE - Enable heartbeats on the
2163 * specified address. Note that if the address
2164 * field is empty all addresses for the association
2165 * have heartbeats enabled upon them.
2167 * SPP_HB_DISABLE - Disable heartbeats on the
2168 * speicifed address. Note that if the address
2169 * field is empty all addresses for the association
2170 * will have their heartbeats disabled. Note also
2171 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2172 * mutually exclusive, only one of these two should
2173 * be specified. Enabling both fields will have
2174 * undetermined results.
2176 * SPP_HB_DEMAND - Request a user initiated heartbeat
2177 * to be made immediately.
2179 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2180 * heartbeat delayis to be set to the value of 0
2181 * milliseconds.
2183 * SPP_PMTUD_ENABLE - This field will enable PMTU
2184 * discovery upon the specified address. Note that
2185 * if the address feild is empty then all addresses
2186 * on the association are effected.
2188 * SPP_PMTUD_DISABLE - This field will disable PMTU
2189 * discovery upon the specified address. Note that
2190 * if the address feild is empty then all addresses
2191 * on the association are effected. Not also that
2192 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2193 * exclusive. Enabling both will have undetermined
2194 * results.
2196 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2197 * on delayed sack. The time specified in spp_sackdelay
2198 * is used to specify the sack delay for this address. Note
2199 * that if spp_address is empty then all addresses will
2200 * enable delayed sack and take on the sack delay
2201 * value specified in spp_sackdelay.
2202 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2203 * off delayed sack. If the spp_address field is blank then
2204 * delayed sack is disabled for the entire association. Note
2205 * also that this field is mutually exclusive to
2206 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2207 * results.
2209 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2210 struct sctp_transport *trans,
2211 struct sctp_association *asoc,
2212 struct sctp_sock *sp,
2213 int hb_change,
2214 int pmtud_change,
2215 int sackdelay_change)
2217 int error;
2219 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2220 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2221 if (error)
2222 return error;
2225 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2226 * this field is ignored. Note also that a value of zero indicates
2227 * the current setting should be left unchanged.
2229 if (params->spp_flags & SPP_HB_ENABLE) {
2231 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2232 * set. This lets us use 0 value when this flag
2233 * is set.
2235 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2236 params->spp_hbinterval = 0;
2238 if (params->spp_hbinterval ||
2239 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2240 if (trans) {
2241 trans->hbinterval =
2242 msecs_to_jiffies(params->spp_hbinterval);
2243 } else if (asoc) {
2244 asoc->hbinterval =
2245 msecs_to_jiffies(params->spp_hbinterval);
2246 } else {
2247 sp->hbinterval = params->spp_hbinterval;
2252 if (hb_change) {
2253 if (trans) {
2254 trans->param_flags =
2255 (trans->param_flags & ~SPP_HB) | hb_change;
2256 } else if (asoc) {
2257 asoc->param_flags =
2258 (asoc->param_flags & ~SPP_HB) | hb_change;
2259 } else {
2260 sp->param_flags =
2261 (sp->param_flags & ~SPP_HB) | hb_change;
2265 /* When Path MTU discovery is disabled the value specified here will
2266 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2267 * include the flag SPP_PMTUD_DISABLE for this field to have any
2268 * effect).
2270 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2271 if (trans) {
2272 trans->pathmtu = params->spp_pathmtu;
2273 sctp_assoc_sync_pmtu(asoc);
2274 } else if (asoc) {
2275 asoc->pathmtu = params->spp_pathmtu;
2276 sctp_frag_point(asoc, params->spp_pathmtu);
2277 } else {
2278 sp->pathmtu = params->spp_pathmtu;
2282 if (pmtud_change) {
2283 if (trans) {
2284 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2285 (params->spp_flags & SPP_PMTUD_ENABLE);
2286 trans->param_flags =
2287 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2288 if (update) {
2289 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2290 sctp_assoc_sync_pmtu(asoc);
2292 } else if (asoc) {
2293 asoc->param_flags =
2294 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2295 } else {
2296 sp->param_flags =
2297 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2301 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2302 * value of this field is ignored. Note also that a value of zero
2303 * indicates the current setting should be left unchanged.
2305 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2306 if (trans) {
2307 trans->sackdelay =
2308 msecs_to_jiffies(params->spp_sackdelay);
2309 } else if (asoc) {
2310 asoc->sackdelay =
2311 msecs_to_jiffies(params->spp_sackdelay);
2312 } else {
2313 sp->sackdelay = params->spp_sackdelay;
2317 if (sackdelay_change) {
2318 if (trans) {
2319 trans->param_flags =
2320 (trans->param_flags & ~SPP_SACKDELAY) |
2321 sackdelay_change;
2322 } else if (asoc) {
2323 asoc->param_flags =
2324 (asoc->param_flags & ~SPP_SACKDELAY) |
2325 sackdelay_change;
2326 } else {
2327 sp->param_flags =
2328 (sp->param_flags & ~SPP_SACKDELAY) |
2329 sackdelay_change;
2333 /* Note that a value of zero indicates the current setting should be
2334 left unchanged.
2336 if (params->spp_pathmaxrxt) {
2337 if (trans) {
2338 trans->pathmaxrxt = params->spp_pathmaxrxt;
2339 } else if (asoc) {
2340 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2341 } else {
2342 sp->pathmaxrxt = params->spp_pathmaxrxt;
2346 return 0;
2349 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2350 char __user *optval,
2351 unsigned int optlen)
2353 struct sctp_paddrparams params;
2354 struct sctp_transport *trans = NULL;
2355 struct sctp_association *asoc = NULL;
2356 struct sctp_sock *sp = sctp_sk(sk);
2357 int error;
2358 int hb_change, pmtud_change, sackdelay_change;
2360 if (optlen != sizeof(struct sctp_paddrparams))
2361 return - EINVAL;
2363 if (copy_from_user(&params, optval, optlen))
2364 return -EFAULT;
2366 /* Validate flags and value parameters. */
2367 hb_change = params.spp_flags & SPP_HB;
2368 pmtud_change = params.spp_flags & SPP_PMTUD;
2369 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2371 if (hb_change == SPP_HB ||
2372 pmtud_change == SPP_PMTUD ||
2373 sackdelay_change == SPP_SACKDELAY ||
2374 params.spp_sackdelay > 500 ||
2375 (params.spp_pathmtu &&
2376 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2377 return -EINVAL;
2379 /* If an address other than INADDR_ANY is specified, and
2380 * no transport is found, then the request is invalid.
2382 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2383 trans = sctp_addr_id2transport(sk, &params.spp_address,
2384 params.spp_assoc_id);
2385 if (!trans)
2386 return -EINVAL;
2389 /* Get association, if assoc_id != 0 and the socket is a one
2390 * to many style socket, and an association was not found, then
2391 * the id was invalid.
2393 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2394 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2395 return -EINVAL;
2397 /* Heartbeat demand can only be sent on a transport or
2398 * association, but not a socket.
2400 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2401 return -EINVAL;
2403 /* Process parameters. */
2404 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2405 hb_change, pmtud_change,
2406 sackdelay_change);
2408 if (error)
2409 return error;
2411 /* If changes are for association, also apply parameters to each
2412 * transport.
2414 if (!trans && asoc) {
2415 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2416 transports) {
2417 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2418 hb_change, pmtud_change,
2419 sackdelay_change);
2423 return 0;
2427 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2429 * This option will effect the way delayed acks are performed. This
2430 * option allows you to get or set the delayed ack time, in
2431 * milliseconds. It also allows changing the delayed ack frequency.
2432 * Changing the frequency to 1 disables the delayed sack algorithm. If
2433 * the assoc_id is 0, then this sets or gets the endpoints default
2434 * values. If the assoc_id field is non-zero, then the set or get
2435 * effects the specified association for the one to many model (the
2436 * assoc_id field is ignored by the one to one model). Note that if
2437 * sack_delay or sack_freq are 0 when setting this option, then the
2438 * current values will remain unchanged.
2440 * struct sctp_sack_info {
2441 * sctp_assoc_t sack_assoc_id;
2442 * uint32_t sack_delay;
2443 * uint32_t sack_freq;
2444 * };
2446 * sack_assoc_id - This parameter, indicates which association the user
2447 * is performing an action upon. Note that if this field's value is
2448 * zero then the endpoints default value is changed (effecting future
2449 * associations only).
2451 * sack_delay - This parameter contains the number of milliseconds that
2452 * the user is requesting the delayed ACK timer be set to. Note that
2453 * this value is defined in the standard to be between 200 and 500
2454 * milliseconds.
2456 * sack_freq - This parameter contains the number of packets that must
2457 * be received before a sack is sent without waiting for the delay
2458 * timer to expire. The default value for this is 2, setting this
2459 * value to 1 will disable the delayed sack algorithm.
2462 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2463 char __user *optval, unsigned int optlen)
2465 struct sctp_sack_info params;
2466 struct sctp_transport *trans = NULL;
2467 struct sctp_association *asoc = NULL;
2468 struct sctp_sock *sp = sctp_sk(sk);
2470 if (optlen == sizeof(struct sctp_sack_info)) {
2471 if (copy_from_user(&params, optval, optlen))
2472 return -EFAULT;
2474 if (params.sack_delay == 0 && params.sack_freq == 0)
2475 return 0;
2476 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2477 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2478 pr_warn("Use struct sctp_sack_info instead\n");
2479 if (copy_from_user(&params, optval, optlen))
2480 return -EFAULT;
2482 if (params.sack_delay == 0)
2483 params.sack_freq = 1;
2484 else
2485 params.sack_freq = 0;
2486 } else
2487 return - EINVAL;
2489 /* Validate value parameter. */
2490 if (params.sack_delay > 500)
2491 return -EINVAL;
2493 /* Get association, if sack_assoc_id != 0 and the socket is a one
2494 * to many style socket, and an association was not found, then
2495 * the id was invalid.
2497 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2498 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2499 return -EINVAL;
2501 if (params.sack_delay) {
2502 if (asoc) {
2503 asoc->sackdelay =
2504 msecs_to_jiffies(params.sack_delay);
2505 asoc->param_flags =
2506 (asoc->param_flags & ~SPP_SACKDELAY) |
2507 SPP_SACKDELAY_ENABLE;
2508 } else {
2509 sp->sackdelay = params.sack_delay;
2510 sp->param_flags =
2511 (sp->param_flags & ~SPP_SACKDELAY) |
2512 SPP_SACKDELAY_ENABLE;
2516 if (params.sack_freq == 1) {
2517 if (asoc) {
2518 asoc->param_flags =
2519 (asoc->param_flags & ~SPP_SACKDELAY) |
2520 SPP_SACKDELAY_DISABLE;
2521 } else {
2522 sp->param_flags =
2523 (sp->param_flags & ~SPP_SACKDELAY) |
2524 SPP_SACKDELAY_DISABLE;
2526 } else if (params.sack_freq > 1) {
2527 if (asoc) {
2528 asoc->sackfreq = params.sack_freq;
2529 asoc->param_flags =
2530 (asoc->param_flags & ~SPP_SACKDELAY) |
2531 SPP_SACKDELAY_ENABLE;
2532 } else {
2533 sp->sackfreq = params.sack_freq;
2534 sp->param_flags =
2535 (sp->param_flags & ~SPP_SACKDELAY) |
2536 SPP_SACKDELAY_ENABLE;
2540 /* If change is for association, also apply to each transport. */
2541 if (asoc) {
2542 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2543 transports) {
2544 if (params.sack_delay) {
2545 trans->sackdelay =
2546 msecs_to_jiffies(params.sack_delay);
2547 trans->param_flags =
2548 (trans->param_flags & ~SPP_SACKDELAY) |
2549 SPP_SACKDELAY_ENABLE;
2551 if (params.sack_freq == 1) {
2552 trans->param_flags =
2553 (trans->param_flags & ~SPP_SACKDELAY) |
2554 SPP_SACKDELAY_DISABLE;
2555 } else if (params.sack_freq > 1) {
2556 trans->sackfreq = params.sack_freq;
2557 trans->param_flags =
2558 (trans->param_flags & ~SPP_SACKDELAY) |
2559 SPP_SACKDELAY_ENABLE;
2564 return 0;
2567 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2569 * Applications can specify protocol parameters for the default association
2570 * initialization. The option name argument to setsockopt() and getsockopt()
2571 * is SCTP_INITMSG.
2573 * Setting initialization parameters is effective only on an unconnected
2574 * socket (for UDP-style sockets only future associations are effected
2575 * by the change). With TCP-style sockets, this option is inherited by
2576 * sockets derived from a listener socket.
2578 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2580 struct sctp_initmsg sinit;
2581 struct sctp_sock *sp = sctp_sk(sk);
2583 if (optlen != sizeof(struct sctp_initmsg))
2584 return -EINVAL;
2585 if (copy_from_user(&sinit, optval, optlen))
2586 return -EFAULT;
2588 if (sinit.sinit_num_ostreams)
2589 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2590 if (sinit.sinit_max_instreams)
2591 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2592 if (sinit.sinit_max_attempts)
2593 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2594 if (sinit.sinit_max_init_timeo)
2595 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2597 return 0;
2601 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2603 * Applications that wish to use the sendto() system call may wish to
2604 * specify a default set of parameters that would normally be supplied
2605 * through the inclusion of ancillary data. This socket option allows
2606 * such an application to set the default sctp_sndrcvinfo structure.
2607 * The application that wishes to use this socket option simply passes
2608 * in to this call the sctp_sndrcvinfo structure defined in Section
2609 * 5.2.2) The input parameters accepted by this call include
2610 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2611 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2612 * to this call if the caller is using the UDP model.
2614 static int sctp_setsockopt_default_send_param(struct sock *sk,
2615 char __user *optval,
2616 unsigned int optlen)
2618 struct sctp_sndrcvinfo info;
2619 struct sctp_association *asoc;
2620 struct sctp_sock *sp = sctp_sk(sk);
2622 if (optlen != sizeof(struct sctp_sndrcvinfo))
2623 return -EINVAL;
2624 if (copy_from_user(&info, optval, optlen))
2625 return -EFAULT;
2627 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2628 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2629 return -EINVAL;
2631 if (asoc) {
2632 asoc->default_stream = info.sinfo_stream;
2633 asoc->default_flags = info.sinfo_flags;
2634 asoc->default_ppid = info.sinfo_ppid;
2635 asoc->default_context = info.sinfo_context;
2636 asoc->default_timetolive = info.sinfo_timetolive;
2637 } else {
2638 sp->default_stream = info.sinfo_stream;
2639 sp->default_flags = info.sinfo_flags;
2640 sp->default_ppid = info.sinfo_ppid;
2641 sp->default_context = info.sinfo_context;
2642 sp->default_timetolive = info.sinfo_timetolive;
2645 return 0;
2648 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2650 * Requests that the local SCTP stack use the enclosed peer address as
2651 * the association primary. The enclosed address must be one of the
2652 * association peer's addresses.
2654 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2655 unsigned int optlen)
2657 struct sctp_prim prim;
2658 struct sctp_transport *trans;
2660 if (optlen != sizeof(struct sctp_prim))
2661 return -EINVAL;
2663 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2664 return -EFAULT;
2666 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2667 if (!trans)
2668 return -EINVAL;
2670 sctp_assoc_set_primary(trans->asoc, trans);
2672 return 0;
2676 * 7.1.5 SCTP_NODELAY
2678 * Turn on/off any Nagle-like algorithm. This means that packets are
2679 * generally sent as soon as possible and no unnecessary delays are
2680 * introduced, at the cost of more packets in the network. Expects an
2681 * integer boolean flag.
2683 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2684 unsigned int optlen)
2686 int val;
2688 if (optlen < sizeof(int))
2689 return -EINVAL;
2690 if (get_user(val, (int __user *)optval))
2691 return -EFAULT;
2693 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2694 return 0;
2699 * 7.1.1 SCTP_RTOINFO
2701 * The protocol parameters used to initialize and bound retransmission
2702 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2703 * and modify these parameters.
2704 * All parameters are time values, in milliseconds. A value of 0, when
2705 * modifying the parameters, indicates that the current value should not
2706 * be changed.
2709 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2711 struct sctp_rtoinfo rtoinfo;
2712 struct sctp_association *asoc;
2714 if (optlen != sizeof (struct sctp_rtoinfo))
2715 return -EINVAL;
2717 if (copy_from_user(&rtoinfo, optval, optlen))
2718 return -EFAULT;
2720 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2722 /* Set the values to the specific association */
2723 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2724 return -EINVAL;
2726 if (asoc) {
2727 if (rtoinfo.srto_initial != 0)
2728 asoc->rto_initial =
2729 msecs_to_jiffies(rtoinfo.srto_initial);
2730 if (rtoinfo.srto_max != 0)
2731 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2732 if (rtoinfo.srto_min != 0)
2733 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2734 } else {
2735 /* If there is no association or the association-id = 0
2736 * set the values to the endpoint.
2738 struct sctp_sock *sp = sctp_sk(sk);
2740 if (rtoinfo.srto_initial != 0)
2741 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2742 if (rtoinfo.srto_max != 0)
2743 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2744 if (rtoinfo.srto_min != 0)
2745 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2748 return 0;
2753 * 7.1.2 SCTP_ASSOCINFO
2755 * This option is used to tune the maximum retransmission attempts
2756 * of the association.
2757 * Returns an error if the new association retransmission value is
2758 * greater than the sum of the retransmission value of the peer.
2759 * See [SCTP] for more information.
2762 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2765 struct sctp_assocparams assocparams;
2766 struct sctp_association *asoc;
2768 if (optlen != sizeof(struct sctp_assocparams))
2769 return -EINVAL;
2770 if (copy_from_user(&assocparams, optval, optlen))
2771 return -EFAULT;
2773 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2775 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2776 return -EINVAL;
2778 /* Set the values to the specific association */
2779 if (asoc) {
2780 if (assocparams.sasoc_asocmaxrxt != 0) {
2781 __u32 path_sum = 0;
2782 int paths = 0;
2783 struct sctp_transport *peer_addr;
2785 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2786 transports) {
2787 path_sum += peer_addr->pathmaxrxt;
2788 paths++;
2791 /* Only validate asocmaxrxt if we have more than
2792 * one path/transport. We do this because path
2793 * retransmissions are only counted when we have more
2794 * then one path.
2796 if (paths > 1 &&
2797 assocparams.sasoc_asocmaxrxt > path_sum)
2798 return -EINVAL;
2800 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2803 if (assocparams.sasoc_cookie_life != 0) {
2804 asoc->cookie_life.tv_sec =
2805 assocparams.sasoc_cookie_life / 1000;
2806 asoc->cookie_life.tv_usec =
2807 (assocparams.sasoc_cookie_life % 1000)
2808 * 1000;
2810 } else {
2811 /* Set the values to the endpoint */
2812 struct sctp_sock *sp = sctp_sk(sk);
2814 if (assocparams.sasoc_asocmaxrxt != 0)
2815 sp->assocparams.sasoc_asocmaxrxt =
2816 assocparams.sasoc_asocmaxrxt;
2817 if (assocparams.sasoc_cookie_life != 0)
2818 sp->assocparams.sasoc_cookie_life =
2819 assocparams.sasoc_cookie_life;
2821 return 0;
2825 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2827 * This socket option is a boolean flag which turns on or off mapped V4
2828 * addresses. If this option is turned on and the socket is type
2829 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2830 * If this option is turned off, then no mapping will be done of V4
2831 * addresses and a user will receive both PF_INET6 and PF_INET type
2832 * addresses on the socket.
2834 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2836 int val;
2837 struct sctp_sock *sp = sctp_sk(sk);
2839 if (optlen < sizeof(int))
2840 return -EINVAL;
2841 if (get_user(val, (int __user *)optval))
2842 return -EFAULT;
2843 if (val)
2844 sp->v4mapped = 1;
2845 else
2846 sp->v4mapped = 0;
2848 return 0;
2852 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2853 * This option will get or set the maximum size to put in any outgoing
2854 * SCTP DATA chunk. If a message is larger than this size it will be
2855 * fragmented by SCTP into the specified size. Note that the underlying
2856 * SCTP implementation may fragment into smaller sized chunks when the
2857 * PMTU of the underlying association is smaller than the value set by
2858 * the user. The default value for this option is '0' which indicates
2859 * the user is NOT limiting fragmentation and only the PMTU will effect
2860 * SCTP's choice of DATA chunk size. Note also that values set larger
2861 * than the maximum size of an IP datagram will effectively let SCTP
2862 * control fragmentation (i.e. the same as setting this option to 0).
2864 * The following structure is used to access and modify this parameter:
2866 * struct sctp_assoc_value {
2867 * sctp_assoc_t assoc_id;
2868 * uint32_t assoc_value;
2869 * };
2871 * assoc_id: This parameter is ignored for one-to-one style sockets.
2872 * For one-to-many style sockets this parameter indicates which
2873 * association the user is performing an action upon. Note that if
2874 * this field's value is zero then the endpoints default value is
2875 * changed (effecting future associations only).
2876 * assoc_value: This parameter specifies the maximum size in bytes.
2878 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2880 struct sctp_assoc_value params;
2881 struct sctp_association *asoc;
2882 struct sctp_sock *sp = sctp_sk(sk);
2883 int val;
2885 if (optlen == sizeof(int)) {
2886 pr_warn("Use of int in maxseg socket option deprecated\n");
2887 pr_warn("Use struct sctp_assoc_value instead\n");
2888 if (copy_from_user(&val, optval, optlen))
2889 return -EFAULT;
2890 params.assoc_id = 0;
2891 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2892 if (copy_from_user(&params, optval, optlen))
2893 return -EFAULT;
2894 val = params.assoc_value;
2895 } else
2896 return -EINVAL;
2898 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2899 return -EINVAL;
2901 asoc = sctp_id2assoc(sk, params.assoc_id);
2902 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2903 return -EINVAL;
2905 if (asoc) {
2906 if (val == 0) {
2907 val = asoc->pathmtu;
2908 val -= sp->pf->af->net_header_len;
2909 val -= sizeof(struct sctphdr) +
2910 sizeof(struct sctp_data_chunk);
2912 asoc->user_frag = val;
2913 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2914 } else {
2915 sp->user_frag = val;
2918 return 0;
2923 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2925 * Requests that the peer mark the enclosed address as the association
2926 * primary. The enclosed address must be one of the association's
2927 * locally bound addresses. The following structure is used to make a
2928 * set primary request:
2930 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2931 unsigned int optlen)
2933 struct sctp_sock *sp;
2934 struct sctp_association *asoc = NULL;
2935 struct sctp_setpeerprim prim;
2936 struct sctp_chunk *chunk;
2937 struct sctp_af *af;
2938 int err;
2940 sp = sctp_sk(sk);
2942 if (!sctp_addip_enable)
2943 return -EPERM;
2945 if (optlen != sizeof(struct sctp_setpeerprim))
2946 return -EINVAL;
2948 if (copy_from_user(&prim, optval, optlen))
2949 return -EFAULT;
2951 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2952 if (!asoc)
2953 return -EINVAL;
2955 if (!asoc->peer.asconf_capable)
2956 return -EPERM;
2958 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2959 return -EPERM;
2961 if (!sctp_state(asoc, ESTABLISHED))
2962 return -ENOTCONN;
2964 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
2965 if (!af)
2966 return -EINVAL;
2968 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
2969 return -EADDRNOTAVAIL;
2971 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2972 return -EADDRNOTAVAIL;
2974 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2975 chunk = sctp_make_asconf_set_prim(asoc,
2976 (union sctp_addr *)&prim.sspp_addr);
2977 if (!chunk)
2978 return -ENOMEM;
2980 err = sctp_send_asconf(asoc, chunk);
2982 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2984 return err;
2987 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2988 unsigned int optlen)
2990 struct sctp_setadaptation adaptation;
2992 if (optlen != sizeof(struct sctp_setadaptation))
2993 return -EINVAL;
2994 if (copy_from_user(&adaptation, optval, optlen))
2995 return -EFAULT;
2997 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2999 return 0;
3003 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3005 * The context field in the sctp_sndrcvinfo structure is normally only
3006 * used when a failed message is retrieved holding the value that was
3007 * sent down on the actual send call. This option allows the setting of
3008 * a default context on an association basis that will be received on
3009 * reading messages from the peer. This is especially helpful in the
3010 * one-2-many model for an application to keep some reference to an
3011 * internal state machine that is processing messages on the
3012 * association. Note that the setting of this value only effects
3013 * received messages from the peer and does not effect the value that is
3014 * saved with outbound messages.
3016 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3017 unsigned int optlen)
3019 struct sctp_assoc_value params;
3020 struct sctp_sock *sp;
3021 struct sctp_association *asoc;
3023 if (optlen != sizeof(struct sctp_assoc_value))
3024 return -EINVAL;
3025 if (copy_from_user(&params, optval, optlen))
3026 return -EFAULT;
3028 sp = sctp_sk(sk);
3030 if (params.assoc_id != 0) {
3031 asoc = sctp_id2assoc(sk, params.assoc_id);
3032 if (!asoc)
3033 return -EINVAL;
3034 asoc->default_rcv_context = params.assoc_value;
3035 } else {
3036 sp->default_rcv_context = params.assoc_value;
3039 return 0;
3043 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3045 * This options will at a minimum specify if the implementation is doing
3046 * fragmented interleave. Fragmented interleave, for a one to many
3047 * socket, is when subsequent calls to receive a message may return
3048 * parts of messages from different associations. Some implementations
3049 * may allow you to turn this value on or off. If so, when turned off,
3050 * no fragment interleave will occur (which will cause a head of line
3051 * blocking amongst multiple associations sharing the same one to many
3052 * socket). When this option is turned on, then each receive call may
3053 * come from a different association (thus the user must receive data
3054 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3055 * association each receive belongs to.
3057 * This option takes a boolean value. A non-zero value indicates that
3058 * fragmented interleave is on. A value of zero indicates that
3059 * fragmented interleave is off.
3061 * Note that it is important that an implementation that allows this
3062 * option to be turned on, have it off by default. Otherwise an unaware
3063 * application using the one to many model may become confused and act
3064 * incorrectly.
3066 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3067 char __user *optval,
3068 unsigned int optlen)
3070 int val;
3072 if (optlen != sizeof(int))
3073 return -EINVAL;
3074 if (get_user(val, (int __user *)optval))
3075 return -EFAULT;
3077 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3079 return 0;
3083 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3084 * (SCTP_PARTIAL_DELIVERY_POINT)
3086 * This option will set or get the SCTP partial delivery point. This
3087 * point is the size of a message where the partial delivery API will be
3088 * invoked to help free up rwnd space for the peer. Setting this to a
3089 * lower value will cause partial deliveries to happen more often. The
3090 * calls argument is an integer that sets or gets the partial delivery
3091 * point. Note also that the call will fail if the user attempts to set
3092 * this value larger than the socket receive buffer size.
3094 * Note that any single message having a length smaller than or equal to
3095 * the SCTP partial delivery point will be delivered in one single read
3096 * call as long as the user provided buffer is large enough to hold the
3097 * message.
3099 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3100 char __user *optval,
3101 unsigned int optlen)
3103 u32 val;
3105 if (optlen != sizeof(u32))
3106 return -EINVAL;
3107 if (get_user(val, (int __user *)optval))
3108 return -EFAULT;
3110 /* Note: We double the receive buffer from what the user sets
3111 * it to be, also initial rwnd is based on rcvbuf/2.
3113 if (val > (sk->sk_rcvbuf >> 1))
3114 return -EINVAL;
3116 sctp_sk(sk)->pd_point = val;
3118 return 0; /* is this the right error code? */
3122 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3124 * This option will allow a user to change the maximum burst of packets
3125 * that can be emitted by this association. Note that the default value
3126 * is 4, and some implementations may restrict this setting so that it
3127 * can only be lowered.
3129 * NOTE: This text doesn't seem right. Do this on a socket basis with
3130 * future associations inheriting the socket value.
3132 static int sctp_setsockopt_maxburst(struct sock *sk,
3133 char __user *optval,
3134 unsigned int optlen)
3136 struct sctp_assoc_value params;
3137 struct sctp_sock *sp;
3138 struct sctp_association *asoc;
3139 int val;
3140 int assoc_id = 0;
3142 if (optlen == sizeof(int)) {
3143 pr_warn("Use of int in max_burst socket option deprecated\n");
3144 pr_warn("Use struct sctp_assoc_value instead\n");
3145 if (copy_from_user(&val, optval, optlen))
3146 return -EFAULT;
3147 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3148 if (copy_from_user(&params, optval, optlen))
3149 return -EFAULT;
3150 val = params.assoc_value;
3151 assoc_id = params.assoc_id;
3152 } else
3153 return -EINVAL;
3155 sp = sctp_sk(sk);
3157 if (assoc_id != 0) {
3158 asoc = sctp_id2assoc(sk, assoc_id);
3159 if (!asoc)
3160 return -EINVAL;
3161 asoc->max_burst = val;
3162 } else
3163 sp->max_burst = val;
3165 return 0;
3169 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3171 * This set option adds a chunk type that the user is requesting to be
3172 * received only in an authenticated way. Changes to the list of chunks
3173 * will only effect future associations on the socket.
3175 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3176 char __user *optval,
3177 unsigned int optlen)
3179 struct sctp_authchunk val;
3181 if (!sctp_auth_enable)
3182 return -EACCES;
3184 if (optlen != sizeof(struct sctp_authchunk))
3185 return -EINVAL;
3186 if (copy_from_user(&val, optval, optlen))
3187 return -EFAULT;
3189 switch (val.sauth_chunk) {
3190 case SCTP_CID_INIT:
3191 case SCTP_CID_INIT_ACK:
3192 case SCTP_CID_SHUTDOWN_COMPLETE:
3193 case SCTP_CID_AUTH:
3194 return -EINVAL;
3197 /* add this chunk id to the endpoint */
3198 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3202 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3204 * This option gets or sets the list of HMAC algorithms that the local
3205 * endpoint requires the peer to use.
3207 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3208 char __user *optval,
3209 unsigned int optlen)
3211 struct sctp_hmacalgo *hmacs;
3212 u32 idents;
3213 int err;
3215 if (!sctp_auth_enable)
3216 return -EACCES;
3218 if (optlen < sizeof(struct sctp_hmacalgo))
3219 return -EINVAL;
3221 hmacs= memdup_user(optval, optlen);
3222 if (IS_ERR(hmacs))
3223 return PTR_ERR(hmacs);
3225 idents = hmacs->shmac_num_idents;
3226 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3227 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3228 err = -EINVAL;
3229 goto out;
3232 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3233 out:
3234 kfree(hmacs);
3235 return err;
3239 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3241 * This option will set a shared secret key which is used to build an
3242 * association shared key.
3244 static int sctp_setsockopt_auth_key(struct sock *sk,
3245 char __user *optval,
3246 unsigned int optlen)
3248 struct sctp_authkey *authkey;
3249 struct sctp_association *asoc;
3250 int ret;
3252 if (!sctp_auth_enable)
3253 return -EACCES;
3255 if (optlen <= sizeof(struct sctp_authkey))
3256 return -EINVAL;
3258 authkey= memdup_user(optval, optlen);
3259 if (IS_ERR(authkey))
3260 return PTR_ERR(authkey);
3262 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3263 ret = -EINVAL;
3264 goto out;
3267 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3268 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3269 ret = -EINVAL;
3270 goto out;
3273 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3274 out:
3275 kfree(authkey);
3276 return ret;
3280 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3282 * This option will get or set the active shared key to be used to build
3283 * the association shared key.
3285 static int sctp_setsockopt_active_key(struct sock *sk,
3286 char __user *optval,
3287 unsigned int optlen)
3289 struct sctp_authkeyid val;
3290 struct sctp_association *asoc;
3292 if (!sctp_auth_enable)
3293 return -EACCES;
3295 if (optlen != sizeof(struct sctp_authkeyid))
3296 return -EINVAL;
3297 if (copy_from_user(&val, optval, optlen))
3298 return -EFAULT;
3300 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3301 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3302 return -EINVAL;
3304 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3305 val.scact_keynumber);
3309 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3311 * This set option will delete a shared secret key from use.
3313 static int sctp_setsockopt_del_key(struct sock *sk,
3314 char __user *optval,
3315 unsigned int optlen)
3317 struct sctp_authkeyid val;
3318 struct sctp_association *asoc;
3320 if (!sctp_auth_enable)
3321 return -EACCES;
3323 if (optlen != sizeof(struct sctp_authkeyid))
3324 return -EINVAL;
3325 if (copy_from_user(&val, optval, optlen))
3326 return -EFAULT;
3328 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3329 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3330 return -EINVAL;
3332 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3333 val.scact_keynumber);
3338 /* API 6.2 setsockopt(), getsockopt()
3340 * Applications use setsockopt() and getsockopt() to set or retrieve
3341 * socket options. Socket options are used to change the default
3342 * behavior of sockets calls. They are described in Section 7.
3344 * The syntax is:
3346 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3347 * int __user *optlen);
3348 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3349 * int optlen);
3351 * sd - the socket descript.
3352 * level - set to IPPROTO_SCTP for all SCTP options.
3353 * optname - the option name.
3354 * optval - the buffer to store the value of the option.
3355 * optlen - the size of the buffer.
3357 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3358 char __user *optval, unsigned int optlen)
3360 int retval = 0;
3362 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3363 sk, optname);
3365 /* I can hardly begin to describe how wrong this is. This is
3366 * so broken as to be worse than useless. The API draft
3367 * REALLY is NOT helpful here... I am not convinced that the
3368 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3369 * are at all well-founded.
3371 if (level != SOL_SCTP) {
3372 struct sctp_af *af = sctp_sk(sk)->pf->af;
3373 retval = af->setsockopt(sk, level, optname, optval, optlen);
3374 goto out_nounlock;
3377 sctp_lock_sock(sk);
3379 switch (optname) {
3380 case SCTP_SOCKOPT_BINDX_ADD:
3381 /* 'optlen' is the size of the addresses buffer. */
3382 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3383 optlen, SCTP_BINDX_ADD_ADDR);
3384 break;
3386 case SCTP_SOCKOPT_BINDX_REM:
3387 /* 'optlen' is the size of the addresses buffer. */
3388 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3389 optlen, SCTP_BINDX_REM_ADDR);
3390 break;
3392 case SCTP_SOCKOPT_CONNECTX_OLD:
3393 /* 'optlen' is the size of the addresses buffer. */
3394 retval = sctp_setsockopt_connectx_old(sk,
3395 (struct sockaddr __user *)optval,
3396 optlen);
3397 break;
3399 case SCTP_SOCKOPT_CONNECTX:
3400 /* 'optlen' is the size of the addresses buffer. */
3401 retval = sctp_setsockopt_connectx(sk,
3402 (struct sockaddr __user *)optval,
3403 optlen);
3404 break;
3406 case SCTP_DISABLE_FRAGMENTS:
3407 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3408 break;
3410 case SCTP_EVENTS:
3411 retval = sctp_setsockopt_events(sk, optval, optlen);
3412 break;
3414 case SCTP_AUTOCLOSE:
3415 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3416 break;
3418 case SCTP_PEER_ADDR_PARAMS:
3419 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3420 break;
3422 case SCTP_DELAYED_SACK:
3423 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3424 break;
3425 case SCTP_PARTIAL_DELIVERY_POINT:
3426 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3427 break;
3429 case SCTP_INITMSG:
3430 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3431 break;
3432 case SCTP_DEFAULT_SEND_PARAM:
3433 retval = sctp_setsockopt_default_send_param(sk, optval,
3434 optlen);
3435 break;
3436 case SCTP_PRIMARY_ADDR:
3437 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3438 break;
3439 case SCTP_SET_PEER_PRIMARY_ADDR:
3440 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3441 break;
3442 case SCTP_NODELAY:
3443 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3444 break;
3445 case SCTP_RTOINFO:
3446 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3447 break;
3448 case SCTP_ASSOCINFO:
3449 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3450 break;
3451 case SCTP_I_WANT_MAPPED_V4_ADDR:
3452 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3453 break;
3454 case SCTP_MAXSEG:
3455 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3456 break;
3457 case SCTP_ADAPTATION_LAYER:
3458 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3459 break;
3460 case SCTP_CONTEXT:
3461 retval = sctp_setsockopt_context(sk, optval, optlen);
3462 break;
3463 case SCTP_FRAGMENT_INTERLEAVE:
3464 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3465 break;
3466 case SCTP_MAX_BURST:
3467 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3468 break;
3469 case SCTP_AUTH_CHUNK:
3470 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3471 break;
3472 case SCTP_HMAC_IDENT:
3473 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3474 break;
3475 case SCTP_AUTH_KEY:
3476 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3477 break;
3478 case SCTP_AUTH_ACTIVE_KEY:
3479 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3480 break;
3481 case SCTP_AUTH_DELETE_KEY:
3482 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3483 break;
3484 default:
3485 retval = -ENOPROTOOPT;
3486 break;
3489 sctp_release_sock(sk);
3491 out_nounlock:
3492 return retval;
3495 /* API 3.1.6 connect() - UDP Style Syntax
3497 * An application may use the connect() call in the UDP model to initiate an
3498 * association without sending data.
3500 * The syntax is:
3502 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3504 * sd: the socket descriptor to have a new association added to.
3506 * nam: the address structure (either struct sockaddr_in or struct
3507 * sockaddr_in6 defined in RFC2553 [7]).
3509 * len: the size of the address.
3511 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3512 int addr_len)
3514 int err = 0;
3515 struct sctp_af *af;
3517 sctp_lock_sock(sk);
3519 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3520 __func__, sk, addr, addr_len);
3522 /* Validate addr_len before calling common connect/connectx routine. */
3523 af = sctp_get_af_specific(addr->sa_family);
3524 if (!af || addr_len < af->sockaddr_len) {
3525 err = -EINVAL;
3526 } else {
3527 /* Pass correct addr len to common routine (so it knows there
3528 * is only one address being passed.
3530 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3533 sctp_release_sock(sk);
3534 return err;
3537 /* FIXME: Write comments. */
3538 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3540 return -EOPNOTSUPP; /* STUB */
3543 /* 4.1.4 accept() - TCP Style Syntax
3545 * Applications use accept() call to remove an established SCTP
3546 * association from the accept queue of the endpoint. A new socket
3547 * descriptor will be returned from accept() to represent the newly
3548 * formed association.
3550 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3552 struct sctp_sock *sp;
3553 struct sctp_endpoint *ep;
3554 struct sock *newsk = NULL;
3555 struct sctp_association *asoc;
3556 long timeo;
3557 int error = 0;
3559 sctp_lock_sock(sk);
3561 sp = sctp_sk(sk);
3562 ep = sp->ep;
3564 if (!sctp_style(sk, TCP)) {
3565 error = -EOPNOTSUPP;
3566 goto out;
3569 if (!sctp_sstate(sk, LISTENING)) {
3570 error = -EINVAL;
3571 goto out;
3574 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3576 error = sctp_wait_for_accept(sk, timeo);
3577 if (error)
3578 goto out;
3580 /* We treat the list of associations on the endpoint as the accept
3581 * queue and pick the first association on the list.
3583 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3585 newsk = sp->pf->create_accept_sk(sk, asoc);
3586 if (!newsk) {
3587 error = -ENOMEM;
3588 goto out;
3591 /* Populate the fields of the newsk from the oldsk and migrate the
3592 * asoc to the newsk.
3594 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3596 out:
3597 sctp_release_sock(sk);
3598 *err = error;
3599 return newsk;
3602 /* The SCTP ioctl handler. */
3603 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3605 int rc = -ENOTCONN;
3607 sctp_lock_sock(sk);
3610 * SEQPACKET-style sockets in LISTENING state are valid, for
3611 * SCTP, so only discard TCP-style sockets in LISTENING state.
3613 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3614 goto out;
3616 switch (cmd) {
3617 case SIOCINQ: {
3618 struct sk_buff *skb;
3619 unsigned int amount = 0;
3621 skb = skb_peek(&sk->sk_receive_queue);
3622 if (skb != NULL) {
3624 * We will only return the amount of this packet since
3625 * that is all that will be read.
3627 amount = skb->len;
3629 rc = put_user(amount, (int __user *)arg);
3630 break;
3632 default:
3633 rc = -ENOIOCTLCMD;
3634 break;
3636 out:
3637 sctp_release_sock(sk);
3638 return rc;
3641 /* This is the function which gets called during socket creation to
3642 * initialized the SCTP-specific portion of the sock.
3643 * The sock structure should already be zero-filled memory.
3645 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3647 struct sctp_endpoint *ep;
3648 struct sctp_sock *sp;
3650 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3652 sp = sctp_sk(sk);
3654 /* Initialize the SCTP per socket area. */
3655 switch (sk->sk_type) {
3656 case SOCK_SEQPACKET:
3657 sp->type = SCTP_SOCKET_UDP;
3658 break;
3659 case SOCK_STREAM:
3660 sp->type = SCTP_SOCKET_TCP;
3661 break;
3662 default:
3663 return -ESOCKTNOSUPPORT;
3666 /* Initialize default send parameters. These parameters can be
3667 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3669 sp->default_stream = 0;
3670 sp->default_ppid = 0;
3671 sp->default_flags = 0;
3672 sp->default_context = 0;
3673 sp->default_timetolive = 0;
3675 sp->default_rcv_context = 0;
3676 sp->max_burst = sctp_max_burst;
3678 /* Initialize default setup parameters. These parameters
3679 * can be modified with the SCTP_INITMSG socket option or
3680 * overridden by the SCTP_INIT CMSG.
3682 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3683 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3684 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3685 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3687 /* Initialize default RTO related parameters. These parameters can
3688 * be modified for with the SCTP_RTOINFO socket option.
3690 sp->rtoinfo.srto_initial = sctp_rto_initial;
3691 sp->rtoinfo.srto_max = sctp_rto_max;
3692 sp->rtoinfo.srto_min = sctp_rto_min;
3694 /* Initialize default association related parameters. These parameters
3695 * can be modified with the SCTP_ASSOCINFO socket option.
3697 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3698 sp->assocparams.sasoc_number_peer_destinations = 0;
3699 sp->assocparams.sasoc_peer_rwnd = 0;
3700 sp->assocparams.sasoc_local_rwnd = 0;
3701 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3703 /* Initialize default event subscriptions. By default, all the
3704 * options are off.
3706 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3708 /* Default Peer Address Parameters. These defaults can
3709 * be modified via SCTP_PEER_ADDR_PARAMS
3711 sp->hbinterval = sctp_hb_interval;
3712 sp->pathmaxrxt = sctp_max_retrans_path;
3713 sp->pathmtu = 0; // allow default discovery
3714 sp->sackdelay = sctp_sack_timeout;
3715 sp->sackfreq = 2;
3716 sp->param_flags = SPP_HB_ENABLE |
3717 SPP_PMTUD_ENABLE |
3718 SPP_SACKDELAY_ENABLE;
3720 /* If enabled no SCTP message fragmentation will be performed.
3721 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3723 sp->disable_fragments = 0;
3725 /* Enable Nagle algorithm by default. */
3726 sp->nodelay = 0;
3728 /* Enable by default. */
3729 sp->v4mapped = 1;
3731 /* Auto-close idle associations after the configured
3732 * number of seconds. A value of 0 disables this
3733 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3734 * for UDP-style sockets only.
3736 sp->autoclose = 0;
3738 /* User specified fragmentation limit. */
3739 sp->user_frag = 0;
3741 sp->adaptation_ind = 0;
3743 sp->pf = sctp_get_pf_specific(sk->sk_family);
3745 /* Control variables for partial data delivery. */
3746 atomic_set(&sp->pd_mode, 0);
3747 skb_queue_head_init(&sp->pd_lobby);
3748 sp->frag_interleave = 0;
3750 /* Create a per socket endpoint structure. Even if we
3751 * change the data structure relationships, this may still
3752 * be useful for storing pre-connect address information.
3754 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3755 if (!ep)
3756 return -ENOMEM;
3758 sp->ep = ep;
3759 sp->hmac = NULL;
3761 SCTP_DBG_OBJCNT_INC(sock);
3763 local_bh_disable();
3764 percpu_counter_inc(&sctp_sockets_allocated);
3765 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3766 local_bh_enable();
3768 return 0;
3771 /* Cleanup any SCTP per socket resources. */
3772 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3774 struct sctp_endpoint *ep;
3776 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3778 /* Release our hold on the endpoint. */
3779 ep = sctp_sk(sk)->ep;
3780 sctp_endpoint_free(ep);
3781 local_bh_disable();
3782 percpu_counter_dec(&sctp_sockets_allocated);
3783 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3784 local_bh_enable();
3787 /* API 4.1.7 shutdown() - TCP Style Syntax
3788 * int shutdown(int socket, int how);
3790 * sd - the socket descriptor of the association to be closed.
3791 * how - Specifies the type of shutdown. The values are
3792 * as follows:
3793 * SHUT_RD
3794 * Disables further receive operations. No SCTP
3795 * protocol action is taken.
3796 * SHUT_WR
3797 * Disables further send operations, and initiates
3798 * the SCTP shutdown sequence.
3799 * SHUT_RDWR
3800 * Disables further send and receive operations
3801 * and initiates the SCTP shutdown sequence.
3803 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3805 struct sctp_endpoint *ep;
3806 struct sctp_association *asoc;
3808 if (!sctp_style(sk, TCP))
3809 return;
3811 if (how & SEND_SHUTDOWN) {
3812 ep = sctp_sk(sk)->ep;
3813 if (!list_empty(&ep->asocs)) {
3814 asoc = list_entry(ep->asocs.next,
3815 struct sctp_association, asocs);
3816 sctp_primitive_SHUTDOWN(asoc, NULL);
3821 /* 7.2.1 Association Status (SCTP_STATUS)
3823 * Applications can retrieve current status information about an
3824 * association, including association state, peer receiver window size,
3825 * number of unacked data chunks, and number of data chunks pending
3826 * receipt. This information is read-only.
3828 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3829 char __user *optval,
3830 int __user *optlen)
3832 struct sctp_status status;
3833 struct sctp_association *asoc = NULL;
3834 struct sctp_transport *transport;
3835 sctp_assoc_t associd;
3836 int retval = 0;
3838 if (len < sizeof(status)) {
3839 retval = -EINVAL;
3840 goto out;
3843 len = sizeof(status);
3844 if (copy_from_user(&status, optval, len)) {
3845 retval = -EFAULT;
3846 goto out;
3849 associd = status.sstat_assoc_id;
3850 asoc = sctp_id2assoc(sk, associd);
3851 if (!asoc) {
3852 retval = -EINVAL;
3853 goto out;
3856 transport = asoc->peer.primary_path;
3858 status.sstat_assoc_id = sctp_assoc2id(asoc);
3859 status.sstat_state = asoc->state;
3860 status.sstat_rwnd = asoc->peer.rwnd;
3861 status.sstat_unackdata = asoc->unack_data;
3863 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3864 status.sstat_instrms = asoc->c.sinit_max_instreams;
3865 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3866 status.sstat_fragmentation_point = asoc->frag_point;
3867 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3868 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3869 transport->af_specific->sockaddr_len);
3870 /* Map ipv4 address into v4-mapped-on-v6 address. */
3871 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3872 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3873 status.sstat_primary.spinfo_state = transport->state;
3874 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3875 status.sstat_primary.spinfo_srtt = transport->srtt;
3876 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3877 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3879 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3880 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3882 if (put_user(len, optlen)) {
3883 retval = -EFAULT;
3884 goto out;
3887 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3888 len, status.sstat_state, status.sstat_rwnd,
3889 status.sstat_assoc_id);
3891 if (copy_to_user(optval, &status, len)) {
3892 retval = -EFAULT;
3893 goto out;
3896 out:
3897 return retval;
3901 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3903 * Applications can retrieve information about a specific peer address
3904 * of an association, including its reachability state, congestion
3905 * window, and retransmission timer values. This information is
3906 * read-only.
3908 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3909 char __user *optval,
3910 int __user *optlen)
3912 struct sctp_paddrinfo pinfo;
3913 struct sctp_transport *transport;
3914 int retval = 0;
3916 if (len < sizeof(pinfo)) {
3917 retval = -EINVAL;
3918 goto out;
3921 len = sizeof(pinfo);
3922 if (copy_from_user(&pinfo, optval, len)) {
3923 retval = -EFAULT;
3924 goto out;
3927 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3928 pinfo.spinfo_assoc_id);
3929 if (!transport)
3930 return -EINVAL;
3932 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3933 pinfo.spinfo_state = transport->state;
3934 pinfo.spinfo_cwnd = transport->cwnd;
3935 pinfo.spinfo_srtt = transport->srtt;
3936 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3937 pinfo.spinfo_mtu = transport->pathmtu;
3939 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3940 pinfo.spinfo_state = SCTP_ACTIVE;
3942 if (put_user(len, optlen)) {
3943 retval = -EFAULT;
3944 goto out;
3947 if (copy_to_user(optval, &pinfo, len)) {
3948 retval = -EFAULT;
3949 goto out;
3952 out:
3953 return retval;
3956 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3958 * This option is a on/off flag. If enabled no SCTP message
3959 * fragmentation will be performed. Instead if a message being sent
3960 * exceeds the current PMTU size, the message will NOT be sent and
3961 * instead a error will be indicated to the user.
3963 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3964 char __user *optval, int __user *optlen)
3966 int val;
3968 if (len < sizeof(int))
3969 return -EINVAL;
3971 len = sizeof(int);
3972 val = (sctp_sk(sk)->disable_fragments == 1);
3973 if (put_user(len, optlen))
3974 return -EFAULT;
3975 if (copy_to_user(optval, &val, len))
3976 return -EFAULT;
3977 return 0;
3980 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3982 * This socket option is used to specify various notifications and
3983 * ancillary data the user wishes to receive.
3985 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3986 int __user *optlen)
3988 if (len < sizeof(struct sctp_event_subscribe))
3989 return -EINVAL;
3990 len = sizeof(struct sctp_event_subscribe);
3991 if (put_user(len, optlen))
3992 return -EFAULT;
3993 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
3994 return -EFAULT;
3995 return 0;
3998 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4000 * This socket option is applicable to the UDP-style socket only. When
4001 * set it will cause associations that are idle for more than the
4002 * specified number of seconds to automatically close. An association
4003 * being idle is defined an association that has NOT sent or received
4004 * user data. The special value of '0' indicates that no automatic
4005 * close of any associations should be performed. The option expects an
4006 * integer defining the number of seconds of idle time before an
4007 * association is closed.
4009 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4011 /* Applicable to UDP-style socket only */
4012 if (sctp_style(sk, TCP))
4013 return -EOPNOTSUPP;
4014 if (len < sizeof(int))
4015 return -EINVAL;
4016 len = sizeof(int);
4017 if (put_user(len, optlen))
4018 return -EFAULT;
4019 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4020 return -EFAULT;
4021 return 0;
4024 /* Helper routine to branch off an association to a new socket. */
4025 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4026 struct socket **sockp)
4028 struct sock *sk = asoc->base.sk;
4029 struct socket *sock;
4030 struct sctp_af *af;
4031 int err = 0;
4033 /* An association cannot be branched off from an already peeled-off
4034 * socket, nor is this supported for tcp style sockets.
4036 if (!sctp_style(sk, UDP))
4037 return -EINVAL;
4039 /* Create a new socket. */
4040 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4041 if (err < 0)
4042 return err;
4044 sctp_copy_sock(sock->sk, sk, asoc);
4046 /* Make peeled-off sockets more like 1-1 accepted sockets.
4047 * Set the daddr and initialize id to something more random
4049 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4050 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4052 /* Populate the fields of the newsk from the oldsk and migrate the
4053 * asoc to the newsk.
4055 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4057 *sockp = sock;
4059 return err;
4062 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4064 sctp_peeloff_arg_t peeloff;
4065 struct socket *newsock;
4066 int retval = 0;
4067 struct sctp_association *asoc;
4069 if (len < sizeof(sctp_peeloff_arg_t))
4070 return -EINVAL;
4071 len = sizeof(sctp_peeloff_arg_t);
4072 if (copy_from_user(&peeloff, optval, len))
4073 return -EFAULT;
4075 asoc = sctp_id2assoc(sk, peeloff.associd);
4076 if (!asoc) {
4077 retval = -EINVAL;
4078 goto out;
4081 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4083 retval = sctp_do_peeloff(asoc, &newsock);
4084 if (retval < 0)
4085 goto out;
4087 /* Map the socket to an unused fd that can be returned to the user. */
4088 retval = sock_map_fd(newsock, 0);
4089 if (retval < 0) {
4090 sock_release(newsock);
4091 goto out;
4094 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4095 __func__, sk, asoc, newsock->sk, retval);
4097 /* Return the fd mapped to the new socket. */
4098 peeloff.sd = retval;
4099 if (put_user(len, optlen))
4100 return -EFAULT;
4101 if (copy_to_user(optval, &peeloff, len))
4102 retval = -EFAULT;
4104 out:
4105 return retval;
4108 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4110 * Applications can enable or disable heartbeats for any peer address of
4111 * an association, modify an address's heartbeat interval, force a
4112 * heartbeat to be sent immediately, and adjust the address's maximum
4113 * number of retransmissions sent before an address is considered
4114 * unreachable. The following structure is used to access and modify an
4115 * address's parameters:
4117 * struct sctp_paddrparams {
4118 * sctp_assoc_t spp_assoc_id;
4119 * struct sockaddr_storage spp_address;
4120 * uint32_t spp_hbinterval;
4121 * uint16_t spp_pathmaxrxt;
4122 * uint32_t spp_pathmtu;
4123 * uint32_t spp_sackdelay;
4124 * uint32_t spp_flags;
4125 * };
4127 * spp_assoc_id - (one-to-many style socket) This is filled in the
4128 * application, and identifies the association for
4129 * this query.
4130 * spp_address - This specifies which address is of interest.
4131 * spp_hbinterval - This contains the value of the heartbeat interval,
4132 * in milliseconds. If a value of zero
4133 * is present in this field then no changes are to
4134 * be made to this parameter.
4135 * spp_pathmaxrxt - This contains the maximum number of
4136 * retransmissions before this address shall be
4137 * considered unreachable. If a value of zero
4138 * is present in this field then no changes are to
4139 * be made to this parameter.
4140 * spp_pathmtu - When Path MTU discovery is disabled the value
4141 * specified here will be the "fixed" path mtu.
4142 * Note that if the spp_address field is empty
4143 * then all associations on this address will
4144 * have this fixed path mtu set upon them.
4146 * spp_sackdelay - When delayed sack is enabled, this value specifies
4147 * the number of milliseconds that sacks will be delayed
4148 * for. This value will apply to all addresses of an
4149 * association if the spp_address field is empty. Note
4150 * also, that if delayed sack is enabled and this
4151 * value is set to 0, no change is made to the last
4152 * recorded delayed sack timer value.
4154 * spp_flags - These flags are used to control various features
4155 * on an association. The flag field may contain
4156 * zero or more of the following options.
4158 * SPP_HB_ENABLE - Enable heartbeats on the
4159 * specified address. Note that if the address
4160 * field is empty all addresses for the association
4161 * have heartbeats enabled upon them.
4163 * SPP_HB_DISABLE - Disable heartbeats on the
4164 * speicifed address. Note that if the address
4165 * field is empty all addresses for the association
4166 * will have their heartbeats disabled. Note also
4167 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4168 * mutually exclusive, only one of these two should
4169 * be specified. Enabling both fields will have
4170 * undetermined results.
4172 * SPP_HB_DEMAND - Request a user initiated heartbeat
4173 * to be made immediately.
4175 * SPP_PMTUD_ENABLE - This field will enable PMTU
4176 * discovery upon the specified address. Note that
4177 * if the address feild is empty then all addresses
4178 * on the association are effected.
4180 * SPP_PMTUD_DISABLE - This field will disable PMTU
4181 * discovery upon the specified address. Note that
4182 * if the address feild is empty then all addresses
4183 * on the association are effected. Not also that
4184 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4185 * exclusive. Enabling both will have undetermined
4186 * results.
4188 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4189 * on delayed sack. The time specified in spp_sackdelay
4190 * is used to specify the sack delay for this address. Note
4191 * that if spp_address is empty then all addresses will
4192 * enable delayed sack and take on the sack delay
4193 * value specified in spp_sackdelay.
4194 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4195 * off delayed sack. If the spp_address field is blank then
4196 * delayed sack is disabled for the entire association. Note
4197 * also that this field is mutually exclusive to
4198 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4199 * results.
4201 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4202 char __user *optval, int __user *optlen)
4204 struct sctp_paddrparams params;
4205 struct sctp_transport *trans = NULL;
4206 struct sctp_association *asoc = NULL;
4207 struct sctp_sock *sp = sctp_sk(sk);
4209 if (len < sizeof(struct sctp_paddrparams))
4210 return -EINVAL;
4211 len = sizeof(struct sctp_paddrparams);
4212 if (copy_from_user(&params, optval, len))
4213 return -EFAULT;
4215 /* If an address other than INADDR_ANY is specified, and
4216 * no transport is found, then the request is invalid.
4218 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4219 trans = sctp_addr_id2transport(sk, &params.spp_address,
4220 params.spp_assoc_id);
4221 if (!trans) {
4222 SCTP_DEBUG_PRINTK("Failed no transport\n");
4223 return -EINVAL;
4227 /* Get association, if assoc_id != 0 and the socket is a one
4228 * to many style socket, and an association was not found, then
4229 * the id was invalid.
4231 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4232 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4233 SCTP_DEBUG_PRINTK("Failed no association\n");
4234 return -EINVAL;
4237 if (trans) {
4238 /* Fetch transport values. */
4239 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4240 params.spp_pathmtu = trans->pathmtu;
4241 params.spp_pathmaxrxt = trans->pathmaxrxt;
4242 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4244 /*draft-11 doesn't say what to return in spp_flags*/
4245 params.spp_flags = trans->param_flags;
4246 } else if (asoc) {
4247 /* Fetch association values. */
4248 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4249 params.spp_pathmtu = asoc->pathmtu;
4250 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4251 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4253 /*draft-11 doesn't say what to return in spp_flags*/
4254 params.spp_flags = asoc->param_flags;
4255 } else {
4256 /* Fetch socket values. */
4257 params.spp_hbinterval = sp->hbinterval;
4258 params.spp_pathmtu = sp->pathmtu;
4259 params.spp_sackdelay = sp->sackdelay;
4260 params.spp_pathmaxrxt = sp->pathmaxrxt;
4262 /*draft-11 doesn't say what to return in spp_flags*/
4263 params.spp_flags = sp->param_flags;
4266 if (copy_to_user(optval, &params, len))
4267 return -EFAULT;
4269 if (put_user(len, optlen))
4270 return -EFAULT;
4272 return 0;
4276 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4278 * This option will effect the way delayed acks are performed. This
4279 * option allows you to get or set the delayed ack time, in
4280 * milliseconds. It also allows changing the delayed ack frequency.
4281 * Changing the frequency to 1 disables the delayed sack algorithm. If
4282 * the assoc_id is 0, then this sets or gets the endpoints default
4283 * values. If the assoc_id field is non-zero, then the set or get
4284 * effects the specified association for the one to many model (the
4285 * assoc_id field is ignored by the one to one model). Note that if
4286 * sack_delay or sack_freq are 0 when setting this option, then the
4287 * current values will remain unchanged.
4289 * struct sctp_sack_info {
4290 * sctp_assoc_t sack_assoc_id;
4291 * uint32_t sack_delay;
4292 * uint32_t sack_freq;
4293 * };
4295 * sack_assoc_id - This parameter, indicates which association the user
4296 * is performing an action upon. Note that if this field's value is
4297 * zero then the endpoints default value is changed (effecting future
4298 * associations only).
4300 * sack_delay - This parameter contains the number of milliseconds that
4301 * the user is requesting the delayed ACK timer be set to. Note that
4302 * this value is defined in the standard to be between 200 and 500
4303 * milliseconds.
4305 * sack_freq - This parameter contains the number of packets that must
4306 * be received before a sack is sent without waiting for the delay
4307 * timer to expire. The default value for this is 2, setting this
4308 * value to 1 will disable the delayed sack algorithm.
4310 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4311 char __user *optval,
4312 int __user *optlen)
4314 struct sctp_sack_info params;
4315 struct sctp_association *asoc = NULL;
4316 struct sctp_sock *sp = sctp_sk(sk);
4318 if (len >= sizeof(struct sctp_sack_info)) {
4319 len = sizeof(struct sctp_sack_info);
4321 if (copy_from_user(&params, optval, len))
4322 return -EFAULT;
4323 } else if (len == sizeof(struct sctp_assoc_value)) {
4324 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4325 pr_warn("Use struct sctp_sack_info instead\n");
4326 if (copy_from_user(&params, optval, len))
4327 return -EFAULT;
4328 } else
4329 return - EINVAL;
4331 /* Get association, if sack_assoc_id != 0 and the socket is a one
4332 * to many style socket, and an association was not found, then
4333 * the id was invalid.
4335 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4336 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4337 return -EINVAL;
4339 if (asoc) {
4340 /* Fetch association values. */
4341 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4342 params.sack_delay = jiffies_to_msecs(
4343 asoc->sackdelay);
4344 params.sack_freq = asoc->sackfreq;
4346 } else {
4347 params.sack_delay = 0;
4348 params.sack_freq = 1;
4350 } else {
4351 /* Fetch socket values. */
4352 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4353 params.sack_delay = sp->sackdelay;
4354 params.sack_freq = sp->sackfreq;
4355 } else {
4356 params.sack_delay = 0;
4357 params.sack_freq = 1;
4361 if (copy_to_user(optval, &params, len))
4362 return -EFAULT;
4364 if (put_user(len, optlen))
4365 return -EFAULT;
4367 return 0;
4370 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4372 * Applications can specify protocol parameters for the default association
4373 * initialization. The option name argument to setsockopt() and getsockopt()
4374 * is SCTP_INITMSG.
4376 * Setting initialization parameters is effective only on an unconnected
4377 * socket (for UDP-style sockets only future associations are effected
4378 * by the change). With TCP-style sockets, this option is inherited by
4379 * sockets derived from a listener socket.
4381 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4383 if (len < sizeof(struct sctp_initmsg))
4384 return -EINVAL;
4385 len = sizeof(struct sctp_initmsg);
4386 if (put_user(len, optlen))
4387 return -EFAULT;
4388 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4389 return -EFAULT;
4390 return 0;
4394 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4395 char __user *optval, int __user *optlen)
4397 struct sctp_association *asoc;
4398 int cnt = 0;
4399 struct sctp_getaddrs getaddrs;
4400 struct sctp_transport *from;
4401 void __user *to;
4402 union sctp_addr temp;
4403 struct sctp_sock *sp = sctp_sk(sk);
4404 int addrlen;
4405 size_t space_left;
4406 int bytes_copied;
4408 if (len < sizeof(struct sctp_getaddrs))
4409 return -EINVAL;
4411 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4412 return -EFAULT;
4414 /* For UDP-style sockets, id specifies the association to query. */
4415 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4416 if (!asoc)
4417 return -EINVAL;
4419 to = optval + offsetof(struct sctp_getaddrs,addrs);
4420 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4422 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4423 transports) {
4424 memcpy(&temp, &from->ipaddr, sizeof(temp));
4425 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4426 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4427 if (space_left < addrlen)
4428 return -ENOMEM;
4429 if (copy_to_user(to, &temp, addrlen))
4430 return -EFAULT;
4431 to += addrlen;
4432 cnt++;
4433 space_left -= addrlen;
4436 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4437 return -EFAULT;
4438 bytes_copied = ((char __user *)to) - optval;
4439 if (put_user(bytes_copied, optlen))
4440 return -EFAULT;
4442 return 0;
4445 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4446 size_t space_left, int *bytes_copied)
4448 struct sctp_sockaddr_entry *addr;
4449 union sctp_addr temp;
4450 int cnt = 0;
4451 int addrlen;
4453 rcu_read_lock();
4454 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4455 if (!addr->valid)
4456 continue;
4458 if ((PF_INET == sk->sk_family) &&
4459 (AF_INET6 == addr->a.sa.sa_family))
4460 continue;
4461 if ((PF_INET6 == sk->sk_family) &&
4462 inet_v6_ipv6only(sk) &&
4463 (AF_INET == addr->a.sa.sa_family))
4464 continue;
4465 memcpy(&temp, &addr->a, sizeof(temp));
4466 if (!temp.v4.sin_port)
4467 temp.v4.sin_port = htons(port);
4469 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4470 &temp);
4471 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4472 if (space_left < addrlen) {
4473 cnt = -ENOMEM;
4474 break;
4476 memcpy(to, &temp, addrlen);
4478 to += addrlen;
4479 cnt ++;
4480 space_left -= addrlen;
4481 *bytes_copied += addrlen;
4483 rcu_read_unlock();
4485 return cnt;
4489 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4490 char __user *optval, int __user *optlen)
4492 struct sctp_bind_addr *bp;
4493 struct sctp_association *asoc;
4494 int cnt = 0;
4495 struct sctp_getaddrs getaddrs;
4496 struct sctp_sockaddr_entry *addr;
4497 void __user *to;
4498 union sctp_addr temp;
4499 struct sctp_sock *sp = sctp_sk(sk);
4500 int addrlen;
4501 int err = 0;
4502 size_t space_left;
4503 int bytes_copied = 0;
4504 void *addrs;
4505 void *buf;
4507 if (len < sizeof(struct sctp_getaddrs))
4508 return -EINVAL;
4510 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4511 return -EFAULT;
4514 * For UDP-style sockets, id specifies the association to query.
4515 * If the id field is set to the value '0' then the locally bound
4516 * addresses are returned without regard to any particular
4517 * association.
4519 if (0 == getaddrs.assoc_id) {
4520 bp = &sctp_sk(sk)->ep->base.bind_addr;
4521 } else {
4522 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4523 if (!asoc)
4524 return -EINVAL;
4525 bp = &asoc->base.bind_addr;
4528 to = optval + offsetof(struct sctp_getaddrs,addrs);
4529 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4531 addrs = kmalloc(space_left, GFP_KERNEL);
4532 if (!addrs)
4533 return -ENOMEM;
4535 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4536 * addresses from the global local address list.
4538 if (sctp_list_single_entry(&bp->address_list)) {
4539 addr = list_entry(bp->address_list.next,
4540 struct sctp_sockaddr_entry, list);
4541 if (sctp_is_any(sk, &addr->a)) {
4542 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4543 space_left, &bytes_copied);
4544 if (cnt < 0) {
4545 err = cnt;
4546 goto out;
4548 goto copy_getaddrs;
4552 buf = addrs;
4553 /* Protection on the bound address list is not needed since
4554 * in the socket option context we hold a socket lock and
4555 * thus the bound address list can't change.
4557 list_for_each_entry(addr, &bp->address_list, list) {
4558 memcpy(&temp, &addr->a, sizeof(temp));
4559 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4560 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4561 if (space_left < addrlen) {
4562 err = -ENOMEM; /*fixme: right error?*/
4563 goto out;
4565 memcpy(buf, &temp, addrlen);
4566 buf += addrlen;
4567 bytes_copied += addrlen;
4568 cnt ++;
4569 space_left -= addrlen;
4572 copy_getaddrs:
4573 if (copy_to_user(to, addrs, bytes_copied)) {
4574 err = -EFAULT;
4575 goto out;
4577 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4578 err = -EFAULT;
4579 goto out;
4581 if (put_user(bytes_copied, optlen))
4582 err = -EFAULT;
4583 out:
4584 kfree(addrs);
4585 return err;
4588 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4590 * Requests that the local SCTP stack use the enclosed peer address as
4591 * the association primary. The enclosed address must be one of the
4592 * association peer's addresses.
4594 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4595 char __user *optval, int __user *optlen)
4597 struct sctp_prim prim;
4598 struct sctp_association *asoc;
4599 struct sctp_sock *sp = sctp_sk(sk);
4601 if (len < sizeof(struct sctp_prim))
4602 return -EINVAL;
4604 len = sizeof(struct sctp_prim);
4606 if (copy_from_user(&prim, optval, len))
4607 return -EFAULT;
4609 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4610 if (!asoc)
4611 return -EINVAL;
4613 if (!asoc->peer.primary_path)
4614 return -ENOTCONN;
4616 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4617 asoc->peer.primary_path->af_specific->sockaddr_len);
4619 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4620 (union sctp_addr *)&prim.ssp_addr);
4622 if (put_user(len, optlen))
4623 return -EFAULT;
4624 if (copy_to_user(optval, &prim, len))
4625 return -EFAULT;
4627 return 0;
4631 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4633 * Requests that the local endpoint set the specified Adaptation Layer
4634 * Indication parameter for all future INIT and INIT-ACK exchanges.
4636 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4637 char __user *optval, int __user *optlen)
4639 struct sctp_setadaptation adaptation;
4641 if (len < sizeof(struct sctp_setadaptation))
4642 return -EINVAL;
4644 len = sizeof(struct sctp_setadaptation);
4646 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4648 if (put_user(len, optlen))
4649 return -EFAULT;
4650 if (copy_to_user(optval, &adaptation, len))
4651 return -EFAULT;
4653 return 0;
4658 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4660 * Applications that wish to use the sendto() system call may wish to
4661 * specify a default set of parameters that would normally be supplied
4662 * through the inclusion of ancillary data. This socket option allows
4663 * such an application to set the default sctp_sndrcvinfo structure.
4666 * The application that wishes to use this socket option simply passes
4667 * in to this call the sctp_sndrcvinfo structure defined in Section
4668 * 5.2.2) The input parameters accepted by this call include
4669 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4670 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4671 * to this call if the caller is using the UDP model.
4673 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4675 static int sctp_getsockopt_default_send_param(struct sock *sk,
4676 int len, char __user *optval,
4677 int __user *optlen)
4679 struct sctp_sndrcvinfo info;
4680 struct sctp_association *asoc;
4681 struct sctp_sock *sp = sctp_sk(sk);
4683 if (len < sizeof(struct sctp_sndrcvinfo))
4684 return -EINVAL;
4686 len = sizeof(struct sctp_sndrcvinfo);
4688 if (copy_from_user(&info, optval, len))
4689 return -EFAULT;
4691 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4692 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4693 return -EINVAL;
4695 if (asoc) {
4696 info.sinfo_stream = asoc->default_stream;
4697 info.sinfo_flags = asoc->default_flags;
4698 info.sinfo_ppid = asoc->default_ppid;
4699 info.sinfo_context = asoc->default_context;
4700 info.sinfo_timetolive = asoc->default_timetolive;
4701 } else {
4702 info.sinfo_stream = sp->default_stream;
4703 info.sinfo_flags = sp->default_flags;
4704 info.sinfo_ppid = sp->default_ppid;
4705 info.sinfo_context = sp->default_context;
4706 info.sinfo_timetolive = sp->default_timetolive;
4709 if (put_user(len, optlen))
4710 return -EFAULT;
4711 if (copy_to_user(optval, &info, len))
4712 return -EFAULT;
4714 return 0;
4719 * 7.1.5 SCTP_NODELAY
4721 * Turn on/off any Nagle-like algorithm. This means that packets are
4722 * generally sent as soon as possible and no unnecessary delays are
4723 * introduced, at the cost of more packets in the network. Expects an
4724 * integer boolean flag.
4727 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4728 char __user *optval, int __user *optlen)
4730 int val;
4732 if (len < sizeof(int))
4733 return -EINVAL;
4735 len = sizeof(int);
4736 val = (sctp_sk(sk)->nodelay == 1);
4737 if (put_user(len, optlen))
4738 return -EFAULT;
4739 if (copy_to_user(optval, &val, len))
4740 return -EFAULT;
4741 return 0;
4746 * 7.1.1 SCTP_RTOINFO
4748 * The protocol parameters used to initialize and bound retransmission
4749 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4750 * and modify these parameters.
4751 * All parameters are time values, in milliseconds. A value of 0, when
4752 * modifying the parameters, indicates that the current value should not
4753 * be changed.
4756 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4757 char __user *optval,
4758 int __user *optlen) {
4759 struct sctp_rtoinfo rtoinfo;
4760 struct sctp_association *asoc;
4762 if (len < sizeof (struct sctp_rtoinfo))
4763 return -EINVAL;
4765 len = sizeof(struct sctp_rtoinfo);
4767 if (copy_from_user(&rtoinfo, optval, len))
4768 return -EFAULT;
4770 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4772 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4773 return -EINVAL;
4775 /* Values corresponding to the specific association. */
4776 if (asoc) {
4777 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4778 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4779 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4780 } else {
4781 /* Values corresponding to the endpoint. */
4782 struct sctp_sock *sp = sctp_sk(sk);
4784 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4785 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4786 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4789 if (put_user(len, optlen))
4790 return -EFAULT;
4792 if (copy_to_user(optval, &rtoinfo, len))
4793 return -EFAULT;
4795 return 0;
4800 * 7.1.2 SCTP_ASSOCINFO
4802 * This option is used to tune the maximum retransmission attempts
4803 * of the association.
4804 * Returns an error if the new association retransmission value is
4805 * greater than the sum of the retransmission value of the peer.
4806 * See [SCTP] for more information.
4809 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4810 char __user *optval,
4811 int __user *optlen)
4814 struct sctp_assocparams assocparams;
4815 struct sctp_association *asoc;
4816 struct list_head *pos;
4817 int cnt = 0;
4819 if (len < sizeof (struct sctp_assocparams))
4820 return -EINVAL;
4822 len = sizeof(struct sctp_assocparams);
4824 if (copy_from_user(&assocparams, optval, len))
4825 return -EFAULT;
4827 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4829 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4830 return -EINVAL;
4832 /* Values correspoinding to the specific association */
4833 if (asoc) {
4834 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4835 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4836 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4837 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4838 * 1000) +
4839 (asoc->cookie_life.tv_usec
4840 / 1000);
4842 list_for_each(pos, &asoc->peer.transport_addr_list) {
4843 cnt ++;
4846 assocparams.sasoc_number_peer_destinations = cnt;
4847 } else {
4848 /* Values corresponding to the endpoint */
4849 struct sctp_sock *sp = sctp_sk(sk);
4851 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4852 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4853 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4854 assocparams.sasoc_cookie_life =
4855 sp->assocparams.sasoc_cookie_life;
4856 assocparams.sasoc_number_peer_destinations =
4857 sp->assocparams.
4858 sasoc_number_peer_destinations;
4861 if (put_user(len, optlen))
4862 return -EFAULT;
4864 if (copy_to_user(optval, &assocparams, len))
4865 return -EFAULT;
4867 return 0;
4871 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4873 * This socket option is a boolean flag which turns on or off mapped V4
4874 * addresses. If this option is turned on and the socket is type
4875 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4876 * If this option is turned off, then no mapping will be done of V4
4877 * addresses and a user will receive both PF_INET6 and PF_INET type
4878 * addresses on the socket.
4880 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4881 char __user *optval, int __user *optlen)
4883 int val;
4884 struct sctp_sock *sp = sctp_sk(sk);
4886 if (len < sizeof(int))
4887 return -EINVAL;
4889 len = sizeof(int);
4890 val = sp->v4mapped;
4891 if (put_user(len, optlen))
4892 return -EFAULT;
4893 if (copy_to_user(optval, &val, len))
4894 return -EFAULT;
4896 return 0;
4900 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4901 * (chapter and verse is quoted at sctp_setsockopt_context())
4903 static int sctp_getsockopt_context(struct sock *sk, int len,
4904 char __user *optval, int __user *optlen)
4906 struct sctp_assoc_value params;
4907 struct sctp_sock *sp;
4908 struct sctp_association *asoc;
4910 if (len < sizeof(struct sctp_assoc_value))
4911 return -EINVAL;
4913 len = sizeof(struct sctp_assoc_value);
4915 if (copy_from_user(&params, optval, len))
4916 return -EFAULT;
4918 sp = sctp_sk(sk);
4920 if (params.assoc_id != 0) {
4921 asoc = sctp_id2assoc(sk, params.assoc_id);
4922 if (!asoc)
4923 return -EINVAL;
4924 params.assoc_value = asoc->default_rcv_context;
4925 } else {
4926 params.assoc_value = sp->default_rcv_context;
4929 if (put_user(len, optlen))
4930 return -EFAULT;
4931 if (copy_to_user(optval, &params, len))
4932 return -EFAULT;
4934 return 0;
4938 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4939 * This option will get or set the maximum size to put in any outgoing
4940 * SCTP DATA chunk. If a message is larger than this size it will be
4941 * fragmented by SCTP into the specified size. Note that the underlying
4942 * SCTP implementation may fragment into smaller sized chunks when the
4943 * PMTU of the underlying association is smaller than the value set by
4944 * the user. The default value for this option is '0' which indicates
4945 * the user is NOT limiting fragmentation and only the PMTU will effect
4946 * SCTP's choice of DATA chunk size. Note also that values set larger
4947 * than the maximum size of an IP datagram will effectively let SCTP
4948 * control fragmentation (i.e. the same as setting this option to 0).
4950 * The following structure is used to access and modify this parameter:
4952 * struct sctp_assoc_value {
4953 * sctp_assoc_t assoc_id;
4954 * uint32_t assoc_value;
4955 * };
4957 * assoc_id: This parameter is ignored for one-to-one style sockets.
4958 * For one-to-many style sockets this parameter indicates which
4959 * association the user is performing an action upon. Note that if
4960 * this field's value is zero then the endpoints default value is
4961 * changed (effecting future associations only).
4962 * assoc_value: This parameter specifies the maximum size in bytes.
4964 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4965 char __user *optval, int __user *optlen)
4967 struct sctp_assoc_value params;
4968 struct sctp_association *asoc;
4970 if (len == sizeof(int)) {
4971 pr_warn("Use of int in maxseg socket option deprecated\n");
4972 pr_warn("Use struct sctp_assoc_value instead\n");
4973 params.assoc_id = 0;
4974 } else if (len >= sizeof(struct sctp_assoc_value)) {
4975 len = sizeof(struct sctp_assoc_value);
4976 if (copy_from_user(&params, optval, sizeof(params)))
4977 return -EFAULT;
4978 } else
4979 return -EINVAL;
4981 asoc = sctp_id2assoc(sk, params.assoc_id);
4982 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4983 return -EINVAL;
4985 if (asoc)
4986 params.assoc_value = asoc->frag_point;
4987 else
4988 params.assoc_value = sctp_sk(sk)->user_frag;
4990 if (put_user(len, optlen))
4991 return -EFAULT;
4992 if (len == sizeof(int)) {
4993 if (copy_to_user(optval, &params.assoc_value, len))
4994 return -EFAULT;
4995 } else {
4996 if (copy_to_user(optval, &params, len))
4997 return -EFAULT;
5000 return 0;
5004 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5005 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5007 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5008 char __user *optval, int __user *optlen)
5010 int val;
5012 if (len < sizeof(int))
5013 return -EINVAL;
5015 len = sizeof(int);
5017 val = sctp_sk(sk)->frag_interleave;
5018 if (put_user(len, optlen))
5019 return -EFAULT;
5020 if (copy_to_user(optval, &val, len))
5021 return -EFAULT;
5023 return 0;
5027 * 7.1.25. Set or Get the sctp partial delivery point
5028 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5030 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5031 char __user *optval,
5032 int __user *optlen)
5034 u32 val;
5036 if (len < sizeof(u32))
5037 return -EINVAL;
5039 len = sizeof(u32);
5041 val = sctp_sk(sk)->pd_point;
5042 if (put_user(len, optlen))
5043 return -EFAULT;
5044 if (copy_to_user(optval, &val, len))
5045 return -EFAULT;
5047 return 0;
5051 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5052 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5054 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5055 char __user *optval,
5056 int __user *optlen)
5058 struct sctp_assoc_value params;
5059 struct sctp_sock *sp;
5060 struct sctp_association *asoc;
5062 if (len == sizeof(int)) {
5063 pr_warn("Use of int in max_burst socket option deprecated\n");
5064 pr_warn("Use struct sctp_assoc_value instead\n");
5065 params.assoc_id = 0;
5066 } else if (len >= sizeof(struct sctp_assoc_value)) {
5067 len = sizeof(struct sctp_assoc_value);
5068 if (copy_from_user(&params, optval, len))
5069 return -EFAULT;
5070 } else
5071 return -EINVAL;
5073 sp = sctp_sk(sk);
5075 if (params.assoc_id != 0) {
5076 asoc = sctp_id2assoc(sk, params.assoc_id);
5077 if (!asoc)
5078 return -EINVAL;
5079 params.assoc_value = asoc->max_burst;
5080 } else
5081 params.assoc_value = sp->max_burst;
5083 if (len == sizeof(int)) {
5084 if (copy_to_user(optval, &params.assoc_value, len))
5085 return -EFAULT;
5086 } else {
5087 if (copy_to_user(optval, &params, len))
5088 return -EFAULT;
5091 return 0;
5095 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5096 char __user *optval, int __user *optlen)
5098 struct sctp_hmacalgo __user *p = (void __user *)optval;
5099 struct sctp_hmac_algo_param *hmacs;
5100 __u16 data_len = 0;
5101 u32 num_idents;
5103 if (!sctp_auth_enable)
5104 return -EACCES;
5106 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5107 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5109 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5110 return -EINVAL;
5112 len = sizeof(struct sctp_hmacalgo) + data_len;
5113 num_idents = data_len / sizeof(u16);
5115 if (put_user(len, optlen))
5116 return -EFAULT;
5117 if (put_user(num_idents, &p->shmac_num_idents))
5118 return -EFAULT;
5119 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5120 return -EFAULT;
5121 return 0;
5124 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5125 char __user *optval, int __user *optlen)
5127 struct sctp_authkeyid val;
5128 struct sctp_association *asoc;
5130 if (!sctp_auth_enable)
5131 return -EACCES;
5133 if (len < sizeof(struct sctp_authkeyid))
5134 return -EINVAL;
5135 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5136 return -EFAULT;
5138 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5139 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5140 return -EINVAL;
5142 if (asoc)
5143 val.scact_keynumber = asoc->active_key_id;
5144 else
5145 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5147 len = sizeof(struct sctp_authkeyid);
5148 if (put_user(len, optlen))
5149 return -EFAULT;
5150 if (copy_to_user(optval, &val, len))
5151 return -EFAULT;
5153 return 0;
5156 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5157 char __user *optval, int __user *optlen)
5159 struct sctp_authchunks __user *p = (void __user *)optval;
5160 struct sctp_authchunks val;
5161 struct sctp_association *asoc;
5162 struct sctp_chunks_param *ch;
5163 u32 num_chunks = 0;
5164 char __user *to;
5166 if (!sctp_auth_enable)
5167 return -EACCES;
5169 if (len < sizeof(struct sctp_authchunks))
5170 return -EINVAL;
5172 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5173 return -EFAULT;
5175 to = p->gauth_chunks;
5176 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5177 if (!asoc)
5178 return -EINVAL;
5180 ch = asoc->peer.peer_chunks;
5181 if (!ch)
5182 goto num;
5184 /* See if the user provided enough room for all the data */
5185 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5186 if (len < num_chunks)
5187 return -EINVAL;
5189 if (copy_to_user(to, ch->chunks, num_chunks))
5190 return -EFAULT;
5191 num:
5192 len = sizeof(struct sctp_authchunks) + num_chunks;
5193 if (put_user(len, optlen)) return -EFAULT;
5194 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5195 return -EFAULT;
5196 return 0;
5199 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5200 char __user *optval, int __user *optlen)
5202 struct sctp_authchunks __user *p = (void __user *)optval;
5203 struct sctp_authchunks val;
5204 struct sctp_association *asoc;
5205 struct sctp_chunks_param *ch;
5206 u32 num_chunks = 0;
5207 char __user *to;
5209 if (!sctp_auth_enable)
5210 return -EACCES;
5212 if (len < sizeof(struct sctp_authchunks))
5213 return -EINVAL;
5215 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5216 return -EFAULT;
5218 to = p->gauth_chunks;
5219 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5220 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5221 return -EINVAL;
5223 if (asoc)
5224 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5225 else
5226 ch = sctp_sk(sk)->ep->auth_chunk_list;
5228 if (!ch)
5229 goto num;
5231 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5232 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5233 return -EINVAL;
5235 if (copy_to_user(to, ch->chunks, num_chunks))
5236 return -EFAULT;
5237 num:
5238 len = sizeof(struct sctp_authchunks) + num_chunks;
5239 if (put_user(len, optlen))
5240 return -EFAULT;
5241 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5242 return -EFAULT;
5244 return 0;
5248 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5249 * This option gets the current number of associations that are attached
5250 * to a one-to-many style socket. The option value is an uint32_t.
5252 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5253 char __user *optval, int __user *optlen)
5255 struct sctp_sock *sp = sctp_sk(sk);
5256 struct sctp_association *asoc;
5257 u32 val = 0;
5259 if (sctp_style(sk, TCP))
5260 return -EOPNOTSUPP;
5262 if (len < sizeof(u32))
5263 return -EINVAL;
5265 len = sizeof(u32);
5267 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5268 val++;
5271 if (put_user(len, optlen))
5272 return -EFAULT;
5273 if (copy_to_user(optval, &val, len))
5274 return -EFAULT;
5276 return 0;
5280 * 8.2.6. Get the Current Identifiers of Associations
5281 * (SCTP_GET_ASSOC_ID_LIST)
5283 * This option gets the current list of SCTP association identifiers of
5284 * the SCTP associations handled by a one-to-many style socket.
5286 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5287 char __user *optval, int __user *optlen)
5289 struct sctp_sock *sp = sctp_sk(sk);
5290 struct sctp_association *asoc;
5291 struct sctp_assoc_ids *ids;
5292 u32 num = 0;
5294 if (sctp_style(sk, TCP))
5295 return -EOPNOTSUPP;
5297 if (len < sizeof(struct sctp_assoc_ids))
5298 return -EINVAL;
5300 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5301 num++;
5304 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5305 return -EINVAL;
5307 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5309 ids = kmalloc(len, GFP_KERNEL);
5310 if (unlikely(!ids))
5311 return -ENOMEM;
5313 ids->gaids_number_of_ids = num;
5314 num = 0;
5315 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5316 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5319 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5320 kfree(ids);
5321 return -EFAULT;
5324 kfree(ids);
5325 return 0;
5328 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5329 char __user *optval, int __user *optlen)
5331 int retval = 0;
5332 int len;
5334 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5335 sk, optname);
5337 /* I can hardly begin to describe how wrong this is. This is
5338 * so broken as to be worse than useless. The API draft
5339 * REALLY is NOT helpful here... I am not convinced that the
5340 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5341 * are at all well-founded.
5343 if (level != SOL_SCTP) {
5344 struct sctp_af *af = sctp_sk(sk)->pf->af;
5346 retval = af->getsockopt(sk, level, optname, optval, optlen);
5347 return retval;
5350 if (get_user(len, optlen))
5351 return -EFAULT;
5353 sctp_lock_sock(sk);
5355 switch (optname) {
5356 case SCTP_STATUS:
5357 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5358 break;
5359 case SCTP_DISABLE_FRAGMENTS:
5360 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5361 optlen);
5362 break;
5363 case SCTP_EVENTS:
5364 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5365 break;
5366 case SCTP_AUTOCLOSE:
5367 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5368 break;
5369 case SCTP_SOCKOPT_PEELOFF:
5370 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5371 break;
5372 case SCTP_PEER_ADDR_PARAMS:
5373 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5374 optlen);
5375 break;
5376 case SCTP_DELAYED_SACK:
5377 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5378 optlen);
5379 break;
5380 case SCTP_INITMSG:
5381 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5382 break;
5383 case SCTP_GET_PEER_ADDRS:
5384 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5385 optlen);
5386 break;
5387 case SCTP_GET_LOCAL_ADDRS:
5388 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5389 optlen);
5390 break;
5391 case SCTP_SOCKOPT_CONNECTX3:
5392 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5393 break;
5394 case SCTP_DEFAULT_SEND_PARAM:
5395 retval = sctp_getsockopt_default_send_param(sk, len,
5396 optval, optlen);
5397 break;
5398 case SCTP_PRIMARY_ADDR:
5399 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5400 break;
5401 case SCTP_NODELAY:
5402 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5403 break;
5404 case SCTP_RTOINFO:
5405 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5406 break;
5407 case SCTP_ASSOCINFO:
5408 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5409 break;
5410 case SCTP_I_WANT_MAPPED_V4_ADDR:
5411 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5412 break;
5413 case SCTP_MAXSEG:
5414 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5415 break;
5416 case SCTP_GET_PEER_ADDR_INFO:
5417 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5418 optlen);
5419 break;
5420 case SCTP_ADAPTATION_LAYER:
5421 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5422 optlen);
5423 break;
5424 case SCTP_CONTEXT:
5425 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5426 break;
5427 case SCTP_FRAGMENT_INTERLEAVE:
5428 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5429 optlen);
5430 break;
5431 case SCTP_PARTIAL_DELIVERY_POINT:
5432 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5433 optlen);
5434 break;
5435 case SCTP_MAX_BURST:
5436 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5437 break;
5438 case SCTP_AUTH_KEY:
5439 case SCTP_AUTH_CHUNK:
5440 case SCTP_AUTH_DELETE_KEY:
5441 retval = -EOPNOTSUPP;
5442 break;
5443 case SCTP_HMAC_IDENT:
5444 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5445 break;
5446 case SCTP_AUTH_ACTIVE_KEY:
5447 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5448 break;
5449 case SCTP_PEER_AUTH_CHUNKS:
5450 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5451 optlen);
5452 break;
5453 case SCTP_LOCAL_AUTH_CHUNKS:
5454 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5455 optlen);
5456 break;
5457 case SCTP_GET_ASSOC_NUMBER:
5458 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5459 break;
5460 case SCTP_GET_ASSOC_ID_LIST:
5461 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5462 break;
5463 default:
5464 retval = -ENOPROTOOPT;
5465 break;
5468 sctp_release_sock(sk);
5469 return retval;
5472 static void sctp_hash(struct sock *sk)
5474 /* STUB */
5477 static void sctp_unhash(struct sock *sk)
5479 /* STUB */
5482 /* Check if port is acceptable. Possibly find first available port.
5484 * The port hash table (contained in the 'global' SCTP protocol storage
5485 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5486 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5487 * list (the list number is the port number hashed out, so as you
5488 * would expect from a hash function, all the ports in a given list have
5489 * such a number that hashes out to the same list number; you were
5490 * expecting that, right?); so each list has a set of ports, with a
5491 * link to the socket (struct sock) that uses it, the port number and
5492 * a fastreuse flag (FIXME: NPI ipg).
5494 static struct sctp_bind_bucket *sctp_bucket_create(
5495 struct sctp_bind_hashbucket *head, unsigned short snum);
5497 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5499 struct sctp_bind_hashbucket *head; /* hash list */
5500 struct sctp_bind_bucket *pp; /* hash list port iterator */
5501 struct hlist_node *node;
5502 unsigned short snum;
5503 int ret;
5505 snum = ntohs(addr->v4.sin_port);
5507 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5508 sctp_local_bh_disable();
5510 if (snum == 0) {
5511 /* Search for an available port. */
5512 int low, high, remaining, index;
5513 unsigned int rover;
5515 inet_get_local_port_range(&low, &high);
5516 remaining = (high - low) + 1;
5517 rover = net_random() % remaining + low;
5519 do {
5520 rover++;
5521 if ((rover < low) || (rover > high))
5522 rover = low;
5523 if (inet_is_reserved_local_port(rover))
5524 continue;
5525 index = sctp_phashfn(rover);
5526 head = &sctp_port_hashtable[index];
5527 sctp_spin_lock(&head->lock);
5528 sctp_for_each_hentry(pp, node, &head->chain)
5529 if (pp->port == rover)
5530 goto next;
5531 break;
5532 next:
5533 sctp_spin_unlock(&head->lock);
5534 } while (--remaining > 0);
5536 /* Exhausted local port range during search? */
5537 ret = 1;
5538 if (remaining <= 0)
5539 goto fail;
5541 /* OK, here is the one we will use. HEAD (the port
5542 * hash table list entry) is non-NULL and we hold it's
5543 * mutex.
5545 snum = rover;
5546 } else {
5547 /* We are given an specific port number; we verify
5548 * that it is not being used. If it is used, we will
5549 * exahust the search in the hash list corresponding
5550 * to the port number (snum) - we detect that with the
5551 * port iterator, pp being NULL.
5553 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5554 sctp_spin_lock(&head->lock);
5555 sctp_for_each_hentry(pp, node, &head->chain) {
5556 if (pp->port == snum)
5557 goto pp_found;
5560 pp = NULL;
5561 goto pp_not_found;
5562 pp_found:
5563 if (!hlist_empty(&pp->owner)) {
5564 /* We had a port hash table hit - there is an
5565 * available port (pp != NULL) and it is being
5566 * used by other socket (pp->owner not empty); that other
5567 * socket is going to be sk2.
5569 int reuse = sk->sk_reuse;
5570 struct sock *sk2;
5572 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5573 if (pp->fastreuse && sk->sk_reuse &&
5574 sk->sk_state != SCTP_SS_LISTENING)
5575 goto success;
5577 /* Run through the list of sockets bound to the port
5578 * (pp->port) [via the pointers bind_next and
5579 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5580 * we get the endpoint they describe and run through
5581 * the endpoint's list of IP (v4 or v6) addresses,
5582 * comparing each of the addresses with the address of
5583 * the socket sk. If we find a match, then that means
5584 * that this port/socket (sk) combination are already
5585 * in an endpoint.
5587 sk_for_each_bound(sk2, node, &pp->owner) {
5588 struct sctp_endpoint *ep2;
5589 ep2 = sctp_sk(sk2)->ep;
5591 if (sk == sk2 ||
5592 (reuse && sk2->sk_reuse &&
5593 sk2->sk_state != SCTP_SS_LISTENING))
5594 continue;
5596 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5597 sctp_sk(sk2), sctp_sk(sk))) {
5598 ret = (long)sk2;
5599 goto fail_unlock;
5602 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5604 pp_not_found:
5605 /* If there was a hash table miss, create a new port. */
5606 ret = 1;
5607 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5608 goto fail_unlock;
5610 /* In either case (hit or miss), make sure fastreuse is 1 only
5611 * if sk->sk_reuse is too (that is, if the caller requested
5612 * SO_REUSEADDR on this socket -sk-).
5614 if (hlist_empty(&pp->owner)) {
5615 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5616 pp->fastreuse = 1;
5617 else
5618 pp->fastreuse = 0;
5619 } else if (pp->fastreuse &&
5620 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5621 pp->fastreuse = 0;
5623 /* We are set, so fill up all the data in the hash table
5624 * entry, tie the socket list information with the rest of the
5625 * sockets FIXME: Blurry, NPI (ipg).
5627 success:
5628 if (!sctp_sk(sk)->bind_hash) {
5629 inet_sk(sk)->inet_num = snum;
5630 sk_add_bind_node(sk, &pp->owner);
5631 sctp_sk(sk)->bind_hash = pp;
5633 ret = 0;
5635 fail_unlock:
5636 sctp_spin_unlock(&head->lock);
5638 fail:
5639 sctp_local_bh_enable();
5640 return ret;
5643 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5644 * port is requested.
5646 static int sctp_get_port(struct sock *sk, unsigned short snum)
5648 long ret;
5649 union sctp_addr addr;
5650 struct sctp_af *af = sctp_sk(sk)->pf->af;
5652 /* Set up a dummy address struct from the sk. */
5653 af->from_sk(&addr, sk);
5654 addr.v4.sin_port = htons(snum);
5656 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5657 ret = sctp_get_port_local(sk, &addr);
5659 return ret ? 1 : 0;
5663 * Move a socket to LISTENING state.
5665 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5667 struct sctp_sock *sp = sctp_sk(sk);
5668 struct sctp_endpoint *ep = sp->ep;
5669 struct crypto_hash *tfm = NULL;
5671 /* Allocate HMAC for generating cookie. */
5672 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5673 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5674 if (IS_ERR(tfm)) {
5675 if (net_ratelimit()) {
5676 pr_info("failed to load transform for %s: %ld\n",
5677 sctp_hmac_alg, PTR_ERR(tfm));
5679 return -ENOSYS;
5681 sctp_sk(sk)->hmac = tfm;
5685 * If a bind() or sctp_bindx() is not called prior to a listen()
5686 * call that allows new associations to be accepted, the system
5687 * picks an ephemeral port and will choose an address set equivalent
5688 * to binding with a wildcard address.
5690 * This is not currently spelled out in the SCTP sockets
5691 * extensions draft, but follows the practice as seen in TCP
5692 * sockets.
5695 sk->sk_state = SCTP_SS_LISTENING;
5696 if (!ep->base.bind_addr.port) {
5697 if (sctp_autobind(sk))
5698 return -EAGAIN;
5699 } else {
5700 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5701 sk->sk_state = SCTP_SS_CLOSED;
5702 return -EADDRINUSE;
5706 sk->sk_max_ack_backlog = backlog;
5707 sctp_hash_endpoint(ep);
5708 return 0;
5712 * 4.1.3 / 5.1.3 listen()
5714 * By default, new associations are not accepted for UDP style sockets.
5715 * An application uses listen() to mark a socket as being able to
5716 * accept new associations.
5718 * On TCP style sockets, applications use listen() to ready the SCTP
5719 * endpoint for accepting inbound associations.
5721 * On both types of endpoints a backlog of '0' disables listening.
5723 * Move a socket to LISTENING state.
5725 int sctp_inet_listen(struct socket *sock, int backlog)
5727 struct sock *sk = sock->sk;
5728 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5729 int err = -EINVAL;
5731 if (unlikely(backlog < 0))
5732 return err;
5734 sctp_lock_sock(sk);
5736 /* Peeled-off sockets are not allowed to listen(). */
5737 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5738 goto out;
5740 if (sock->state != SS_UNCONNECTED)
5741 goto out;
5743 /* If backlog is zero, disable listening. */
5744 if (!backlog) {
5745 if (sctp_sstate(sk, CLOSED))
5746 goto out;
5748 err = 0;
5749 sctp_unhash_endpoint(ep);
5750 sk->sk_state = SCTP_SS_CLOSED;
5751 if (sk->sk_reuse)
5752 sctp_sk(sk)->bind_hash->fastreuse = 1;
5753 goto out;
5756 /* If we are already listening, just update the backlog */
5757 if (sctp_sstate(sk, LISTENING))
5758 sk->sk_max_ack_backlog = backlog;
5759 else {
5760 err = sctp_listen_start(sk, backlog);
5761 if (err)
5762 goto out;
5765 err = 0;
5766 out:
5767 sctp_release_sock(sk);
5768 return err;
5772 * This function is done by modeling the current datagram_poll() and the
5773 * tcp_poll(). Note that, based on these implementations, we don't
5774 * lock the socket in this function, even though it seems that,
5775 * ideally, locking or some other mechanisms can be used to ensure
5776 * the integrity of the counters (sndbuf and wmem_alloc) used
5777 * in this place. We assume that we don't need locks either until proven
5778 * otherwise.
5780 * Another thing to note is that we include the Async I/O support
5781 * here, again, by modeling the current TCP/UDP code. We don't have
5782 * a good way to test with it yet.
5784 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5786 struct sock *sk = sock->sk;
5787 struct sctp_sock *sp = sctp_sk(sk);
5788 unsigned int mask;
5790 poll_wait(file, sk_sleep(sk), wait);
5792 /* A TCP-style listening socket becomes readable when the accept queue
5793 * is not empty.
5795 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5796 return (!list_empty(&sp->ep->asocs)) ?
5797 (POLLIN | POLLRDNORM) : 0;
5799 mask = 0;
5801 /* Is there any exceptional events? */
5802 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5803 mask |= POLLERR;
5804 if (sk->sk_shutdown & RCV_SHUTDOWN)
5805 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5806 if (sk->sk_shutdown == SHUTDOWN_MASK)
5807 mask |= POLLHUP;
5809 /* Is it readable? Reconsider this code with TCP-style support. */
5810 if (!skb_queue_empty(&sk->sk_receive_queue))
5811 mask |= POLLIN | POLLRDNORM;
5813 /* The association is either gone or not ready. */
5814 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5815 return mask;
5817 /* Is it writable? */
5818 if (sctp_writeable(sk)) {
5819 mask |= POLLOUT | POLLWRNORM;
5820 } else {
5821 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5823 * Since the socket is not locked, the buffer
5824 * might be made available after the writeable check and
5825 * before the bit is set. This could cause a lost I/O
5826 * signal. tcp_poll() has a race breaker for this race
5827 * condition. Based on their implementation, we put
5828 * in the following code to cover it as well.
5830 if (sctp_writeable(sk))
5831 mask |= POLLOUT | POLLWRNORM;
5833 return mask;
5836 /********************************************************************
5837 * 2nd Level Abstractions
5838 ********************************************************************/
5840 static struct sctp_bind_bucket *sctp_bucket_create(
5841 struct sctp_bind_hashbucket *head, unsigned short snum)
5843 struct sctp_bind_bucket *pp;
5845 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5846 if (pp) {
5847 SCTP_DBG_OBJCNT_INC(bind_bucket);
5848 pp->port = snum;
5849 pp->fastreuse = 0;
5850 INIT_HLIST_HEAD(&pp->owner);
5851 hlist_add_head(&pp->node, &head->chain);
5853 return pp;
5856 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5857 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5859 if (pp && hlist_empty(&pp->owner)) {
5860 __hlist_del(&pp->node);
5861 kmem_cache_free(sctp_bucket_cachep, pp);
5862 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5866 /* Release this socket's reference to a local port. */
5867 static inline void __sctp_put_port(struct sock *sk)
5869 struct sctp_bind_hashbucket *head =
5870 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5871 struct sctp_bind_bucket *pp;
5873 sctp_spin_lock(&head->lock);
5874 pp = sctp_sk(sk)->bind_hash;
5875 __sk_del_bind_node(sk);
5876 sctp_sk(sk)->bind_hash = NULL;
5877 inet_sk(sk)->inet_num = 0;
5878 sctp_bucket_destroy(pp);
5879 sctp_spin_unlock(&head->lock);
5882 void sctp_put_port(struct sock *sk)
5884 sctp_local_bh_disable();
5885 __sctp_put_port(sk);
5886 sctp_local_bh_enable();
5890 * The system picks an ephemeral port and choose an address set equivalent
5891 * to binding with a wildcard address.
5892 * One of those addresses will be the primary address for the association.
5893 * This automatically enables the multihoming capability of SCTP.
5895 static int sctp_autobind(struct sock *sk)
5897 union sctp_addr autoaddr;
5898 struct sctp_af *af;
5899 __be16 port;
5901 /* Initialize a local sockaddr structure to INADDR_ANY. */
5902 af = sctp_sk(sk)->pf->af;
5904 port = htons(inet_sk(sk)->inet_num);
5905 af->inaddr_any(&autoaddr, port);
5907 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5910 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5912 * From RFC 2292
5913 * 4.2 The cmsghdr Structure *
5915 * When ancillary data is sent or received, any number of ancillary data
5916 * objects can be specified by the msg_control and msg_controllen members of
5917 * the msghdr structure, because each object is preceded by
5918 * a cmsghdr structure defining the object's length (the cmsg_len member).
5919 * Historically Berkeley-derived implementations have passed only one object
5920 * at a time, but this API allows multiple objects to be
5921 * passed in a single call to sendmsg() or recvmsg(). The following example
5922 * shows two ancillary data objects in a control buffer.
5924 * |<--------------------------- msg_controllen -------------------------->|
5925 * | |
5927 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5929 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5930 * | | |
5932 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5934 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5935 * | | | | |
5937 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5938 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5940 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5942 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5946 * msg_control
5947 * points here
5949 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5950 sctp_cmsgs_t *cmsgs)
5952 struct cmsghdr *cmsg;
5953 struct msghdr *my_msg = (struct msghdr *)msg;
5955 for (cmsg = CMSG_FIRSTHDR(msg);
5956 cmsg != NULL;
5957 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5958 if (!CMSG_OK(my_msg, cmsg))
5959 return -EINVAL;
5961 /* Should we parse this header or ignore? */
5962 if (cmsg->cmsg_level != IPPROTO_SCTP)
5963 continue;
5965 /* Strictly check lengths following example in SCM code. */
5966 switch (cmsg->cmsg_type) {
5967 case SCTP_INIT:
5968 /* SCTP Socket API Extension
5969 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5971 * This cmsghdr structure provides information for
5972 * initializing new SCTP associations with sendmsg().
5973 * The SCTP_INITMSG socket option uses this same data
5974 * structure. This structure is not used for
5975 * recvmsg().
5977 * cmsg_level cmsg_type cmsg_data[]
5978 * ------------ ------------ ----------------------
5979 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5981 if (cmsg->cmsg_len !=
5982 CMSG_LEN(sizeof(struct sctp_initmsg)))
5983 return -EINVAL;
5984 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5985 break;
5987 case SCTP_SNDRCV:
5988 /* SCTP Socket API Extension
5989 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5991 * This cmsghdr structure specifies SCTP options for
5992 * sendmsg() and describes SCTP header information
5993 * about a received message through recvmsg().
5995 * cmsg_level cmsg_type cmsg_data[]
5996 * ------------ ------------ ----------------------
5997 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5999 if (cmsg->cmsg_len !=
6000 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6001 return -EINVAL;
6003 cmsgs->info =
6004 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6006 /* Minimally, validate the sinfo_flags. */
6007 if (cmsgs->info->sinfo_flags &
6008 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6009 SCTP_ABORT | SCTP_EOF))
6010 return -EINVAL;
6011 break;
6013 default:
6014 return -EINVAL;
6017 return 0;
6021 * Wait for a packet..
6022 * Note: This function is the same function as in core/datagram.c
6023 * with a few modifications to make lksctp work.
6025 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6027 int error;
6028 DEFINE_WAIT(wait);
6030 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6032 /* Socket errors? */
6033 error = sock_error(sk);
6034 if (error)
6035 goto out;
6037 if (!skb_queue_empty(&sk->sk_receive_queue))
6038 goto ready;
6040 /* Socket shut down? */
6041 if (sk->sk_shutdown & RCV_SHUTDOWN)
6042 goto out;
6044 /* Sequenced packets can come disconnected. If so we report the
6045 * problem.
6047 error = -ENOTCONN;
6049 /* Is there a good reason to think that we may receive some data? */
6050 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6051 goto out;
6053 /* Handle signals. */
6054 if (signal_pending(current))
6055 goto interrupted;
6057 /* Let another process have a go. Since we are going to sleep
6058 * anyway. Note: This may cause odd behaviors if the message
6059 * does not fit in the user's buffer, but this seems to be the
6060 * only way to honor MSG_DONTWAIT realistically.
6062 sctp_release_sock(sk);
6063 *timeo_p = schedule_timeout(*timeo_p);
6064 sctp_lock_sock(sk);
6066 ready:
6067 finish_wait(sk_sleep(sk), &wait);
6068 return 0;
6070 interrupted:
6071 error = sock_intr_errno(*timeo_p);
6073 out:
6074 finish_wait(sk_sleep(sk), &wait);
6075 *err = error;
6076 return error;
6079 /* Receive a datagram.
6080 * Note: This is pretty much the same routine as in core/datagram.c
6081 * with a few changes to make lksctp work.
6083 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6084 int noblock, int *err)
6086 int error;
6087 struct sk_buff *skb;
6088 long timeo;
6090 timeo = sock_rcvtimeo(sk, noblock);
6092 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6093 timeo, MAX_SCHEDULE_TIMEOUT);
6095 do {
6096 /* Again only user level code calls this function,
6097 * so nothing interrupt level
6098 * will suddenly eat the receive_queue.
6100 * Look at current nfs client by the way...
6101 * However, this function was correct in any case. 8)
6103 if (flags & MSG_PEEK) {
6104 spin_lock_bh(&sk->sk_receive_queue.lock);
6105 skb = skb_peek(&sk->sk_receive_queue);
6106 if (skb)
6107 atomic_inc(&skb->users);
6108 spin_unlock_bh(&sk->sk_receive_queue.lock);
6109 } else {
6110 skb = skb_dequeue(&sk->sk_receive_queue);
6113 if (skb)
6114 return skb;
6116 /* Caller is allowed not to check sk->sk_err before calling. */
6117 error = sock_error(sk);
6118 if (error)
6119 goto no_packet;
6121 if (sk->sk_shutdown & RCV_SHUTDOWN)
6122 break;
6124 /* User doesn't want to wait. */
6125 error = -EAGAIN;
6126 if (!timeo)
6127 goto no_packet;
6128 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6130 return NULL;
6132 no_packet:
6133 *err = error;
6134 return NULL;
6137 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6138 static void __sctp_write_space(struct sctp_association *asoc)
6140 struct sock *sk = asoc->base.sk;
6141 struct socket *sock = sk->sk_socket;
6143 if ((sctp_wspace(asoc) > 0) && sock) {
6144 if (waitqueue_active(&asoc->wait))
6145 wake_up_interruptible(&asoc->wait);
6147 if (sctp_writeable(sk)) {
6148 wait_queue_head_t *wq = sk_sleep(sk);
6150 if (wq && waitqueue_active(wq))
6151 wake_up_interruptible(wq);
6153 /* Note that we try to include the Async I/O support
6154 * here by modeling from the current TCP/UDP code.
6155 * We have not tested with it yet.
6157 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6158 sock_wake_async(sock,
6159 SOCK_WAKE_SPACE, POLL_OUT);
6164 /* Do accounting for the sndbuf space.
6165 * Decrement the used sndbuf space of the corresponding association by the
6166 * data size which was just transmitted(freed).
6168 static void sctp_wfree(struct sk_buff *skb)
6170 struct sctp_association *asoc;
6171 struct sctp_chunk *chunk;
6172 struct sock *sk;
6174 /* Get the saved chunk pointer. */
6175 chunk = *((struct sctp_chunk **)(skb->cb));
6176 asoc = chunk->asoc;
6177 sk = asoc->base.sk;
6178 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6179 sizeof(struct sk_buff) +
6180 sizeof(struct sctp_chunk);
6182 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6185 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6187 sk->sk_wmem_queued -= skb->truesize;
6188 sk_mem_uncharge(sk, skb->truesize);
6190 sock_wfree(skb);
6191 __sctp_write_space(asoc);
6193 sctp_association_put(asoc);
6196 /* Do accounting for the receive space on the socket.
6197 * Accounting for the association is done in ulpevent.c
6198 * We set this as a destructor for the cloned data skbs so that
6199 * accounting is done at the correct time.
6201 void sctp_sock_rfree(struct sk_buff *skb)
6203 struct sock *sk = skb->sk;
6204 struct sctp_ulpevent *event = sctp_skb2event(skb);
6206 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6209 * Mimic the behavior of sock_rfree
6211 sk_mem_uncharge(sk, event->rmem_len);
6215 /* Helper function to wait for space in the sndbuf. */
6216 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6217 size_t msg_len)
6219 struct sock *sk = asoc->base.sk;
6220 int err = 0;
6221 long current_timeo = *timeo_p;
6222 DEFINE_WAIT(wait);
6224 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6225 asoc, (long)(*timeo_p), msg_len);
6227 /* Increment the association's refcnt. */
6228 sctp_association_hold(asoc);
6230 /* Wait on the association specific sndbuf space. */
6231 for (;;) {
6232 prepare_to_wait_exclusive(&asoc->wait, &wait,
6233 TASK_INTERRUPTIBLE);
6234 if (!*timeo_p)
6235 goto do_nonblock;
6236 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6237 asoc->base.dead)
6238 goto do_error;
6239 if (signal_pending(current))
6240 goto do_interrupted;
6241 if (msg_len <= sctp_wspace(asoc))
6242 break;
6244 /* Let another process have a go. Since we are going
6245 * to sleep anyway.
6247 sctp_release_sock(sk);
6248 current_timeo = schedule_timeout(current_timeo);
6249 BUG_ON(sk != asoc->base.sk);
6250 sctp_lock_sock(sk);
6252 *timeo_p = current_timeo;
6255 out:
6256 finish_wait(&asoc->wait, &wait);
6258 /* Release the association's refcnt. */
6259 sctp_association_put(asoc);
6261 return err;
6263 do_error:
6264 err = -EPIPE;
6265 goto out;
6267 do_interrupted:
6268 err = sock_intr_errno(*timeo_p);
6269 goto out;
6271 do_nonblock:
6272 err = -EAGAIN;
6273 goto out;
6276 void sctp_data_ready(struct sock *sk, int len)
6278 struct socket_wq *wq;
6280 rcu_read_lock();
6281 wq = rcu_dereference(sk->sk_wq);
6282 if (wq_has_sleeper(wq))
6283 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6284 POLLRDNORM | POLLRDBAND);
6285 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6286 rcu_read_unlock();
6289 /* If socket sndbuf has changed, wake up all per association waiters. */
6290 void sctp_write_space(struct sock *sk)
6292 struct sctp_association *asoc;
6294 /* Wake up the tasks in each wait queue. */
6295 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6296 __sctp_write_space(asoc);
6300 /* Is there any sndbuf space available on the socket?
6302 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6303 * associations on the same socket. For a UDP-style socket with
6304 * multiple associations, it is possible for it to be "unwriteable"
6305 * prematurely. I assume that this is acceptable because
6306 * a premature "unwriteable" is better than an accidental "writeable" which
6307 * would cause an unwanted block under certain circumstances. For the 1-1
6308 * UDP-style sockets or TCP-style sockets, this code should work.
6309 * - Daisy
6311 static int sctp_writeable(struct sock *sk)
6313 int amt = 0;
6315 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6316 if (amt < 0)
6317 amt = 0;
6318 return amt;
6321 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6322 * returns immediately with EINPROGRESS.
6324 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6326 struct sock *sk = asoc->base.sk;
6327 int err = 0;
6328 long current_timeo = *timeo_p;
6329 DEFINE_WAIT(wait);
6331 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6332 (long)(*timeo_p));
6334 /* Increment the association's refcnt. */
6335 sctp_association_hold(asoc);
6337 for (;;) {
6338 prepare_to_wait_exclusive(&asoc->wait, &wait,
6339 TASK_INTERRUPTIBLE);
6340 if (!*timeo_p)
6341 goto do_nonblock;
6342 if (sk->sk_shutdown & RCV_SHUTDOWN)
6343 break;
6344 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6345 asoc->base.dead)
6346 goto do_error;
6347 if (signal_pending(current))
6348 goto do_interrupted;
6350 if (sctp_state(asoc, ESTABLISHED))
6351 break;
6353 /* Let another process have a go. Since we are going
6354 * to sleep anyway.
6356 sctp_release_sock(sk);
6357 current_timeo = schedule_timeout(current_timeo);
6358 sctp_lock_sock(sk);
6360 *timeo_p = current_timeo;
6363 out:
6364 finish_wait(&asoc->wait, &wait);
6366 /* Release the association's refcnt. */
6367 sctp_association_put(asoc);
6369 return err;
6371 do_error:
6372 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6373 err = -ETIMEDOUT;
6374 else
6375 err = -ECONNREFUSED;
6376 goto out;
6378 do_interrupted:
6379 err = sock_intr_errno(*timeo_p);
6380 goto out;
6382 do_nonblock:
6383 err = -EINPROGRESS;
6384 goto out;
6387 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6389 struct sctp_endpoint *ep;
6390 int err = 0;
6391 DEFINE_WAIT(wait);
6393 ep = sctp_sk(sk)->ep;
6396 for (;;) {
6397 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6398 TASK_INTERRUPTIBLE);
6400 if (list_empty(&ep->asocs)) {
6401 sctp_release_sock(sk);
6402 timeo = schedule_timeout(timeo);
6403 sctp_lock_sock(sk);
6406 err = -EINVAL;
6407 if (!sctp_sstate(sk, LISTENING))
6408 break;
6410 err = 0;
6411 if (!list_empty(&ep->asocs))
6412 break;
6414 err = sock_intr_errno(timeo);
6415 if (signal_pending(current))
6416 break;
6418 err = -EAGAIN;
6419 if (!timeo)
6420 break;
6423 finish_wait(sk_sleep(sk), &wait);
6425 return err;
6428 static void sctp_wait_for_close(struct sock *sk, long timeout)
6430 DEFINE_WAIT(wait);
6432 do {
6433 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6434 if (list_empty(&sctp_sk(sk)->ep->asocs))
6435 break;
6436 sctp_release_sock(sk);
6437 timeout = schedule_timeout(timeout);
6438 sctp_lock_sock(sk);
6439 } while (!signal_pending(current) && timeout);
6441 finish_wait(sk_sleep(sk), &wait);
6444 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6446 struct sk_buff *frag;
6448 if (!skb->data_len)
6449 goto done;
6451 /* Don't forget the fragments. */
6452 skb_walk_frags(skb, frag)
6453 sctp_skb_set_owner_r_frag(frag, sk);
6455 done:
6456 sctp_skb_set_owner_r(skb, sk);
6459 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6460 struct sctp_association *asoc)
6462 struct inet_sock *inet = inet_sk(sk);
6463 struct inet_sock *newinet;
6465 newsk->sk_type = sk->sk_type;
6466 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6467 newsk->sk_flags = sk->sk_flags;
6468 newsk->sk_no_check = sk->sk_no_check;
6469 newsk->sk_reuse = sk->sk_reuse;
6471 newsk->sk_shutdown = sk->sk_shutdown;
6472 newsk->sk_destruct = inet_sock_destruct;
6473 newsk->sk_family = sk->sk_family;
6474 newsk->sk_protocol = IPPROTO_SCTP;
6475 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6476 newsk->sk_sndbuf = sk->sk_sndbuf;
6477 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6478 newsk->sk_lingertime = sk->sk_lingertime;
6479 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6480 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6482 newinet = inet_sk(newsk);
6484 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6485 * getsockname() and getpeername()
6487 newinet->inet_sport = inet->inet_sport;
6488 newinet->inet_saddr = inet->inet_saddr;
6489 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6490 newinet->inet_dport = htons(asoc->peer.port);
6491 newinet->pmtudisc = inet->pmtudisc;
6492 newinet->inet_id = asoc->next_tsn ^ jiffies;
6494 newinet->uc_ttl = inet->uc_ttl;
6495 newinet->mc_loop = 1;
6496 newinet->mc_ttl = 1;
6497 newinet->mc_index = 0;
6498 newinet->mc_list = NULL;
6501 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6502 * and its messages to the newsk.
6504 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6505 struct sctp_association *assoc,
6506 sctp_socket_type_t type)
6508 struct sctp_sock *oldsp = sctp_sk(oldsk);
6509 struct sctp_sock *newsp = sctp_sk(newsk);
6510 struct sctp_bind_bucket *pp; /* hash list port iterator */
6511 struct sctp_endpoint *newep = newsp->ep;
6512 struct sk_buff *skb, *tmp;
6513 struct sctp_ulpevent *event;
6514 struct sctp_bind_hashbucket *head;
6516 /* Migrate socket buffer sizes and all the socket level options to the
6517 * new socket.
6519 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6520 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6521 /* Brute force copy old sctp opt. */
6522 inet_sk_copy_descendant(newsk, oldsk);
6524 /* Restore the ep value that was overwritten with the above structure
6525 * copy.
6527 newsp->ep = newep;
6528 newsp->hmac = NULL;
6530 /* Hook this new socket in to the bind_hash list. */
6531 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6532 sctp_local_bh_disable();
6533 sctp_spin_lock(&head->lock);
6534 pp = sctp_sk(oldsk)->bind_hash;
6535 sk_add_bind_node(newsk, &pp->owner);
6536 sctp_sk(newsk)->bind_hash = pp;
6537 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6538 sctp_spin_unlock(&head->lock);
6539 sctp_local_bh_enable();
6541 /* Copy the bind_addr list from the original endpoint to the new
6542 * endpoint so that we can handle restarts properly
6544 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6545 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6547 /* Move any messages in the old socket's receive queue that are for the
6548 * peeled off association to the new socket's receive queue.
6550 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6551 event = sctp_skb2event(skb);
6552 if (event->asoc == assoc) {
6553 __skb_unlink(skb, &oldsk->sk_receive_queue);
6554 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6555 sctp_skb_set_owner_r_frag(skb, newsk);
6559 /* Clean up any messages pending delivery due to partial
6560 * delivery. Three cases:
6561 * 1) No partial deliver; no work.
6562 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6563 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6565 skb_queue_head_init(&newsp->pd_lobby);
6566 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6568 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6569 struct sk_buff_head *queue;
6571 /* Decide which queue to move pd_lobby skbs to. */
6572 if (assoc->ulpq.pd_mode) {
6573 queue = &newsp->pd_lobby;
6574 } else
6575 queue = &newsk->sk_receive_queue;
6577 /* Walk through the pd_lobby, looking for skbs that
6578 * need moved to the new socket.
6580 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6581 event = sctp_skb2event(skb);
6582 if (event->asoc == assoc) {
6583 __skb_unlink(skb, &oldsp->pd_lobby);
6584 __skb_queue_tail(queue, skb);
6585 sctp_skb_set_owner_r_frag(skb, newsk);
6589 /* Clear up any skbs waiting for the partial
6590 * delivery to finish.
6592 if (assoc->ulpq.pd_mode)
6593 sctp_clear_pd(oldsk, NULL);
6597 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6598 sctp_skb_set_owner_r_frag(skb, newsk);
6600 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6601 sctp_skb_set_owner_r_frag(skb, newsk);
6603 /* Set the type of socket to indicate that it is peeled off from the
6604 * original UDP-style socket or created with the accept() call on a
6605 * TCP-style socket..
6607 newsp->type = type;
6609 /* Mark the new socket "in-use" by the user so that any packets
6610 * that may arrive on the association after we've moved it are
6611 * queued to the backlog. This prevents a potential race between
6612 * backlog processing on the old socket and new-packet processing
6613 * on the new socket.
6615 * The caller has just allocated newsk so we can guarantee that other
6616 * paths won't try to lock it and then oldsk.
6618 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6619 sctp_assoc_migrate(assoc, newsk);
6621 /* If the association on the newsk is already closed before accept()
6622 * is called, set RCV_SHUTDOWN flag.
6624 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6625 newsk->sk_shutdown |= RCV_SHUTDOWN;
6627 newsk->sk_state = SCTP_SS_ESTABLISHED;
6628 sctp_release_sock(newsk);
6632 /* This proto struct describes the ULP interface for SCTP. */
6633 struct proto sctp_prot = {
6634 .name = "SCTP",
6635 .owner = THIS_MODULE,
6636 .close = sctp_close,
6637 .connect = sctp_connect,
6638 .disconnect = sctp_disconnect,
6639 .accept = sctp_accept,
6640 .ioctl = sctp_ioctl,
6641 .init = sctp_init_sock,
6642 .destroy = sctp_destroy_sock,
6643 .shutdown = sctp_shutdown,
6644 .setsockopt = sctp_setsockopt,
6645 .getsockopt = sctp_getsockopt,
6646 .sendmsg = sctp_sendmsg,
6647 .recvmsg = sctp_recvmsg,
6648 .bind = sctp_bind,
6649 .backlog_rcv = sctp_backlog_rcv,
6650 .hash = sctp_hash,
6651 .unhash = sctp_unhash,
6652 .get_port = sctp_get_port,
6653 .obj_size = sizeof(struct sctp_sock),
6654 .sysctl_mem = sysctl_sctp_mem,
6655 .sysctl_rmem = sysctl_sctp_rmem,
6656 .sysctl_wmem = sysctl_sctp_wmem,
6657 .memory_pressure = &sctp_memory_pressure,
6658 .enter_memory_pressure = sctp_enter_memory_pressure,
6659 .memory_allocated = &sctp_memory_allocated,
6660 .sockets_allocated = &sctp_sockets_allocated,
6663 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6665 struct proto sctpv6_prot = {
6666 .name = "SCTPv6",
6667 .owner = THIS_MODULE,
6668 .close = sctp_close,
6669 .connect = sctp_connect,
6670 .disconnect = sctp_disconnect,
6671 .accept = sctp_accept,
6672 .ioctl = sctp_ioctl,
6673 .init = sctp_init_sock,
6674 .destroy = sctp_destroy_sock,
6675 .shutdown = sctp_shutdown,
6676 .setsockopt = sctp_setsockopt,
6677 .getsockopt = sctp_getsockopt,
6678 .sendmsg = sctp_sendmsg,
6679 .recvmsg = sctp_recvmsg,
6680 .bind = sctp_bind,
6681 .backlog_rcv = sctp_backlog_rcv,
6682 .hash = sctp_hash,
6683 .unhash = sctp_unhash,
6684 .get_port = sctp_get_port,
6685 .obj_size = sizeof(struct sctp6_sock),
6686 .sysctl_mem = sysctl_sctp_mem,
6687 .sysctl_rmem = sysctl_sctp_rmem,
6688 .sysctl_wmem = sysctl_sctp_wmem,
6689 .memory_pressure = &sctp_memory_pressure,
6690 .enter_memory_pressure = sctp_enter_memory_pressure,
6691 .memory_allocated = &sctp_memory_allocated,
6692 .sockets_allocated = &sctp_sockets_allocated,
6694 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */