at91: Fix uhpck clock rate in upll case
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / sctp / socket.c
blob0b9ee34ad35ceb31ff764baf9cec73fba010f55b
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 != htons(bp->port)) {
662 retval = -EINVAL;
663 goto err_bindx_rem;
666 /* FIXME - There is probably a need to check if sk->sk_saddr and
667 * sk->sk_rcv_addr are currently set to one of the addresses to
668 * be removed. This is something which needs to be looked into
669 * when we are fixing the outstanding issues with multi-homing
670 * socket routing and failover schemes. Refer to comments in
671 * sctp_do_bind(). -daisy
673 retval = sctp_del_bind_addr(bp, sa_addr);
675 addr_buf += af->sockaddr_len;
676 err_bindx_rem:
677 if (retval < 0) {
678 /* Failed. Add the ones that has been removed back */
679 if (cnt > 0)
680 sctp_bindx_add(sk, addrs, cnt);
681 return retval;
685 return retval;
688 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
689 * the associations that are part of the endpoint indicating that a list of
690 * local addresses are removed from the endpoint.
692 * If any of the addresses is already in the bind address list of the
693 * association, we do not send the chunk for that association. But it will not
694 * affect other associations.
696 * Only sctp_setsockopt_bindx() is supposed to call this function.
698 static int sctp_send_asconf_del_ip(struct sock *sk,
699 struct sockaddr *addrs,
700 int addrcnt)
702 struct sctp_sock *sp;
703 struct sctp_endpoint *ep;
704 struct sctp_association *asoc;
705 struct sctp_transport *transport;
706 struct sctp_bind_addr *bp;
707 struct sctp_chunk *chunk;
708 union sctp_addr *laddr;
709 void *addr_buf;
710 struct sctp_af *af;
711 struct sctp_sockaddr_entry *saddr;
712 int i;
713 int retval = 0;
715 if (!sctp_addip_enable)
716 return retval;
718 sp = sctp_sk(sk);
719 ep = sp->ep;
721 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
722 __func__, sk, addrs, addrcnt);
724 list_for_each_entry(asoc, &ep->asocs, asocs) {
726 if (!asoc->peer.asconf_capable)
727 continue;
729 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
730 continue;
732 if (!sctp_state(asoc, ESTABLISHED))
733 continue;
735 /* Check if any address in the packed array of addresses is
736 * not present in the bind address list of the association.
737 * If so, do not send the asconf chunk to its peer, but
738 * continue with other associations.
740 addr_buf = addrs;
741 for (i = 0; i < addrcnt; i++) {
742 laddr = (union sctp_addr *)addr_buf;
743 af = sctp_get_af_specific(laddr->v4.sin_family);
744 if (!af) {
745 retval = -EINVAL;
746 goto out;
749 if (!sctp_assoc_lookup_laddr(asoc, laddr))
750 break;
752 addr_buf += af->sockaddr_len;
754 if (i < addrcnt)
755 continue;
757 /* Find one address in the association's bind address list
758 * that is not in the packed array of addresses. This is to
759 * make sure that we do not delete all the addresses in the
760 * association.
762 bp = &asoc->base.bind_addr;
763 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
764 addrcnt, sp);
765 if (!laddr)
766 continue;
768 /* We do not need RCU protection throughout this loop
769 * because this is done under a socket lock from the
770 * setsockopt call.
772 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
773 SCTP_PARAM_DEL_IP);
774 if (!chunk) {
775 retval = -ENOMEM;
776 goto out;
779 /* Reset use_as_src flag for the addresses in the bind address
780 * list that are to be deleted.
782 addr_buf = addrs;
783 for (i = 0; i < addrcnt; i++) {
784 laddr = (union sctp_addr *)addr_buf;
785 af = sctp_get_af_specific(laddr->v4.sin_family);
786 list_for_each_entry(saddr, &bp->address_list, list) {
787 if (sctp_cmp_addr_exact(&saddr->a, laddr))
788 saddr->state = SCTP_ADDR_DEL;
790 addr_buf += af->sockaddr_len;
793 /* Update the route and saddr entries for all the transports
794 * as some of the addresses in the bind address list are
795 * about to be deleted and cannot be used as source addresses.
797 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
798 transports) {
799 dst_release(transport->dst);
800 sctp_transport_route(transport, NULL,
801 sctp_sk(asoc->base.sk));
804 retval = sctp_send_asconf(asoc, chunk);
806 out:
807 return retval;
810 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
812 * API 8.1
813 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
814 * int flags);
816 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
817 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
818 * or IPv6 addresses.
820 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
821 * Section 3.1.2 for this usage.
823 * addrs is a pointer to an array of one or more socket addresses. Each
824 * address is contained in its appropriate structure (i.e. struct
825 * sockaddr_in or struct sockaddr_in6) the family of the address type
826 * must be used to distinguish the address length (note that this
827 * representation is termed a "packed array" of addresses). The caller
828 * specifies the number of addresses in the array with addrcnt.
830 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
831 * -1, and sets errno to the appropriate error code.
833 * For SCTP, the port given in each socket address must be the same, or
834 * sctp_bindx() will fail, setting errno to EINVAL.
836 * The flags parameter is formed from the bitwise OR of zero or more of
837 * the following currently defined flags:
839 * SCTP_BINDX_ADD_ADDR
841 * SCTP_BINDX_REM_ADDR
843 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
844 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
845 * addresses from the association. The two flags are mutually exclusive;
846 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
847 * not remove all addresses from an association; sctp_bindx() will
848 * reject such an attempt with EINVAL.
850 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
851 * additional addresses with an endpoint after calling bind(). Or use
852 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
853 * socket is associated with so that no new association accepted will be
854 * associated with those addresses. If the endpoint supports dynamic
855 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
856 * endpoint to send the appropriate message to the peer to change the
857 * peers address lists.
859 * Adding and removing addresses from a connected association is
860 * optional functionality. Implementations that do not support this
861 * functionality should return EOPNOTSUPP.
863 * Basically do nothing but copying the addresses from user to kernel
864 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
865 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
866 * from userspace.
868 * We don't use copy_from_user() for optimization: we first do the
869 * sanity checks (buffer size -fast- and access check-healthy
870 * pointer); if all of those succeed, then we can alloc the memory
871 * (expensive operation) needed to copy the data to kernel. Then we do
872 * the copying without checking the user space area
873 * (__copy_from_user()).
875 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
876 * it.
878 * sk The sk of the socket
879 * addrs The pointer to the addresses in user land
880 * addrssize Size of the addrs buffer
881 * op Operation to perform (add or remove, see the flags of
882 * sctp_bindx)
884 * Returns 0 if ok, <0 errno code on error.
886 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
887 struct sockaddr __user *addrs,
888 int addrs_size, int op)
890 struct sockaddr *kaddrs;
891 int err;
892 int addrcnt = 0;
893 int walk_size = 0;
894 struct sockaddr *sa_addr;
895 void *addr_buf;
896 struct sctp_af *af;
898 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
899 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
901 if (unlikely(addrs_size <= 0))
902 return -EINVAL;
904 /* Check the user passed a healthy pointer. */
905 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
906 return -EFAULT;
908 /* Alloc space for the address array in kernel memory. */
909 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
910 if (unlikely(!kaddrs))
911 return -ENOMEM;
913 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
914 kfree(kaddrs);
915 return -EFAULT;
918 /* Walk through the addrs buffer and count the number of addresses. */
919 addr_buf = kaddrs;
920 while (walk_size < addrs_size) {
921 if (walk_size + sizeof(sa_family_t) > addrs_size) {
922 kfree(kaddrs);
923 return -EINVAL;
926 sa_addr = (struct sockaddr *)addr_buf;
927 af = sctp_get_af_specific(sa_addr->sa_family);
929 /* If the address family is not supported or if this address
930 * causes the address buffer to overflow return EINVAL.
932 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
933 kfree(kaddrs);
934 return -EINVAL;
936 addrcnt++;
937 addr_buf += af->sockaddr_len;
938 walk_size += af->sockaddr_len;
941 /* Do the work. */
942 switch (op) {
943 case SCTP_BINDX_ADD_ADDR:
944 err = sctp_bindx_add(sk, kaddrs, addrcnt);
945 if (err)
946 goto out;
947 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
948 break;
950 case SCTP_BINDX_REM_ADDR:
951 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
952 if (err)
953 goto out;
954 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
955 break;
957 default:
958 err = -EINVAL;
959 break;
962 out:
963 kfree(kaddrs);
965 return err;
968 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
970 * Common routine for handling connect() and sctp_connectx().
971 * Connect will come in with just a single address.
973 static int __sctp_connect(struct sock* sk,
974 struct sockaddr *kaddrs,
975 int addrs_size,
976 sctp_assoc_t *assoc_id)
978 struct sctp_sock *sp;
979 struct sctp_endpoint *ep;
980 struct sctp_association *asoc = NULL;
981 struct sctp_association *asoc2;
982 struct sctp_transport *transport;
983 union sctp_addr to;
984 struct sctp_af *af;
985 sctp_scope_t scope;
986 long timeo;
987 int err = 0;
988 int addrcnt = 0;
989 int walk_size = 0;
990 union sctp_addr *sa_addr = NULL;
991 void *addr_buf;
992 unsigned short port;
993 unsigned int f_flags = 0;
995 sp = sctp_sk(sk);
996 ep = sp->ep;
998 /* connect() cannot be done on a socket that is already in ESTABLISHED
999 * state - UDP-style peeled off socket or a TCP-style socket that
1000 * is already connected.
1001 * It cannot be done even on a TCP-style listening socket.
1003 if (sctp_sstate(sk, ESTABLISHED) ||
1004 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1005 err = -EISCONN;
1006 goto out_free;
1009 /* Walk through the addrs buffer and count the number of addresses. */
1010 addr_buf = kaddrs;
1011 while (walk_size < addrs_size) {
1012 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1013 err = -EINVAL;
1014 goto out_free;
1017 sa_addr = (union sctp_addr *)addr_buf;
1018 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1020 /* If the address family is not supported or if this address
1021 * causes the address buffer to overflow return EINVAL.
1023 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1024 err = -EINVAL;
1025 goto out_free;
1028 port = ntohs(sa_addr->v4.sin_port);
1030 /* Save current address so we can work with it */
1031 memcpy(&to, sa_addr, af->sockaddr_len);
1033 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1034 if (err)
1035 goto out_free;
1037 /* Make sure the destination port is correctly set
1038 * in all addresses.
1040 if (asoc && asoc->peer.port && asoc->peer.port != port)
1041 goto out_free;
1044 /* Check if there already is a matching association on the
1045 * endpoint (other than the one created here).
1047 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1048 if (asoc2 && asoc2 != asoc) {
1049 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1050 err = -EISCONN;
1051 else
1052 err = -EALREADY;
1053 goto out_free;
1056 /* If we could not find a matching association on the endpoint,
1057 * make sure that there is no peeled-off association matching
1058 * the peer address even on another socket.
1060 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1061 err = -EADDRNOTAVAIL;
1062 goto out_free;
1065 if (!asoc) {
1066 /* If a bind() or sctp_bindx() is not called prior to
1067 * an sctp_connectx() call, the system picks an
1068 * ephemeral port and will choose an address set
1069 * equivalent to binding with a wildcard address.
1071 if (!ep->base.bind_addr.port) {
1072 if (sctp_autobind(sk)) {
1073 err = -EAGAIN;
1074 goto out_free;
1076 } else {
1078 * If an unprivileged user inherits a 1-many
1079 * style socket with open associations on a
1080 * privileged port, it MAY be permitted to
1081 * accept new associations, but it SHOULD NOT
1082 * be permitted to open new associations.
1084 if (ep->base.bind_addr.port < PROT_SOCK &&
1085 !capable(CAP_NET_BIND_SERVICE)) {
1086 err = -EACCES;
1087 goto out_free;
1091 scope = sctp_scope(&to);
1092 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1093 if (!asoc) {
1094 err = -ENOMEM;
1095 goto out_free;
1098 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1099 GFP_KERNEL);
1100 if (err < 0) {
1101 goto out_free;
1106 /* Prime the peer's transport structures. */
1107 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1108 SCTP_UNKNOWN);
1109 if (!transport) {
1110 err = -ENOMEM;
1111 goto out_free;
1114 addrcnt++;
1115 addr_buf += af->sockaddr_len;
1116 walk_size += af->sockaddr_len;
1119 /* In case the user of sctp_connectx() wants an association
1120 * id back, assign one now.
1122 if (assoc_id) {
1123 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1124 if (err < 0)
1125 goto out_free;
1128 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1129 if (err < 0) {
1130 goto out_free;
1133 /* Initialize sk's dport and daddr for getpeername() */
1134 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1135 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1136 af->to_sk_daddr(sa_addr, sk);
1137 sk->sk_err = 0;
1139 /* in-kernel sockets don't generally have a file allocated to them
1140 * if all they do is call sock_create_kern().
1142 if (sk->sk_socket->file)
1143 f_flags = sk->sk_socket->file->f_flags;
1145 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1147 err = sctp_wait_for_connect(asoc, &timeo);
1148 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1149 *assoc_id = asoc->assoc_id;
1151 /* Don't free association on exit. */
1152 asoc = NULL;
1154 out_free:
1156 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1157 " kaddrs: %p err: %d\n",
1158 asoc, kaddrs, err);
1159 if (asoc)
1160 sctp_association_free(asoc);
1161 return err;
1164 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1166 * API 8.9
1167 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1168 * sctp_assoc_t *asoc);
1170 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1171 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1172 * or IPv6 addresses.
1174 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1175 * Section 3.1.2 for this usage.
1177 * addrs is a pointer to an array of one or more socket addresses. Each
1178 * address is contained in its appropriate structure (i.e. struct
1179 * sockaddr_in or struct sockaddr_in6) the family of the address type
1180 * must be used to distengish the address length (note that this
1181 * representation is termed a "packed array" of addresses). The caller
1182 * specifies the number of addresses in the array with addrcnt.
1184 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1185 * the association id of the new association. On failure, sctp_connectx()
1186 * returns -1, and sets errno to the appropriate error code. The assoc_id
1187 * is not touched by the kernel.
1189 * For SCTP, the port given in each socket address must be the same, or
1190 * sctp_connectx() will fail, setting errno to EINVAL.
1192 * An application can use sctp_connectx to initiate an association with
1193 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1194 * allows a caller to specify multiple addresses at which a peer can be
1195 * reached. The way the SCTP stack uses the list of addresses to set up
1196 * the association is implementation dependant. This function only
1197 * specifies that the stack will try to make use of all the addresses in
1198 * the list when needed.
1200 * Note that the list of addresses passed in is only used for setting up
1201 * the association. It does not necessarily equal the set of addresses
1202 * the peer uses for the resulting association. If the caller wants to
1203 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1204 * retrieve them after the association has been set up.
1206 * Basically do nothing but copying the addresses from user to kernel
1207 * land and invoking either sctp_connectx(). This is used for tunneling
1208 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1210 * We don't use copy_from_user() for optimization: we first do the
1211 * sanity checks (buffer size -fast- and access check-healthy
1212 * pointer); if all of those succeed, then we can alloc the memory
1213 * (expensive operation) needed to copy the data to kernel. Then we do
1214 * the copying without checking the user space area
1215 * (__copy_from_user()).
1217 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1218 * it.
1220 * sk The sk of the socket
1221 * addrs The pointer to the addresses in user land
1222 * addrssize Size of the addrs buffer
1224 * Returns >=0 if ok, <0 errno code on error.
1226 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1227 struct sockaddr __user *addrs,
1228 int addrs_size,
1229 sctp_assoc_t *assoc_id)
1231 int err = 0;
1232 struct sockaddr *kaddrs;
1234 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1235 __func__, sk, addrs, addrs_size);
1237 if (unlikely(addrs_size <= 0))
1238 return -EINVAL;
1240 /* Check the user passed a healthy pointer. */
1241 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1242 return -EFAULT;
1244 /* Alloc space for the address array in kernel memory. */
1245 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1246 if (unlikely(!kaddrs))
1247 return -ENOMEM;
1249 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1250 err = -EFAULT;
1251 } else {
1252 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1255 kfree(kaddrs);
1257 return err;
1261 * This is an older interface. It's kept for backward compatibility
1262 * to the option that doesn't provide association id.
1264 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1265 struct sockaddr __user *addrs,
1266 int addrs_size)
1268 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1272 * New interface for the API. The since the API is done with a socket
1273 * option, to make it simple we feed back the association id is as a return
1274 * indication to the call. Error is always negative and association id is
1275 * always positive.
1277 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1278 struct sockaddr __user *addrs,
1279 int addrs_size)
1281 sctp_assoc_t assoc_id = 0;
1282 int err = 0;
1284 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1286 if (err)
1287 return err;
1288 else
1289 return assoc_id;
1293 * New (hopefully final) interface for the API.
1294 * We use the sctp_getaddrs_old structure so that use-space library
1295 * can avoid any unnecessary allocations. The only defferent part
1296 * is that we store the actual length of the address buffer into the
1297 * addrs_num structure member. That way we can re-use the existing
1298 * code.
1300 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1301 char __user *optval,
1302 int __user *optlen)
1304 struct sctp_getaddrs_old param;
1305 sctp_assoc_t assoc_id = 0;
1306 int err = 0;
1308 if (len < sizeof(param))
1309 return -EINVAL;
1311 if (copy_from_user(&param, optval, sizeof(param)))
1312 return -EFAULT;
1314 err = __sctp_setsockopt_connectx(sk,
1315 (struct sockaddr __user *)param.addrs,
1316 param.addr_num, &assoc_id);
1318 if (err == 0 || err == -EINPROGRESS) {
1319 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1320 return -EFAULT;
1321 if (put_user(sizeof(assoc_id), optlen))
1322 return -EFAULT;
1325 return err;
1328 /* API 3.1.4 close() - UDP Style Syntax
1329 * Applications use close() to perform graceful shutdown (as described in
1330 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1331 * by a UDP-style socket.
1333 * The syntax is
1335 * ret = close(int sd);
1337 * sd - the socket descriptor of the associations to be closed.
1339 * To gracefully shutdown a specific association represented by the
1340 * UDP-style socket, an application should use the sendmsg() call,
1341 * passing no user data, but including the appropriate flag in the
1342 * ancillary data (see Section xxxx).
1344 * If sd in the close() call is a branched-off socket representing only
1345 * one association, the shutdown is performed on that association only.
1347 * 4.1.6 close() - TCP Style Syntax
1349 * Applications use close() to gracefully close down an association.
1351 * The syntax is:
1353 * int close(int sd);
1355 * sd - the socket descriptor of the association to be closed.
1357 * After an application calls close() on a socket descriptor, no further
1358 * socket operations will succeed on that descriptor.
1360 * API 7.1.4 SO_LINGER
1362 * An application using the TCP-style socket can use this option to
1363 * perform the SCTP ABORT primitive. The linger option structure is:
1365 * struct linger {
1366 * int l_onoff; // option on/off
1367 * int l_linger; // linger time
1368 * };
1370 * To enable the option, set l_onoff to 1. If the l_linger value is set
1371 * to 0, calling close() is the same as the ABORT primitive. If the
1372 * value is set to a negative value, the setsockopt() call will return
1373 * an error. If the value is set to a positive value linger_time, the
1374 * close() can be blocked for at most linger_time ms. If the graceful
1375 * shutdown phase does not finish during this period, close() will
1376 * return but the graceful shutdown phase continues in the system.
1378 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1380 struct sctp_endpoint *ep;
1381 struct sctp_association *asoc;
1382 struct list_head *pos, *temp;
1384 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1386 sctp_lock_sock(sk);
1387 sk->sk_shutdown = SHUTDOWN_MASK;
1388 sk->sk_state = SCTP_SS_CLOSING;
1390 ep = sctp_sk(sk)->ep;
1392 /* Walk all associations on an endpoint. */
1393 list_for_each_safe(pos, temp, &ep->asocs) {
1394 asoc = list_entry(pos, struct sctp_association, asocs);
1396 if (sctp_style(sk, TCP)) {
1397 /* A closed association can still be in the list if
1398 * it belongs to a TCP-style listening socket that is
1399 * not yet accepted. If so, free it. If not, send an
1400 * ABORT or SHUTDOWN based on the linger options.
1402 if (sctp_state(asoc, CLOSED)) {
1403 sctp_unhash_established(asoc);
1404 sctp_association_free(asoc);
1405 continue;
1409 if (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime) {
1410 struct sctp_chunk *chunk;
1412 chunk = sctp_make_abort_user(asoc, NULL, 0);
1413 if (chunk)
1414 sctp_primitive_ABORT(asoc, chunk);
1415 } else
1416 sctp_primitive_SHUTDOWN(asoc, NULL);
1419 /* Clean up any skbs sitting on the receive queue. */
1420 sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1421 sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1423 /* On a TCP-style socket, block for at most linger_time if set. */
1424 if (sctp_style(sk, TCP) && timeout)
1425 sctp_wait_for_close(sk, timeout);
1427 /* This will run the backlog queue. */
1428 sctp_release_sock(sk);
1430 /* Supposedly, no process has access to the socket, but
1431 * the net layers still may.
1433 sctp_local_bh_disable();
1434 sctp_bh_lock_sock(sk);
1436 /* Hold the sock, since sk_common_release() will put sock_put()
1437 * and we have just a little more cleanup.
1439 sock_hold(sk);
1440 sk_common_release(sk);
1442 sctp_bh_unlock_sock(sk);
1443 sctp_local_bh_enable();
1445 sock_put(sk);
1447 SCTP_DBG_OBJCNT_DEC(sock);
1450 /* Handle EPIPE error. */
1451 static int sctp_error(struct sock *sk, int flags, int err)
1453 if (err == -EPIPE)
1454 err = sock_error(sk) ? : -EPIPE;
1455 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1456 send_sig(SIGPIPE, current, 0);
1457 return err;
1460 /* API 3.1.3 sendmsg() - UDP Style Syntax
1462 * An application uses sendmsg() and recvmsg() calls to transmit data to
1463 * and receive data from its peer.
1465 * ssize_t sendmsg(int socket, const struct msghdr *message,
1466 * int flags);
1468 * socket - the socket descriptor of the endpoint.
1469 * message - pointer to the msghdr structure which contains a single
1470 * user message and possibly some ancillary data.
1472 * See Section 5 for complete description of the data
1473 * structures.
1475 * flags - flags sent or received with the user message, see Section
1476 * 5 for complete description of the flags.
1478 * Note: This function could use a rewrite especially when explicit
1479 * connect support comes in.
1481 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1483 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1485 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1486 struct msghdr *msg, size_t msg_len)
1488 struct sctp_sock *sp;
1489 struct sctp_endpoint *ep;
1490 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1491 struct sctp_transport *transport, *chunk_tp;
1492 struct sctp_chunk *chunk;
1493 union sctp_addr to;
1494 struct sockaddr *msg_name = NULL;
1495 struct sctp_sndrcvinfo default_sinfo = { 0 };
1496 struct sctp_sndrcvinfo *sinfo;
1497 struct sctp_initmsg *sinit;
1498 sctp_assoc_t associd = 0;
1499 sctp_cmsgs_t cmsgs = { NULL };
1500 int err;
1501 sctp_scope_t scope;
1502 long timeo;
1503 __u16 sinfo_flags = 0;
1504 struct sctp_datamsg *datamsg;
1505 int msg_flags = msg->msg_flags;
1507 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1508 sk, msg, msg_len);
1510 err = 0;
1511 sp = sctp_sk(sk);
1512 ep = sp->ep;
1514 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1516 /* We cannot send a message over a TCP-style listening socket. */
1517 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1518 err = -EPIPE;
1519 goto out_nounlock;
1522 /* Parse out the SCTP CMSGs. */
1523 err = sctp_msghdr_parse(msg, &cmsgs);
1525 if (err) {
1526 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1527 goto out_nounlock;
1530 /* Fetch the destination address for this packet. This
1531 * address only selects the association--it is not necessarily
1532 * the address we will send to.
1533 * For a peeled-off socket, msg_name is ignored.
1535 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1536 int msg_namelen = msg->msg_namelen;
1538 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1539 msg_namelen);
1540 if (err)
1541 return err;
1543 if (msg_namelen > sizeof(to))
1544 msg_namelen = sizeof(to);
1545 memcpy(&to, msg->msg_name, msg_namelen);
1546 msg_name = msg->msg_name;
1549 sinfo = cmsgs.info;
1550 sinit = cmsgs.init;
1552 /* Did the user specify SNDRCVINFO? */
1553 if (sinfo) {
1554 sinfo_flags = sinfo->sinfo_flags;
1555 associd = sinfo->sinfo_assoc_id;
1558 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1559 msg_len, sinfo_flags);
1561 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1562 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1563 err = -EINVAL;
1564 goto out_nounlock;
1567 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1568 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1569 * If SCTP_ABORT is set, the message length could be non zero with
1570 * the msg_iov set to the user abort reason.
1572 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1573 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1574 err = -EINVAL;
1575 goto out_nounlock;
1578 /* If SCTP_ADDR_OVER is set, there must be an address
1579 * specified in msg_name.
1581 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1582 err = -EINVAL;
1583 goto out_nounlock;
1586 transport = NULL;
1588 SCTP_DEBUG_PRINTK("About to look up association.\n");
1590 sctp_lock_sock(sk);
1592 /* If a msg_name has been specified, assume this is to be used. */
1593 if (msg_name) {
1594 /* Look for a matching association on the endpoint. */
1595 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1596 if (!asoc) {
1597 /* If we could not find a matching association on the
1598 * endpoint, make sure that it is not a TCP-style
1599 * socket that already has an association or there is
1600 * no peeled-off association on another socket.
1602 if ((sctp_style(sk, TCP) &&
1603 sctp_sstate(sk, ESTABLISHED)) ||
1604 sctp_endpoint_is_peeled_off(ep, &to)) {
1605 err = -EADDRNOTAVAIL;
1606 goto out_unlock;
1609 } else {
1610 asoc = sctp_id2assoc(sk, associd);
1611 if (!asoc) {
1612 err = -EPIPE;
1613 goto out_unlock;
1617 if (asoc) {
1618 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1620 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1621 * socket that has an association in CLOSED state. This can
1622 * happen when an accepted socket has an association that is
1623 * already CLOSED.
1625 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1626 err = -EPIPE;
1627 goto out_unlock;
1630 if (sinfo_flags & SCTP_EOF) {
1631 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1632 asoc);
1633 sctp_primitive_SHUTDOWN(asoc, NULL);
1634 err = 0;
1635 goto out_unlock;
1637 if (sinfo_flags & SCTP_ABORT) {
1639 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1640 if (!chunk) {
1641 err = -ENOMEM;
1642 goto out_unlock;
1645 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1646 sctp_primitive_ABORT(asoc, chunk);
1647 err = 0;
1648 goto out_unlock;
1652 /* Do we need to create the association? */
1653 if (!asoc) {
1654 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1656 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1657 err = -EINVAL;
1658 goto out_unlock;
1661 /* Check for invalid stream against the stream counts,
1662 * either the default or the user specified stream counts.
1664 if (sinfo) {
1665 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1666 /* Check against the defaults. */
1667 if (sinfo->sinfo_stream >=
1668 sp->initmsg.sinit_num_ostreams) {
1669 err = -EINVAL;
1670 goto out_unlock;
1672 } else {
1673 /* Check against the requested. */
1674 if (sinfo->sinfo_stream >=
1675 sinit->sinit_num_ostreams) {
1676 err = -EINVAL;
1677 goto out_unlock;
1683 * API 3.1.2 bind() - UDP Style Syntax
1684 * If a bind() or sctp_bindx() is not called prior to a
1685 * sendmsg() call that initiates a new association, the
1686 * system picks an ephemeral port and will choose an address
1687 * set equivalent to binding with a wildcard address.
1689 if (!ep->base.bind_addr.port) {
1690 if (sctp_autobind(sk)) {
1691 err = -EAGAIN;
1692 goto out_unlock;
1694 } else {
1696 * If an unprivileged user inherits a one-to-many
1697 * style socket with open associations on a privileged
1698 * port, it MAY be permitted to accept new associations,
1699 * but it SHOULD NOT be permitted to open new
1700 * associations.
1702 if (ep->base.bind_addr.port < PROT_SOCK &&
1703 !capable(CAP_NET_BIND_SERVICE)) {
1704 err = -EACCES;
1705 goto out_unlock;
1709 scope = sctp_scope(&to);
1710 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1711 if (!new_asoc) {
1712 err = -ENOMEM;
1713 goto out_unlock;
1715 asoc = new_asoc;
1716 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1717 if (err < 0) {
1718 err = -ENOMEM;
1719 goto out_free;
1722 /* If the SCTP_INIT ancillary data is specified, set all
1723 * the association init values accordingly.
1725 if (sinit) {
1726 if (sinit->sinit_num_ostreams) {
1727 asoc->c.sinit_num_ostreams =
1728 sinit->sinit_num_ostreams;
1730 if (sinit->sinit_max_instreams) {
1731 asoc->c.sinit_max_instreams =
1732 sinit->sinit_max_instreams;
1734 if (sinit->sinit_max_attempts) {
1735 asoc->max_init_attempts
1736 = sinit->sinit_max_attempts;
1738 if (sinit->sinit_max_init_timeo) {
1739 asoc->max_init_timeo =
1740 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1744 /* Prime the peer's transport structures. */
1745 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1746 if (!transport) {
1747 err = -ENOMEM;
1748 goto out_free;
1752 /* ASSERT: we have a valid association at this point. */
1753 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1755 if (!sinfo) {
1756 /* If the user didn't specify SNDRCVINFO, make up one with
1757 * some defaults.
1759 default_sinfo.sinfo_stream = asoc->default_stream;
1760 default_sinfo.sinfo_flags = asoc->default_flags;
1761 default_sinfo.sinfo_ppid = asoc->default_ppid;
1762 default_sinfo.sinfo_context = asoc->default_context;
1763 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1764 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1765 sinfo = &default_sinfo;
1768 /* API 7.1.7, the sndbuf size per association bounds the
1769 * maximum size of data that can be sent in a single send call.
1771 if (msg_len > sk->sk_sndbuf) {
1772 err = -EMSGSIZE;
1773 goto out_free;
1776 if (asoc->pmtu_pending)
1777 sctp_assoc_pending_pmtu(asoc);
1779 /* If fragmentation is disabled and the message length exceeds the
1780 * association fragmentation point, return EMSGSIZE. The I-D
1781 * does not specify what this error is, but this looks like
1782 * a great fit.
1784 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1785 err = -EMSGSIZE;
1786 goto out_free;
1789 if (sinfo) {
1790 /* Check for invalid stream. */
1791 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1792 err = -EINVAL;
1793 goto out_free;
1797 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1798 if (!sctp_wspace(asoc)) {
1799 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1800 if (err)
1801 goto out_free;
1804 /* If an address is passed with the sendto/sendmsg call, it is used
1805 * to override the primary destination address in the TCP model, or
1806 * when SCTP_ADDR_OVER flag is set in the UDP model.
1808 if ((sctp_style(sk, TCP) && msg_name) ||
1809 (sinfo_flags & SCTP_ADDR_OVER)) {
1810 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1811 if (!chunk_tp) {
1812 err = -EINVAL;
1813 goto out_free;
1815 } else
1816 chunk_tp = NULL;
1818 /* Auto-connect, if we aren't connected already. */
1819 if (sctp_state(asoc, CLOSED)) {
1820 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1821 if (err < 0)
1822 goto out_free;
1823 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1826 /* Break the message into multiple chunks of maximum size. */
1827 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1828 if (!datamsg) {
1829 err = -ENOMEM;
1830 goto out_free;
1833 /* Now send the (possibly) fragmented message. */
1834 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1835 sctp_chunk_hold(chunk);
1837 /* Do accounting for the write space. */
1838 sctp_set_owner_w(chunk);
1840 chunk->transport = chunk_tp;
1843 /* Send it to the lower layers. Note: all chunks
1844 * must either fail or succeed. The lower layer
1845 * works that way today. Keep it that way or this
1846 * breaks.
1848 err = sctp_primitive_SEND(asoc, datamsg);
1849 /* Did the lower layer accept the chunk? */
1850 if (err)
1851 sctp_datamsg_free(datamsg);
1852 else
1853 sctp_datamsg_put(datamsg);
1855 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1857 if (err)
1858 goto out_free;
1859 else
1860 err = msg_len;
1862 /* If we are already past ASSOCIATE, the lower
1863 * layers are responsible for association cleanup.
1865 goto out_unlock;
1867 out_free:
1868 if (new_asoc)
1869 sctp_association_free(asoc);
1870 out_unlock:
1871 sctp_release_sock(sk);
1873 out_nounlock:
1874 return sctp_error(sk, msg_flags, err);
1876 #if 0
1877 do_sock_err:
1878 if (msg_len)
1879 err = msg_len;
1880 else
1881 err = sock_error(sk);
1882 goto out;
1884 do_interrupted:
1885 if (msg_len)
1886 err = msg_len;
1887 goto out;
1888 #endif /* 0 */
1891 /* This is an extended version of skb_pull() that removes the data from the
1892 * start of a skb even when data is spread across the list of skb's in the
1893 * frag_list. len specifies the total amount of data that needs to be removed.
1894 * when 'len' bytes could be removed from the skb, it returns 0.
1895 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1896 * could not be removed.
1898 static int sctp_skb_pull(struct sk_buff *skb, int len)
1900 struct sk_buff *list;
1901 int skb_len = skb_headlen(skb);
1902 int rlen;
1904 if (len <= skb_len) {
1905 __skb_pull(skb, len);
1906 return 0;
1908 len -= skb_len;
1909 __skb_pull(skb, skb_len);
1911 skb_walk_frags(skb, list) {
1912 rlen = sctp_skb_pull(list, len);
1913 skb->len -= (len-rlen);
1914 skb->data_len -= (len-rlen);
1916 if (!rlen)
1917 return 0;
1919 len = rlen;
1922 return len;
1925 /* API 3.1.3 recvmsg() - UDP Style Syntax
1927 * ssize_t recvmsg(int socket, struct msghdr *message,
1928 * int flags);
1930 * socket - the socket descriptor of the endpoint.
1931 * message - pointer to the msghdr structure which contains a single
1932 * user message and possibly some ancillary data.
1934 * See Section 5 for complete description of the data
1935 * structures.
1937 * flags - flags sent or received with the user message, see Section
1938 * 5 for complete description of the flags.
1940 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
1942 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
1943 struct msghdr *msg, size_t len, int noblock,
1944 int flags, int *addr_len)
1946 struct sctp_ulpevent *event = NULL;
1947 struct sctp_sock *sp = sctp_sk(sk);
1948 struct sk_buff *skb;
1949 int copied;
1950 int err = 0;
1951 int skb_len;
1953 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
1954 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
1955 "len", len, "knoblauch", noblock,
1956 "flags", flags, "addr_len", addr_len);
1958 sctp_lock_sock(sk);
1960 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
1961 err = -ENOTCONN;
1962 goto out;
1965 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
1966 if (!skb)
1967 goto out;
1969 /* Get the total length of the skb including any skb's in the
1970 * frag_list.
1972 skb_len = skb->len;
1974 copied = skb_len;
1975 if (copied > len)
1976 copied = len;
1978 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
1980 event = sctp_skb2event(skb);
1982 if (err)
1983 goto out_free;
1985 sock_recv_ts_and_drops(msg, sk, skb);
1986 if (sctp_ulpevent_is_notification(event)) {
1987 msg->msg_flags |= MSG_NOTIFICATION;
1988 sp->pf->event_msgname(event, msg->msg_name, addr_len);
1989 } else {
1990 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
1993 /* Check if we allow SCTP_SNDRCVINFO. */
1994 if (sp->subscribe.sctp_data_io_event)
1995 sctp_ulpevent_read_sndrcvinfo(event, msg);
1996 #if 0
1997 /* FIXME: we should be calling IP/IPv6 layers. */
1998 if (sk->sk_protinfo.af_inet.cmsg_flags)
1999 ip_cmsg_recv(msg, skb);
2000 #endif
2002 err = copied;
2004 /* If skb's length exceeds the user's buffer, update the skb and
2005 * push it back to the receive_queue so that the next call to
2006 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2008 if (skb_len > copied) {
2009 msg->msg_flags &= ~MSG_EOR;
2010 if (flags & MSG_PEEK)
2011 goto out_free;
2012 sctp_skb_pull(skb, copied);
2013 skb_queue_head(&sk->sk_receive_queue, skb);
2015 /* When only partial message is copied to the user, increase
2016 * rwnd by that amount. If all the data in the skb is read,
2017 * rwnd is updated when the event is freed.
2019 if (!sctp_ulpevent_is_notification(event))
2020 sctp_assoc_rwnd_increase(event->asoc, copied);
2021 goto out;
2022 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2023 (event->msg_flags & MSG_EOR))
2024 msg->msg_flags |= MSG_EOR;
2025 else
2026 msg->msg_flags &= ~MSG_EOR;
2028 out_free:
2029 if (flags & MSG_PEEK) {
2030 /* Release the skb reference acquired after peeking the skb in
2031 * sctp_skb_recv_datagram().
2033 kfree_skb(skb);
2034 } else {
2035 /* Free the event which includes releasing the reference to
2036 * the owner of the skb, freeing the skb and updating the
2037 * rwnd.
2039 sctp_ulpevent_free(event);
2041 out:
2042 sctp_release_sock(sk);
2043 return err;
2046 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2048 * This option is a on/off flag. If enabled no SCTP message
2049 * fragmentation will be performed. Instead if a message being sent
2050 * exceeds the current PMTU size, the message will NOT be sent and
2051 * instead a error will be indicated to the user.
2053 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2054 char __user *optval,
2055 unsigned int optlen)
2057 int val;
2059 if (optlen < sizeof(int))
2060 return -EINVAL;
2062 if (get_user(val, (int __user *)optval))
2063 return -EFAULT;
2065 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2067 return 0;
2070 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2071 unsigned int optlen)
2073 if (optlen > sizeof(struct sctp_event_subscribe))
2074 return -EINVAL;
2075 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2076 return -EFAULT;
2077 return 0;
2080 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2082 * This socket option is applicable to the UDP-style socket only. When
2083 * set it will cause associations that are idle for more than the
2084 * specified number of seconds to automatically close. An association
2085 * being idle is defined an association that has NOT sent or received
2086 * user data. The special value of '0' indicates that no automatic
2087 * close of any associations should be performed. The option expects an
2088 * integer defining the number of seconds of idle time before an
2089 * association is closed.
2091 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2092 unsigned int optlen)
2094 struct sctp_sock *sp = sctp_sk(sk);
2096 /* Applicable to UDP-style socket only */
2097 if (sctp_style(sk, TCP))
2098 return -EOPNOTSUPP;
2099 if (optlen != sizeof(int))
2100 return -EINVAL;
2101 if (copy_from_user(&sp->autoclose, optval, optlen))
2102 return -EFAULT;
2103 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2104 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2106 return 0;
2109 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2111 * Applications can enable or disable heartbeats for any peer address of
2112 * an association, modify an address's heartbeat interval, force a
2113 * heartbeat to be sent immediately, and adjust the address's maximum
2114 * number of retransmissions sent before an address is considered
2115 * unreachable. The following structure is used to access and modify an
2116 * address's parameters:
2118 * struct sctp_paddrparams {
2119 * sctp_assoc_t spp_assoc_id;
2120 * struct sockaddr_storage spp_address;
2121 * uint32_t spp_hbinterval;
2122 * uint16_t spp_pathmaxrxt;
2123 * uint32_t spp_pathmtu;
2124 * uint32_t spp_sackdelay;
2125 * uint32_t spp_flags;
2126 * };
2128 * spp_assoc_id - (one-to-many style socket) This is filled in the
2129 * application, and identifies the association for
2130 * this query.
2131 * spp_address - This specifies which address is of interest.
2132 * spp_hbinterval - This contains the value of the heartbeat interval,
2133 * in milliseconds. If a value of zero
2134 * is present in this field then no changes are to
2135 * be made to this parameter.
2136 * spp_pathmaxrxt - This contains the maximum number of
2137 * retransmissions before this address shall be
2138 * considered unreachable. If a value of zero
2139 * is present in this field then no changes are to
2140 * be made to this parameter.
2141 * spp_pathmtu - When Path MTU discovery is disabled the value
2142 * specified here will be the "fixed" path mtu.
2143 * Note that if the spp_address field is empty
2144 * then all associations on this address will
2145 * have this fixed path mtu set upon them.
2147 * spp_sackdelay - When delayed sack is enabled, this value specifies
2148 * the number of milliseconds that sacks will be delayed
2149 * for. This value will apply to all addresses of an
2150 * association if the spp_address field is empty. Note
2151 * also, that if delayed sack is enabled and this
2152 * value is set to 0, no change is made to the last
2153 * recorded delayed sack timer value.
2155 * spp_flags - These flags are used to control various features
2156 * on an association. The flag field may contain
2157 * zero or more of the following options.
2159 * SPP_HB_ENABLE - Enable heartbeats on the
2160 * specified address. Note that if the address
2161 * field is empty all addresses for the association
2162 * have heartbeats enabled upon them.
2164 * SPP_HB_DISABLE - Disable heartbeats on the
2165 * speicifed address. Note that if the address
2166 * field is empty all addresses for the association
2167 * will have their heartbeats disabled. Note also
2168 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2169 * mutually exclusive, only one of these two should
2170 * be specified. Enabling both fields will have
2171 * undetermined results.
2173 * SPP_HB_DEMAND - Request a user initiated heartbeat
2174 * to be made immediately.
2176 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2177 * heartbeat delayis to be set to the value of 0
2178 * milliseconds.
2180 * SPP_PMTUD_ENABLE - This field will enable PMTU
2181 * discovery upon the specified address. Note that
2182 * if the address feild is empty then all addresses
2183 * on the association are effected.
2185 * SPP_PMTUD_DISABLE - This field will disable PMTU
2186 * discovery upon the specified address. Note that
2187 * if the address feild is empty then all addresses
2188 * on the association are effected. Not also that
2189 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2190 * exclusive. Enabling both will have undetermined
2191 * results.
2193 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2194 * on delayed sack. The time specified in spp_sackdelay
2195 * is used to specify the sack delay for this address. Note
2196 * that if spp_address is empty then all addresses will
2197 * enable delayed sack and take on the sack delay
2198 * value specified in spp_sackdelay.
2199 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2200 * off delayed sack. If the spp_address field is blank then
2201 * delayed sack is disabled for the entire association. Note
2202 * also that this field is mutually exclusive to
2203 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2204 * results.
2206 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2207 struct sctp_transport *trans,
2208 struct sctp_association *asoc,
2209 struct sctp_sock *sp,
2210 int hb_change,
2211 int pmtud_change,
2212 int sackdelay_change)
2214 int error;
2216 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2217 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2218 if (error)
2219 return error;
2222 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2223 * this field is ignored. Note also that a value of zero indicates
2224 * the current setting should be left unchanged.
2226 if (params->spp_flags & SPP_HB_ENABLE) {
2228 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2229 * set. This lets us use 0 value when this flag
2230 * is set.
2232 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2233 params->spp_hbinterval = 0;
2235 if (params->spp_hbinterval ||
2236 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2237 if (trans) {
2238 trans->hbinterval =
2239 msecs_to_jiffies(params->spp_hbinterval);
2240 } else if (asoc) {
2241 asoc->hbinterval =
2242 msecs_to_jiffies(params->spp_hbinterval);
2243 } else {
2244 sp->hbinterval = params->spp_hbinterval;
2249 if (hb_change) {
2250 if (trans) {
2251 trans->param_flags =
2252 (trans->param_flags & ~SPP_HB) | hb_change;
2253 } else if (asoc) {
2254 asoc->param_flags =
2255 (asoc->param_flags & ~SPP_HB) | hb_change;
2256 } else {
2257 sp->param_flags =
2258 (sp->param_flags & ~SPP_HB) | hb_change;
2262 /* When Path MTU discovery is disabled the value specified here will
2263 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2264 * include the flag SPP_PMTUD_DISABLE for this field to have any
2265 * effect).
2267 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2268 if (trans) {
2269 trans->pathmtu = params->spp_pathmtu;
2270 sctp_assoc_sync_pmtu(asoc);
2271 } else if (asoc) {
2272 asoc->pathmtu = params->spp_pathmtu;
2273 sctp_frag_point(asoc, params->spp_pathmtu);
2274 } else {
2275 sp->pathmtu = params->spp_pathmtu;
2279 if (pmtud_change) {
2280 if (trans) {
2281 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2282 (params->spp_flags & SPP_PMTUD_ENABLE);
2283 trans->param_flags =
2284 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2285 if (update) {
2286 sctp_transport_pmtu(trans);
2287 sctp_assoc_sync_pmtu(asoc);
2289 } else if (asoc) {
2290 asoc->param_flags =
2291 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2292 } else {
2293 sp->param_flags =
2294 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2298 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2299 * value of this field is ignored. Note also that a value of zero
2300 * indicates the current setting should be left unchanged.
2302 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2303 if (trans) {
2304 trans->sackdelay =
2305 msecs_to_jiffies(params->spp_sackdelay);
2306 } else if (asoc) {
2307 asoc->sackdelay =
2308 msecs_to_jiffies(params->spp_sackdelay);
2309 } else {
2310 sp->sackdelay = params->spp_sackdelay;
2314 if (sackdelay_change) {
2315 if (trans) {
2316 trans->param_flags =
2317 (trans->param_flags & ~SPP_SACKDELAY) |
2318 sackdelay_change;
2319 } else if (asoc) {
2320 asoc->param_flags =
2321 (asoc->param_flags & ~SPP_SACKDELAY) |
2322 sackdelay_change;
2323 } else {
2324 sp->param_flags =
2325 (sp->param_flags & ~SPP_SACKDELAY) |
2326 sackdelay_change;
2330 /* Note that a value of zero indicates the current setting should be
2331 left unchanged.
2333 if (params->spp_pathmaxrxt) {
2334 if (trans) {
2335 trans->pathmaxrxt = params->spp_pathmaxrxt;
2336 } else if (asoc) {
2337 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2338 } else {
2339 sp->pathmaxrxt = params->spp_pathmaxrxt;
2343 return 0;
2346 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2347 char __user *optval,
2348 unsigned int optlen)
2350 struct sctp_paddrparams params;
2351 struct sctp_transport *trans = NULL;
2352 struct sctp_association *asoc = NULL;
2353 struct sctp_sock *sp = sctp_sk(sk);
2354 int error;
2355 int hb_change, pmtud_change, sackdelay_change;
2357 if (optlen != sizeof(struct sctp_paddrparams))
2358 return - EINVAL;
2360 if (copy_from_user(&params, optval, optlen))
2361 return -EFAULT;
2363 /* Validate flags and value parameters. */
2364 hb_change = params.spp_flags & SPP_HB;
2365 pmtud_change = params.spp_flags & SPP_PMTUD;
2366 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2368 if (hb_change == SPP_HB ||
2369 pmtud_change == SPP_PMTUD ||
2370 sackdelay_change == SPP_SACKDELAY ||
2371 params.spp_sackdelay > 500 ||
2372 (params.spp_pathmtu &&
2373 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2374 return -EINVAL;
2376 /* If an address other than INADDR_ANY is specified, and
2377 * no transport is found, then the request is invalid.
2379 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2380 trans = sctp_addr_id2transport(sk, &params.spp_address,
2381 params.spp_assoc_id);
2382 if (!trans)
2383 return -EINVAL;
2386 /* Get association, if assoc_id != 0 and the socket is a one
2387 * to many style socket, and an association was not found, then
2388 * the id was invalid.
2390 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2391 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2392 return -EINVAL;
2394 /* Heartbeat demand can only be sent on a transport or
2395 * association, but not a socket.
2397 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2398 return -EINVAL;
2400 /* Process parameters. */
2401 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2402 hb_change, pmtud_change,
2403 sackdelay_change);
2405 if (error)
2406 return error;
2408 /* If changes are for association, also apply parameters to each
2409 * transport.
2411 if (!trans && asoc) {
2412 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2413 transports) {
2414 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2415 hb_change, pmtud_change,
2416 sackdelay_change);
2420 return 0;
2424 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2426 * This option will effect the way delayed acks are performed. This
2427 * option allows you to get or set the delayed ack time, in
2428 * milliseconds. It also allows changing the delayed ack frequency.
2429 * Changing the frequency to 1 disables the delayed sack algorithm. If
2430 * the assoc_id is 0, then this sets or gets the endpoints default
2431 * values. If the assoc_id field is non-zero, then the set or get
2432 * effects the specified association for the one to many model (the
2433 * assoc_id field is ignored by the one to one model). Note that if
2434 * sack_delay or sack_freq are 0 when setting this option, then the
2435 * current values will remain unchanged.
2437 * struct sctp_sack_info {
2438 * sctp_assoc_t sack_assoc_id;
2439 * uint32_t sack_delay;
2440 * uint32_t sack_freq;
2441 * };
2443 * sack_assoc_id - This parameter, indicates which association the user
2444 * is performing an action upon. Note that if this field's value is
2445 * zero then the endpoints default value is changed (effecting future
2446 * associations only).
2448 * sack_delay - This parameter contains the number of milliseconds that
2449 * the user is requesting the delayed ACK timer be set to. Note that
2450 * this value is defined in the standard to be between 200 and 500
2451 * milliseconds.
2453 * sack_freq - This parameter contains the number of packets that must
2454 * be received before a sack is sent without waiting for the delay
2455 * timer to expire. The default value for this is 2, setting this
2456 * value to 1 will disable the delayed sack algorithm.
2459 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2460 char __user *optval, unsigned int optlen)
2462 struct sctp_sack_info params;
2463 struct sctp_transport *trans = NULL;
2464 struct sctp_association *asoc = NULL;
2465 struct sctp_sock *sp = sctp_sk(sk);
2467 if (optlen == sizeof(struct sctp_sack_info)) {
2468 if (copy_from_user(&params, optval, optlen))
2469 return -EFAULT;
2471 if (params.sack_delay == 0 && params.sack_freq == 0)
2472 return 0;
2473 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2474 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2475 pr_warn("Use struct sctp_sack_info instead\n");
2476 if (copy_from_user(&params, optval, optlen))
2477 return -EFAULT;
2479 if (params.sack_delay == 0)
2480 params.sack_freq = 1;
2481 else
2482 params.sack_freq = 0;
2483 } else
2484 return - EINVAL;
2486 /* Validate value parameter. */
2487 if (params.sack_delay > 500)
2488 return -EINVAL;
2490 /* Get association, if sack_assoc_id != 0 and the socket is a one
2491 * to many style socket, and an association was not found, then
2492 * the id was invalid.
2494 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2495 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2496 return -EINVAL;
2498 if (params.sack_delay) {
2499 if (asoc) {
2500 asoc->sackdelay =
2501 msecs_to_jiffies(params.sack_delay);
2502 asoc->param_flags =
2503 (asoc->param_flags & ~SPP_SACKDELAY) |
2504 SPP_SACKDELAY_ENABLE;
2505 } else {
2506 sp->sackdelay = params.sack_delay;
2507 sp->param_flags =
2508 (sp->param_flags & ~SPP_SACKDELAY) |
2509 SPP_SACKDELAY_ENABLE;
2513 if (params.sack_freq == 1) {
2514 if (asoc) {
2515 asoc->param_flags =
2516 (asoc->param_flags & ~SPP_SACKDELAY) |
2517 SPP_SACKDELAY_DISABLE;
2518 } else {
2519 sp->param_flags =
2520 (sp->param_flags & ~SPP_SACKDELAY) |
2521 SPP_SACKDELAY_DISABLE;
2523 } else if (params.sack_freq > 1) {
2524 if (asoc) {
2525 asoc->sackfreq = params.sack_freq;
2526 asoc->param_flags =
2527 (asoc->param_flags & ~SPP_SACKDELAY) |
2528 SPP_SACKDELAY_ENABLE;
2529 } else {
2530 sp->sackfreq = params.sack_freq;
2531 sp->param_flags =
2532 (sp->param_flags & ~SPP_SACKDELAY) |
2533 SPP_SACKDELAY_ENABLE;
2537 /* If change is for association, also apply to each transport. */
2538 if (asoc) {
2539 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2540 transports) {
2541 if (params.sack_delay) {
2542 trans->sackdelay =
2543 msecs_to_jiffies(params.sack_delay);
2544 trans->param_flags =
2545 (trans->param_flags & ~SPP_SACKDELAY) |
2546 SPP_SACKDELAY_ENABLE;
2548 if (params.sack_freq == 1) {
2549 trans->param_flags =
2550 (trans->param_flags & ~SPP_SACKDELAY) |
2551 SPP_SACKDELAY_DISABLE;
2552 } else if (params.sack_freq > 1) {
2553 trans->sackfreq = params.sack_freq;
2554 trans->param_flags =
2555 (trans->param_flags & ~SPP_SACKDELAY) |
2556 SPP_SACKDELAY_ENABLE;
2561 return 0;
2564 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2566 * Applications can specify protocol parameters for the default association
2567 * initialization. The option name argument to setsockopt() and getsockopt()
2568 * is SCTP_INITMSG.
2570 * Setting initialization parameters is effective only on an unconnected
2571 * socket (for UDP-style sockets only future associations are effected
2572 * by the change). With TCP-style sockets, this option is inherited by
2573 * sockets derived from a listener socket.
2575 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2577 struct sctp_initmsg sinit;
2578 struct sctp_sock *sp = sctp_sk(sk);
2580 if (optlen != sizeof(struct sctp_initmsg))
2581 return -EINVAL;
2582 if (copy_from_user(&sinit, optval, optlen))
2583 return -EFAULT;
2585 if (sinit.sinit_num_ostreams)
2586 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2587 if (sinit.sinit_max_instreams)
2588 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2589 if (sinit.sinit_max_attempts)
2590 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2591 if (sinit.sinit_max_init_timeo)
2592 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2594 return 0;
2598 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2600 * Applications that wish to use the sendto() system call may wish to
2601 * specify a default set of parameters that would normally be supplied
2602 * through the inclusion of ancillary data. This socket option allows
2603 * such an application to set the default sctp_sndrcvinfo structure.
2604 * The application that wishes to use this socket option simply passes
2605 * in to this call the sctp_sndrcvinfo structure defined in Section
2606 * 5.2.2) The input parameters accepted by this call include
2607 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2608 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2609 * to this call if the caller is using the UDP model.
2611 static int sctp_setsockopt_default_send_param(struct sock *sk,
2612 char __user *optval,
2613 unsigned int optlen)
2615 struct sctp_sndrcvinfo info;
2616 struct sctp_association *asoc;
2617 struct sctp_sock *sp = sctp_sk(sk);
2619 if (optlen != sizeof(struct sctp_sndrcvinfo))
2620 return -EINVAL;
2621 if (copy_from_user(&info, optval, optlen))
2622 return -EFAULT;
2624 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2625 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2626 return -EINVAL;
2628 if (asoc) {
2629 asoc->default_stream = info.sinfo_stream;
2630 asoc->default_flags = info.sinfo_flags;
2631 asoc->default_ppid = info.sinfo_ppid;
2632 asoc->default_context = info.sinfo_context;
2633 asoc->default_timetolive = info.sinfo_timetolive;
2634 } else {
2635 sp->default_stream = info.sinfo_stream;
2636 sp->default_flags = info.sinfo_flags;
2637 sp->default_ppid = info.sinfo_ppid;
2638 sp->default_context = info.sinfo_context;
2639 sp->default_timetolive = info.sinfo_timetolive;
2642 return 0;
2645 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2647 * Requests that the local SCTP stack use the enclosed peer address as
2648 * the association primary. The enclosed address must be one of the
2649 * association peer's addresses.
2651 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2652 unsigned int optlen)
2654 struct sctp_prim prim;
2655 struct sctp_transport *trans;
2657 if (optlen != sizeof(struct sctp_prim))
2658 return -EINVAL;
2660 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2661 return -EFAULT;
2663 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2664 if (!trans)
2665 return -EINVAL;
2667 sctp_assoc_set_primary(trans->asoc, trans);
2669 return 0;
2673 * 7.1.5 SCTP_NODELAY
2675 * Turn on/off any Nagle-like algorithm. This means that packets are
2676 * generally sent as soon as possible and no unnecessary delays are
2677 * introduced, at the cost of more packets in the network. Expects an
2678 * integer boolean flag.
2680 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2681 unsigned int optlen)
2683 int val;
2685 if (optlen < sizeof(int))
2686 return -EINVAL;
2687 if (get_user(val, (int __user *)optval))
2688 return -EFAULT;
2690 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2691 return 0;
2696 * 7.1.1 SCTP_RTOINFO
2698 * The protocol parameters used to initialize and bound retransmission
2699 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2700 * and modify these parameters.
2701 * All parameters are time values, in milliseconds. A value of 0, when
2702 * modifying the parameters, indicates that the current value should not
2703 * be changed.
2706 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2708 struct sctp_rtoinfo rtoinfo;
2709 struct sctp_association *asoc;
2711 if (optlen != sizeof (struct sctp_rtoinfo))
2712 return -EINVAL;
2714 if (copy_from_user(&rtoinfo, optval, optlen))
2715 return -EFAULT;
2717 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2719 /* Set the values to the specific association */
2720 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2721 return -EINVAL;
2723 if (asoc) {
2724 if (rtoinfo.srto_initial != 0)
2725 asoc->rto_initial =
2726 msecs_to_jiffies(rtoinfo.srto_initial);
2727 if (rtoinfo.srto_max != 0)
2728 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2729 if (rtoinfo.srto_min != 0)
2730 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2731 } else {
2732 /* If there is no association or the association-id = 0
2733 * set the values to the endpoint.
2735 struct sctp_sock *sp = sctp_sk(sk);
2737 if (rtoinfo.srto_initial != 0)
2738 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2739 if (rtoinfo.srto_max != 0)
2740 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2741 if (rtoinfo.srto_min != 0)
2742 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2745 return 0;
2750 * 7.1.2 SCTP_ASSOCINFO
2752 * This option is used to tune the maximum retransmission attempts
2753 * of the association.
2754 * Returns an error if the new association retransmission value is
2755 * greater than the sum of the retransmission value of the peer.
2756 * See [SCTP] for more information.
2759 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2762 struct sctp_assocparams assocparams;
2763 struct sctp_association *asoc;
2765 if (optlen != sizeof(struct sctp_assocparams))
2766 return -EINVAL;
2767 if (copy_from_user(&assocparams, optval, optlen))
2768 return -EFAULT;
2770 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2772 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2773 return -EINVAL;
2775 /* Set the values to the specific association */
2776 if (asoc) {
2777 if (assocparams.sasoc_asocmaxrxt != 0) {
2778 __u32 path_sum = 0;
2779 int paths = 0;
2780 struct sctp_transport *peer_addr;
2782 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2783 transports) {
2784 path_sum += peer_addr->pathmaxrxt;
2785 paths++;
2788 /* Only validate asocmaxrxt if we have more than
2789 * one path/transport. We do this because path
2790 * retransmissions are only counted when we have more
2791 * then one path.
2793 if (paths > 1 &&
2794 assocparams.sasoc_asocmaxrxt > path_sum)
2795 return -EINVAL;
2797 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2800 if (assocparams.sasoc_cookie_life != 0) {
2801 asoc->cookie_life.tv_sec =
2802 assocparams.sasoc_cookie_life / 1000;
2803 asoc->cookie_life.tv_usec =
2804 (assocparams.sasoc_cookie_life % 1000)
2805 * 1000;
2807 } else {
2808 /* Set the values to the endpoint */
2809 struct sctp_sock *sp = sctp_sk(sk);
2811 if (assocparams.sasoc_asocmaxrxt != 0)
2812 sp->assocparams.sasoc_asocmaxrxt =
2813 assocparams.sasoc_asocmaxrxt;
2814 if (assocparams.sasoc_cookie_life != 0)
2815 sp->assocparams.sasoc_cookie_life =
2816 assocparams.sasoc_cookie_life;
2818 return 0;
2822 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2824 * This socket option is a boolean flag which turns on or off mapped V4
2825 * addresses. If this option is turned on and the socket is type
2826 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2827 * If this option is turned off, then no mapping will be done of V4
2828 * addresses and a user will receive both PF_INET6 and PF_INET type
2829 * addresses on the socket.
2831 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2833 int val;
2834 struct sctp_sock *sp = sctp_sk(sk);
2836 if (optlen < sizeof(int))
2837 return -EINVAL;
2838 if (get_user(val, (int __user *)optval))
2839 return -EFAULT;
2840 if (val)
2841 sp->v4mapped = 1;
2842 else
2843 sp->v4mapped = 0;
2845 return 0;
2849 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2850 * This option will get or set the maximum size to put in any outgoing
2851 * SCTP DATA chunk. If a message is larger than this size it will be
2852 * fragmented by SCTP into the specified size. Note that the underlying
2853 * SCTP implementation may fragment into smaller sized chunks when the
2854 * PMTU of the underlying association is smaller than the value set by
2855 * the user. The default value for this option is '0' which indicates
2856 * the user is NOT limiting fragmentation and only the PMTU will effect
2857 * SCTP's choice of DATA chunk size. Note also that values set larger
2858 * than the maximum size of an IP datagram will effectively let SCTP
2859 * control fragmentation (i.e. the same as setting this option to 0).
2861 * The following structure is used to access and modify this parameter:
2863 * struct sctp_assoc_value {
2864 * sctp_assoc_t assoc_id;
2865 * uint32_t assoc_value;
2866 * };
2868 * assoc_id: This parameter is ignored for one-to-one style sockets.
2869 * For one-to-many style sockets this parameter indicates which
2870 * association the user is performing an action upon. Note that if
2871 * this field's value is zero then the endpoints default value is
2872 * changed (effecting future associations only).
2873 * assoc_value: This parameter specifies the maximum size in bytes.
2875 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2877 struct sctp_assoc_value params;
2878 struct sctp_association *asoc;
2879 struct sctp_sock *sp = sctp_sk(sk);
2880 int val;
2882 if (optlen == sizeof(int)) {
2883 pr_warn("Use of int in maxseg socket option deprecated\n");
2884 pr_warn("Use struct sctp_assoc_value instead\n");
2885 if (copy_from_user(&val, optval, optlen))
2886 return -EFAULT;
2887 params.assoc_id = 0;
2888 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2889 if (copy_from_user(&params, optval, optlen))
2890 return -EFAULT;
2891 val = params.assoc_value;
2892 } else
2893 return -EINVAL;
2895 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2896 return -EINVAL;
2898 asoc = sctp_id2assoc(sk, params.assoc_id);
2899 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2900 return -EINVAL;
2902 if (asoc) {
2903 if (val == 0) {
2904 val = asoc->pathmtu;
2905 val -= sp->pf->af->net_header_len;
2906 val -= sizeof(struct sctphdr) +
2907 sizeof(struct sctp_data_chunk);
2909 asoc->user_frag = val;
2910 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
2911 } else {
2912 sp->user_frag = val;
2915 return 0;
2920 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
2922 * Requests that the peer mark the enclosed address as the association
2923 * primary. The enclosed address must be one of the association's
2924 * locally bound addresses. The following structure is used to make a
2925 * set primary request:
2927 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
2928 unsigned int optlen)
2930 struct sctp_sock *sp;
2931 struct sctp_endpoint *ep;
2932 struct sctp_association *asoc = NULL;
2933 struct sctp_setpeerprim prim;
2934 struct sctp_chunk *chunk;
2935 struct sctp_af *af;
2936 int err;
2938 sp = sctp_sk(sk);
2939 ep = sp->ep;
2941 if (!sctp_addip_enable)
2942 return -EPERM;
2944 if (optlen != sizeof(struct sctp_setpeerprim))
2945 return -EINVAL;
2947 if (copy_from_user(&prim, optval, optlen))
2948 return -EFAULT;
2950 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2951 if (!asoc)
2952 return -EINVAL;
2954 if (!asoc->peer.asconf_capable)
2955 return -EPERM;
2957 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2958 return -EPERM;
2960 if (!sctp_state(asoc, ESTABLISHED))
2961 return -ENOTCONN;
2963 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
2964 if (!af)
2965 return -EINVAL;
2967 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
2968 return -EADDRNOTAVAIL;
2970 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2971 return -EADDRNOTAVAIL;
2973 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2974 chunk = sctp_make_asconf_set_prim(asoc,
2975 (union sctp_addr *)&prim.sspp_addr);
2976 if (!chunk)
2977 return -ENOMEM;
2979 err = sctp_send_asconf(asoc, chunk);
2981 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2983 return err;
2986 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2987 unsigned int optlen)
2989 struct sctp_setadaptation adaptation;
2991 if (optlen != sizeof(struct sctp_setadaptation))
2992 return -EINVAL;
2993 if (copy_from_user(&adaptation, optval, optlen))
2994 return -EFAULT;
2996 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2998 return 0;
3002 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3004 * The context field in the sctp_sndrcvinfo structure is normally only
3005 * used when a failed message is retrieved holding the value that was
3006 * sent down on the actual send call. This option allows the setting of
3007 * a default context on an association basis that will be received on
3008 * reading messages from the peer. This is especially helpful in the
3009 * one-2-many model for an application to keep some reference to an
3010 * internal state machine that is processing messages on the
3011 * association. Note that the setting of this value only effects
3012 * received messages from the peer and does not effect the value that is
3013 * saved with outbound messages.
3015 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3016 unsigned int optlen)
3018 struct sctp_assoc_value params;
3019 struct sctp_sock *sp;
3020 struct sctp_association *asoc;
3022 if (optlen != sizeof(struct sctp_assoc_value))
3023 return -EINVAL;
3024 if (copy_from_user(&params, optval, optlen))
3025 return -EFAULT;
3027 sp = sctp_sk(sk);
3029 if (params.assoc_id != 0) {
3030 asoc = sctp_id2assoc(sk, params.assoc_id);
3031 if (!asoc)
3032 return -EINVAL;
3033 asoc->default_rcv_context = params.assoc_value;
3034 } else {
3035 sp->default_rcv_context = params.assoc_value;
3038 return 0;
3042 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3044 * This options will at a minimum specify if the implementation is doing
3045 * fragmented interleave. Fragmented interleave, for a one to many
3046 * socket, is when subsequent calls to receive a message may return
3047 * parts of messages from different associations. Some implementations
3048 * may allow you to turn this value on or off. If so, when turned off,
3049 * no fragment interleave will occur (which will cause a head of line
3050 * blocking amongst multiple associations sharing the same one to many
3051 * socket). When this option is turned on, then each receive call may
3052 * come from a different association (thus the user must receive data
3053 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3054 * association each receive belongs to.
3056 * This option takes a boolean value. A non-zero value indicates that
3057 * fragmented interleave is on. A value of zero indicates that
3058 * fragmented interleave is off.
3060 * Note that it is important that an implementation that allows this
3061 * option to be turned on, have it off by default. Otherwise an unaware
3062 * application using the one to many model may become confused and act
3063 * incorrectly.
3065 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3066 char __user *optval,
3067 unsigned int optlen)
3069 int val;
3071 if (optlen != sizeof(int))
3072 return -EINVAL;
3073 if (get_user(val, (int __user *)optval))
3074 return -EFAULT;
3076 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3078 return 0;
3082 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3083 * (SCTP_PARTIAL_DELIVERY_POINT)
3085 * This option will set or get the SCTP partial delivery point. This
3086 * point is the size of a message where the partial delivery API will be
3087 * invoked to help free up rwnd space for the peer. Setting this to a
3088 * lower value will cause partial deliveries to happen more often. The
3089 * calls argument is an integer that sets or gets the partial delivery
3090 * point. Note also that the call will fail if the user attempts to set
3091 * this value larger than the socket receive buffer size.
3093 * Note that any single message having a length smaller than or equal to
3094 * the SCTP partial delivery point will be delivered in one single read
3095 * call as long as the user provided buffer is large enough to hold the
3096 * message.
3098 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3099 char __user *optval,
3100 unsigned int optlen)
3102 u32 val;
3104 if (optlen != sizeof(u32))
3105 return -EINVAL;
3106 if (get_user(val, (int __user *)optval))
3107 return -EFAULT;
3109 /* Note: We double the receive buffer from what the user sets
3110 * it to be, also initial rwnd is based on rcvbuf/2.
3112 if (val > (sk->sk_rcvbuf >> 1))
3113 return -EINVAL;
3115 sctp_sk(sk)->pd_point = val;
3117 return 0; /* is this the right error code? */
3121 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3123 * This option will allow a user to change the maximum burst of packets
3124 * that can be emitted by this association. Note that the default value
3125 * is 4, and some implementations may restrict this setting so that it
3126 * can only be lowered.
3128 * NOTE: This text doesn't seem right. Do this on a socket basis with
3129 * future associations inheriting the socket value.
3131 static int sctp_setsockopt_maxburst(struct sock *sk,
3132 char __user *optval,
3133 unsigned int optlen)
3135 struct sctp_assoc_value params;
3136 struct sctp_sock *sp;
3137 struct sctp_association *asoc;
3138 int val;
3139 int assoc_id = 0;
3141 if (optlen == sizeof(int)) {
3142 pr_warn("Use of int in max_burst socket option deprecated\n");
3143 pr_warn("Use struct sctp_assoc_value instead\n");
3144 if (copy_from_user(&val, optval, optlen))
3145 return -EFAULT;
3146 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3147 if (copy_from_user(&params, optval, optlen))
3148 return -EFAULT;
3149 val = params.assoc_value;
3150 assoc_id = params.assoc_id;
3151 } else
3152 return -EINVAL;
3154 sp = sctp_sk(sk);
3156 if (assoc_id != 0) {
3157 asoc = sctp_id2assoc(sk, assoc_id);
3158 if (!asoc)
3159 return -EINVAL;
3160 asoc->max_burst = val;
3161 } else
3162 sp->max_burst = val;
3164 return 0;
3168 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3170 * This set option adds a chunk type that the user is requesting to be
3171 * received only in an authenticated way. Changes to the list of chunks
3172 * will only effect future associations on the socket.
3174 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3175 char __user *optval,
3176 unsigned int optlen)
3178 struct sctp_authchunk val;
3180 if (!sctp_auth_enable)
3181 return -EACCES;
3183 if (optlen != sizeof(struct sctp_authchunk))
3184 return -EINVAL;
3185 if (copy_from_user(&val, optval, optlen))
3186 return -EFAULT;
3188 switch (val.sauth_chunk) {
3189 case SCTP_CID_INIT:
3190 case SCTP_CID_INIT_ACK:
3191 case SCTP_CID_SHUTDOWN_COMPLETE:
3192 case SCTP_CID_AUTH:
3193 return -EINVAL;
3196 /* add this chunk id to the endpoint */
3197 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3201 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3203 * This option gets or sets the list of HMAC algorithms that the local
3204 * endpoint requires the peer to use.
3206 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3207 char __user *optval,
3208 unsigned int optlen)
3210 struct sctp_hmacalgo *hmacs;
3211 u32 idents;
3212 int err;
3214 if (!sctp_auth_enable)
3215 return -EACCES;
3217 if (optlen < sizeof(struct sctp_hmacalgo))
3218 return -EINVAL;
3220 hmacs = kmalloc(optlen, GFP_KERNEL);
3221 if (!hmacs)
3222 return -ENOMEM;
3224 if (copy_from_user(hmacs, optval, optlen)) {
3225 err = -EFAULT;
3226 goto out;
3229 idents = hmacs->shmac_num_idents;
3230 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3231 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3232 err = -EINVAL;
3233 goto out;
3236 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3237 out:
3238 kfree(hmacs);
3239 return err;
3243 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3245 * This option will set a shared secret key which is used to build an
3246 * association shared key.
3248 static int sctp_setsockopt_auth_key(struct sock *sk,
3249 char __user *optval,
3250 unsigned int optlen)
3252 struct sctp_authkey *authkey;
3253 struct sctp_association *asoc;
3254 int ret;
3256 if (!sctp_auth_enable)
3257 return -EACCES;
3259 if (optlen <= sizeof(struct sctp_authkey))
3260 return -EINVAL;
3262 authkey = kmalloc(optlen, GFP_KERNEL);
3263 if (!authkey)
3264 return -ENOMEM;
3266 if (copy_from_user(authkey, optval, optlen)) {
3267 ret = -EFAULT;
3268 goto out;
3271 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3272 ret = -EINVAL;
3273 goto out;
3276 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3277 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3278 ret = -EINVAL;
3279 goto out;
3282 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3283 out:
3284 kfree(authkey);
3285 return ret;
3289 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3291 * This option will get or set the active shared key to be used to build
3292 * the association shared key.
3294 static int sctp_setsockopt_active_key(struct sock *sk,
3295 char __user *optval,
3296 unsigned int optlen)
3298 struct sctp_authkeyid val;
3299 struct sctp_association *asoc;
3301 if (!sctp_auth_enable)
3302 return -EACCES;
3304 if (optlen != sizeof(struct sctp_authkeyid))
3305 return -EINVAL;
3306 if (copy_from_user(&val, optval, optlen))
3307 return -EFAULT;
3309 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3310 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3311 return -EINVAL;
3313 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3314 val.scact_keynumber);
3318 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3320 * This set option will delete a shared secret key from use.
3322 static int sctp_setsockopt_del_key(struct sock *sk,
3323 char __user *optval,
3324 unsigned int optlen)
3326 struct sctp_authkeyid val;
3327 struct sctp_association *asoc;
3329 if (!sctp_auth_enable)
3330 return -EACCES;
3332 if (optlen != sizeof(struct sctp_authkeyid))
3333 return -EINVAL;
3334 if (copy_from_user(&val, optval, optlen))
3335 return -EFAULT;
3337 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3338 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3339 return -EINVAL;
3341 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3342 val.scact_keynumber);
3347 /* API 6.2 setsockopt(), getsockopt()
3349 * Applications use setsockopt() and getsockopt() to set or retrieve
3350 * socket options. Socket options are used to change the default
3351 * behavior of sockets calls. They are described in Section 7.
3353 * The syntax is:
3355 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3356 * int __user *optlen);
3357 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3358 * int optlen);
3360 * sd - the socket descript.
3361 * level - set to IPPROTO_SCTP for all SCTP options.
3362 * optname - the option name.
3363 * optval - the buffer to store the value of the option.
3364 * optlen - the size of the buffer.
3366 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3367 char __user *optval, unsigned int optlen)
3369 int retval = 0;
3371 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3372 sk, optname);
3374 /* I can hardly begin to describe how wrong this is. This is
3375 * so broken as to be worse than useless. The API draft
3376 * REALLY is NOT helpful here... I am not convinced that the
3377 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3378 * are at all well-founded.
3380 if (level != SOL_SCTP) {
3381 struct sctp_af *af = sctp_sk(sk)->pf->af;
3382 retval = af->setsockopt(sk, level, optname, optval, optlen);
3383 goto out_nounlock;
3386 sctp_lock_sock(sk);
3388 switch (optname) {
3389 case SCTP_SOCKOPT_BINDX_ADD:
3390 /* 'optlen' is the size of the addresses buffer. */
3391 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3392 optlen, SCTP_BINDX_ADD_ADDR);
3393 break;
3395 case SCTP_SOCKOPT_BINDX_REM:
3396 /* 'optlen' is the size of the addresses buffer. */
3397 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3398 optlen, SCTP_BINDX_REM_ADDR);
3399 break;
3401 case SCTP_SOCKOPT_CONNECTX_OLD:
3402 /* 'optlen' is the size of the addresses buffer. */
3403 retval = sctp_setsockopt_connectx_old(sk,
3404 (struct sockaddr __user *)optval,
3405 optlen);
3406 break;
3408 case SCTP_SOCKOPT_CONNECTX:
3409 /* 'optlen' is the size of the addresses buffer. */
3410 retval = sctp_setsockopt_connectx(sk,
3411 (struct sockaddr __user *)optval,
3412 optlen);
3413 break;
3415 case SCTP_DISABLE_FRAGMENTS:
3416 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3417 break;
3419 case SCTP_EVENTS:
3420 retval = sctp_setsockopt_events(sk, optval, optlen);
3421 break;
3423 case SCTP_AUTOCLOSE:
3424 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3425 break;
3427 case SCTP_PEER_ADDR_PARAMS:
3428 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3429 break;
3431 case SCTP_DELAYED_ACK:
3432 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3433 break;
3434 case SCTP_PARTIAL_DELIVERY_POINT:
3435 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3436 break;
3438 case SCTP_INITMSG:
3439 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3440 break;
3441 case SCTP_DEFAULT_SEND_PARAM:
3442 retval = sctp_setsockopt_default_send_param(sk, optval,
3443 optlen);
3444 break;
3445 case SCTP_PRIMARY_ADDR:
3446 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3447 break;
3448 case SCTP_SET_PEER_PRIMARY_ADDR:
3449 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3450 break;
3451 case SCTP_NODELAY:
3452 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3453 break;
3454 case SCTP_RTOINFO:
3455 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3456 break;
3457 case SCTP_ASSOCINFO:
3458 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3459 break;
3460 case SCTP_I_WANT_MAPPED_V4_ADDR:
3461 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3462 break;
3463 case SCTP_MAXSEG:
3464 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3465 break;
3466 case SCTP_ADAPTATION_LAYER:
3467 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3468 break;
3469 case SCTP_CONTEXT:
3470 retval = sctp_setsockopt_context(sk, optval, optlen);
3471 break;
3472 case SCTP_FRAGMENT_INTERLEAVE:
3473 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3474 break;
3475 case SCTP_MAX_BURST:
3476 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3477 break;
3478 case SCTP_AUTH_CHUNK:
3479 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3480 break;
3481 case SCTP_HMAC_IDENT:
3482 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3483 break;
3484 case SCTP_AUTH_KEY:
3485 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3486 break;
3487 case SCTP_AUTH_ACTIVE_KEY:
3488 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3489 break;
3490 case SCTP_AUTH_DELETE_KEY:
3491 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3492 break;
3493 default:
3494 retval = -ENOPROTOOPT;
3495 break;
3498 sctp_release_sock(sk);
3500 out_nounlock:
3501 return retval;
3504 /* API 3.1.6 connect() - UDP Style Syntax
3506 * An application may use the connect() call in the UDP model to initiate an
3507 * association without sending data.
3509 * The syntax is:
3511 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3513 * sd: the socket descriptor to have a new association added to.
3515 * nam: the address structure (either struct sockaddr_in or struct
3516 * sockaddr_in6 defined in RFC2553 [7]).
3518 * len: the size of the address.
3520 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3521 int addr_len)
3523 int err = 0;
3524 struct sctp_af *af;
3526 sctp_lock_sock(sk);
3528 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3529 __func__, sk, addr, addr_len);
3531 /* Validate addr_len before calling common connect/connectx routine. */
3532 af = sctp_get_af_specific(addr->sa_family);
3533 if (!af || addr_len < af->sockaddr_len) {
3534 err = -EINVAL;
3535 } else {
3536 /* Pass correct addr len to common routine (so it knows there
3537 * is only one address being passed.
3539 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3542 sctp_release_sock(sk);
3543 return err;
3546 /* FIXME: Write comments. */
3547 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3549 return -EOPNOTSUPP; /* STUB */
3552 /* 4.1.4 accept() - TCP Style Syntax
3554 * Applications use accept() call to remove an established SCTP
3555 * association from the accept queue of the endpoint. A new socket
3556 * descriptor will be returned from accept() to represent the newly
3557 * formed association.
3559 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3561 struct sctp_sock *sp;
3562 struct sctp_endpoint *ep;
3563 struct sock *newsk = NULL;
3564 struct sctp_association *asoc;
3565 long timeo;
3566 int error = 0;
3568 sctp_lock_sock(sk);
3570 sp = sctp_sk(sk);
3571 ep = sp->ep;
3573 if (!sctp_style(sk, TCP)) {
3574 error = -EOPNOTSUPP;
3575 goto out;
3578 if (!sctp_sstate(sk, LISTENING)) {
3579 error = -EINVAL;
3580 goto out;
3583 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3585 error = sctp_wait_for_accept(sk, timeo);
3586 if (error)
3587 goto out;
3589 /* We treat the list of associations on the endpoint as the accept
3590 * queue and pick the first association on the list.
3592 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3594 newsk = sp->pf->create_accept_sk(sk, asoc);
3595 if (!newsk) {
3596 error = -ENOMEM;
3597 goto out;
3600 /* Populate the fields of the newsk from the oldsk and migrate the
3601 * asoc to the newsk.
3603 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3605 out:
3606 sctp_release_sock(sk);
3607 *err = error;
3608 return newsk;
3611 /* The SCTP ioctl handler. */
3612 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3614 int rc = -ENOTCONN;
3616 sctp_lock_sock(sk);
3619 * SEQPACKET-style sockets in LISTENING state are valid, for
3620 * SCTP, so only discard TCP-style sockets in LISTENING state.
3622 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3623 goto out;
3625 switch (cmd) {
3626 case SIOCINQ: {
3627 struct sk_buff *skb;
3628 unsigned int amount = 0;
3630 skb = skb_peek(&sk->sk_receive_queue);
3631 if (skb != NULL) {
3633 * We will only return the amount of this packet since
3634 * that is all that will be read.
3636 amount = skb->len;
3638 rc = put_user(amount, (int __user *)arg);
3639 break;
3641 default:
3642 rc = -ENOIOCTLCMD;
3643 break;
3645 out:
3646 sctp_release_sock(sk);
3647 return rc;
3650 /* This is the function which gets called during socket creation to
3651 * initialized the SCTP-specific portion of the sock.
3652 * The sock structure should already be zero-filled memory.
3654 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3656 struct sctp_endpoint *ep;
3657 struct sctp_sock *sp;
3659 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3661 sp = sctp_sk(sk);
3663 /* Initialize the SCTP per socket area. */
3664 switch (sk->sk_type) {
3665 case SOCK_SEQPACKET:
3666 sp->type = SCTP_SOCKET_UDP;
3667 break;
3668 case SOCK_STREAM:
3669 sp->type = SCTP_SOCKET_TCP;
3670 break;
3671 default:
3672 return -ESOCKTNOSUPPORT;
3675 /* Initialize default send parameters. These parameters can be
3676 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3678 sp->default_stream = 0;
3679 sp->default_ppid = 0;
3680 sp->default_flags = 0;
3681 sp->default_context = 0;
3682 sp->default_timetolive = 0;
3684 sp->default_rcv_context = 0;
3685 sp->max_burst = sctp_max_burst;
3687 /* Initialize default setup parameters. These parameters
3688 * can be modified with the SCTP_INITMSG socket option or
3689 * overridden by the SCTP_INIT CMSG.
3691 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3692 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3693 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3694 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3696 /* Initialize default RTO related parameters. These parameters can
3697 * be modified for with the SCTP_RTOINFO socket option.
3699 sp->rtoinfo.srto_initial = sctp_rto_initial;
3700 sp->rtoinfo.srto_max = sctp_rto_max;
3701 sp->rtoinfo.srto_min = sctp_rto_min;
3703 /* Initialize default association related parameters. These parameters
3704 * can be modified with the SCTP_ASSOCINFO socket option.
3706 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3707 sp->assocparams.sasoc_number_peer_destinations = 0;
3708 sp->assocparams.sasoc_peer_rwnd = 0;
3709 sp->assocparams.sasoc_local_rwnd = 0;
3710 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3712 /* Initialize default event subscriptions. By default, all the
3713 * options are off.
3715 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3717 /* Default Peer Address Parameters. These defaults can
3718 * be modified via SCTP_PEER_ADDR_PARAMS
3720 sp->hbinterval = sctp_hb_interval;
3721 sp->pathmaxrxt = sctp_max_retrans_path;
3722 sp->pathmtu = 0; // allow default discovery
3723 sp->sackdelay = sctp_sack_timeout;
3724 sp->sackfreq = 2;
3725 sp->param_flags = SPP_HB_ENABLE |
3726 SPP_PMTUD_ENABLE |
3727 SPP_SACKDELAY_ENABLE;
3729 /* If enabled no SCTP message fragmentation will be performed.
3730 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3732 sp->disable_fragments = 0;
3734 /* Enable Nagle algorithm by default. */
3735 sp->nodelay = 0;
3737 /* Enable by default. */
3738 sp->v4mapped = 1;
3740 /* Auto-close idle associations after the configured
3741 * number of seconds. A value of 0 disables this
3742 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3743 * for UDP-style sockets only.
3745 sp->autoclose = 0;
3747 /* User specified fragmentation limit. */
3748 sp->user_frag = 0;
3750 sp->adaptation_ind = 0;
3752 sp->pf = sctp_get_pf_specific(sk->sk_family);
3754 /* Control variables for partial data delivery. */
3755 atomic_set(&sp->pd_mode, 0);
3756 skb_queue_head_init(&sp->pd_lobby);
3757 sp->frag_interleave = 0;
3759 /* Create a per socket endpoint structure. Even if we
3760 * change the data structure relationships, this may still
3761 * be useful for storing pre-connect address information.
3763 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3764 if (!ep)
3765 return -ENOMEM;
3767 sp->ep = ep;
3768 sp->hmac = NULL;
3770 SCTP_DBG_OBJCNT_INC(sock);
3772 local_bh_disable();
3773 percpu_counter_inc(&sctp_sockets_allocated);
3774 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3775 local_bh_enable();
3777 return 0;
3780 /* Cleanup any SCTP per socket resources. */
3781 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3783 struct sctp_endpoint *ep;
3785 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3787 /* Release our hold on the endpoint. */
3788 ep = sctp_sk(sk)->ep;
3789 sctp_endpoint_free(ep);
3790 local_bh_disable();
3791 percpu_counter_dec(&sctp_sockets_allocated);
3792 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3793 local_bh_enable();
3796 /* API 4.1.7 shutdown() - TCP Style Syntax
3797 * int shutdown(int socket, int how);
3799 * sd - the socket descriptor of the association to be closed.
3800 * how - Specifies the type of shutdown. The values are
3801 * as follows:
3802 * SHUT_RD
3803 * Disables further receive operations. No SCTP
3804 * protocol action is taken.
3805 * SHUT_WR
3806 * Disables further send operations, and initiates
3807 * the SCTP shutdown sequence.
3808 * SHUT_RDWR
3809 * Disables further send and receive operations
3810 * and initiates the SCTP shutdown sequence.
3812 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3814 struct sctp_endpoint *ep;
3815 struct sctp_association *asoc;
3817 if (!sctp_style(sk, TCP))
3818 return;
3820 if (how & SEND_SHUTDOWN) {
3821 ep = sctp_sk(sk)->ep;
3822 if (!list_empty(&ep->asocs)) {
3823 asoc = list_entry(ep->asocs.next,
3824 struct sctp_association, asocs);
3825 sctp_primitive_SHUTDOWN(asoc, NULL);
3830 /* 7.2.1 Association Status (SCTP_STATUS)
3832 * Applications can retrieve current status information about an
3833 * association, including association state, peer receiver window size,
3834 * number of unacked data chunks, and number of data chunks pending
3835 * receipt. This information is read-only.
3837 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3838 char __user *optval,
3839 int __user *optlen)
3841 struct sctp_status status;
3842 struct sctp_association *asoc = NULL;
3843 struct sctp_transport *transport;
3844 sctp_assoc_t associd;
3845 int retval = 0;
3847 if (len < sizeof(status)) {
3848 retval = -EINVAL;
3849 goto out;
3852 len = sizeof(status);
3853 if (copy_from_user(&status, optval, len)) {
3854 retval = -EFAULT;
3855 goto out;
3858 associd = status.sstat_assoc_id;
3859 asoc = sctp_id2assoc(sk, associd);
3860 if (!asoc) {
3861 retval = -EINVAL;
3862 goto out;
3865 transport = asoc->peer.primary_path;
3867 status.sstat_assoc_id = sctp_assoc2id(asoc);
3868 status.sstat_state = asoc->state;
3869 status.sstat_rwnd = asoc->peer.rwnd;
3870 status.sstat_unackdata = asoc->unack_data;
3872 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3873 status.sstat_instrms = asoc->c.sinit_max_instreams;
3874 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3875 status.sstat_fragmentation_point = asoc->frag_point;
3876 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3877 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3878 transport->af_specific->sockaddr_len);
3879 /* Map ipv4 address into v4-mapped-on-v6 address. */
3880 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3881 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3882 status.sstat_primary.spinfo_state = transport->state;
3883 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3884 status.sstat_primary.spinfo_srtt = transport->srtt;
3885 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3886 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3888 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3889 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3891 if (put_user(len, optlen)) {
3892 retval = -EFAULT;
3893 goto out;
3896 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3897 len, status.sstat_state, status.sstat_rwnd,
3898 status.sstat_assoc_id);
3900 if (copy_to_user(optval, &status, len)) {
3901 retval = -EFAULT;
3902 goto out;
3905 out:
3906 return retval;
3910 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3912 * Applications can retrieve information about a specific peer address
3913 * of an association, including its reachability state, congestion
3914 * window, and retransmission timer values. This information is
3915 * read-only.
3917 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3918 char __user *optval,
3919 int __user *optlen)
3921 struct sctp_paddrinfo pinfo;
3922 struct sctp_transport *transport;
3923 int retval = 0;
3925 if (len < sizeof(pinfo)) {
3926 retval = -EINVAL;
3927 goto out;
3930 len = sizeof(pinfo);
3931 if (copy_from_user(&pinfo, optval, len)) {
3932 retval = -EFAULT;
3933 goto out;
3936 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3937 pinfo.spinfo_assoc_id);
3938 if (!transport)
3939 return -EINVAL;
3941 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3942 pinfo.spinfo_state = transport->state;
3943 pinfo.spinfo_cwnd = transport->cwnd;
3944 pinfo.spinfo_srtt = transport->srtt;
3945 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3946 pinfo.spinfo_mtu = transport->pathmtu;
3948 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3949 pinfo.spinfo_state = SCTP_ACTIVE;
3951 if (put_user(len, optlen)) {
3952 retval = -EFAULT;
3953 goto out;
3956 if (copy_to_user(optval, &pinfo, len)) {
3957 retval = -EFAULT;
3958 goto out;
3961 out:
3962 return retval;
3965 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3967 * This option is a on/off flag. If enabled no SCTP message
3968 * fragmentation will be performed. Instead if a message being sent
3969 * exceeds the current PMTU size, the message will NOT be sent and
3970 * instead a error will be indicated to the user.
3972 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3973 char __user *optval, int __user *optlen)
3975 int val;
3977 if (len < sizeof(int))
3978 return -EINVAL;
3980 len = sizeof(int);
3981 val = (sctp_sk(sk)->disable_fragments == 1);
3982 if (put_user(len, optlen))
3983 return -EFAULT;
3984 if (copy_to_user(optval, &val, len))
3985 return -EFAULT;
3986 return 0;
3989 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3991 * This socket option is used to specify various notifications and
3992 * ancillary data the user wishes to receive.
3994 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3995 int __user *optlen)
3997 if (len < sizeof(struct sctp_event_subscribe))
3998 return -EINVAL;
3999 len = sizeof(struct sctp_event_subscribe);
4000 if (put_user(len, optlen))
4001 return -EFAULT;
4002 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4003 return -EFAULT;
4004 return 0;
4007 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4009 * This socket option is applicable to the UDP-style socket only. When
4010 * set it will cause associations that are idle for more than the
4011 * specified number of seconds to automatically close. An association
4012 * being idle is defined an association that has NOT sent or received
4013 * user data. The special value of '0' indicates that no automatic
4014 * close of any associations should be performed. The option expects an
4015 * integer defining the number of seconds of idle time before an
4016 * association is closed.
4018 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4020 /* Applicable to UDP-style socket only */
4021 if (sctp_style(sk, TCP))
4022 return -EOPNOTSUPP;
4023 if (len < sizeof(int))
4024 return -EINVAL;
4025 len = sizeof(int);
4026 if (put_user(len, optlen))
4027 return -EFAULT;
4028 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4029 return -EFAULT;
4030 return 0;
4033 /* Helper routine to branch off an association to a new socket. */
4034 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4035 struct socket **sockp)
4037 struct sock *sk = asoc->base.sk;
4038 struct socket *sock;
4039 struct sctp_af *af;
4040 int err = 0;
4042 /* An association cannot be branched off from an already peeled-off
4043 * socket, nor is this supported for tcp style sockets.
4045 if (!sctp_style(sk, UDP))
4046 return -EINVAL;
4048 /* Create a new socket. */
4049 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4050 if (err < 0)
4051 return err;
4053 sctp_copy_sock(sock->sk, sk, asoc);
4055 /* Make peeled-off sockets more like 1-1 accepted sockets.
4056 * Set the daddr and initialize id to something more random
4058 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4059 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4061 /* Populate the fields of the newsk from the oldsk and migrate the
4062 * asoc to the newsk.
4064 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4066 *sockp = sock;
4068 return err;
4071 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4073 sctp_peeloff_arg_t peeloff;
4074 struct socket *newsock;
4075 int retval = 0;
4076 struct sctp_association *asoc;
4078 if (len < sizeof(sctp_peeloff_arg_t))
4079 return -EINVAL;
4080 len = sizeof(sctp_peeloff_arg_t);
4081 if (copy_from_user(&peeloff, optval, len))
4082 return -EFAULT;
4084 asoc = sctp_id2assoc(sk, peeloff.associd);
4085 if (!asoc) {
4086 retval = -EINVAL;
4087 goto out;
4090 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4092 retval = sctp_do_peeloff(asoc, &newsock);
4093 if (retval < 0)
4094 goto out;
4096 /* Map the socket to an unused fd that can be returned to the user. */
4097 retval = sock_map_fd(newsock, 0);
4098 if (retval < 0) {
4099 sock_release(newsock);
4100 goto out;
4103 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4104 __func__, sk, asoc, newsock->sk, retval);
4106 /* Return the fd mapped to the new socket. */
4107 peeloff.sd = retval;
4108 if (put_user(len, optlen))
4109 return -EFAULT;
4110 if (copy_to_user(optval, &peeloff, len))
4111 retval = -EFAULT;
4113 out:
4114 return retval;
4117 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4119 * Applications can enable or disable heartbeats for any peer address of
4120 * an association, modify an address's heartbeat interval, force a
4121 * heartbeat to be sent immediately, and adjust the address's maximum
4122 * number of retransmissions sent before an address is considered
4123 * unreachable. The following structure is used to access and modify an
4124 * address's parameters:
4126 * struct sctp_paddrparams {
4127 * sctp_assoc_t spp_assoc_id;
4128 * struct sockaddr_storage spp_address;
4129 * uint32_t spp_hbinterval;
4130 * uint16_t spp_pathmaxrxt;
4131 * uint32_t spp_pathmtu;
4132 * uint32_t spp_sackdelay;
4133 * uint32_t spp_flags;
4134 * };
4136 * spp_assoc_id - (one-to-many style socket) This is filled in the
4137 * application, and identifies the association for
4138 * this query.
4139 * spp_address - This specifies which address is of interest.
4140 * spp_hbinterval - This contains the value of the heartbeat interval,
4141 * in milliseconds. If a value of zero
4142 * is present in this field then no changes are to
4143 * be made to this parameter.
4144 * spp_pathmaxrxt - This contains the maximum number of
4145 * retransmissions before this address shall be
4146 * considered unreachable. If a value of zero
4147 * is present in this field then no changes are to
4148 * be made to this parameter.
4149 * spp_pathmtu - When Path MTU discovery is disabled the value
4150 * specified here will be the "fixed" path mtu.
4151 * Note that if the spp_address field is empty
4152 * then all associations on this address will
4153 * have this fixed path mtu set upon them.
4155 * spp_sackdelay - When delayed sack is enabled, this value specifies
4156 * the number of milliseconds that sacks will be delayed
4157 * for. This value will apply to all addresses of an
4158 * association if the spp_address field is empty. Note
4159 * also, that if delayed sack is enabled and this
4160 * value is set to 0, no change is made to the last
4161 * recorded delayed sack timer value.
4163 * spp_flags - These flags are used to control various features
4164 * on an association. The flag field may contain
4165 * zero or more of the following options.
4167 * SPP_HB_ENABLE - Enable heartbeats on the
4168 * specified address. Note that if the address
4169 * field is empty all addresses for the association
4170 * have heartbeats enabled upon them.
4172 * SPP_HB_DISABLE - Disable heartbeats on the
4173 * speicifed address. Note that if the address
4174 * field is empty all addresses for the association
4175 * will have their heartbeats disabled. Note also
4176 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4177 * mutually exclusive, only one of these two should
4178 * be specified. Enabling both fields will have
4179 * undetermined results.
4181 * SPP_HB_DEMAND - Request a user initiated heartbeat
4182 * to be made immediately.
4184 * SPP_PMTUD_ENABLE - This field will enable PMTU
4185 * discovery upon the specified address. Note that
4186 * if the address feild is empty then all addresses
4187 * on the association are effected.
4189 * SPP_PMTUD_DISABLE - This field will disable PMTU
4190 * discovery upon the specified address. Note that
4191 * if the address feild is empty then all addresses
4192 * on the association are effected. Not also that
4193 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4194 * exclusive. Enabling both will have undetermined
4195 * results.
4197 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4198 * on delayed sack. The time specified in spp_sackdelay
4199 * is used to specify the sack delay for this address. Note
4200 * that if spp_address is empty then all addresses will
4201 * enable delayed sack and take on the sack delay
4202 * value specified in spp_sackdelay.
4203 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4204 * off delayed sack. If the spp_address field is blank then
4205 * delayed sack is disabled for the entire association. Note
4206 * also that this field is mutually exclusive to
4207 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4208 * results.
4210 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4211 char __user *optval, int __user *optlen)
4213 struct sctp_paddrparams params;
4214 struct sctp_transport *trans = NULL;
4215 struct sctp_association *asoc = NULL;
4216 struct sctp_sock *sp = sctp_sk(sk);
4218 if (len < sizeof(struct sctp_paddrparams))
4219 return -EINVAL;
4220 len = sizeof(struct sctp_paddrparams);
4221 if (copy_from_user(&params, optval, len))
4222 return -EFAULT;
4224 /* If an address other than INADDR_ANY is specified, and
4225 * no transport is found, then the request is invalid.
4227 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4228 trans = sctp_addr_id2transport(sk, &params.spp_address,
4229 params.spp_assoc_id);
4230 if (!trans) {
4231 SCTP_DEBUG_PRINTK("Failed no transport\n");
4232 return -EINVAL;
4236 /* Get association, if assoc_id != 0 and the socket is a one
4237 * to many style socket, and an association was not found, then
4238 * the id was invalid.
4240 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4241 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4242 SCTP_DEBUG_PRINTK("Failed no association\n");
4243 return -EINVAL;
4246 if (trans) {
4247 /* Fetch transport values. */
4248 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4249 params.spp_pathmtu = trans->pathmtu;
4250 params.spp_pathmaxrxt = trans->pathmaxrxt;
4251 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4253 /*draft-11 doesn't say what to return in spp_flags*/
4254 params.spp_flags = trans->param_flags;
4255 } else if (asoc) {
4256 /* Fetch association values. */
4257 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4258 params.spp_pathmtu = asoc->pathmtu;
4259 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4260 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4262 /*draft-11 doesn't say what to return in spp_flags*/
4263 params.spp_flags = asoc->param_flags;
4264 } else {
4265 /* Fetch socket values. */
4266 params.spp_hbinterval = sp->hbinterval;
4267 params.spp_pathmtu = sp->pathmtu;
4268 params.spp_sackdelay = sp->sackdelay;
4269 params.spp_pathmaxrxt = sp->pathmaxrxt;
4271 /*draft-11 doesn't say what to return in spp_flags*/
4272 params.spp_flags = sp->param_flags;
4275 if (copy_to_user(optval, &params, len))
4276 return -EFAULT;
4278 if (put_user(len, optlen))
4279 return -EFAULT;
4281 return 0;
4285 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4287 * This option will effect the way delayed acks are performed. This
4288 * option allows you to get or set the delayed ack time, in
4289 * milliseconds. It also allows changing the delayed ack frequency.
4290 * Changing the frequency to 1 disables the delayed sack algorithm. If
4291 * the assoc_id is 0, then this sets or gets the endpoints default
4292 * values. If the assoc_id field is non-zero, then the set or get
4293 * effects the specified association for the one to many model (the
4294 * assoc_id field is ignored by the one to one model). Note that if
4295 * sack_delay or sack_freq are 0 when setting this option, then the
4296 * current values will remain unchanged.
4298 * struct sctp_sack_info {
4299 * sctp_assoc_t sack_assoc_id;
4300 * uint32_t sack_delay;
4301 * uint32_t sack_freq;
4302 * };
4304 * sack_assoc_id - This parameter, indicates which association the user
4305 * is performing an action upon. Note that if this field's value is
4306 * zero then the endpoints default value is changed (effecting future
4307 * associations only).
4309 * sack_delay - This parameter contains the number of milliseconds that
4310 * the user is requesting the delayed ACK timer be set to. Note that
4311 * this value is defined in the standard to be between 200 and 500
4312 * milliseconds.
4314 * sack_freq - This parameter contains the number of packets that must
4315 * be received before a sack is sent without waiting for the delay
4316 * timer to expire. The default value for this is 2, setting this
4317 * value to 1 will disable the delayed sack algorithm.
4319 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4320 char __user *optval,
4321 int __user *optlen)
4323 struct sctp_sack_info params;
4324 struct sctp_association *asoc = NULL;
4325 struct sctp_sock *sp = sctp_sk(sk);
4327 if (len >= sizeof(struct sctp_sack_info)) {
4328 len = sizeof(struct sctp_sack_info);
4330 if (copy_from_user(&params, optval, len))
4331 return -EFAULT;
4332 } else if (len == sizeof(struct sctp_assoc_value)) {
4333 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4334 pr_warn("Use struct sctp_sack_info instead\n");
4335 if (copy_from_user(&params, optval, len))
4336 return -EFAULT;
4337 } else
4338 return - EINVAL;
4340 /* Get association, if sack_assoc_id != 0 and the socket is a one
4341 * to many style socket, and an association was not found, then
4342 * the id was invalid.
4344 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4345 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4346 return -EINVAL;
4348 if (asoc) {
4349 /* Fetch association values. */
4350 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4351 params.sack_delay = jiffies_to_msecs(
4352 asoc->sackdelay);
4353 params.sack_freq = asoc->sackfreq;
4355 } else {
4356 params.sack_delay = 0;
4357 params.sack_freq = 1;
4359 } else {
4360 /* Fetch socket values. */
4361 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4362 params.sack_delay = sp->sackdelay;
4363 params.sack_freq = sp->sackfreq;
4364 } else {
4365 params.sack_delay = 0;
4366 params.sack_freq = 1;
4370 if (copy_to_user(optval, &params, len))
4371 return -EFAULT;
4373 if (put_user(len, optlen))
4374 return -EFAULT;
4376 return 0;
4379 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4381 * Applications can specify protocol parameters for the default association
4382 * initialization. The option name argument to setsockopt() and getsockopt()
4383 * is SCTP_INITMSG.
4385 * Setting initialization parameters is effective only on an unconnected
4386 * socket (for UDP-style sockets only future associations are effected
4387 * by the change). With TCP-style sockets, this option is inherited by
4388 * sockets derived from a listener socket.
4390 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4392 if (len < sizeof(struct sctp_initmsg))
4393 return -EINVAL;
4394 len = sizeof(struct sctp_initmsg);
4395 if (put_user(len, optlen))
4396 return -EFAULT;
4397 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4398 return -EFAULT;
4399 return 0;
4403 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4404 char __user *optval, int __user *optlen)
4406 struct sctp_association *asoc;
4407 int cnt = 0;
4408 struct sctp_getaddrs getaddrs;
4409 struct sctp_transport *from;
4410 void __user *to;
4411 union sctp_addr temp;
4412 struct sctp_sock *sp = sctp_sk(sk);
4413 int addrlen;
4414 size_t space_left;
4415 int bytes_copied;
4417 if (len < sizeof(struct sctp_getaddrs))
4418 return -EINVAL;
4420 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4421 return -EFAULT;
4423 /* For UDP-style sockets, id specifies the association to query. */
4424 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4425 if (!asoc)
4426 return -EINVAL;
4428 to = optval + offsetof(struct sctp_getaddrs,addrs);
4429 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4431 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4432 transports) {
4433 memcpy(&temp, &from->ipaddr, sizeof(temp));
4434 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4435 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4436 if (space_left < addrlen)
4437 return -ENOMEM;
4438 if (copy_to_user(to, &temp, addrlen))
4439 return -EFAULT;
4440 to += addrlen;
4441 cnt++;
4442 space_left -= addrlen;
4445 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4446 return -EFAULT;
4447 bytes_copied = ((char __user *)to) - optval;
4448 if (put_user(bytes_copied, optlen))
4449 return -EFAULT;
4451 return 0;
4454 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4455 size_t space_left, int *bytes_copied)
4457 struct sctp_sockaddr_entry *addr;
4458 union sctp_addr temp;
4459 int cnt = 0;
4460 int addrlen;
4462 rcu_read_lock();
4463 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4464 if (!addr->valid)
4465 continue;
4467 if ((PF_INET == sk->sk_family) &&
4468 (AF_INET6 == addr->a.sa.sa_family))
4469 continue;
4470 if ((PF_INET6 == sk->sk_family) &&
4471 inet_v6_ipv6only(sk) &&
4472 (AF_INET == addr->a.sa.sa_family))
4473 continue;
4474 memcpy(&temp, &addr->a, sizeof(temp));
4475 if (!temp.v4.sin_port)
4476 temp.v4.sin_port = htons(port);
4478 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4479 &temp);
4480 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4481 if (space_left < addrlen) {
4482 cnt = -ENOMEM;
4483 break;
4485 memcpy(to, &temp, addrlen);
4487 to += addrlen;
4488 cnt ++;
4489 space_left -= addrlen;
4490 *bytes_copied += addrlen;
4492 rcu_read_unlock();
4494 return cnt;
4498 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4499 char __user *optval, int __user *optlen)
4501 struct sctp_bind_addr *bp;
4502 struct sctp_association *asoc;
4503 int cnt = 0;
4504 struct sctp_getaddrs getaddrs;
4505 struct sctp_sockaddr_entry *addr;
4506 void __user *to;
4507 union sctp_addr temp;
4508 struct sctp_sock *sp = sctp_sk(sk);
4509 int addrlen;
4510 int err = 0;
4511 size_t space_left;
4512 int bytes_copied = 0;
4513 void *addrs;
4514 void *buf;
4516 if (len < sizeof(struct sctp_getaddrs))
4517 return -EINVAL;
4519 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4520 return -EFAULT;
4523 * For UDP-style sockets, id specifies the association to query.
4524 * If the id field is set to the value '0' then the locally bound
4525 * addresses are returned without regard to any particular
4526 * association.
4528 if (0 == getaddrs.assoc_id) {
4529 bp = &sctp_sk(sk)->ep->base.bind_addr;
4530 } else {
4531 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4532 if (!asoc)
4533 return -EINVAL;
4534 bp = &asoc->base.bind_addr;
4537 to = optval + offsetof(struct sctp_getaddrs,addrs);
4538 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4540 addrs = kmalloc(space_left, GFP_KERNEL);
4541 if (!addrs)
4542 return -ENOMEM;
4544 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4545 * addresses from the global local address list.
4547 if (sctp_list_single_entry(&bp->address_list)) {
4548 addr = list_entry(bp->address_list.next,
4549 struct sctp_sockaddr_entry, list);
4550 if (sctp_is_any(sk, &addr->a)) {
4551 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4552 space_left, &bytes_copied);
4553 if (cnt < 0) {
4554 err = cnt;
4555 goto out;
4557 goto copy_getaddrs;
4561 buf = addrs;
4562 /* Protection on the bound address list is not needed since
4563 * in the socket option context we hold a socket lock and
4564 * thus the bound address list can't change.
4566 list_for_each_entry(addr, &bp->address_list, list) {
4567 memcpy(&temp, &addr->a, sizeof(temp));
4568 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4569 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4570 if (space_left < addrlen) {
4571 err = -ENOMEM; /*fixme: right error?*/
4572 goto out;
4574 memcpy(buf, &temp, addrlen);
4575 buf += addrlen;
4576 bytes_copied += addrlen;
4577 cnt ++;
4578 space_left -= addrlen;
4581 copy_getaddrs:
4582 if (copy_to_user(to, addrs, bytes_copied)) {
4583 err = -EFAULT;
4584 goto out;
4586 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4587 err = -EFAULT;
4588 goto out;
4590 if (put_user(bytes_copied, optlen))
4591 err = -EFAULT;
4592 out:
4593 kfree(addrs);
4594 return err;
4597 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4599 * Requests that the local SCTP stack use the enclosed peer address as
4600 * the association primary. The enclosed address must be one of the
4601 * association peer's addresses.
4603 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4604 char __user *optval, int __user *optlen)
4606 struct sctp_prim prim;
4607 struct sctp_association *asoc;
4608 struct sctp_sock *sp = sctp_sk(sk);
4610 if (len < sizeof(struct sctp_prim))
4611 return -EINVAL;
4613 len = sizeof(struct sctp_prim);
4615 if (copy_from_user(&prim, optval, len))
4616 return -EFAULT;
4618 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4619 if (!asoc)
4620 return -EINVAL;
4622 if (!asoc->peer.primary_path)
4623 return -ENOTCONN;
4625 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4626 asoc->peer.primary_path->af_specific->sockaddr_len);
4628 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4629 (union sctp_addr *)&prim.ssp_addr);
4631 if (put_user(len, optlen))
4632 return -EFAULT;
4633 if (copy_to_user(optval, &prim, len))
4634 return -EFAULT;
4636 return 0;
4640 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4642 * Requests that the local endpoint set the specified Adaptation Layer
4643 * Indication parameter for all future INIT and INIT-ACK exchanges.
4645 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4646 char __user *optval, int __user *optlen)
4648 struct sctp_setadaptation adaptation;
4650 if (len < sizeof(struct sctp_setadaptation))
4651 return -EINVAL;
4653 len = sizeof(struct sctp_setadaptation);
4655 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4657 if (put_user(len, optlen))
4658 return -EFAULT;
4659 if (copy_to_user(optval, &adaptation, len))
4660 return -EFAULT;
4662 return 0;
4667 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4669 * Applications that wish to use the sendto() system call may wish to
4670 * specify a default set of parameters that would normally be supplied
4671 * through the inclusion of ancillary data. This socket option allows
4672 * such an application to set the default sctp_sndrcvinfo structure.
4675 * The application that wishes to use this socket option simply passes
4676 * in to this call the sctp_sndrcvinfo structure defined in Section
4677 * 5.2.2) The input parameters accepted by this call include
4678 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4679 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4680 * to this call if the caller is using the UDP model.
4682 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4684 static int sctp_getsockopt_default_send_param(struct sock *sk,
4685 int len, char __user *optval,
4686 int __user *optlen)
4688 struct sctp_sndrcvinfo info;
4689 struct sctp_association *asoc;
4690 struct sctp_sock *sp = sctp_sk(sk);
4692 if (len < sizeof(struct sctp_sndrcvinfo))
4693 return -EINVAL;
4695 len = sizeof(struct sctp_sndrcvinfo);
4697 if (copy_from_user(&info, optval, len))
4698 return -EFAULT;
4700 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4701 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4702 return -EINVAL;
4704 if (asoc) {
4705 info.sinfo_stream = asoc->default_stream;
4706 info.sinfo_flags = asoc->default_flags;
4707 info.sinfo_ppid = asoc->default_ppid;
4708 info.sinfo_context = asoc->default_context;
4709 info.sinfo_timetolive = asoc->default_timetolive;
4710 } else {
4711 info.sinfo_stream = sp->default_stream;
4712 info.sinfo_flags = sp->default_flags;
4713 info.sinfo_ppid = sp->default_ppid;
4714 info.sinfo_context = sp->default_context;
4715 info.sinfo_timetolive = sp->default_timetolive;
4718 if (put_user(len, optlen))
4719 return -EFAULT;
4720 if (copy_to_user(optval, &info, len))
4721 return -EFAULT;
4723 return 0;
4728 * 7.1.5 SCTP_NODELAY
4730 * Turn on/off any Nagle-like algorithm. This means that packets are
4731 * generally sent as soon as possible and no unnecessary delays are
4732 * introduced, at the cost of more packets in the network. Expects an
4733 * integer boolean flag.
4736 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4737 char __user *optval, int __user *optlen)
4739 int val;
4741 if (len < sizeof(int))
4742 return -EINVAL;
4744 len = sizeof(int);
4745 val = (sctp_sk(sk)->nodelay == 1);
4746 if (put_user(len, optlen))
4747 return -EFAULT;
4748 if (copy_to_user(optval, &val, len))
4749 return -EFAULT;
4750 return 0;
4755 * 7.1.1 SCTP_RTOINFO
4757 * The protocol parameters used to initialize and bound retransmission
4758 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4759 * and modify these parameters.
4760 * All parameters are time values, in milliseconds. A value of 0, when
4761 * modifying the parameters, indicates that the current value should not
4762 * be changed.
4765 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4766 char __user *optval,
4767 int __user *optlen) {
4768 struct sctp_rtoinfo rtoinfo;
4769 struct sctp_association *asoc;
4771 if (len < sizeof (struct sctp_rtoinfo))
4772 return -EINVAL;
4774 len = sizeof(struct sctp_rtoinfo);
4776 if (copy_from_user(&rtoinfo, optval, len))
4777 return -EFAULT;
4779 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4781 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4782 return -EINVAL;
4784 /* Values corresponding to the specific association. */
4785 if (asoc) {
4786 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4787 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4788 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4789 } else {
4790 /* Values corresponding to the endpoint. */
4791 struct sctp_sock *sp = sctp_sk(sk);
4793 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4794 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4795 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4798 if (put_user(len, optlen))
4799 return -EFAULT;
4801 if (copy_to_user(optval, &rtoinfo, len))
4802 return -EFAULT;
4804 return 0;
4809 * 7.1.2 SCTP_ASSOCINFO
4811 * This option is used to tune the maximum retransmission attempts
4812 * of the association.
4813 * Returns an error if the new association retransmission value is
4814 * greater than the sum of the retransmission value of the peer.
4815 * See [SCTP] for more information.
4818 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4819 char __user *optval,
4820 int __user *optlen)
4823 struct sctp_assocparams assocparams;
4824 struct sctp_association *asoc;
4825 struct list_head *pos;
4826 int cnt = 0;
4828 if (len < sizeof (struct sctp_assocparams))
4829 return -EINVAL;
4831 len = sizeof(struct sctp_assocparams);
4833 if (copy_from_user(&assocparams, optval, len))
4834 return -EFAULT;
4836 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4838 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4839 return -EINVAL;
4841 /* Values correspoinding to the specific association */
4842 if (asoc) {
4843 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4844 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4845 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4846 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4847 * 1000) +
4848 (asoc->cookie_life.tv_usec
4849 / 1000);
4851 list_for_each(pos, &asoc->peer.transport_addr_list) {
4852 cnt ++;
4855 assocparams.sasoc_number_peer_destinations = cnt;
4856 } else {
4857 /* Values corresponding to the endpoint */
4858 struct sctp_sock *sp = sctp_sk(sk);
4860 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4861 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4862 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4863 assocparams.sasoc_cookie_life =
4864 sp->assocparams.sasoc_cookie_life;
4865 assocparams.sasoc_number_peer_destinations =
4866 sp->assocparams.
4867 sasoc_number_peer_destinations;
4870 if (put_user(len, optlen))
4871 return -EFAULT;
4873 if (copy_to_user(optval, &assocparams, len))
4874 return -EFAULT;
4876 return 0;
4880 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4882 * This socket option is a boolean flag which turns on or off mapped V4
4883 * addresses. If this option is turned on and the socket is type
4884 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4885 * If this option is turned off, then no mapping will be done of V4
4886 * addresses and a user will receive both PF_INET6 and PF_INET type
4887 * addresses on the socket.
4889 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4890 char __user *optval, int __user *optlen)
4892 int val;
4893 struct sctp_sock *sp = sctp_sk(sk);
4895 if (len < sizeof(int))
4896 return -EINVAL;
4898 len = sizeof(int);
4899 val = sp->v4mapped;
4900 if (put_user(len, optlen))
4901 return -EFAULT;
4902 if (copy_to_user(optval, &val, len))
4903 return -EFAULT;
4905 return 0;
4909 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4910 * (chapter and verse is quoted at sctp_setsockopt_context())
4912 static int sctp_getsockopt_context(struct sock *sk, int len,
4913 char __user *optval, int __user *optlen)
4915 struct sctp_assoc_value params;
4916 struct sctp_sock *sp;
4917 struct sctp_association *asoc;
4919 if (len < sizeof(struct sctp_assoc_value))
4920 return -EINVAL;
4922 len = sizeof(struct sctp_assoc_value);
4924 if (copy_from_user(&params, optval, len))
4925 return -EFAULT;
4927 sp = sctp_sk(sk);
4929 if (params.assoc_id != 0) {
4930 asoc = sctp_id2assoc(sk, params.assoc_id);
4931 if (!asoc)
4932 return -EINVAL;
4933 params.assoc_value = asoc->default_rcv_context;
4934 } else {
4935 params.assoc_value = sp->default_rcv_context;
4938 if (put_user(len, optlen))
4939 return -EFAULT;
4940 if (copy_to_user(optval, &params, len))
4941 return -EFAULT;
4943 return 0;
4947 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4948 * This option will get or set the maximum size to put in any outgoing
4949 * SCTP DATA chunk. If a message is larger than this size it will be
4950 * fragmented by SCTP into the specified size. Note that the underlying
4951 * SCTP implementation may fragment into smaller sized chunks when the
4952 * PMTU of the underlying association is smaller than the value set by
4953 * the user. The default value for this option is '0' which indicates
4954 * the user is NOT limiting fragmentation and only the PMTU will effect
4955 * SCTP's choice of DATA chunk size. Note also that values set larger
4956 * than the maximum size of an IP datagram will effectively let SCTP
4957 * control fragmentation (i.e. the same as setting this option to 0).
4959 * The following structure is used to access and modify this parameter:
4961 * struct sctp_assoc_value {
4962 * sctp_assoc_t assoc_id;
4963 * uint32_t assoc_value;
4964 * };
4966 * assoc_id: This parameter is ignored for one-to-one style sockets.
4967 * For one-to-many style sockets this parameter indicates which
4968 * association the user is performing an action upon. Note that if
4969 * this field's value is zero then the endpoints default value is
4970 * changed (effecting future associations only).
4971 * assoc_value: This parameter specifies the maximum size in bytes.
4973 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4974 char __user *optval, int __user *optlen)
4976 struct sctp_assoc_value params;
4977 struct sctp_association *asoc;
4979 if (len == sizeof(int)) {
4980 pr_warn("Use of int in maxseg socket option deprecated\n");
4981 pr_warn("Use struct sctp_assoc_value instead\n");
4982 params.assoc_id = 0;
4983 } else if (len >= sizeof(struct sctp_assoc_value)) {
4984 len = sizeof(struct sctp_assoc_value);
4985 if (copy_from_user(&params, optval, sizeof(params)))
4986 return -EFAULT;
4987 } else
4988 return -EINVAL;
4990 asoc = sctp_id2assoc(sk, params.assoc_id);
4991 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4992 return -EINVAL;
4994 if (asoc)
4995 params.assoc_value = asoc->frag_point;
4996 else
4997 params.assoc_value = sctp_sk(sk)->user_frag;
4999 if (put_user(len, optlen))
5000 return -EFAULT;
5001 if (len == sizeof(int)) {
5002 if (copy_to_user(optval, &params.assoc_value, len))
5003 return -EFAULT;
5004 } else {
5005 if (copy_to_user(optval, &params, len))
5006 return -EFAULT;
5009 return 0;
5013 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5014 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5016 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5017 char __user *optval, int __user *optlen)
5019 int val;
5021 if (len < sizeof(int))
5022 return -EINVAL;
5024 len = sizeof(int);
5026 val = sctp_sk(sk)->frag_interleave;
5027 if (put_user(len, optlen))
5028 return -EFAULT;
5029 if (copy_to_user(optval, &val, len))
5030 return -EFAULT;
5032 return 0;
5036 * 7.1.25. Set or Get the sctp partial delivery point
5037 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5039 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5040 char __user *optval,
5041 int __user *optlen)
5043 u32 val;
5045 if (len < sizeof(u32))
5046 return -EINVAL;
5048 len = sizeof(u32);
5050 val = sctp_sk(sk)->pd_point;
5051 if (put_user(len, optlen))
5052 return -EFAULT;
5053 if (copy_to_user(optval, &val, len))
5054 return -EFAULT;
5056 return -ENOTSUPP;
5060 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5061 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5063 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5064 char __user *optval,
5065 int __user *optlen)
5067 struct sctp_assoc_value params;
5068 struct sctp_sock *sp;
5069 struct sctp_association *asoc;
5071 if (len == sizeof(int)) {
5072 pr_warn("Use of int in max_burst socket option deprecated\n");
5073 pr_warn("Use struct sctp_assoc_value instead\n");
5074 params.assoc_id = 0;
5075 } else if (len >= sizeof(struct sctp_assoc_value)) {
5076 len = sizeof(struct sctp_assoc_value);
5077 if (copy_from_user(&params, optval, len))
5078 return -EFAULT;
5079 } else
5080 return -EINVAL;
5082 sp = sctp_sk(sk);
5084 if (params.assoc_id != 0) {
5085 asoc = sctp_id2assoc(sk, params.assoc_id);
5086 if (!asoc)
5087 return -EINVAL;
5088 params.assoc_value = asoc->max_burst;
5089 } else
5090 params.assoc_value = sp->max_burst;
5092 if (len == sizeof(int)) {
5093 if (copy_to_user(optval, &params.assoc_value, len))
5094 return -EFAULT;
5095 } else {
5096 if (copy_to_user(optval, &params, len))
5097 return -EFAULT;
5100 return 0;
5104 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5105 char __user *optval, int __user *optlen)
5107 struct sctp_hmacalgo __user *p = (void __user *)optval;
5108 struct sctp_hmac_algo_param *hmacs;
5109 __u16 data_len = 0;
5110 u32 num_idents;
5112 if (!sctp_auth_enable)
5113 return -EACCES;
5115 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5116 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5118 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5119 return -EINVAL;
5121 len = sizeof(struct sctp_hmacalgo) + data_len;
5122 num_idents = data_len / sizeof(u16);
5124 if (put_user(len, optlen))
5125 return -EFAULT;
5126 if (put_user(num_idents, &p->shmac_num_idents))
5127 return -EFAULT;
5128 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5129 return -EFAULT;
5130 return 0;
5133 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5134 char __user *optval, int __user *optlen)
5136 struct sctp_authkeyid val;
5137 struct sctp_association *asoc;
5139 if (!sctp_auth_enable)
5140 return -EACCES;
5142 if (len < sizeof(struct sctp_authkeyid))
5143 return -EINVAL;
5144 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5145 return -EFAULT;
5147 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5148 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5149 return -EINVAL;
5151 if (asoc)
5152 val.scact_keynumber = asoc->active_key_id;
5153 else
5154 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5156 len = sizeof(struct sctp_authkeyid);
5157 if (put_user(len, optlen))
5158 return -EFAULT;
5159 if (copy_to_user(optval, &val, len))
5160 return -EFAULT;
5162 return 0;
5165 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5166 char __user *optval, int __user *optlen)
5168 struct sctp_authchunks __user *p = (void __user *)optval;
5169 struct sctp_authchunks val;
5170 struct sctp_association *asoc;
5171 struct sctp_chunks_param *ch;
5172 u32 num_chunks = 0;
5173 char __user *to;
5175 if (!sctp_auth_enable)
5176 return -EACCES;
5178 if (len < sizeof(struct sctp_authchunks))
5179 return -EINVAL;
5181 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5182 return -EFAULT;
5184 to = p->gauth_chunks;
5185 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5186 if (!asoc)
5187 return -EINVAL;
5189 ch = asoc->peer.peer_chunks;
5190 if (!ch)
5191 goto num;
5193 /* See if the user provided enough room for all the data */
5194 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5195 if (len < num_chunks)
5196 return -EINVAL;
5198 if (copy_to_user(to, ch->chunks, num_chunks))
5199 return -EFAULT;
5200 num:
5201 len = sizeof(struct sctp_authchunks) + num_chunks;
5202 if (put_user(len, optlen)) return -EFAULT;
5203 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5204 return -EFAULT;
5205 return 0;
5208 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5209 char __user *optval, int __user *optlen)
5211 struct sctp_authchunks __user *p = (void __user *)optval;
5212 struct sctp_authchunks val;
5213 struct sctp_association *asoc;
5214 struct sctp_chunks_param *ch;
5215 u32 num_chunks = 0;
5216 char __user *to;
5218 if (!sctp_auth_enable)
5219 return -EACCES;
5221 if (len < sizeof(struct sctp_authchunks))
5222 return -EINVAL;
5224 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5225 return -EFAULT;
5227 to = p->gauth_chunks;
5228 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5229 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5230 return -EINVAL;
5232 if (asoc)
5233 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5234 else
5235 ch = sctp_sk(sk)->ep->auth_chunk_list;
5237 if (!ch)
5238 goto num;
5240 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5241 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5242 return -EINVAL;
5244 if (copy_to_user(to, ch->chunks, num_chunks))
5245 return -EFAULT;
5246 num:
5247 len = sizeof(struct sctp_authchunks) + num_chunks;
5248 if (put_user(len, optlen))
5249 return -EFAULT;
5250 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5251 return -EFAULT;
5253 return 0;
5257 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5258 * This option gets the current number of associations that are attached
5259 * to a one-to-many style socket. The option value is an uint32_t.
5261 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5262 char __user *optval, int __user *optlen)
5264 struct sctp_sock *sp = sctp_sk(sk);
5265 struct sctp_association *asoc;
5266 u32 val = 0;
5268 if (sctp_style(sk, TCP))
5269 return -EOPNOTSUPP;
5271 if (len < sizeof(u32))
5272 return -EINVAL;
5274 len = sizeof(u32);
5276 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5277 val++;
5280 if (put_user(len, optlen))
5281 return -EFAULT;
5282 if (copy_to_user(optval, &val, len))
5283 return -EFAULT;
5285 return 0;
5288 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5289 char __user *optval, int __user *optlen)
5291 int retval = 0;
5292 int len;
5294 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5295 sk, optname);
5297 /* I can hardly begin to describe how wrong this is. This is
5298 * so broken as to be worse than useless. The API draft
5299 * REALLY is NOT helpful here... I am not convinced that the
5300 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5301 * are at all well-founded.
5303 if (level != SOL_SCTP) {
5304 struct sctp_af *af = sctp_sk(sk)->pf->af;
5306 retval = af->getsockopt(sk, level, optname, optval, optlen);
5307 return retval;
5310 if (get_user(len, optlen))
5311 return -EFAULT;
5313 sctp_lock_sock(sk);
5315 switch (optname) {
5316 case SCTP_STATUS:
5317 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5318 break;
5319 case SCTP_DISABLE_FRAGMENTS:
5320 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5321 optlen);
5322 break;
5323 case SCTP_EVENTS:
5324 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5325 break;
5326 case SCTP_AUTOCLOSE:
5327 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5328 break;
5329 case SCTP_SOCKOPT_PEELOFF:
5330 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5331 break;
5332 case SCTP_PEER_ADDR_PARAMS:
5333 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5334 optlen);
5335 break;
5336 case SCTP_DELAYED_ACK:
5337 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5338 optlen);
5339 break;
5340 case SCTP_INITMSG:
5341 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5342 break;
5343 case SCTP_GET_PEER_ADDRS:
5344 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5345 optlen);
5346 break;
5347 case SCTP_GET_LOCAL_ADDRS:
5348 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5349 optlen);
5350 break;
5351 case SCTP_SOCKOPT_CONNECTX3:
5352 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5353 break;
5354 case SCTP_DEFAULT_SEND_PARAM:
5355 retval = sctp_getsockopt_default_send_param(sk, len,
5356 optval, optlen);
5357 break;
5358 case SCTP_PRIMARY_ADDR:
5359 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5360 break;
5361 case SCTP_NODELAY:
5362 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5363 break;
5364 case SCTP_RTOINFO:
5365 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5366 break;
5367 case SCTP_ASSOCINFO:
5368 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5369 break;
5370 case SCTP_I_WANT_MAPPED_V4_ADDR:
5371 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5372 break;
5373 case SCTP_MAXSEG:
5374 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5375 break;
5376 case SCTP_GET_PEER_ADDR_INFO:
5377 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5378 optlen);
5379 break;
5380 case SCTP_ADAPTATION_LAYER:
5381 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5382 optlen);
5383 break;
5384 case SCTP_CONTEXT:
5385 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5386 break;
5387 case SCTP_FRAGMENT_INTERLEAVE:
5388 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5389 optlen);
5390 break;
5391 case SCTP_PARTIAL_DELIVERY_POINT:
5392 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5393 optlen);
5394 break;
5395 case SCTP_MAX_BURST:
5396 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5397 break;
5398 case SCTP_AUTH_KEY:
5399 case SCTP_AUTH_CHUNK:
5400 case SCTP_AUTH_DELETE_KEY:
5401 retval = -EOPNOTSUPP;
5402 break;
5403 case SCTP_HMAC_IDENT:
5404 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5405 break;
5406 case SCTP_AUTH_ACTIVE_KEY:
5407 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5408 break;
5409 case SCTP_PEER_AUTH_CHUNKS:
5410 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5411 optlen);
5412 break;
5413 case SCTP_LOCAL_AUTH_CHUNKS:
5414 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5415 optlen);
5416 break;
5417 case SCTP_GET_ASSOC_NUMBER:
5418 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5419 break;
5420 default:
5421 retval = -ENOPROTOOPT;
5422 break;
5425 sctp_release_sock(sk);
5426 return retval;
5429 static void sctp_hash(struct sock *sk)
5431 /* STUB */
5434 static void sctp_unhash(struct sock *sk)
5436 /* STUB */
5439 /* Check if port is acceptable. Possibly find first available port.
5441 * The port hash table (contained in the 'global' SCTP protocol storage
5442 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5443 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5444 * list (the list number is the port number hashed out, so as you
5445 * would expect from a hash function, all the ports in a given list have
5446 * such a number that hashes out to the same list number; you were
5447 * expecting that, right?); so each list has a set of ports, with a
5448 * link to the socket (struct sock) that uses it, the port number and
5449 * a fastreuse flag (FIXME: NPI ipg).
5451 static struct sctp_bind_bucket *sctp_bucket_create(
5452 struct sctp_bind_hashbucket *head, unsigned short snum);
5454 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5456 struct sctp_bind_hashbucket *head; /* hash list */
5457 struct sctp_bind_bucket *pp; /* hash list port iterator */
5458 struct hlist_node *node;
5459 unsigned short snum;
5460 int ret;
5462 snum = ntohs(addr->v4.sin_port);
5464 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5465 sctp_local_bh_disable();
5467 if (snum == 0) {
5468 /* Search for an available port. */
5469 int low, high, remaining, index;
5470 unsigned int rover;
5472 inet_get_local_port_range(&low, &high);
5473 remaining = (high - low) + 1;
5474 rover = net_random() % remaining + low;
5476 do {
5477 rover++;
5478 if ((rover < low) || (rover > high))
5479 rover = low;
5480 if (inet_is_reserved_local_port(rover))
5481 continue;
5482 index = sctp_phashfn(rover);
5483 head = &sctp_port_hashtable[index];
5484 sctp_spin_lock(&head->lock);
5485 sctp_for_each_hentry(pp, node, &head->chain)
5486 if (pp->port == rover)
5487 goto next;
5488 break;
5489 next:
5490 sctp_spin_unlock(&head->lock);
5491 } while (--remaining > 0);
5493 /* Exhausted local port range during search? */
5494 ret = 1;
5495 if (remaining <= 0)
5496 goto fail;
5498 /* OK, here is the one we will use. HEAD (the port
5499 * hash table list entry) is non-NULL and we hold it's
5500 * mutex.
5502 snum = rover;
5503 } else {
5504 /* We are given an specific port number; we verify
5505 * that it is not being used. If it is used, we will
5506 * exahust the search in the hash list corresponding
5507 * to the port number (snum) - we detect that with the
5508 * port iterator, pp being NULL.
5510 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5511 sctp_spin_lock(&head->lock);
5512 sctp_for_each_hentry(pp, node, &head->chain) {
5513 if (pp->port == snum)
5514 goto pp_found;
5517 pp = NULL;
5518 goto pp_not_found;
5519 pp_found:
5520 if (!hlist_empty(&pp->owner)) {
5521 /* We had a port hash table hit - there is an
5522 * available port (pp != NULL) and it is being
5523 * used by other socket (pp->owner not empty); that other
5524 * socket is going to be sk2.
5526 int reuse = sk->sk_reuse;
5527 struct sock *sk2;
5529 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5530 if (pp->fastreuse && sk->sk_reuse &&
5531 sk->sk_state != SCTP_SS_LISTENING)
5532 goto success;
5534 /* Run through the list of sockets bound to the port
5535 * (pp->port) [via the pointers bind_next and
5536 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5537 * we get the endpoint they describe and run through
5538 * the endpoint's list of IP (v4 or v6) addresses,
5539 * comparing each of the addresses with the address of
5540 * the socket sk. If we find a match, then that means
5541 * that this port/socket (sk) combination are already
5542 * in an endpoint.
5544 sk_for_each_bound(sk2, node, &pp->owner) {
5545 struct sctp_endpoint *ep2;
5546 ep2 = sctp_sk(sk2)->ep;
5548 if (sk == sk2 ||
5549 (reuse && sk2->sk_reuse &&
5550 sk2->sk_state != SCTP_SS_LISTENING))
5551 continue;
5553 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5554 sctp_sk(sk2), sctp_sk(sk))) {
5555 ret = (long)sk2;
5556 goto fail_unlock;
5559 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5561 pp_not_found:
5562 /* If there was a hash table miss, create a new port. */
5563 ret = 1;
5564 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5565 goto fail_unlock;
5567 /* In either case (hit or miss), make sure fastreuse is 1 only
5568 * if sk->sk_reuse is too (that is, if the caller requested
5569 * SO_REUSEADDR on this socket -sk-).
5571 if (hlist_empty(&pp->owner)) {
5572 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5573 pp->fastreuse = 1;
5574 else
5575 pp->fastreuse = 0;
5576 } else if (pp->fastreuse &&
5577 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5578 pp->fastreuse = 0;
5580 /* We are set, so fill up all the data in the hash table
5581 * entry, tie the socket list information with the rest of the
5582 * sockets FIXME: Blurry, NPI (ipg).
5584 success:
5585 if (!sctp_sk(sk)->bind_hash) {
5586 inet_sk(sk)->inet_num = snum;
5587 sk_add_bind_node(sk, &pp->owner);
5588 sctp_sk(sk)->bind_hash = pp;
5590 ret = 0;
5592 fail_unlock:
5593 sctp_spin_unlock(&head->lock);
5595 fail:
5596 sctp_local_bh_enable();
5597 return ret;
5600 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5601 * port is requested.
5603 static int sctp_get_port(struct sock *sk, unsigned short snum)
5605 long ret;
5606 union sctp_addr addr;
5607 struct sctp_af *af = sctp_sk(sk)->pf->af;
5609 /* Set up a dummy address struct from the sk. */
5610 af->from_sk(&addr, sk);
5611 addr.v4.sin_port = htons(snum);
5613 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5614 ret = sctp_get_port_local(sk, &addr);
5616 return ret ? 1 : 0;
5620 * Move a socket to LISTENING state.
5622 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5624 struct sctp_sock *sp = sctp_sk(sk);
5625 struct sctp_endpoint *ep = sp->ep;
5626 struct crypto_hash *tfm = NULL;
5628 /* Allocate HMAC for generating cookie. */
5629 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5630 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5631 if (IS_ERR(tfm)) {
5632 if (net_ratelimit()) {
5633 pr_info("failed to load transform for %s: %ld\n",
5634 sctp_hmac_alg, PTR_ERR(tfm));
5636 return -ENOSYS;
5638 sctp_sk(sk)->hmac = tfm;
5642 * If a bind() or sctp_bindx() is not called prior to a listen()
5643 * call that allows new associations to be accepted, the system
5644 * picks an ephemeral port and will choose an address set equivalent
5645 * to binding with a wildcard address.
5647 * This is not currently spelled out in the SCTP sockets
5648 * extensions draft, but follows the practice as seen in TCP
5649 * sockets.
5652 sk->sk_state = SCTP_SS_LISTENING;
5653 if (!ep->base.bind_addr.port) {
5654 if (sctp_autobind(sk))
5655 return -EAGAIN;
5656 } else {
5657 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5658 sk->sk_state = SCTP_SS_CLOSED;
5659 return -EADDRINUSE;
5663 sk->sk_max_ack_backlog = backlog;
5664 sctp_hash_endpoint(ep);
5665 return 0;
5669 * 4.1.3 / 5.1.3 listen()
5671 * By default, new associations are not accepted for UDP style sockets.
5672 * An application uses listen() to mark a socket as being able to
5673 * accept new associations.
5675 * On TCP style sockets, applications use listen() to ready the SCTP
5676 * endpoint for accepting inbound associations.
5678 * On both types of endpoints a backlog of '0' disables listening.
5680 * Move a socket to LISTENING state.
5682 int sctp_inet_listen(struct socket *sock, int backlog)
5684 struct sock *sk = sock->sk;
5685 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5686 int err = -EINVAL;
5688 if (unlikely(backlog < 0))
5689 return err;
5691 sctp_lock_sock(sk);
5693 /* Peeled-off sockets are not allowed to listen(). */
5694 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5695 goto out;
5697 if (sock->state != SS_UNCONNECTED)
5698 goto out;
5700 /* If backlog is zero, disable listening. */
5701 if (!backlog) {
5702 if (sctp_sstate(sk, CLOSED))
5703 goto out;
5705 err = 0;
5706 sctp_unhash_endpoint(ep);
5707 sk->sk_state = SCTP_SS_CLOSED;
5708 if (sk->sk_reuse)
5709 sctp_sk(sk)->bind_hash->fastreuse = 1;
5710 goto out;
5713 /* If we are already listening, just update the backlog */
5714 if (sctp_sstate(sk, LISTENING))
5715 sk->sk_max_ack_backlog = backlog;
5716 else {
5717 err = sctp_listen_start(sk, backlog);
5718 if (err)
5719 goto out;
5722 err = 0;
5723 out:
5724 sctp_release_sock(sk);
5725 return err;
5729 * This function is done by modeling the current datagram_poll() and the
5730 * tcp_poll(). Note that, based on these implementations, we don't
5731 * lock the socket in this function, even though it seems that,
5732 * ideally, locking or some other mechanisms can be used to ensure
5733 * the integrity of the counters (sndbuf and wmem_alloc) used
5734 * in this place. We assume that we don't need locks either until proven
5735 * otherwise.
5737 * Another thing to note is that we include the Async I/O support
5738 * here, again, by modeling the current TCP/UDP code. We don't have
5739 * a good way to test with it yet.
5741 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5743 struct sock *sk = sock->sk;
5744 struct sctp_sock *sp = sctp_sk(sk);
5745 unsigned int mask;
5747 poll_wait(file, sk_sleep(sk), wait);
5749 /* A TCP-style listening socket becomes readable when the accept queue
5750 * is not empty.
5752 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5753 return (!list_empty(&sp->ep->asocs)) ?
5754 (POLLIN | POLLRDNORM) : 0;
5756 mask = 0;
5758 /* Is there any exceptional events? */
5759 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5760 mask |= POLLERR;
5761 if (sk->sk_shutdown & RCV_SHUTDOWN)
5762 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5763 if (sk->sk_shutdown == SHUTDOWN_MASK)
5764 mask |= POLLHUP;
5766 /* Is it readable? Reconsider this code with TCP-style support. */
5767 if (!skb_queue_empty(&sk->sk_receive_queue))
5768 mask |= POLLIN | POLLRDNORM;
5770 /* The association is either gone or not ready. */
5771 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5772 return mask;
5774 /* Is it writable? */
5775 if (sctp_writeable(sk)) {
5776 mask |= POLLOUT | POLLWRNORM;
5777 } else {
5778 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5780 * Since the socket is not locked, the buffer
5781 * might be made available after the writeable check and
5782 * before the bit is set. This could cause a lost I/O
5783 * signal. tcp_poll() has a race breaker for this race
5784 * condition. Based on their implementation, we put
5785 * in the following code to cover it as well.
5787 if (sctp_writeable(sk))
5788 mask |= POLLOUT | POLLWRNORM;
5790 return mask;
5793 /********************************************************************
5794 * 2nd Level Abstractions
5795 ********************************************************************/
5797 static struct sctp_bind_bucket *sctp_bucket_create(
5798 struct sctp_bind_hashbucket *head, unsigned short snum)
5800 struct sctp_bind_bucket *pp;
5802 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5803 if (pp) {
5804 SCTP_DBG_OBJCNT_INC(bind_bucket);
5805 pp->port = snum;
5806 pp->fastreuse = 0;
5807 INIT_HLIST_HEAD(&pp->owner);
5808 hlist_add_head(&pp->node, &head->chain);
5810 return pp;
5813 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5814 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5816 if (pp && hlist_empty(&pp->owner)) {
5817 __hlist_del(&pp->node);
5818 kmem_cache_free(sctp_bucket_cachep, pp);
5819 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5823 /* Release this socket's reference to a local port. */
5824 static inline void __sctp_put_port(struct sock *sk)
5826 struct sctp_bind_hashbucket *head =
5827 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5828 struct sctp_bind_bucket *pp;
5830 sctp_spin_lock(&head->lock);
5831 pp = sctp_sk(sk)->bind_hash;
5832 __sk_del_bind_node(sk);
5833 sctp_sk(sk)->bind_hash = NULL;
5834 inet_sk(sk)->inet_num = 0;
5835 sctp_bucket_destroy(pp);
5836 sctp_spin_unlock(&head->lock);
5839 void sctp_put_port(struct sock *sk)
5841 sctp_local_bh_disable();
5842 __sctp_put_port(sk);
5843 sctp_local_bh_enable();
5847 * The system picks an ephemeral port and choose an address set equivalent
5848 * to binding with a wildcard address.
5849 * One of those addresses will be the primary address for the association.
5850 * This automatically enables the multihoming capability of SCTP.
5852 static int sctp_autobind(struct sock *sk)
5854 union sctp_addr autoaddr;
5855 struct sctp_af *af;
5856 __be16 port;
5858 /* Initialize a local sockaddr structure to INADDR_ANY. */
5859 af = sctp_sk(sk)->pf->af;
5861 port = htons(inet_sk(sk)->inet_num);
5862 af->inaddr_any(&autoaddr, port);
5864 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5867 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5869 * From RFC 2292
5870 * 4.2 The cmsghdr Structure *
5872 * When ancillary data is sent or received, any number of ancillary data
5873 * objects can be specified by the msg_control and msg_controllen members of
5874 * the msghdr structure, because each object is preceded by
5875 * a cmsghdr structure defining the object's length (the cmsg_len member).
5876 * Historically Berkeley-derived implementations have passed only one object
5877 * at a time, but this API allows multiple objects to be
5878 * passed in a single call to sendmsg() or recvmsg(). The following example
5879 * shows two ancillary data objects in a control buffer.
5881 * |<--------------------------- msg_controllen -------------------------->|
5882 * | |
5884 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5886 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5887 * | | |
5889 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5891 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5892 * | | | | |
5894 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5895 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5897 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5899 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5903 * msg_control
5904 * points here
5906 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5907 sctp_cmsgs_t *cmsgs)
5909 struct cmsghdr *cmsg;
5910 struct msghdr *my_msg = (struct msghdr *)msg;
5912 for (cmsg = CMSG_FIRSTHDR(msg);
5913 cmsg != NULL;
5914 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5915 if (!CMSG_OK(my_msg, cmsg))
5916 return -EINVAL;
5918 /* Should we parse this header or ignore? */
5919 if (cmsg->cmsg_level != IPPROTO_SCTP)
5920 continue;
5922 /* Strictly check lengths following example in SCM code. */
5923 switch (cmsg->cmsg_type) {
5924 case SCTP_INIT:
5925 /* SCTP Socket API Extension
5926 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5928 * This cmsghdr structure provides information for
5929 * initializing new SCTP associations with sendmsg().
5930 * The SCTP_INITMSG socket option uses this same data
5931 * structure. This structure is not used for
5932 * recvmsg().
5934 * cmsg_level cmsg_type cmsg_data[]
5935 * ------------ ------------ ----------------------
5936 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5938 if (cmsg->cmsg_len !=
5939 CMSG_LEN(sizeof(struct sctp_initmsg)))
5940 return -EINVAL;
5941 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5942 break;
5944 case SCTP_SNDRCV:
5945 /* SCTP Socket API Extension
5946 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5948 * This cmsghdr structure specifies SCTP options for
5949 * sendmsg() and describes SCTP header information
5950 * about a received message through recvmsg().
5952 * cmsg_level cmsg_type cmsg_data[]
5953 * ------------ ------------ ----------------------
5954 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5956 if (cmsg->cmsg_len !=
5957 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5958 return -EINVAL;
5960 cmsgs->info =
5961 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5963 /* Minimally, validate the sinfo_flags. */
5964 if (cmsgs->info->sinfo_flags &
5965 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5966 SCTP_ABORT | SCTP_EOF))
5967 return -EINVAL;
5968 break;
5970 default:
5971 return -EINVAL;
5974 return 0;
5978 * Wait for a packet..
5979 * Note: This function is the same function as in core/datagram.c
5980 * with a few modifications to make lksctp work.
5982 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5984 int error;
5985 DEFINE_WAIT(wait);
5987 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
5989 /* Socket errors? */
5990 error = sock_error(sk);
5991 if (error)
5992 goto out;
5994 if (!skb_queue_empty(&sk->sk_receive_queue))
5995 goto ready;
5997 /* Socket shut down? */
5998 if (sk->sk_shutdown & RCV_SHUTDOWN)
5999 goto out;
6001 /* Sequenced packets can come disconnected. If so we report the
6002 * problem.
6004 error = -ENOTCONN;
6006 /* Is there a good reason to think that we may receive some data? */
6007 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6008 goto out;
6010 /* Handle signals. */
6011 if (signal_pending(current))
6012 goto interrupted;
6014 /* Let another process have a go. Since we are going to sleep
6015 * anyway. Note: This may cause odd behaviors if the message
6016 * does not fit in the user's buffer, but this seems to be the
6017 * only way to honor MSG_DONTWAIT realistically.
6019 sctp_release_sock(sk);
6020 *timeo_p = schedule_timeout(*timeo_p);
6021 sctp_lock_sock(sk);
6023 ready:
6024 finish_wait(sk_sleep(sk), &wait);
6025 return 0;
6027 interrupted:
6028 error = sock_intr_errno(*timeo_p);
6030 out:
6031 finish_wait(sk_sleep(sk), &wait);
6032 *err = error;
6033 return error;
6036 /* Receive a datagram.
6037 * Note: This is pretty much the same routine as in core/datagram.c
6038 * with a few changes to make lksctp work.
6040 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6041 int noblock, int *err)
6043 int error;
6044 struct sk_buff *skb;
6045 long timeo;
6047 timeo = sock_rcvtimeo(sk, noblock);
6049 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6050 timeo, MAX_SCHEDULE_TIMEOUT);
6052 do {
6053 /* Again only user level code calls this function,
6054 * so nothing interrupt level
6055 * will suddenly eat the receive_queue.
6057 * Look at current nfs client by the way...
6058 * However, this function was corrent in any case. 8)
6060 if (flags & MSG_PEEK) {
6061 spin_lock_bh(&sk->sk_receive_queue.lock);
6062 skb = skb_peek(&sk->sk_receive_queue);
6063 if (skb)
6064 atomic_inc(&skb->users);
6065 spin_unlock_bh(&sk->sk_receive_queue.lock);
6066 } else {
6067 skb = skb_dequeue(&sk->sk_receive_queue);
6070 if (skb)
6071 return skb;
6073 /* Caller is allowed not to check sk->sk_err before calling. */
6074 error = sock_error(sk);
6075 if (error)
6076 goto no_packet;
6078 if (sk->sk_shutdown & RCV_SHUTDOWN)
6079 break;
6081 /* User doesn't want to wait. */
6082 error = -EAGAIN;
6083 if (!timeo)
6084 goto no_packet;
6085 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6087 return NULL;
6089 no_packet:
6090 *err = error;
6091 return NULL;
6094 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6095 static void __sctp_write_space(struct sctp_association *asoc)
6097 struct sock *sk = asoc->base.sk;
6098 struct socket *sock = sk->sk_socket;
6100 if ((sctp_wspace(asoc) > 0) && sock) {
6101 if (waitqueue_active(&asoc->wait))
6102 wake_up_interruptible(&asoc->wait);
6104 if (sctp_writeable(sk)) {
6105 if (sk_sleep(sk) && waitqueue_active(sk_sleep(sk)))
6106 wake_up_interruptible(sk_sleep(sk));
6108 /* Note that we try to include the Async I/O support
6109 * here by modeling from the current TCP/UDP code.
6110 * We have not tested with it yet.
6112 if (sock->wq->fasync_list &&
6113 !(sk->sk_shutdown & SEND_SHUTDOWN))
6114 sock_wake_async(sock,
6115 SOCK_WAKE_SPACE, POLL_OUT);
6120 /* Do accounting for the sndbuf space.
6121 * Decrement the used sndbuf space of the corresponding association by the
6122 * data size which was just transmitted(freed).
6124 static void sctp_wfree(struct sk_buff *skb)
6126 struct sctp_association *asoc;
6127 struct sctp_chunk *chunk;
6128 struct sock *sk;
6130 /* Get the saved chunk pointer. */
6131 chunk = *((struct sctp_chunk **)(skb->cb));
6132 asoc = chunk->asoc;
6133 sk = asoc->base.sk;
6134 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6135 sizeof(struct sk_buff) +
6136 sizeof(struct sctp_chunk);
6138 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6141 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6143 sk->sk_wmem_queued -= skb->truesize;
6144 sk_mem_uncharge(sk, skb->truesize);
6146 sock_wfree(skb);
6147 __sctp_write_space(asoc);
6149 sctp_association_put(asoc);
6152 /* Do accounting for the receive space on the socket.
6153 * Accounting for the association is done in ulpevent.c
6154 * We set this as a destructor for the cloned data skbs so that
6155 * accounting is done at the correct time.
6157 void sctp_sock_rfree(struct sk_buff *skb)
6159 struct sock *sk = skb->sk;
6160 struct sctp_ulpevent *event = sctp_skb2event(skb);
6162 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6165 * Mimic the behavior of sock_rfree
6167 sk_mem_uncharge(sk, event->rmem_len);
6171 /* Helper function to wait for space in the sndbuf. */
6172 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6173 size_t msg_len)
6175 struct sock *sk = asoc->base.sk;
6176 int err = 0;
6177 long current_timeo = *timeo_p;
6178 DEFINE_WAIT(wait);
6180 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6181 asoc, (long)(*timeo_p), msg_len);
6183 /* Increment the association's refcnt. */
6184 sctp_association_hold(asoc);
6186 /* Wait on the association specific sndbuf space. */
6187 for (;;) {
6188 prepare_to_wait_exclusive(&asoc->wait, &wait,
6189 TASK_INTERRUPTIBLE);
6190 if (!*timeo_p)
6191 goto do_nonblock;
6192 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6193 asoc->base.dead)
6194 goto do_error;
6195 if (signal_pending(current))
6196 goto do_interrupted;
6197 if (msg_len <= sctp_wspace(asoc))
6198 break;
6200 /* Let another process have a go. Since we are going
6201 * to sleep anyway.
6203 sctp_release_sock(sk);
6204 current_timeo = schedule_timeout(current_timeo);
6205 BUG_ON(sk != asoc->base.sk);
6206 sctp_lock_sock(sk);
6208 *timeo_p = current_timeo;
6211 out:
6212 finish_wait(&asoc->wait, &wait);
6214 /* Release the association's refcnt. */
6215 sctp_association_put(asoc);
6217 return err;
6219 do_error:
6220 err = -EPIPE;
6221 goto out;
6223 do_interrupted:
6224 err = sock_intr_errno(*timeo_p);
6225 goto out;
6227 do_nonblock:
6228 err = -EAGAIN;
6229 goto out;
6232 void sctp_data_ready(struct sock *sk, int len)
6234 struct socket_wq *wq;
6236 rcu_read_lock();
6237 wq = rcu_dereference(sk->sk_wq);
6238 if (wq_has_sleeper(wq))
6239 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6240 POLLRDNORM | POLLRDBAND);
6241 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6242 rcu_read_unlock();
6245 /* If socket sndbuf has changed, wake up all per association waiters. */
6246 void sctp_write_space(struct sock *sk)
6248 struct sctp_association *asoc;
6250 /* Wake up the tasks in each wait queue. */
6251 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6252 __sctp_write_space(asoc);
6256 /* Is there any sndbuf space available on the socket?
6258 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6259 * associations on the same socket. For a UDP-style socket with
6260 * multiple associations, it is possible for it to be "unwriteable"
6261 * prematurely. I assume that this is acceptable because
6262 * a premature "unwriteable" is better than an accidental "writeable" which
6263 * would cause an unwanted block under certain circumstances. For the 1-1
6264 * UDP-style sockets or TCP-style sockets, this code should work.
6265 * - Daisy
6267 static int sctp_writeable(struct sock *sk)
6269 int amt = 0;
6271 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6272 if (amt < 0)
6273 amt = 0;
6274 return amt;
6277 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6278 * returns immediately with EINPROGRESS.
6280 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6282 struct sock *sk = asoc->base.sk;
6283 int err = 0;
6284 long current_timeo = *timeo_p;
6285 DEFINE_WAIT(wait);
6287 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6288 (long)(*timeo_p));
6290 /* Increment the association's refcnt. */
6291 sctp_association_hold(asoc);
6293 for (;;) {
6294 prepare_to_wait_exclusive(&asoc->wait, &wait,
6295 TASK_INTERRUPTIBLE);
6296 if (!*timeo_p)
6297 goto do_nonblock;
6298 if (sk->sk_shutdown & RCV_SHUTDOWN)
6299 break;
6300 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6301 asoc->base.dead)
6302 goto do_error;
6303 if (signal_pending(current))
6304 goto do_interrupted;
6306 if (sctp_state(asoc, ESTABLISHED))
6307 break;
6309 /* Let another process have a go. Since we are going
6310 * to sleep anyway.
6312 sctp_release_sock(sk);
6313 current_timeo = schedule_timeout(current_timeo);
6314 sctp_lock_sock(sk);
6316 *timeo_p = current_timeo;
6319 out:
6320 finish_wait(&asoc->wait, &wait);
6322 /* Release the association's refcnt. */
6323 sctp_association_put(asoc);
6325 return err;
6327 do_error:
6328 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6329 err = -ETIMEDOUT;
6330 else
6331 err = -ECONNREFUSED;
6332 goto out;
6334 do_interrupted:
6335 err = sock_intr_errno(*timeo_p);
6336 goto out;
6338 do_nonblock:
6339 err = -EINPROGRESS;
6340 goto out;
6343 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6345 struct sctp_endpoint *ep;
6346 int err = 0;
6347 DEFINE_WAIT(wait);
6349 ep = sctp_sk(sk)->ep;
6352 for (;;) {
6353 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6354 TASK_INTERRUPTIBLE);
6356 if (list_empty(&ep->asocs)) {
6357 sctp_release_sock(sk);
6358 timeo = schedule_timeout(timeo);
6359 sctp_lock_sock(sk);
6362 err = -EINVAL;
6363 if (!sctp_sstate(sk, LISTENING))
6364 break;
6366 err = 0;
6367 if (!list_empty(&ep->asocs))
6368 break;
6370 err = sock_intr_errno(timeo);
6371 if (signal_pending(current))
6372 break;
6374 err = -EAGAIN;
6375 if (!timeo)
6376 break;
6379 finish_wait(sk_sleep(sk), &wait);
6381 return err;
6384 static void sctp_wait_for_close(struct sock *sk, long timeout)
6386 DEFINE_WAIT(wait);
6388 do {
6389 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6390 if (list_empty(&sctp_sk(sk)->ep->asocs))
6391 break;
6392 sctp_release_sock(sk);
6393 timeout = schedule_timeout(timeout);
6394 sctp_lock_sock(sk);
6395 } while (!signal_pending(current) && timeout);
6397 finish_wait(sk_sleep(sk), &wait);
6400 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6402 struct sk_buff *frag;
6404 if (!skb->data_len)
6405 goto done;
6407 /* Don't forget the fragments. */
6408 skb_walk_frags(skb, frag)
6409 sctp_skb_set_owner_r_frag(frag, sk);
6411 done:
6412 sctp_skb_set_owner_r(skb, sk);
6415 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6416 struct sctp_association *asoc)
6418 struct inet_sock *inet = inet_sk(sk);
6419 struct inet_sock *newinet;
6421 newsk->sk_type = sk->sk_type;
6422 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6423 newsk->sk_flags = sk->sk_flags;
6424 newsk->sk_no_check = sk->sk_no_check;
6425 newsk->sk_reuse = sk->sk_reuse;
6427 newsk->sk_shutdown = sk->sk_shutdown;
6428 newsk->sk_destruct = inet_sock_destruct;
6429 newsk->sk_family = sk->sk_family;
6430 newsk->sk_protocol = IPPROTO_SCTP;
6431 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6432 newsk->sk_sndbuf = sk->sk_sndbuf;
6433 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6434 newsk->sk_lingertime = sk->sk_lingertime;
6435 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6436 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6438 newinet = inet_sk(newsk);
6440 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6441 * getsockname() and getpeername()
6443 newinet->inet_sport = inet->inet_sport;
6444 newinet->inet_saddr = inet->inet_saddr;
6445 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6446 newinet->inet_dport = htons(asoc->peer.port);
6447 newinet->pmtudisc = inet->pmtudisc;
6448 newinet->inet_id = asoc->next_tsn ^ jiffies;
6450 newinet->uc_ttl = inet->uc_ttl;
6451 newinet->mc_loop = 1;
6452 newinet->mc_ttl = 1;
6453 newinet->mc_index = 0;
6454 newinet->mc_list = NULL;
6457 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6458 * and its messages to the newsk.
6460 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6461 struct sctp_association *assoc,
6462 sctp_socket_type_t type)
6464 struct sctp_sock *oldsp = sctp_sk(oldsk);
6465 struct sctp_sock *newsp = sctp_sk(newsk);
6466 struct sctp_bind_bucket *pp; /* hash list port iterator */
6467 struct sctp_endpoint *newep = newsp->ep;
6468 struct sk_buff *skb, *tmp;
6469 struct sctp_ulpevent *event;
6470 struct sctp_bind_hashbucket *head;
6472 /* Migrate socket buffer sizes and all the socket level options to the
6473 * new socket.
6475 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6476 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6477 /* Brute force copy old sctp opt. */
6478 inet_sk_copy_descendant(newsk, oldsk);
6480 /* Restore the ep value that was overwritten with the above structure
6481 * copy.
6483 newsp->ep = newep;
6484 newsp->hmac = NULL;
6486 /* Hook this new socket in to the bind_hash list. */
6487 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6488 sctp_local_bh_disable();
6489 sctp_spin_lock(&head->lock);
6490 pp = sctp_sk(oldsk)->bind_hash;
6491 sk_add_bind_node(newsk, &pp->owner);
6492 sctp_sk(newsk)->bind_hash = pp;
6493 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6494 sctp_spin_unlock(&head->lock);
6495 sctp_local_bh_enable();
6497 /* Copy the bind_addr list from the original endpoint to the new
6498 * endpoint so that we can handle restarts properly
6500 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6501 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6503 /* Move any messages in the old socket's receive queue that are for the
6504 * peeled off association to the new socket's receive queue.
6506 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6507 event = sctp_skb2event(skb);
6508 if (event->asoc == assoc) {
6509 __skb_unlink(skb, &oldsk->sk_receive_queue);
6510 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6511 sctp_skb_set_owner_r_frag(skb, newsk);
6515 /* Clean up any messages pending delivery due to partial
6516 * delivery. Three cases:
6517 * 1) No partial deliver; no work.
6518 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6519 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6521 skb_queue_head_init(&newsp->pd_lobby);
6522 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6524 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6525 struct sk_buff_head *queue;
6527 /* Decide which queue to move pd_lobby skbs to. */
6528 if (assoc->ulpq.pd_mode) {
6529 queue = &newsp->pd_lobby;
6530 } else
6531 queue = &newsk->sk_receive_queue;
6533 /* Walk through the pd_lobby, looking for skbs that
6534 * need moved to the new socket.
6536 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6537 event = sctp_skb2event(skb);
6538 if (event->asoc == assoc) {
6539 __skb_unlink(skb, &oldsp->pd_lobby);
6540 __skb_queue_tail(queue, skb);
6541 sctp_skb_set_owner_r_frag(skb, newsk);
6545 /* Clear up any skbs waiting for the partial
6546 * delivery to finish.
6548 if (assoc->ulpq.pd_mode)
6549 sctp_clear_pd(oldsk, NULL);
6553 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6554 sctp_skb_set_owner_r_frag(skb, newsk);
6556 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6557 sctp_skb_set_owner_r_frag(skb, newsk);
6559 /* Set the type of socket to indicate that it is peeled off from the
6560 * original UDP-style socket or created with the accept() call on a
6561 * TCP-style socket..
6563 newsp->type = type;
6565 /* Mark the new socket "in-use" by the user so that any packets
6566 * that may arrive on the association after we've moved it are
6567 * queued to the backlog. This prevents a potential race between
6568 * backlog processing on the old socket and new-packet processing
6569 * on the new socket.
6571 * The caller has just allocated newsk so we can guarantee that other
6572 * paths won't try to lock it and then oldsk.
6574 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6575 sctp_assoc_migrate(assoc, newsk);
6577 /* If the association on the newsk is already closed before accept()
6578 * is called, set RCV_SHUTDOWN flag.
6580 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6581 newsk->sk_shutdown |= RCV_SHUTDOWN;
6583 newsk->sk_state = SCTP_SS_ESTABLISHED;
6584 sctp_release_sock(newsk);
6588 /* This proto struct describes the ULP interface for SCTP. */
6589 struct proto sctp_prot = {
6590 .name = "SCTP",
6591 .owner = THIS_MODULE,
6592 .close = sctp_close,
6593 .connect = sctp_connect,
6594 .disconnect = sctp_disconnect,
6595 .accept = sctp_accept,
6596 .ioctl = sctp_ioctl,
6597 .init = sctp_init_sock,
6598 .destroy = sctp_destroy_sock,
6599 .shutdown = sctp_shutdown,
6600 .setsockopt = sctp_setsockopt,
6601 .getsockopt = sctp_getsockopt,
6602 .sendmsg = sctp_sendmsg,
6603 .recvmsg = sctp_recvmsg,
6604 .bind = sctp_bind,
6605 .backlog_rcv = sctp_backlog_rcv,
6606 .hash = sctp_hash,
6607 .unhash = sctp_unhash,
6608 .get_port = sctp_get_port,
6609 .obj_size = sizeof(struct sctp_sock),
6610 .sysctl_mem = sysctl_sctp_mem,
6611 .sysctl_rmem = sysctl_sctp_rmem,
6612 .sysctl_wmem = sysctl_sctp_wmem,
6613 .memory_pressure = &sctp_memory_pressure,
6614 .enter_memory_pressure = sctp_enter_memory_pressure,
6615 .memory_allocated = &sctp_memory_allocated,
6616 .sockets_allocated = &sctp_sockets_allocated,
6619 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6621 struct proto sctpv6_prot = {
6622 .name = "SCTPv6",
6623 .owner = THIS_MODULE,
6624 .close = sctp_close,
6625 .connect = sctp_connect,
6626 .disconnect = sctp_disconnect,
6627 .accept = sctp_accept,
6628 .ioctl = sctp_ioctl,
6629 .init = sctp_init_sock,
6630 .destroy = sctp_destroy_sock,
6631 .shutdown = sctp_shutdown,
6632 .setsockopt = sctp_setsockopt,
6633 .getsockopt = sctp_getsockopt,
6634 .sendmsg = sctp_sendmsg,
6635 .recvmsg = sctp_recvmsg,
6636 .bind = sctp_bind,
6637 .backlog_rcv = sctp_backlog_rcv,
6638 .hash = sctp_hash,
6639 .unhash = sctp_unhash,
6640 .get_port = sctp_get_port,
6641 .obj_size = sizeof(struct sctp6_sock),
6642 .sysctl_mem = sysctl_sctp_mem,
6643 .sysctl_rmem = sysctl_sctp_rmem,
6644 .sysctl_wmem = sysctl_sctp_wmem,
6645 .memory_pressure = &sctp_memory_pressure,
6646 .enter_memory_pressure = sctp_enter_memory_pressure,
6647 .memory_allocated = &sctp_memory_allocated,
6648 .sockets_allocated = &sctp_sockets_allocated,
6650 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */