vfs: get rid of insane dentry hashing rules
[linux-2.6/cjktty.git] / net / sctp / socket.c
blobdeb82e35a107e89b9a2328445c46bc8935a143d6
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 dependent. 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_association *asoc = NULL;
2932 struct sctp_setpeerprim prim;
2933 struct sctp_chunk *chunk;
2934 struct sctp_af *af;
2935 int err;
2937 sp = sctp_sk(sk);
2939 if (!sctp_addip_enable)
2940 return -EPERM;
2942 if (optlen != sizeof(struct sctp_setpeerprim))
2943 return -EINVAL;
2945 if (copy_from_user(&prim, optval, optlen))
2946 return -EFAULT;
2948 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
2949 if (!asoc)
2950 return -EINVAL;
2952 if (!asoc->peer.asconf_capable)
2953 return -EPERM;
2955 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
2956 return -EPERM;
2958 if (!sctp_state(asoc, ESTABLISHED))
2959 return -ENOTCONN;
2961 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
2962 if (!af)
2963 return -EINVAL;
2965 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
2966 return -EADDRNOTAVAIL;
2968 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
2969 return -EADDRNOTAVAIL;
2971 /* Create an ASCONF chunk with SET_PRIMARY parameter */
2972 chunk = sctp_make_asconf_set_prim(asoc,
2973 (union sctp_addr *)&prim.sspp_addr);
2974 if (!chunk)
2975 return -ENOMEM;
2977 err = sctp_send_asconf(asoc, chunk);
2979 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
2981 return err;
2984 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
2985 unsigned int optlen)
2987 struct sctp_setadaptation adaptation;
2989 if (optlen != sizeof(struct sctp_setadaptation))
2990 return -EINVAL;
2991 if (copy_from_user(&adaptation, optval, optlen))
2992 return -EFAULT;
2994 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
2996 return 0;
3000 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3002 * The context field in the sctp_sndrcvinfo structure is normally only
3003 * used when a failed message is retrieved holding the value that was
3004 * sent down on the actual send call. This option allows the setting of
3005 * a default context on an association basis that will be received on
3006 * reading messages from the peer. This is especially helpful in the
3007 * one-2-many model for an application to keep some reference to an
3008 * internal state machine that is processing messages on the
3009 * association. Note that the setting of this value only effects
3010 * received messages from the peer and does not effect the value that is
3011 * saved with outbound messages.
3013 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3014 unsigned int optlen)
3016 struct sctp_assoc_value params;
3017 struct sctp_sock *sp;
3018 struct sctp_association *asoc;
3020 if (optlen != sizeof(struct sctp_assoc_value))
3021 return -EINVAL;
3022 if (copy_from_user(&params, optval, optlen))
3023 return -EFAULT;
3025 sp = sctp_sk(sk);
3027 if (params.assoc_id != 0) {
3028 asoc = sctp_id2assoc(sk, params.assoc_id);
3029 if (!asoc)
3030 return -EINVAL;
3031 asoc->default_rcv_context = params.assoc_value;
3032 } else {
3033 sp->default_rcv_context = params.assoc_value;
3036 return 0;
3040 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3042 * This options will at a minimum specify if the implementation is doing
3043 * fragmented interleave. Fragmented interleave, for a one to many
3044 * socket, is when subsequent calls to receive a message may return
3045 * parts of messages from different associations. Some implementations
3046 * may allow you to turn this value on or off. If so, when turned off,
3047 * no fragment interleave will occur (which will cause a head of line
3048 * blocking amongst multiple associations sharing the same one to many
3049 * socket). When this option is turned on, then each receive call may
3050 * come from a different association (thus the user must receive data
3051 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3052 * association each receive belongs to.
3054 * This option takes a boolean value. A non-zero value indicates that
3055 * fragmented interleave is on. A value of zero indicates that
3056 * fragmented interleave is off.
3058 * Note that it is important that an implementation that allows this
3059 * option to be turned on, have it off by default. Otherwise an unaware
3060 * application using the one to many model may become confused and act
3061 * incorrectly.
3063 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3064 char __user *optval,
3065 unsigned int optlen)
3067 int val;
3069 if (optlen != sizeof(int))
3070 return -EINVAL;
3071 if (get_user(val, (int __user *)optval))
3072 return -EFAULT;
3074 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3076 return 0;
3080 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3081 * (SCTP_PARTIAL_DELIVERY_POINT)
3083 * This option will set or get the SCTP partial delivery point. This
3084 * point is the size of a message where the partial delivery API will be
3085 * invoked to help free up rwnd space for the peer. Setting this to a
3086 * lower value will cause partial deliveries to happen more often. The
3087 * calls argument is an integer that sets or gets the partial delivery
3088 * point. Note also that the call will fail if the user attempts to set
3089 * this value larger than the socket receive buffer size.
3091 * Note that any single message having a length smaller than or equal to
3092 * the SCTP partial delivery point will be delivered in one single read
3093 * call as long as the user provided buffer is large enough to hold the
3094 * message.
3096 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3097 char __user *optval,
3098 unsigned int optlen)
3100 u32 val;
3102 if (optlen != sizeof(u32))
3103 return -EINVAL;
3104 if (get_user(val, (int __user *)optval))
3105 return -EFAULT;
3107 /* Note: We double the receive buffer from what the user sets
3108 * it to be, also initial rwnd is based on rcvbuf/2.
3110 if (val > (sk->sk_rcvbuf >> 1))
3111 return -EINVAL;
3113 sctp_sk(sk)->pd_point = val;
3115 return 0; /* is this the right error code? */
3119 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3121 * This option will allow a user to change the maximum burst of packets
3122 * that can be emitted by this association. Note that the default value
3123 * is 4, and some implementations may restrict this setting so that it
3124 * can only be lowered.
3126 * NOTE: This text doesn't seem right. Do this on a socket basis with
3127 * future associations inheriting the socket value.
3129 static int sctp_setsockopt_maxburst(struct sock *sk,
3130 char __user *optval,
3131 unsigned int optlen)
3133 struct sctp_assoc_value params;
3134 struct sctp_sock *sp;
3135 struct sctp_association *asoc;
3136 int val;
3137 int assoc_id = 0;
3139 if (optlen == sizeof(int)) {
3140 pr_warn("Use of int in max_burst socket option deprecated\n");
3141 pr_warn("Use struct sctp_assoc_value instead\n");
3142 if (copy_from_user(&val, optval, optlen))
3143 return -EFAULT;
3144 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3145 if (copy_from_user(&params, optval, optlen))
3146 return -EFAULT;
3147 val = params.assoc_value;
3148 assoc_id = params.assoc_id;
3149 } else
3150 return -EINVAL;
3152 sp = sctp_sk(sk);
3154 if (assoc_id != 0) {
3155 asoc = sctp_id2assoc(sk, assoc_id);
3156 if (!asoc)
3157 return -EINVAL;
3158 asoc->max_burst = val;
3159 } else
3160 sp->max_burst = val;
3162 return 0;
3166 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3168 * This set option adds a chunk type that the user is requesting to be
3169 * received only in an authenticated way. Changes to the list of chunks
3170 * will only effect future associations on the socket.
3172 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3173 char __user *optval,
3174 unsigned int optlen)
3176 struct sctp_authchunk val;
3178 if (!sctp_auth_enable)
3179 return -EACCES;
3181 if (optlen != sizeof(struct sctp_authchunk))
3182 return -EINVAL;
3183 if (copy_from_user(&val, optval, optlen))
3184 return -EFAULT;
3186 switch (val.sauth_chunk) {
3187 case SCTP_CID_INIT:
3188 case SCTP_CID_INIT_ACK:
3189 case SCTP_CID_SHUTDOWN_COMPLETE:
3190 case SCTP_CID_AUTH:
3191 return -EINVAL;
3194 /* add this chunk id to the endpoint */
3195 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3199 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3201 * This option gets or sets the list of HMAC algorithms that the local
3202 * endpoint requires the peer to use.
3204 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3205 char __user *optval,
3206 unsigned int optlen)
3208 struct sctp_hmacalgo *hmacs;
3209 u32 idents;
3210 int err;
3212 if (!sctp_auth_enable)
3213 return -EACCES;
3215 if (optlen < sizeof(struct sctp_hmacalgo))
3216 return -EINVAL;
3218 hmacs = kmalloc(optlen, GFP_KERNEL);
3219 if (!hmacs)
3220 return -ENOMEM;
3222 if (copy_from_user(hmacs, optval, optlen)) {
3223 err = -EFAULT;
3224 goto out;
3227 idents = hmacs->shmac_num_idents;
3228 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3229 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3230 err = -EINVAL;
3231 goto out;
3234 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3235 out:
3236 kfree(hmacs);
3237 return err;
3241 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3243 * This option will set a shared secret key which is used to build an
3244 * association shared key.
3246 static int sctp_setsockopt_auth_key(struct sock *sk,
3247 char __user *optval,
3248 unsigned int optlen)
3250 struct sctp_authkey *authkey;
3251 struct sctp_association *asoc;
3252 int ret;
3254 if (!sctp_auth_enable)
3255 return -EACCES;
3257 if (optlen <= sizeof(struct sctp_authkey))
3258 return -EINVAL;
3260 authkey = kmalloc(optlen, GFP_KERNEL);
3261 if (!authkey)
3262 return -ENOMEM;
3264 if (copy_from_user(authkey, optval, optlen)) {
3265 ret = -EFAULT;
3266 goto out;
3269 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3270 ret = -EINVAL;
3271 goto out;
3274 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3275 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3276 ret = -EINVAL;
3277 goto out;
3280 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3281 out:
3282 kfree(authkey);
3283 return ret;
3287 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3289 * This option will get or set the active shared key to be used to build
3290 * the association shared key.
3292 static int sctp_setsockopt_active_key(struct sock *sk,
3293 char __user *optval,
3294 unsigned int optlen)
3296 struct sctp_authkeyid val;
3297 struct sctp_association *asoc;
3299 if (!sctp_auth_enable)
3300 return -EACCES;
3302 if (optlen != sizeof(struct sctp_authkeyid))
3303 return -EINVAL;
3304 if (copy_from_user(&val, optval, optlen))
3305 return -EFAULT;
3307 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3308 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3309 return -EINVAL;
3311 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3312 val.scact_keynumber);
3316 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3318 * This set option will delete a shared secret key from use.
3320 static int sctp_setsockopt_del_key(struct sock *sk,
3321 char __user *optval,
3322 unsigned int optlen)
3324 struct sctp_authkeyid val;
3325 struct sctp_association *asoc;
3327 if (!sctp_auth_enable)
3328 return -EACCES;
3330 if (optlen != sizeof(struct sctp_authkeyid))
3331 return -EINVAL;
3332 if (copy_from_user(&val, optval, optlen))
3333 return -EFAULT;
3335 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3336 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3337 return -EINVAL;
3339 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3340 val.scact_keynumber);
3345 /* API 6.2 setsockopt(), getsockopt()
3347 * Applications use setsockopt() and getsockopt() to set or retrieve
3348 * socket options. Socket options are used to change the default
3349 * behavior of sockets calls. They are described in Section 7.
3351 * The syntax is:
3353 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3354 * int __user *optlen);
3355 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3356 * int optlen);
3358 * sd - the socket descript.
3359 * level - set to IPPROTO_SCTP for all SCTP options.
3360 * optname - the option name.
3361 * optval - the buffer to store the value of the option.
3362 * optlen - the size of the buffer.
3364 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3365 char __user *optval, unsigned int optlen)
3367 int retval = 0;
3369 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3370 sk, optname);
3372 /* I can hardly begin to describe how wrong this is. This is
3373 * so broken as to be worse than useless. The API draft
3374 * REALLY is NOT helpful here... I am not convinced that the
3375 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3376 * are at all well-founded.
3378 if (level != SOL_SCTP) {
3379 struct sctp_af *af = sctp_sk(sk)->pf->af;
3380 retval = af->setsockopt(sk, level, optname, optval, optlen);
3381 goto out_nounlock;
3384 sctp_lock_sock(sk);
3386 switch (optname) {
3387 case SCTP_SOCKOPT_BINDX_ADD:
3388 /* 'optlen' is the size of the addresses buffer. */
3389 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3390 optlen, SCTP_BINDX_ADD_ADDR);
3391 break;
3393 case SCTP_SOCKOPT_BINDX_REM:
3394 /* 'optlen' is the size of the addresses buffer. */
3395 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3396 optlen, SCTP_BINDX_REM_ADDR);
3397 break;
3399 case SCTP_SOCKOPT_CONNECTX_OLD:
3400 /* 'optlen' is the size of the addresses buffer. */
3401 retval = sctp_setsockopt_connectx_old(sk,
3402 (struct sockaddr __user *)optval,
3403 optlen);
3404 break;
3406 case SCTP_SOCKOPT_CONNECTX:
3407 /* 'optlen' is the size of the addresses buffer. */
3408 retval = sctp_setsockopt_connectx(sk,
3409 (struct sockaddr __user *)optval,
3410 optlen);
3411 break;
3413 case SCTP_DISABLE_FRAGMENTS:
3414 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3415 break;
3417 case SCTP_EVENTS:
3418 retval = sctp_setsockopt_events(sk, optval, optlen);
3419 break;
3421 case SCTP_AUTOCLOSE:
3422 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3423 break;
3425 case SCTP_PEER_ADDR_PARAMS:
3426 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3427 break;
3429 case SCTP_DELAYED_SACK:
3430 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3431 break;
3432 case SCTP_PARTIAL_DELIVERY_POINT:
3433 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3434 break;
3436 case SCTP_INITMSG:
3437 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3438 break;
3439 case SCTP_DEFAULT_SEND_PARAM:
3440 retval = sctp_setsockopt_default_send_param(sk, optval,
3441 optlen);
3442 break;
3443 case SCTP_PRIMARY_ADDR:
3444 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3445 break;
3446 case SCTP_SET_PEER_PRIMARY_ADDR:
3447 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3448 break;
3449 case SCTP_NODELAY:
3450 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3451 break;
3452 case SCTP_RTOINFO:
3453 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3454 break;
3455 case SCTP_ASSOCINFO:
3456 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3457 break;
3458 case SCTP_I_WANT_MAPPED_V4_ADDR:
3459 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3460 break;
3461 case SCTP_MAXSEG:
3462 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3463 break;
3464 case SCTP_ADAPTATION_LAYER:
3465 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3466 break;
3467 case SCTP_CONTEXT:
3468 retval = sctp_setsockopt_context(sk, optval, optlen);
3469 break;
3470 case SCTP_FRAGMENT_INTERLEAVE:
3471 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3472 break;
3473 case SCTP_MAX_BURST:
3474 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3475 break;
3476 case SCTP_AUTH_CHUNK:
3477 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3478 break;
3479 case SCTP_HMAC_IDENT:
3480 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3481 break;
3482 case SCTP_AUTH_KEY:
3483 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3484 break;
3485 case SCTP_AUTH_ACTIVE_KEY:
3486 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3487 break;
3488 case SCTP_AUTH_DELETE_KEY:
3489 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3490 break;
3491 default:
3492 retval = -ENOPROTOOPT;
3493 break;
3496 sctp_release_sock(sk);
3498 out_nounlock:
3499 return retval;
3502 /* API 3.1.6 connect() - UDP Style Syntax
3504 * An application may use the connect() call in the UDP model to initiate an
3505 * association without sending data.
3507 * The syntax is:
3509 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3511 * sd: the socket descriptor to have a new association added to.
3513 * nam: the address structure (either struct sockaddr_in or struct
3514 * sockaddr_in6 defined in RFC2553 [7]).
3516 * len: the size of the address.
3518 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3519 int addr_len)
3521 int err = 0;
3522 struct sctp_af *af;
3524 sctp_lock_sock(sk);
3526 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3527 __func__, sk, addr, addr_len);
3529 /* Validate addr_len before calling common connect/connectx routine. */
3530 af = sctp_get_af_specific(addr->sa_family);
3531 if (!af || addr_len < af->sockaddr_len) {
3532 err = -EINVAL;
3533 } else {
3534 /* Pass correct addr len to common routine (so it knows there
3535 * is only one address being passed.
3537 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3540 sctp_release_sock(sk);
3541 return err;
3544 /* FIXME: Write comments. */
3545 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3547 return -EOPNOTSUPP; /* STUB */
3550 /* 4.1.4 accept() - TCP Style Syntax
3552 * Applications use accept() call to remove an established SCTP
3553 * association from the accept queue of the endpoint. A new socket
3554 * descriptor will be returned from accept() to represent the newly
3555 * formed association.
3557 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3559 struct sctp_sock *sp;
3560 struct sctp_endpoint *ep;
3561 struct sock *newsk = NULL;
3562 struct sctp_association *asoc;
3563 long timeo;
3564 int error = 0;
3566 sctp_lock_sock(sk);
3568 sp = sctp_sk(sk);
3569 ep = sp->ep;
3571 if (!sctp_style(sk, TCP)) {
3572 error = -EOPNOTSUPP;
3573 goto out;
3576 if (!sctp_sstate(sk, LISTENING)) {
3577 error = -EINVAL;
3578 goto out;
3581 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3583 error = sctp_wait_for_accept(sk, timeo);
3584 if (error)
3585 goto out;
3587 /* We treat the list of associations on the endpoint as the accept
3588 * queue and pick the first association on the list.
3590 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3592 newsk = sp->pf->create_accept_sk(sk, asoc);
3593 if (!newsk) {
3594 error = -ENOMEM;
3595 goto out;
3598 /* Populate the fields of the newsk from the oldsk and migrate the
3599 * asoc to the newsk.
3601 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3603 out:
3604 sctp_release_sock(sk);
3605 *err = error;
3606 return newsk;
3609 /* The SCTP ioctl handler. */
3610 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3612 int rc = -ENOTCONN;
3614 sctp_lock_sock(sk);
3617 * SEQPACKET-style sockets in LISTENING state are valid, for
3618 * SCTP, so only discard TCP-style sockets in LISTENING state.
3620 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3621 goto out;
3623 switch (cmd) {
3624 case SIOCINQ: {
3625 struct sk_buff *skb;
3626 unsigned int amount = 0;
3628 skb = skb_peek(&sk->sk_receive_queue);
3629 if (skb != NULL) {
3631 * We will only return the amount of this packet since
3632 * that is all that will be read.
3634 amount = skb->len;
3636 rc = put_user(amount, (int __user *)arg);
3637 break;
3639 default:
3640 rc = -ENOIOCTLCMD;
3641 break;
3643 out:
3644 sctp_release_sock(sk);
3645 return rc;
3648 /* This is the function which gets called during socket creation to
3649 * initialized the SCTP-specific portion of the sock.
3650 * The sock structure should already be zero-filled memory.
3652 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3654 struct sctp_endpoint *ep;
3655 struct sctp_sock *sp;
3657 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3659 sp = sctp_sk(sk);
3661 /* Initialize the SCTP per socket area. */
3662 switch (sk->sk_type) {
3663 case SOCK_SEQPACKET:
3664 sp->type = SCTP_SOCKET_UDP;
3665 break;
3666 case SOCK_STREAM:
3667 sp->type = SCTP_SOCKET_TCP;
3668 break;
3669 default:
3670 return -ESOCKTNOSUPPORT;
3673 /* Initialize default send parameters. These parameters can be
3674 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3676 sp->default_stream = 0;
3677 sp->default_ppid = 0;
3678 sp->default_flags = 0;
3679 sp->default_context = 0;
3680 sp->default_timetolive = 0;
3682 sp->default_rcv_context = 0;
3683 sp->max_burst = sctp_max_burst;
3685 /* Initialize default setup parameters. These parameters
3686 * can be modified with the SCTP_INITMSG socket option or
3687 * overridden by the SCTP_INIT CMSG.
3689 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3690 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3691 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3692 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3694 /* Initialize default RTO related parameters. These parameters can
3695 * be modified for with the SCTP_RTOINFO socket option.
3697 sp->rtoinfo.srto_initial = sctp_rto_initial;
3698 sp->rtoinfo.srto_max = sctp_rto_max;
3699 sp->rtoinfo.srto_min = sctp_rto_min;
3701 /* Initialize default association related parameters. These parameters
3702 * can be modified with the SCTP_ASSOCINFO socket option.
3704 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3705 sp->assocparams.sasoc_number_peer_destinations = 0;
3706 sp->assocparams.sasoc_peer_rwnd = 0;
3707 sp->assocparams.sasoc_local_rwnd = 0;
3708 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3710 /* Initialize default event subscriptions. By default, all the
3711 * options are off.
3713 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3715 /* Default Peer Address Parameters. These defaults can
3716 * be modified via SCTP_PEER_ADDR_PARAMS
3718 sp->hbinterval = sctp_hb_interval;
3719 sp->pathmaxrxt = sctp_max_retrans_path;
3720 sp->pathmtu = 0; // allow default discovery
3721 sp->sackdelay = sctp_sack_timeout;
3722 sp->sackfreq = 2;
3723 sp->param_flags = SPP_HB_ENABLE |
3724 SPP_PMTUD_ENABLE |
3725 SPP_SACKDELAY_ENABLE;
3727 /* If enabled no SCTP message fragmentation will be performed.
3728 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3730 sp->disable_fragments = 0;
3732 /* Enable Nagle algorithm by default. */
3733 sp->nodelay = 0;
3735 /* Enable by default. */
3736 sp->v4mapped = 1;
3738 /* Auto-close idle associations after the configured
3739 * number of seconds. A value of 0 disables this
3740 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3741 * for UDP-style sockets only.
3743 sp->autoclose = 0;
3745 /* User specified fragmentation limit. */
3746 sp->user_frag = 0;
3748 sp->adaptation_ind = 0;
3750 sp->pf = sctp_get_pf_specific(sk->sk_family);
3752 /* Control variables for partial data delivery. */
3753 atomic_set(&sp->pd_mode, 0);
3754 skb_queue_head_init(&sp->pd_lobby);
3755 sp->frag_interleave = 0;
3757 /* Create a per socket endpoint structure. Even if we
3758 * change the data structure relationships, this may still
3759 * be useful for storing pre-connect address information.
3761 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3762 if (!ep)
3763 return -ENOMEM;
3765 sp->ep = ep;
3766 sp->hmac = NULL;
3768 SCTP_DBG_OBJCNT_INC(sock);
3770 local_bh_disable();
3771 percpu_counter_inc(&sctp_sockets_allocated);
3772 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3773 local_bh_enable();
3775 return 0;
3778 /* Cleanup any SCTP per socket resources. */
3779 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3781 struct sctp_endpoint *ep;
3783 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3785 /* Release our hold on the endpoint. */
3786 ep = sctp_sk(sk)->ep;
3787 sctp_endpoint_free(ep);
3788 local_bh_disable();
3789 percpu_counter_dec(&sctp_sockets_allocated);
3790 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3791 local_bh_enable();
3794 /* API 4.1.7 shutdown() - TCP Style Syntax
3795 * int shutdown(int socket, int how);
3797 * sd - the socket descriptor of the association to be closed.
3798 * how - Specifies the type of shutdown. The values are
3799 * as follows:
3800 * SHUT_RD
3801 * Disables further receive operations. No SCTP
3802 * protocol action is taken.
3803 * SHUT_WR
3804 * Disables further send operations, and initiates
3805 * the SCTP shutdown sequence.
3806 * SHUT_RDWR
3807 * Disables further send and receive operations
3808 * and initiates the SCTP shutdown sequence.
3810 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3812 struct sctp_endpoint *ep;
3813 struct sctp_association *asoc;
3815 if (!sctp_style(sk, TCP))
3816 return;
3818 if (how & SEND_SHUTDOWN) {
3819 ep = sctp_sk(sk)->ep;
3820 if (!list_empty(&ep->asocs)) {
3821 asoc = list_entry(ep->asocs.next,
3822 struct sctp_association, asocs);
3823 sctp_primitive_SHUTDOWN(asoc, NULL);
3828 /* 7.2.1 Association Status (SCTP_STATUS)
3830 * Applications can retrieve current status information about an
3831 * association, including association state, peer receiver window size,
3832 * number of unacked data chunks, and number of data chunks pending
3833 * receipt. This information is read-only.
3835 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3836 char __user *optval,
3837 int __user *optlen)
3839 struct sctp_status status;
3840 struct sctp_association *asoc = NULL;
3841 struct sctp_transport *transport;
3842 sctp_assoc_t associd;
3843 int retval = 0;
3845 if (len < sizeof(status)) {
3846 retval = -EINVAL;
3847 goto out;
3850 len = sizeof(status);
3851 if (copy_from_user(&status, optval, len)) {
3852 retval = -EFAULT;
3853 goto out;
3856 associd = status.sstat_assoc_id;
3857 asoc = sctp_id2assoc(sk, associd);
3858 if (!asoc) {
3859 retval = -EINVAL;
3860 goto out;
3863 transport = asoc->peer.primary_path;
3865 status.sstat_assoc_id = sctp_assoc2id(asoc);
3866 status.sstat_state = asoc->state;
3867 status.sstat_rwnd = asoc->peer.rwnd;
3868 status.sstat_unackdata = asoc->unack_data;
3870 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
3871 status.sstat_instrms = asoc->c.sinit_max_instreams;
3872 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
3873 status.sstat_fragmentation_point = asoc->frag_point;
3874 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3875 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
3876 transport->af_specific->sockaddr_len);
3877 /* Map ipv4 address into v4-mapped-on-v6 address. */
3878 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
3879 (union sctp_addr *)&status.sstat_primary.spinfo_address);
3880 status.sstat_primary.spinfo_state = transport->state;
3881 status.sstat_primary.spinfo_cwnd = transport->cwnd;
3882 status.sstat_primary.spinfo_srtt = transport->srtt;
3883 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
3884 status.sstat_primary.spinfo_mtu = transport->pathmtu;
3886 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
3887 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
3889 if (put_user(len, optlen)) {
3890 retval = -EFAULT;
3891 goto out;
3894 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
3895 len, status.sstat_state, status.sstat_rwnd,
3896 status.sstat_assoc_id);
3898 if (copy_to_user(optval, &status, len)) {
3899 retval = -EFAULT;
3900 goto out;
3903 out:
3904 return retval;
3908 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
3910 * Applications can retrieve information about a specific peer address
3911 * of an association, including its reachability state, congestion
3912 * window, and retransmission timer values. This information is
3913 * read-only.
3915 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
3916 char __user *optval,
3917 int __user *optlen)
3919 struct sctp_paddrinfo pinfo;
3920 struct sctp_transport *transport;
3921 int retval = 0;
3923 if (len < sizeof(pinfo)) {
3924 retval = -EINVAL;
3925 goto out;
3928 len = sizeof(pinfo);
3929 if (copy_from_user(&pinfo, optval, len)) {
3930 retval = -EFAULT;
3931 goto out;
3934 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
3935 pinfo.spinfo_assoc_id);
3936 if (!transport)
3937 return -EINVAL;
3939 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
3940 pinfo.spinfo_state = transport->state;
3941 pinfo.spinfo_cwnd = transport->cwnd;
3942 pinfo.spinfo_srtt = transport->srtt;
3943 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
3944 pinfo.spinfo_mtu = transport->pathmtu;
3946 if (pinfo.spinfo_state == SCTP_UNKNOWN)
3947 pinfo.spinfo_state = SCTP_ACTIVE;
3949 if (put_user(len, optlen)) {
3950 retval = -EFAULT;
3951 goto out;
3954 if (copy_to_user(optval, &pinfo, len)) {
3955 retval = -EFAULT;
3956 goto out;
3959 out:
3960 return retval;
3963 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
3965 * This option is a on/off flag. If enabled no SCTP message
3966 * fragmentation will be performed. Instead if a message being sent
3967 * exceeds the current PMTU size, the message will NOT be sent and
3968 * instead a error will be indicated to the user.
3970 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
3971 char __user *optval, int __user *optlen)
3973 int val;
3975 if (len < sizeof(int))
3976 return -EINVAL;
3978 len = sizeof(int);
3979 val = (sctp_sk(sk)->disable_fragments == 1);
3980 if (put_user(len, optlen))
3981 return -EFAULT;
3982 if (copy_to_user(optval, &val, len))
3983 return -EFAULT;
3984 return 0;
3987 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
3989 * This socket option is used to specify various notifications and
3990 * ancillary data the user wishes to receive.
3992 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
3993 int __user *optlen)
3995 if (len < sizeof(struct sctp_event_subscribe))
3996 return -EINVAL;
3997 len = sizeof(struct sctp_event_subscribe);
3998 if (put_user(len, optlen))
3999 return -EFAULT;
4000 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4001 return -EFAULT;
4002 return 0;
4005 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4007 * This socket option is applicable to the UDP-style socket only. When
4008 * set it will cause associations that are idle for more than the
4009 * specified number of seconds to automatically close. An association
4010 * being idle is defined an association that has NOT sent or received
4011 * user data. The special value of '0' indicates that no automatic
4012 * close of any associations should be performed. The option expects an
4013 * integer defining the number of seconds of idle time before an
4014 * association is closed.
4016 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4018 /* Applicable to UDP-style socket only */
4019 if (sctp_style(sk, TCP))
4020 return -EOPNOTSUPP;
4021 if (len < sizeof(int))
4022 return -EINVAL;
4023 len = sizeof(int);
4024 if (put_user(len, optlen))
4025 return -EFAULT;
4026 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4027 return -EFAULT;
4028 return 0;
4031 /* Helper routine to branch off an association to a new socket. */
4032 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4033 struct socket **sockp)
4035 struct sock *sk = asoc->base.sk;
4036 struct socket *sock;
4037 struct sctp_af *af;
4038 int err = 0;
4040 /* An association cannot be branched off from an already peeled-off
4041 * socket, nor is this supported for tcp style sockets.
4043 if (!sctp_style(sk, UDP))
4044 return -EINVAL;
4046 /* Create a new socket. */
4047 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4048 if (err < 0)
4049 return err;
4051 sctp_copy_sock(sock->sk, sk, asoc);
4053 /* Make peeled-off sockets more like 1-1 accepted sockets.
4054 * Set the daddr and initialize id to something more random
4056 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4057 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4059 /* Populate the fields of the newsk from the oldsk and migrate the
4060 * asoc to the newsk.
4062 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4064 *sockp = sock;
4066 return err;
4069 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4071 sctp_peeloff_arg_t peeloff;
4072 struct socket *newsock;
4073 int retval = 0;
4074 struct sctp_association *asoc;
4076 if (len < sizeof(sctp_peeloff_arg_t))
4077 return -EINVAL;
4078 len = sizeof(sctp_peeloff_arg_t);
4079 if (copy_from_user(&peeloff, optval, len))
4080 return -EFAULT;
4082 asoc = sctp_id2assoc(sk, peeloff.associd);
4083 if (!asoc) {
4084 retval = -EINVAL;
4085 goto out;
4088 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4090 retval = sctp_do_peeloff(asoc, &newsock);
4091 if (retval < 0)
4092 goto out;
4094 /* Map the socket to an unused fd that can be returned to the user. */
4095 retval = sock_map_fd(newsock, 0);
4096 if (retval < 0) {
4097 sock_release(newsock);
4098 goto out;
4101 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4102 __func__, sk, asoc, newsock->sk, retval);
4104 /* Return the fd mapped to the new socket. */
4105 peeloff.sd = retval;
4106 if (put_user(len, optlen))
4107 return -EFAULT;
4108 if (copy_to_user(optval, &peeloff, len))
4109 retval = -EFAULT;
4111 out:
4112 return retval;
4115 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4117 * Applications can enable or disable heartbeats for any peer address of
4118 * an association, modify an address's heartbeat interval, force a
4119 * heartbeat to be sent immediately, and adjust the address's maximum
4120 * number of retransmissions sent before an address is considered
4121 * unreachable. The following structure is used to access and modify an
4122 * address's parameters:
4124 * struct sctp_paddrparams {
4125 * sctp_assoc_t spp_assoc_id;
4126 * struct sockaddr_storage spp_address;
4127 * uint32_t spp_hbinterval;
4128 * uint16_t spp_pathmaxrxt;
4129 * uint32_t spp_pathmtu;
4130 * uint32_t spp_sackdelay;
4131 * uint32_t spp_flags;
4132 * };
4134 * spp_assoc_id - (one-to-many style socket) This is filled in the
4135 * application, and identifies the association for
4136 * this query.
4137 * spp_address - This specifies which address is of interest.
4138 * spp_hbinterval - This contains the value of the heartbeat interval,
4139 * in milliseconds. If a value of zero
4140 * is present in this field then no changes are to
4141 * be made to this parameter.
4142 * spp_pathmaxrxt - This contains the maximum number of
4143 * retransmissions before this address shall be
4144 * considered unreachable. If a value of zero
4145 * is present in this field then no changes are to
4146 * be made to this parameter.
4147 * spp_pathmtu - When Path MTU discovery is disabled the value
4148 * specified here will be the "fixed" path mtu.
4149 * Note that if the spp_address field is empty
4150 * then all associations on this address will
4151 * have this fixed path mtu set upon them.
4153 * spp_sackdelay - When delayed sack is enabled, this value specifies
4154 * the number of milliseconds that sacks will be delayed
4155 * for. This value will apply to all addresses of an
4156 * association if the spp_address field is empty. Note
4157 * also, that if delayed sack is enabled and this
4158 * value is set to 0, no change is made to the last
4159 * recorded delayed sack timer value.
4161 * spp_flags - These flags are used to control various features
4162 * on an association. The flag field may contain
4163 * zero or more of the following options.
4165 * SPP_HB_ENABLE - Enable heartbeats on the
4166 * specified address. Note that if the address
4167 * field is empty all addresses for the association
4168 * have heartbeats enabled upon them.
4170 * SPP_HB_DISABLE - Disable heartbeats on the
4171 * speicifed address. Note that if the address
4172 * field is empty all addresses for the association
4173 * will have their heartbeats disabled. Note also
4174 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4175 * mutually exclusive, only one of these two should
4176 * be specified. Enabling both fields will have
4177 * undetermined results.
4179 * SPP_HB_DEMAND - Request a user initiated heartbeat
4180 * to be made immediately.
4182 * SPP_PMTUD_ENABLE - This field will enable PMTU
4183 * discovery upon the specified address. Note that
4184 * if the address feild is empty then all addresses
4185 * on the association are effected.
4187 * SPP_PMTUD_DISABLE - This field will disable PMTU
4188 * discovery upon the specified address. Note that
4189 * if the address feild is empty then all addresses
4190 * on the association are effected. Not also that
4191 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4192 * exclusive. Enabling both will have undetermined
4193 * results.
4195 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4196 * on delayed sack. The time specified in spp_sackdelay
4197 * is used to specify the sack delay for this address. Note
4198 * that if spp_address is empty then all addresses will
4199 * enable delayed sack and take on the sack delay
4200 * value specified in spp_sackdelay.
4201 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4202 * off delayed sack. If the spp_address field is blank then
4203 * delayed sack is disabled for the entire association. Note
4204 * also that this field is mutually exclusive to
4205 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4206 * results.
4208 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4209 char __user *optval, int __user *optlen)
4211 struct sctp_paddrparams params;
4212 struct sctp_transport *trans = NULL;
4213 struct sctp_association *asoc = NULL;
4214 struct sctp_sock *sp = sctp_sk(sk);
4216 if (len < sizeof(struct sctp_paddrparams))
4217 return -EINVAL;
4218 len = sizeof(struct sctp_paddrparams);
4219 if (copy_from_user(&params, optval, len))
4220 return -EFAULT;
4222 /* If an address other than INADDR_ANY is specified, and
4223 * no transport is found, then the request is invalid.
4225 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4226 trans = sctp_addr_id2transport(sk, &params.spp_address,
4227 params.spp_assoc_id);
4228 if (!trans) {
4229 SCTP_DEBUG_PRINTK("Failed no transport\n");
4230 return -EINVAL;
4234 /* Get association, if assoc_id != 0 and the socket is a one
4235 * to many style socket, and an association was not found, then
4236 * the id was invalid.
4238 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4239 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4240 SCTP_DEBUG_PRINTK("Failed no association\n");
4241 return -EINVAL;
4244 if (trans) {
4245 /* Fetch transport values. */
4246 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4247 params.spp_pathmtu = trans->pathmtu;
4248 params.spp_pathmaxrxt = trans->pathmaxrxt;
4249 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4251 /*draft-11 doesn't say what to return in spp_flags*/
4252 params.spp_flags = trans->param_flags;
4253 } else if (asoc) {
4254 /* Fetch association values. */
4255 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4256 params.spp_pathmtu = asoc->pathmtu;
4257 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4258 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4260 /*draft-11 doesn't say what to return in spp_flags*/
4261 params.spp_flags = asoc->param_flags;
4262 } else {
4263 /* Fetch socket values. */
4264 params.spp_hbinterval = sp->hbinterval;
4265 params.spp_pathmtu = sp->pathmtu;
4266 params.spp_sackdelay = sp->sackdelay;
4267 params.spp_pathmaxrxt = sp->pathmaxrxt;
4269 /*draft-11 doesn't say what to return in spp_flags*/
4270 params.spp_flags = sp->param_flags;
4273 if (copy_to_user(optval, &params, len))
4274 return -EFAULT;
4276 if (put_user(len, optlen))
4277 return -EFAULT;
4279 return 0;
4283 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4285 * This option will effect the way delayed acks are performed. This
4286 * option allows you to get or set the delayed ack time, in
4287 * milliseconds. It also allows changing the delayed ack frequency.
4288 * Changing the frequency to 1 disables the delayed sack algorithm. If
4289 * the assoc_id is 0, then this sets or gets the endpoints default
4290 * values. If the assoc_id field is non-zero, then the set or get
4291 * effects the specified association for the one to many model (the
4292 * assoc_id field is ignored by the one to one model). Note that if
4293 * sack_delay or sack_freq are 0 when setting this option, then the
4294 * current values will remain unchanged.
4296 * struct sctp_sack_info {
4297 * sctp_assoc_t sack_assoc_id;
4298 * uint32_t sack_delay;
4299 * uint32_t sack_freq;
4300 * };
4302 * sack_assoc_id - This parameter, indicates which association the user
4303 * is performing an action upon. Note that if this field's value is
4304 * zero then the endpoints default value is changed (effecting future
4305 * associations only).
4307 * sack_delay - This parameter contains the number of milliseconds that
4308 * the user is requesting the delayed ACK timer be set to. Note that
4309 * this value is defined in the standard to be between 200 and 500
4310 * milliseconds.
4312 * sack_freq - This parameter contains the number of packets that must
4313 * be received before a sack is sent without waiting for the delay
4314 * timer to expire. The default value for this is 2, setting this
4315 * value to 1 will disable the delayed sack algorithm.
4317 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4318 char __user *optval,
4319 int __user *optlen)
4321 struct sctp_sack_info params;
4322 struct sctp_association *asoc = NULL;
4323 struct sctp_sock *sp = sctp_sk(sk);
4325 if (len >= sizeof(struct sctp_sack_info)) {
4326 len = sizeof(struct sctp_sack_info);
4328 if (copy_from_user(&params, optval, len))
4329 return -EFAULT;
4330 } else if (len == sizeof(struct sctp_assoc_value)) {
4331 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4332 pr_warn("Use struct sctp_sack_info instead\n");
4333 if (copy_from_user(&params, optval, len))
4334 return -EFAULT;
4335 } else
4336 return - EINVAL;
4338 /* Get association, if sack_assoc_id != 0 and the socket is a one
4339 * to many style socket, and an association was not found, then
4340 * the id was invalid.
4342 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4343 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4344 return -EINVAL;
4346 if (asoc) {
4347 /* Fetch association values. */
4348 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4349 params.sack_delay = jiffies_to_msecs(
4350 asoc->sackdelay);
4351 params.sack_freq = asoc->sackfreq;
4353 } else {
4354 params.sack_delay = 0;
4355 params.sack_freq = 1;
4357 } else {
4358 /* Fetch socket values. */
4359 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4360 params.sack_delay = sp->sackdelay;
4361 params.sack_freq = sp->sackfreq;
4362 } else {
4363 params.sack_delay = 0;
4364 params.sack_freq = 1;
4368 if (copy_to_user(optval, &params, len))
4369 return -EFAULT;
4371 if (put_user(len, optlen))
4372 return -EFAULT;
4374 return 0;
4377 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4379 * Applications can specify protocol parameters for the default association
4380 * initialization. The option name argument to setsockopt() and getsockopt()
4381 * is SCTP_INITMSG.
4383 * Setting initialization parameters is effective only on an unconnected
4384 * socket (for UDP-style sockets only future associations are effected
4385 * by the change). With TCP-style sockets, this option is inherited by
4386 * sockets derived from a listener socket.
4388 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4390 if (len < sizeof(struct sctp_initmsg))
4391 return -EINVAL;
4392 len = sizeof(struct sctp_initmsg);
4393 if (put_user(len, optlen))
4394 return -EFAULT;
4395 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4396 return -EFAULT;
4397 return 0;
4401 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4402 char __user *optval, int __user *optlen)
4404 struct sctp_association *asoc;
4405 int cnt = 0;
4406 struct sctp_getaddrs getaddrs;
4407 struct sctp_transport *from;
4408 void __user *to;
4409 union sctp_addr temp;
4410 struct sctp_sock *sp = sctp_sk(sk);
4411 int addrlen;
4412 size_t space_left;
4413 int bytes_copied;
4415 if (len < sizeof(struct sctp_getaddrs))
4416 return -EINVAL;
4418 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4419 return -EFAULT;
4421 /* For UDP-style sockets, id specifies the association to query. */
4422 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4423 if (!asoc)
4424 return -EINVAL;
4426 to = optval + offsetof(struct sctp_getaddrs,addrs);
4427 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4429 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4430 transports) {
4431 memcpy(&temp, &from->ipaddr, sizeof(temp));
4432 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4433 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4434 if (space_left < addrlen)
4435 return -ENOMEM;
4436 if (copy_to_user(to, &temp, addrlen))
4437 return -EFAULT;
4438 to += addrlen;
4439 cnt++;
4440 space_left -= addrlen;
4443 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4444 return -EFAULT;
4445 bytes_copied = ((char __user *)to) - optval;
4446 if (put_user(bytes_copied, optlen))
4447 return -EFAULT;
4449 return 0;
4452 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4453 size_t space_left, int *bytes_copied)
4455 struct sctp_sockaddr_entry *addr;
4456 union sctp_addr temp;
4457 int cnt = 0;
4458 int addrlen;
4460 rcu_read_lock();
4461 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4462 if (!addr->valid)
4463 continue;
4465 if ((PF_INET == sk->sk_family) &&
4466 (AF_INET6 == addr->a.sa.sa_family))
4467 continue;
4468 if ((PF_INET6 == sk->sk_family) &&
4469 inet_v6_ipv6only(sk) &&
4470 (AF_INET == addr->a.sa.sa_family))
4471 continue;
4472 memcpy(&temp, &addr->a, sizeof(temp));
4473 if (!temp.v4.sin_port)
4474 temp.v4.sin_port = htons(port);
4476 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4477 &temp);
4478 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4479 if (space_left < addrlen) {
4480 cnt = -ENOMEM;
4481 break;
4483 memcpy(to, &temp, addrlen);
4485 to += addrlen;
4486 cnt ++;
4487 space_left -= addrlen;
4488 *bytes_copied += addrlen;
4490 rcu_read_unlock();
4492 return cnt;
4496 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4497 char __user *optval, int __user *optlen)
4499 struct sctp_bind_addr *bp;
4500 struct sctp_association *asoc;
4501 int cnt = 0;
4502 struct sctp_getaddrs getaddrs;
4503 struct sctp_sockaddr_entry *addr;
4504 void __user *to;
4505 union sctp_addr temp;
4506 struct sctp_sock *sp = sctp_sk(sk);
4507 int addrlen;
4508 int err = 0;
4509 size_t space_left;
4510 int bytes_copied = 0;
4511 void *addrs;
4512 void *buf;
4514 if (len < sizeof(struct sctp_getaddrs))
4515 return -EINVAL;
4517 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4518 return -EFAULT;
4521 * For UDP-style sockets, id specifies the association to query.
4522 * If the id field is set to the value '0' then the locally bound
4523 * addresses are returned without regard to any particular
4524 * association.
4526 if (0 == getaddrs.assoc_id) {
4527 bp = &sctp_sk(sk)->ep->base.bind_addr;
4528 } else {
4529 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4530 if (!asoc)
4531 return -EINVAL;
4532 bp = &asoc->base.bind_addr;
4535 to = optval + offsetof(struct sctp_getaddrs,addrs);
4536 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4538 addrs = kmalloc(space_left, GFP_KERNEL);
4539 if (!addrs)
4540 return -ENOMEM;
4542 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4543 * addresses from the global local address list.
4545 if (sctp_list_single_entry(&bp->address_list)) {
4546 addr = list_entry(bp->address_list.next,
4547 struct sctp_sockaddr_entry, list);
4548 if (sctp_is_any(sk, &addr->a)) {
4549 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4550 space_left, &bytes_copied);
4551 if (cnt < 0) {
4552 err = cnt;
4553 goto out;
4555 goto copy_getaddrs;
4559 buf = addrs;
4560 /* Protection on the bound address list is not needed since
4561 * in the socket option context we hold a socket lock and
4562 * thus the bound address list can't change.
4564 list_for_each_entry(addr, &bp->address_list, list) {
4565 memcpy(&temp, &addr->a, sizeof(temp));
4566 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4567 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4568 if (space_left < addrlen) {
4569 err = -ENOMEM; /*fixme: right error?*/
4570 goto out;
4572 memcpy(buf, &temp, addrlen);
4573 buf += addrlen;
4574 bytes_copied += addrlen;
4575 cnt ++;
4576 space_left -= addrlen;
4579 copy_getaddrs:
4580 if (copy_to_user(to, addrs, bytes_copied)) {
4581 err = -EFAULT;
4582 goto out;
4584 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4585 err = -EFAULT;
4586 goto out;
4588 if (put_user(bytes_copied, optlen))
4589 err = -EFAULT;
4590 out:
4591 kfree(addrs);
4592 return err;
4595 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4597 * Requests that the local SCTP stack use the enclosed peer address as
4598 * the association primary. The enclosed address must be one of the
4599 * association peer's addresses.
4601 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4602 char __user *optval, int __user *optlen)
4604 struct sctp_prim prim;
4605 struct sctp_association *asoc;
4606 struct sctp_sock *sp = sctp_sk(sk);
4608 if (len < sizeof(struct sctp_prim))
4609 return -EINVAL;
4611 len = sizeof(struct sctp_prim);
4613 if (copy_from_user(&prim, optval, len))
4614 return -EFAULT;
4616 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4617 if (!asoc)
4618 return -EINVAL;
4620 if (!asoc->peer.primary_path)
4621 return -ENOTCONN;
4623 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4624 asoc->peer.primary_path->af_specific->sockaddr_len);
4626 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4627 (union sctp_addr *)&prim.ssp_addr);
4629 if (put_user(len, optlen))
4630 return -EFAULT;
4631 if (copy_to_user(optval, &prim, len))
4632 return -EFAULT;
4634 return 0;
4638 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4640 * Requests that the local endpoint set the specified Adaptation Layer
4641 * Indication parameter for all future INIT and INIT-ACK exchanges.
4643 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4644 char __user *optval, int __user *optlen)
4646 struct sctp_setadaptation adaptation;
4648 if (len < sizeof(struct sctp_setadaptation))
4649 return -EINVAL;
4651 len = sizeof(struct sctp_setadaptation);
4653 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4655 if (put_user(len, optlen))
4656 return -EFAULT;
4657 if (copy_to_user(optval, &adaptation, len))
4658 return -EFAULT;
4660 return 0;
4665 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4667 * Applications that wish to use the sendto() system call may wish to
4668 * specify a default set of parameters that would normally be supplied
4669 * through the inclusion of ancillary data. This socket option allows
4670 * such an application to set the default sctp_sndrcvinfo structure.
4673 * The application that wishes to use this socket option simply passes
4674 * in to this call the sctp_sndrcvinfo structure defined in Section
4675 * 5.2.2) The input parameters accepted by this call include
4676 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4677 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4678 * to this call if the caller is using the UDP model.
4680 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4682 static int sctp_getsockopt_default_send_param(struct sock *sk,
4683 int len, char __user *optval,
4684 int __user *optlen)
4686 struct sctp_sndrcvinfo info;
4687 struct sctp_association *asoc;
4688 struct sctp_sock *sp = sctp_sk(sk);
4690 if (len < sizeof(struct sctp_sndrcvinfo))
4691 return -EINVAL;
4693 len = sizeof(struct sctp_sndrcvinfo);
4695 if (copy_from_user(&info, optval, len))
4696 return -EFAULT;
4698 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4699 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4700 return -EINVAL;
4702 if (asoc) {
4703 info.sinfo_stream = asoc->default_stream;
4704 info.sinfo_flags = asoc->default_flags;
4705 info.sinfo_ppid = asoc->default_ppid;
4706 info.sinfo_context = asoc->default_context;
4707 info.sinfo_timetolive = asoc->default_timetolive;
4708 } else {
4709 info.sinfo_stream = sp->default_stream;
4710 info.sinfo_flags = sp->default_flags;
4711 info.sinfo_ppid = sp->default_ppid;
4712 info.sinfo_context = sp->default_context;
4713 info.sinfo_timetolive = sp->default_timetolive;
4716 if (put_user(len, optlen))
4717 return -EFAULT;
4718 if (copy_to_user(optval, &info, len))
4719 return -EFAULT;
4721 return 0;
4726 * 7.1.5 SCTP_NODELAY
4728 * Turn on/off any Nagle-like algorithm. This means that packets are
4729 * generally sent as soon as possible and no unnecessary delays are
4730 * introduced, at the cost of more packets in the network. Expects an
4731 * integer boolean flag.
4734 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4735 char __user *optval, int __user *optlen)
4737 int val;
4739 if (len < sizeof(int))
4740 return -EINVAL;
4742 len = sizeof(int);
4743 val = (sctp_sk(sk)->nodelay == 1);
4744 if (put_user(len, optlen))
4745 return -EFAULT;
4746 if (copy_to_user(optval, &val, len))
4747 return -EFAULT;
4748 return 0;
4753 * 7.1.1 SCTP_RTOINFO
4755 * The protocol parameters used to initialize and bound retransmission
4756 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4757 * and modify these parameters.
4758 * All parameters are time values, in milliseconds. A value of 0, when
4759 * modifying the parameters, indicates that the current value should not
4760 * be changed.
4763 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4764 char __user *optval,
4765 int __user *optlen) {
4766 struct sctp_rtoinfo rtoinfo;
4767 struct sctp_association *asoc;
4769 if (len < sizeof (struct sctp_rtoinfo))
4770 return -EINVAL;
4772 len = sizeof(struct sctp_rtoinfo);
4774 if (copy_from_user(&rtoinfo, optval, len))
4775 return -EFAULT;
4777 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4779 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4780 return -EINVAL;
4782 /* Values corresponding to the specific association. */
4783 if (asoc) {
4784 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4785 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4786 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4787 } else {
4788 /* Values corresponding to the endpoint. */
4789 struct sctp_sock *sp = sctp_sk(sk);
4791 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4792 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4793 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4796 if (put_user(len, optlen))
4797 return -EFAULT;
4799 if (copy_to_user(optval, &rtoinfo, len))
4800 return -EFAULT;
4802 return 0;
4807 * 7.1.2 SCTP_ASSOCINFO
4809 * This option is used to tune the maximum retransmission attempts
4810 * of the association.
4811 * Returns an error if the new association retransmission value is
4812 * greater than the sum of the retransmission value of the peer.
4813 * See [SCTP] for more information.
4816 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4817 char __user *optval,
4818 int __user *optlen)
4821 struct sctp_assocparams assocparams;
4822 struct sctp_association *asoc;
4823 struct list_head *pos;
4824 int cnt = 0;
4826 if (len < sizeof (struct sctp_assocparams))
4827 return -EINVAL;
4829 len = sizeof(struct sctp_assocparams);
4831 if (copy_from_user(&assocparams, optval, len))
4832 return -EFAULT;
4834 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4836 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4837 return -EINVAL;
4839 /* Values correspoinding to the specific association */
4840 if (asoc) {
4841 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4842 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4843 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4844 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4845 * 1000) +
4846 (asoc->cookie_life.tv_usec
4847 / 1000);
4849 list_for_each(pos, &asoc->peer.transport_addr_list) {
4850 cnt ++;
4853 assocparams.sasoc_number_peer_destinations = cnt;
4854 } else {
4855 /* Values corresponding to the endpoint */
4856 struct sctp_sock *sp = sctp_sk(sk);
4858 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
4859 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
4860 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
4861 assocparams.sasoc_cookie_life =
4862 sp->assocparams.sasoc_cookie_life;
4863 assocparams.sasoc_number_peer_destinations =
4864 sp->assocparams.
4865 sasoc_number_peer_destinations;
4868 if (put_user(len, optlen))
4869 return -EFAULT;
4871 if (copy_to_user(optval, &assocparams, len))
4872 return -EFAULT;
4874 return 0;
4878 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
4880 * This socket option is a boolean flag which turns on or off mapped V4
4881 * addresses. If this option is turned on and the socket is type
4882 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
4883 * If this option is turned off, then no mapping will be done of V4
4884 * addresses and a user will receive both PF_INET6 and PF_INET type
4885 * addresses on the socket.
4887 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
4888 char __user *optval, int __user *optlen)
4890 int val;
4891 struct sctp_sock *sp = sctp_sk(sk);
4893 if (len < sizeof(int))
4894 return -EINVAL;
4896 len = sizeof(int);
4897 val = sp->v4mapped;
4898 if (put_user(len, optlen))
4899 return -EFAULT;
4900 if (copy_to_user(optval, &val, len))
4901 return -EFAULT;
4903 return 0;
4907 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
4908 * (chapter and verse is quoted at sctp_setsockopt_context())
4910 static int sctp_getsockopt_context(struct sock *sk, int len,
4911 char __user *optval, int __user *optlen)
4913 struct sctp_assoc_value params;
4914 struct sctp_sock *sp;
4915 struct sctp_association *asoc;
4917 if (len < sizeof(struct sctp_assoc_value))
4918 return -EINVAL;
4920 len = sizeof(struct sctp_assoc_value);
4922 if (copy_from_user(&params, optval, len))
4923 return -EFAULT;
4925 sp = sctp_sk(sk);
4927 if (params.assoc_id != 0) {
4928 asoc = sctp_id2assoc(sk, params.assoc_id);
4929 if (!asoc)
4930 return -EINVAL;
4931 params.assoc_value = asoc->default_rcv_context;
4932 } else {
4933 params.assoc_value = sp->default_rcv_context;
4936 if (put_user(len, optlen))
4937 return -EFAULT;
4938 if (copy_to_user(optval, &params, len))
4939 return -EFAULT;
4941 return 0;
4945 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
4946 * This option will get or set the maximum size to put in any outgoing
4947 * SCTP DATA chunk. If a message is larger than this size it will be
4948 * fragmented by SCTP into the specified size. Note that the underlying
4949 * SCTP implementation may fragment into smaller sized chunks when the
4950 * PMTU of the underlying association is smaller than the value set by
4951 * the user. The default value for this option is '0' which indicates
4952 * the user is NOT limiting fragmentation and only the PMTU will effect
4953 * SCTP's choice of DATA chunk size. Note also that values set larger
4954 * than the maximum size of an IP datagram will effectively let SCTP
4955 * control fragmentation (i.e. the same as setting this option to 0).
4957 * The following structure is used to access and modify this parameter:
4959 * struct sctp_assoc_value {
4960 * sctp_assoc_t assoc_id;
4961 * uint32_t assoc_value;
4962 * };
4964 * assoc_id: This parameter is ignored for one-to-one style sockets.
4965 * For one-to-many style sockets this parameter indicates which
4966 * association the user is performing an action upon. Note that if
4967 * this field's value is zero then the endpoints default value is
4968 * changed (effecting future associations only).
4969 * assoc_value: This parameter specifies the maximum size in bytes.
4971 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
4972 char __user *optval, int __user *optlen)
4974 struct sctp_assoc_value params;
4975 struct sctp_association *asoc;
4977 if (len == sizeof(int)) {
4978 pr_warn("Use of int in maxseg socket option deprecated\n");
4979 pr_warn("Use struct sctp_assoc_value instead\n");
4980 params.assoc_id = 0;
4981 } else if (len >= sizeof(struct sctp_assoc_value)) {
4982 len = sizeof(struct sctp_assoc_value);
4983 if (copy_from_user(&params, optval, sizeof(params)))
4984 return -EFAULT;
4985 } else
4986 return -EINVAL;
4988 asoc = sctp_id2assoc(sk, params.assoc_id);
4989 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
4990 return -EINVAL;
4992 if (asoc)
4993 params.assoc_value = asoc->frag_point;
4994 else
4995 params.assoc_value = sctp_sk(sk)->user_frag;
4997 if (put_user(len, optlen))
4998 return -EFAULT;
4999 if (len == sizeof(int)) {
5000 if (copy_to_user(optval, &params.assoc_value, len))
5001 return -EFAULT;
5002 } else {
5003 if (copy_to_user(optval, &params, len))
5004 return -EFAULT;
5007 return 0;
5011 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5012 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5014 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5015 char __user *optval, int __user *optlen)
5017 int val;
5019 if (len < sizeof(int))
5020 return -EINVAL;
5022 len = sizeof(int);
5024 val = sctp_sk(sk)->frag_interleave;
5025 if (put_user(len, optlen))
5026 return -EFAULT;
5027 if (copy_to_user(optval, &val, len))
5028 return -EFAULT;
5030 return 0;
5034 * 7.1.25. Set or Get the sctp partial delivery point
5035 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5037 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5038 char __user *optval,
5039 int __user *optlen)
5041 u32 val;
5043 if (len < sizeof(u32))
5044 return -EINVAL;
5046 len = sizeof(u32);
5048 val = sctp_sk(sk)->pd_point;
5049 if (put_user(len, optlen))
5050 return -EFAULT;
5051 if (copy_to_user(optval, &val, len))
5052 return -EFAULT;
5054 return 0;
5058 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5059 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5061 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5062 char __user *optval,
5063 int __user *optlen)
5065 struct sctp_assoc_value params;
5066 struct sctp_sock *sp;
5067 struct sctp_association *asoc;
5069 if (len == sizeof(int)) {
5070 pr_warn("Use of int in max_burst socket option deprecated\n");
5071 pr_warn("Use struct sctp_assoc_value instead\n");
5072 params.assoc_id = 0;
5073 } else if (len >= sizeof(struct sctp_assoc_value)) {
5074 len = sizeof(struct sctp_assoc_value);
5075 if (copy_from_user(&params, optval, len))
5076 return -EFAULT;
5077 } else
5078 return -EINVAL;
5080 sp = sctp_sk(sk);
5082 if (params.assoc_id != 0) {
5083 asoc = sctp_id2assoc(sk, params.assoc_id);
5084 if (!asoc)
5085 return -EINVAL;
5086 params.assoc_value = asoc->max_burst;
5087 } else
5088 params.assoc_value = sp->max_burst;
5090 if (len == sizeof(int)) {
5091 if (copy_to_user(optval, &params.assoc_value, len))
5092 return -EFAULT;
5093 } else {
5094 if (copy_to_user(optval, &params, len))
5095 return -EFAULT;
5098 return 0;
5102 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5103 char __user *optval, int __user *optlen)
5105 struct sctp_hmacalgo __user *p = (void __user *)optval;
5106 struct sctp_hmac_algo_param *hmacs;
5107 __u16 data_len = 0;
5108 u32 num_idents;
5110 if (!sctp_auth_enable)
5111 return -EACCES;
5113 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5114 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5116 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5117 return -EINVAL;
5119 len = sizeof(struct sctp_hmacalgo) + data_len;
5120 num_idents = data_len / sizeof(u16);
5122 if (put_user(len, optlen))
5123 return -EFAULT;
5124 if (put_user(num_idents, &p->shmac_num_idents))
5125 return -EFAULT;
5126 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5127 return -EFAULT;
5128 return 0;
5131 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5132 char __user *optval, int __user *optlen)
5134 struct sctp_authkeyid val;
5135 struct sctp_association *asoc;
5137 if (!sctp_auth_enable)
5138 return -EACCES;
5140 if (len < sizeof(struct sctp_authkeyid))
5141 return -EINVAL;
5142 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5143 return -EFAULT;
5145 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5146 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5147 return -EINVAL;
5149 if (asoc)
5150 val.scact_keynumber = asoc->active_key_id;
5151 else
5152 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5154 len = sizeof(struct sctp_authkeyid);
5155 if (put_user(len, optlen))
5156 return -EFAULT;
5157 if (copy_to_user(optval, &val, len))
5158 return -EFAULT;
5160 return 0;
5163 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5164 char __user *optval, int __user *optlen)
5166 struct sctp_authchunks __user *p = (void __user *)optval;
5167 struct sctp_authchunks val;
5168 struct sctp_association *asoc;
5169 struct sctp_chunks_param *ch;
5170 u32 num_chunks = 0;
5171 char __user *to;
5173 if (!sctp_auth_enable)
5174 return -EACCES;
5176 if (len < sizeof(struct sctp_authchunks))
5177 return -EINVAL;
5179 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5180 return -EFAULT;
5182 to = p->gauth_chunks;
5183 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5184 if (!asoc)
5185 return -EINVAL;
5187 ch = asoc->peer.peer_chunks;
5188 if (!ch)
5189 goto num;
5191 /* See if the user provided enough room for all the data */
5192 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5193 if (len < num_chunks)
5194 return -EINVAL;
5196 if (copy_to_user(to, ch->chunks, num_chunks))
5197 return -EFAULT;
5198 num:
5199 len = sizeof(struct sctp_authchunks) + num_chunks;
5200 if (put_user(len, optlen)) return -EFAULT;
5201 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5202 return -EFAULT;
5203 return 0;
5206 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5207 char __user *optval, int __user *optlen)
5209 struct sctp_authchunks __user *p = (void __user *)optval;
5210 struct sctp_authchunks val;
5211 struct sctp_association *asoc;
5212 struct sctp_chunks_param *ch;
5213 u32 num_chunks = 0;
5214 char __user *to;
5216 if (!sctp_auth_enable)
5217 return -EACCES;
5219 if (len < sizeof(struct sctp_authchunks))
5220 return -EINVAL;
5222 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5223 return -EFAULT;
5225 to = p->gauth_chunks;
5226 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5227 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5228 return -EINVAL;
5230 if (asoc)
5231 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5232 else
5233 ch = sctp_sk(sk)->ep->auth_chunk_list;
5235 if (!ch)
5236 goto num;
5238 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5239 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5240 return -EINVAL;
5242 if (copy_to_user(to, ch->chunks, num_chunks))
5243 return -EFAULT;
5244 num:
5245 len = sizeof(struct sctp_authchunks) + num_chunks;
5246 if (put_user(len, optlen))
5247 return -EFAULT;
5248 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5249 return -EFAULT;
5251 return 0;
5255 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5256 * This option gets the current number of associations that are attached
5257 * to a one-to-many style socket. The option value is an uint32_t.
5259 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5260 char __user *optval, int __user *optlen)
5262 struct sctp_sock *sp = sctp_sk(sk);
5263 struct sctp_association *asoc;
5264 u32 val = 0;
5266 if (sctp_style(sk, TCP))
5267 return -EOPNOTSUPP;
5269 if (len < sizeof(u32))
5270 return -EINVAL;
5272 len = sizeof(u32);
5274 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5275 val++;
5278 if (put_user(len, optlen))
5279 return -EFAULT;
5280 if (copy_to_user(optval, &val, len))
5281 return -EFAULT;
5283 return 0;
5286 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5287 char __user *optval, int __user *optlen)
5289 int retval = 0;
5290 int len;
5292 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5293 sk, optname);
5295 /* I can hardly begin to describe how wrong this is. This is
5296 * so broken as to be worse than useless. The API draft
5297 * REALLY is NOT helpful here... I am not convinced that the
5298 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5299 * are at all well-founded.
5301 if (level != SOL_SCTP) {
5302 struct sctp_af *af = sctp_sk(sk)->pf->af;
5304 retval = af->getsockopt(sk, level, optname, optval, optlen);
5305 return retval;
5308 if (get_user(len, optlen))
5309 return -EFAULT;
5311 sctp_lock_sock(sk);
5313 switch (optname) {
5314 case SCTP_STATUS:
5315 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5316 break;
5317 case SCTP_DISABLE_FRAGMENTS:
5318 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5319 optlen);
5320 break;
5321 case SCTP_EVENTS:
5322 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5323 break;
5324 case SCTP_AUTOCLOSE:
5325 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5326 break;
5327 case SCTP_SOCKOPT_PEELOFF:
5328 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5329 break;
5330 case SCTP_PEER_ADDR_PARAMS:
5331 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5332 optlen);
5333 break;
5334 case SCTP_DELAYED_SACK:
5335 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5336 optlen);
5337 break;
5338 case SCTP_INITMSG:
5339 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5340 break;
5341 case SCTP_GET_PEER_ADDRS:
5342 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5343 optlen);
5344 break;
5345 case SCTP_GET_LOCAL_ADDRS:
5346 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5347 optlen);
5348 break;
5349 case SCTP_SOCKOPT_CONNECTX3:
5350 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5351 break;
5352 case SCTP_DEFAULT_SEND_PARAM:
5353 retval = sctp_getsockopt_default_send_param(sk, len,
5354 optval, optlen);
5355 break;
5356 case SCTP_PRIMARY_ADDR:
5357 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5358 break;
5359 case SCTP_NODELAY:
5360 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5361 break;
5362 case SCTP_RTOINFO:
5363 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5364 break;
5365 case SCTP_ASSOCINFO:
5366 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5367 break;
5368 case SCTP_I_WANT_MAPPED_V4_ADDR:
5369 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5370 break;
5371 case SCTP_MAXSEG:
5372 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5373 break;
5374 case SCTP_GET_PEER_ADDR_INFO:
5375 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5376 optlen);
5377 break;
5378 case SCTP_ADAPTATION_LAYER:
5379 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5380 optlen);
5381 break;
5382 case SCTP_CONTEXT:
5383 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5384 break;
5385 case SCTP_FRAGMENT_INTERLEAVE:
5386 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5387 optlen);
5388 break;
5389 case SCTP_PARTIAL_DELIVERY_POINT:
5390 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5391 optlen);
5392 break;
5393 case SCTP_MAX_BURST:
5394 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5395 break;
5396 case SCTP_AUTH_KEY:
5397 case SCTP_AUTH_CHUNK:
5398 case SCTP_AUTH_DELETE_KEY:
5399 retval = -EOPNOTSUPP;
5400 break;
5401 case SCTP_HMAC_IDENT:
5402 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5403 break;
5404 case SCTP_AUTH_ACTIVE_KEY:
5405 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5406 break;
5407 case SCTP_PEER_AUTH_CHUNKS:
5408 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5409 optlen);
5410 break;
5411 case SCTP_LOCAL_AUTH_CHUNKS:
5412 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5413 optlen);
5414 break;
5415 case SCTP_GET_ASSOC_NUMBER:
5416 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5417 break;
5418 default:
5419 retval = -ENOPROTOOPT;
5420 break;
5423 sctp_release_sock(sk);
5424 return retval;
5427 static void sctp_hash(struct sock *sk)
5429 /* STUB */
5432 static void sctp_unhash(struct sock *sk)
5434 /* STUB */
5437 /* Check if port is acceptable. Possibly find first available port.
5439 * The port hash table (contained in the 'global' SCTP protocol storage
5440 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5441 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5442 * list (the list number is the port number hashed out, so as you
5443 * would expect from a hash function, all the ports in a given list have
5444 * such a number that hashes out to the same list number; you were
5445 * expecting that, right?); so each list has a set of ports, with a
5446 * link to the socket (struct sock) that uses it, the port number and
5447 * a fastreuse flag (FIXME: NPI ipg).
5449 static struct sctp_bind_bucket *sctp_bucket_create(
5450 struct sctp_bind_hashbucket *head, unsigned short snum);
5452 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5454 struct sctp_bind_hashbucket *head; /* hash list */
5455 struct sctp_bind_bucket *pp; /* hash list port iterator */
5456 struct hlist_node *node;
5457 unsigned short snum;
5458 int ret;
5460 snum = ntohs(addr->v4.sin_port);
5462 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5463 sctp_local_bh_disable();
5465 if (snum == 0) {
5466 /* Search for an available port. */
5467 int low, high, remaining, index;
5468 unsigned int rover;
5470 inet_get_local_port_range(&low, &high);
5471 remaining = (high - low) + 1;
5472 rover = net_random() % remaining + low;
5474 do {
5475 rover++;
5476 if ((rover < low) || (rover > high))
5477 rover = low;
5478 if (inet_is_reserved_local_port(rover))
5479 continue;
5480 index = sctp_phashfn(rover);
5481 head = &sctp_port_hashtable[index];
5482 sctp_spin_lock(&head->lock);
5483 sctp_for_each_hentry(pp, node, &head->chain)
5484 if (pp->port == rover)
5485 goto next;
5486 break;
5487 next:
5488 sctp_spin_unlock(&head->lock);
5489 } while (--remaining > 0);
5491 /* Exhausted local port range during search? */
5492 ret = 1;
5493 if (remaining <= 0)
5494 goto fail;
5496 /* OK, here is the one we will use. HEAD (the port
5497 * hash table list entry) is non-NULL and we hold it's
5498 * mutex.
5500 snum = rover;
5501 } else {
5502 /* We are given an specific port number; we verify
5503 * that it is not being used. If it is used, we will
5504 * exahust the search in the hash list corresponding
5505 * to the port number (snum) - we detect that with the
5506 * port iterator, pp being NULL.
5508 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5509 sctp_spin_lock(&head->lock);
5510 sctp_for_each_hentry(pp, node, &head->chain) {
5511 if (pp->port == snum)
5512 goto pp_found;
5515 pp = NULL;
5516 goto pp_not_found;
5517 pp_found:
5518 if (!hlist_empty(&pp->owner)) {
5519 /* We had a port hash table hit - there is an
5520 * available port (pp != NULL) and it is being
5521 * used by other socket (pp->owner not empty); that other
5522 * socket is going to be sk2.
5524 int reuse = sk->sk_reuse;
5525 struct sock *sk2;
5527 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5528 if (pp->fastreuse && sk->sk_reuse &&
5529 sk->sk_state != SCTP_SS_LISTENING)
5530 goto success;
5532 /* Run through the list of sockets bound to the port
5533 * (pp->port) [via the pointers bind_next and
5534 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5535 * we get the endpoint they describe and run through
5536 * the endpoint's list of IP (v4 or v6) addresses,
5537 * comparing each of the addresses with the address of
5538 * the socket sk. If we find a match, then that means
5539 * that this port/socket (sk) combination are already
5540 * in an endpoint.
5542 sk_for_each_bound(sk2, node, &pp->owner) {
5543 struct sctp_endpoint *ep2;
5544 ep2 = sctp_sk(sk2)->ep;
5546 if (sk == sk2 ||
5547 (reuse && sk2->sk_reuse &&
5548 sk2->sk_state != SCTP_SS_LISTENING))
5549 continue;
5551 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5552 sctp_sk(sk2), sctp_sk(sk))) {
5553 ret = (long)sk2;
5554 goto fail_unlock;
5557 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5559 pp_not_found:
5560 /* If there was a hash table miss, create a new port. */
5561 ret = 1;
5562 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5563 goto fail_unlock;
5565 /* In either case (hit or miss), make sure fastreuse is 1 only
5566 * if sk->sk_reuse is too (that is, if the caller requested
5567 * SO_REUSEADDR on this socket -sk-).
5569 if (hlist_empty(&pp->owner)) {
5570 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5571 pp->fastreuse = 1;
5572 else
5573 pp->fastreuse = 0;
5574 } else if (pp->fastreuse &&
5575 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5576 pp->fastreuse = 0;
5578 /* We are set, so fill up all the data in the hash table
5579 * entry, tie the socket list information with the rest of the
5580 * sockets FIXME: Blurry, NPI (ipg).
5582 success:
5583 if (!sctp_sk(sk)->bind_hash) {
5584 inet_sk(sk)->inet_num = snum;
5585 sk_add_bind_node(sk, &pp->owner);
5586 sctp_sk(sk)->bind_hash = pp;
5588 ret = 0;
5590 fail_unlock:
5591 sctp_spin_unlock(&head->lock);
5593 fail:
5594 sctp_local_bh_enable();
5595 return ret;
5598 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5599 * port is requested.
5601 static int sctp_get_port(struct sock *sk, unsigned short snum)
5603 long ret;
5604 union sctp_addr addr;
5605 struct sctp_af *af = sctp_sk(sk)->pf->af;
5607 /* Set up a dummy address struct from the sk. */
5608 af->from_sk(&addr, sk);
5609 addr.v4.sin_port = htons(snum);
5611 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5612 ret = sctp_get_port_local(sk, &addr);
5614 return ret ? 1 : 0;
5618 * Move a socket to LISTENING state.
5620 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5622 struct sctp_sock *sp = sctp_sk(sk);
5623 struct sctp_endpoint *ep = sp->ep;
5624 struct crypto_hash *tfm = NULL;
5626 /* Allocate HMAC for generating cookie. */
5627 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5628 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5629 if (IS_ERR(tfm)) {
5630 if (net_ratelimit()) {
5631 pr_info("failed to load transform for %s: %ld\n",
5632 sctp_hmac_alg, PTR_ERR(tfm));
5634 return -ENOSYS;
5636 sctp_sk(sk)->hmac = tfm;
5640 * If a bind() or sctp_bindx() is not called prior to a listen()
5641 * call that allows new associations to be accepted, the system
5642 * picks an ephemeral port and will choose an address set equivalent
5643 * to binding with a wildcard address.
5645 * This is not currently spelled out in the SCTP sockets
5646 * extensions draft, but follows the practice as seen in TCP
5647 * sockets.
5650 sk->sk_state = SCTP_SS_LISTENING;
5651 if (!ep->base.bind_addr.port) {
5652 if (sctp_autobind(sk))
5653 return -EAGAIN;
5654 } else {
5655 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5656 sk->sk_state = SCTP_SS_CLOSED;
5657 return -EADDRINUSE;
5661 sk->sk_max_ack_backlog = backlog;
5662 sctp_hash_endpoint(ep);
5663 return 0;
5667 * 4.1.3 / 5.1.3 listen()
5669 * By default, new associations are not accepted for UDP style sockets.
5670 * An application uses listen() to mark a socket as being able to
5671 * accept new associations.
5673 * On TCP style sockets, applications use listen() to ready the SCTP
5674 * endpoint for accepting inbound associations.
5676 * On both types of endpoints a backlog of '0' disables listening.
5678 * Move a socket to LISTENING state.
5680 int sctp_inet_listen(struct socket *sock, int backlog)
5682 struct sock *sk = sock->sk;
5683 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5684 int err = -EINVAL;
5686 if (unlikely(backlog < 0))
5687 return err;
5689 sctp_lock_sock(sk);
5691 /* Peeled-off sockets are not allowed to listen(). */
5692 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5693 goto out;
5695 if (sock->state != SS_UNCONNECTED)
5696 goto out;
5698 /* If backlog is zero, disable listening. */
5699 if (!backlog) {
5700 if (sctp_sstate(sk, CLOSED))
5701 goto out;
5703 err = 0;
5704 sctp_unhash_endpoint(ep);
5705 sk->sk_state = SCTP_SS_CLOSED;
5706 if (sk->sk_reuse)
5707 sctp_sk(sk)->bind_hash->fastreuse = 1;
5708 goto out;
5711 /* If we are already listening, just update the backlog */
5712 if (sctp_sstate(sk, LISTENING))
5713 sk->sk_max_ack_backlog = backlog;
5714 else {
5715 err = sctp_listen_start(sk, backlog);
5716 if (err)
5717 goto out;
5720 err = 0;
5721 out:
5722 sctp_release_sock(sk);
5723 return err;
5727 * This function is done by modeling the current datagram_poll() and the
5728 * tcp_poll(). Note that, based on these implementations, we don't
5729 * lock the socket in this function, even though it seems that,
5730 * ideally, locking or some other mechanisms can be used to ensure
5731 * the integrity of the counters (sndbuf and wmem_alloc) used
5732 * in this place. We assume that we don't need locks either until proven
5733 * otherwise.
5735 * Another thing to note is that we include the Async I/O support
5736 * here, again, by modeling the current TCP/UDP code. We don't have
5737 * a good way to test with it yet.
5739 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5741 struct sock *sk = sock->sk;
5742 struct sctp_sock *sp = sctp_sk(sk);
5743 unsigned int mask;
5745 poll_wait(file, sk_sleep(sk), wait);
5747 /* A TCP-style listening socket becomes readable when the accept queue
5748 * is not empty.
5750 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5751 return (!list_empty(&sp->ep->asocs)) ?
5752 (POLLIN | POLLRDNORM) : 0;
5754 mask = 0;
5756 /* Is there any exceptional events? */
5757 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5758 mask |= POLLERR;
5759 if (sk->sk_shutdown & RCV_SHUTDOWN)
5760 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5761 if (sk->sk_shutdown == SHUTDOWN_MASK)
5762 mask |= POLLHUP;
5764 /* Is it readable? Reconsider this code with TCP-style support. */
5765 if (!skb_queue_empty(&sk->sk_receive_queue))
5766 mask |= POLLIN | POLLRDNORM;
5768 /* The association is either gone or not ready. */
5769 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5770 return mask;
5772 /* Is it writable? */
5773 if (sctp_writeable(sk)) {
5774 mask |= POLLOUT | POLLWRNORM;
5775 } else {
5776 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5778 * Since the socket is not locked, the buffer
5779 * might be made available after the writeable check and
5780 * before the bit is set. This could cause a lost I/O
5781 * signal. tcp_poll() has a race breaker for this race
5782 * condition. Based on their implementation, we put
5783 * in the following code to cover it as well.
5785 if (sctp_writeable(sk))
5786 mask |= POLLOUT | POLLWRNORM;
5788 return mask;
5791 /********************************************************************
5792 * 2nd Level Abstractions
5793 ********************************************************************/
5795 static struct sctp_bind_bucket *sctp_bucket_create(
5796 struct sctp_bind_hashbucket *head, unsigned short snum)
5798 struct sctp_bind_bucket *pp;
5800 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
5801 if (pp) {
5802 SCTP_DBG_OBJCNT_INC(bind_bucket);
5803 pp->port = snum;
5804 pp->fastreuse = 0;
5805 INIT_HLIST_HEAD(&pp->owner);
5806 hlist_add_head(&pp->node, &head->chain);
5808 return pp;
5811 /* Caller must hold hashbucket lock for this tb with local BH disabled */
5812 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
5814 if (pp && hlist_empty(&pp->owner)) {
5815 __hlist_del(&pp->node);
5816 kmem_cache_free(sctp_bucket_cachep, pp);
5817 SCTP_DBG_OBJCNT_DEC(bind_bucket);
5821 /* Release this socket's reference to a local port. */
5822 static inline void __sctp_put_port(struct sock *sk)
5824 struct sctp_bind_hashbucket *head =
5825 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
5826 struct sctp_bind_bucket *pp;
5828 sctp_spin_lock(&head->lock);
5829 pp = sctp_sk(sk)->bind_hash;
5830 __sk_del_bind_node(sk);
5831 sctp_sk(sk)->bind_hash = NULL;
5832 inet_sk(sk)->inet_num = 0;
5833 sctp_bucket_destroy(pp);
5834 sctp_spin_unlock(&head->lock);
5837 void sctp_put_port(struct sock *sk)
5839 sctp_local_bh_disable();
5840 __sctp_put_port(sk);
5841 sctp_local_bh_enable();
5845 * The system picks an ephemeral port and choose an address set equivalent
5846 * to binding with a wildcard address.
5847 * One of those addresses will be the primary address for the association.
5848 * This automatically enables the multihoming capability of SCTP.
5850 static int sctp_autobind(struct sock *sk)
5852 union sctp_addr autoaddr;
5853 struct sctp_af *af;
5854 __be16 port;
5856 /* Initialize a local sockaddr structure to INADDR_ANY. */
5857 af = sctp_sk(sk)->pf->af;
5859 port = htons(inet_sk(sk)->inet_num);
5860 af->inaddr_any(&autoaddr, port);
5862 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
5865 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
5867 * From RFC 2292
5868 * 4.2 The cmsghdr Structure *
5870 * When ancillary data is sent or received, any number of ancillary data
5871 * objects can be specified by the msg_control and msg_controllen members of
5872 * the msghdr structure, because each object is preceded by
5873 * a cmsghdr structure defining the object's length (the cmsg_len member).
5874 * Historically Berkeley-derived implementations have passed only one object
5875 * at a time, but this API allows multiple objects to be
5876 * passed in a single call to sendmsg() or recvmsg(). The following example
5877 * shows two ancillary data objects in a control buffer.
5879 * |<--------------------------- msg_controllen -------------------------->|
5880 * | |
5882 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
5884 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
5885 * | | |
5887 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
5889 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
5890 * | | | | |
5892 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5893 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
5895 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
5897 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
5901 * msg_control
5902 * points here
5904 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
5905 sctp_cmsgs_t *cmsgs)
5907 struct cmsghdr *cmsg;
5908 struct msghdr *my_msg = (struct msghdr *)msg;
5910 for (cmsg = CMSG_FIRSTHDR(msg);
5911 cmsg != NULL;
5912 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
5913 if (!CMSG_OK(my_msg, cmsg))
5914 return -EINVAL;
5916 /* Should we parse this header or ignore? */
5917 if (cmsg->cmsg_level != IPPROTO_SCTP)
5918 continue;
5920 /* Strictly check lengths following example in SCM code. */
5921 switch (cmsg->cmsg_type) {
5922 case SCTP_INIT:
5923 /* SCTP Socket API Extension
5924 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
5926 * This cmsghdr structure provides information for
5927 * initializing new SCTP associations with sendmsg().
5928 * The SCTP_INITMSG socket option uses this same data
5929 * structure. This structure is not used for
5930 * recvmsg().
5932 * cmsg_level cmsg_type cmsg_data[]
5933 * ------------ ------------ ----------------------
5934 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
5936 if (cmsg->cmsg_len !=
5937 CMSG_LEN(sizeof(struct sctp_initmsg)))
5938 return -EINVAL;
5939 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
5940 break;
5942 case SCTP_SNDRCV:
5943 /* SCTP Socket API Extension
5944 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
5946 * This cmsghdr structure specifies SCTP options for
5947 * sendmsg() and describes SCTP header information
5948 * about a received message through recvmsg().
5950 * cmsg_level cmsg_type cmsg_data[]
5951 * ------------ ------------ ----------------------
5952 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
5954 if (cmsg->cmsg_len !=
5955 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
5956 return -EINVAL;
5958 cmsgs->info =
5959 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
5961 /* Minimally, validate the sinfo_flags. */
5962 if (cmsgs->info->sinfo_flags &
5963 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
5964 SCTP_ABORT | SCTP_EOF))
5965 return -EINVAL;
5966 break;
5968 default:
5969 return -EINVAL;
5972 return 0;
5976 * Wait for a packet..
5977 * Note: This function is the same function as in core/datagram.c
5978 * with a few modifications to make lksctp work.
5980 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
5982 int error;
5983 DEFINE_WAIT(wait);
5985 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
5987 /* Socket errors? */
5988 error = sock_error(sk);
5989 if (error)
5990 goto out;
5992 if (!skb_queue_empty(&sk->sk_receive_queue))
5993 goto ready;
5995 /* Socket shut down? */
5996 if (sk->sk_shutdown & RCV_SHUTDOWN)
5997 goto out;
5999 /* Sequenced packets can come disconnected. If so we report the
6000 * problem.
6002 error = -ENOTCONN;
6004 /* Is there a good reason to think that we may receive some data? */
6005 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6006 goto out;
6008 /* Handle signals. */
6009 if (signal_pending(current))
6010 goto interrupted;
6012 /* Let another process have a go. Since we are going to sleep
6013 * anyway. Note: This may cause odd behaviors if the message
6014 * does not fit in the user's buffer, but this seems to be the
6015 * only way to honor MSG_DONTWAIT realistically.
6017 sctp_release_sock(sk);
6018 *timeo_p = schedule_timeout(*timeo_p);
6019 sctp_lock_sock(sk);
6021 ready:
6022 finish_wait(sk_sleep(sk), &wait);
6023 return 0;
6025 interrupted:
6026 error = sock_intr_errno(*timeo_p);
6028 out:
6029 finish_wait(sk_sleep(sk), &wait);
6030 *err = error;
6031 return error;
6034 /* Receive a datagram.
6035 * Note: This is pretty much the same routine as in core/datagram.c
6036 * with a few changes to make lksctp work.
6038 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6039 int noblock, int *err)
6041 int error;
6042 struct sk_buff *skb;
6043 long timeo;
6045 timeo = sock_rcvtimeo(sk, noblock);
6047 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6048 timeo, MAX_SCHEDULE_TIMEOUT);
6050 do {
6051 /* Again only user level code calls this function,
6052 * so nothing interrupt level
6053 * will suddenly eat the receive_queue.
6055 * Look at current nfs client by the way...
6056 * However, this function was correct in any case. 8)
6058 if (flags & MSG_PEEK) {
6059 spin_lock_bh(&sk->sk_receive_queue.lock);
6060 skb = skb_peek(&sk->sk_receive_queue);
6061 if (skb)
6062 atomic_inc(&skb->users);
6063 spin_unlock_bh(&sk->sk_receive_queue.lock);
6064 } else {
6065 skb = skb_dequeue(&sk->sk_receive_queue);
6068 if (skb)
6069 return skb;
6071 /* Caller is allowed not to check sk->sk_err before calling. */
6072 error = sock_error(sk);
6073 if (error)
6074 goto no_packet;
6076 if (sk->sk_shutdown & RCV_SHUTDOWN)
6077 break;
6079 /* User doesn't want to wait. */
6080 error = -EAGAIN;
6081 if (!timeo)
6082 goto no_packet;
6083 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6085 return NULL;
6087 no_packet:
6088 *err = error;
6089 return NULL;
6092 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6093 static void __sctp_write_space(struct sctp_association *asoc)
6095 struct sock *sk = asoc->base.sk;
6096 struct socket *sock = sk->sk_socket;
6098 if ((sctp_wspace(asoc) > 0) && sock) {
6099 if (waitqueue_active(&asoc->wait))
6100 wake_up_interruptible(&asoc->wait);
6102 if (sctp_writeable(sk)) {
6103 wait_queue_head_t *wq = sk_sleep(sk);
6105 if (wq && waitqueue_active(wq))
6106 wake_up_interruptible(wq);
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 (!(sk->sk_shutdown & SEND_SHUTDOWN))
6113 sock_wake_async(sock,
6114 SOCK_WAKE_SPACE, POLL_OUT);
6119 /* Do accounting for the sndbuf space.
6120 * Decrement the used sndbuf space of the corresponding association by the
6121 * data size which was just transmitted(freed).
6123 static void sctp_wfree(struct sk_buff *skb)
6125 struct sctp_association *asoc;
6126 struct sctp_chunk *chunk;
6127 struct sock *sk;
6129 /* Get the saved chunk pointer. */
6130 chunk = *((struct sctp_chunk **)(skb->cb));
6131 asoc = chunk->asoc;
6132 sk = asoc->base.sk;
6133 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6134 sizeof(struct sk_buff) +
6135 sizeof(struct sctp_chunk);
6137 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6140 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6142 sk->sk_wmem_queued -= skb->truesize;
6143 sk_mem_uncharge(sk, skb->truesize);
6145 sock_wfree(skb);
6146 __sctp_write_space(asoc);
6148 sctp_association_put(asoc);
6151 /* Do accounting for the receive space on the socket.
6152 * Accounting for the association is done in ulpevent.c
6153 * We set this as a destructor for the cloned data skbs so that
6154 * accounting is done at the correct time.
6156 void sctp_sock_rfree(struct sk_buff *skb)
6158 struct sock *sk = skb->sk;
6159 struct sctp_ulpevent *event = sctp_skb2event(skb);
6161 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6164 * Mimic the behavior of sock_rfree
6166 sk_mem_uncharge(sk, event->rmem_len);
6170 /* Helper function to wait for space in the sndbuf. */
6171 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6172 size_t msg_len)
6174 struct sock *sk = asoc->base.sk;
6175 int err = 0;
6176 long current_timeo = *timeo_p;
6177 DEFINE_WAIT(wait);
6179 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6180 asoc, (long)(*timeo_p), msg_len);
6182 /* Increment the association's refcnt. */
6183 sctp_association_hold(asoc);
6185 /* Wait on the association specific sndbuf space. */
6186 for (;;) {
6187 prepare_to_wait_exclusive(&asoc->wait, &wait,
6188 TASK_INTERRUPTIBLE);
6189 if (!*timeo_p)
6190 goto do_nonblock;
6191 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6192 asoc->base.dead)
6193 goto do_error;
6194 if (signal_pending(current))
6195 goto do_interrupted;
6196 if (msg_len <= sctp_wspace(asoc))
6197 break;
6199 /* Let another process have a go. Since we are going
6200 * to sleep anyway.
6202 sctp_release_sock(sk);
6203 current_timeo = schedule_timeout(current_timeo);
6204 BUG_ON(sk != asoc->base.sk);
6205 sctp_lock_sock(sk);
6207 *timeo_p = current_timeo;
6210 out:
6211 finish_wait(&asoc->wait, &wait);
6213 /* Release the association's refcnt. */
6214 sctp_association_put(asoc);
6216 return err;
6218 do_error:
6219 err = -EPIPE;
6220 goto out;
6222 do_interrupted:
6223 err = sock_intr_errno(*timeo_p);
6224 goto out;
6226 do_nonblock:
6227 err = -EAGAIN;
6228 goto out;
6231 void sctp_data_ready(struct sock *sk, int len)
6233 struct socket_wq *wq;
6235 rcu_read_lock();
6236 wq = rcu_dereference(sk->sk_wq);
6237 if (wq_has_sleeper(wq))
6238 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6239 POLLRDNORM | POLLRDBAND);
6240 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6241 rcu_read_unlock();
6244 /* If socket sndbuf has changed, wake up all per association waiters. */
6245 void sctp_write_space(struct sock *sk)
6247 struct sctp_association *asoc;
6249 /* Wake up the tasks in each wait queue. */
6250 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6251 __sctp_write_space(asoc);
6255 /* Is there any sndbuf space available on the socket?
6257 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6258 * associations on the same socket. For a UDP-style socket with
6259 * multiple associations, it is possible for it to be "unwriteable"
6260 * prematurely. I assume that this is acceptable because
6261 * a premature "unwriteable" is better than an accidental "writeable" which
6262 * would cause an unwanted block under certain circumstances. For the 1-1
6263 * UDP-style sockets or TCP-style sockets, this code should work.
6264 * - Daisy
6266 static int sctp_writeable(struct sock *sk)
6268 int amt = 0;
6270 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6271 if (amt < 0)
6272 amt = 0;
6273 return amt;
6276 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6277 * returns immediately with EINPROGRESS.
6279 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6281 struct sock *sk = asoc->base.sk;
6282 int err = 0;
6283 long current_timeo = *timeo_p;
6284 DEFINE_WAIT(wait);
6286 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6287 (long)(*timeo_p));
6289 /* Increment the association's refcnt. */
6290 sctp_association_hold(asoc);
6292 for (;;) {
6293 prepare_to_wait_exclusive(&asoc->wait, &wait,
6294 TASK_INTERRUPTIBLE);
6295 if (!*timeo_p)
6296 goto do_nonblock;
6297 if (sk->sk_shutdown & RCV_SHUTDOWN)
6298 break;
6299 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6300 asoc->base.dead)
6301 goto do_error;
6302 if (signal_pending(current))
6303 goto do_interrupted;
6305 if (sctp_state(asoc, ESTABLISHED))
6306 break;
6308 /* Let another process have a go. Since we are going
6309 * to sleep anyway.
6311 sctp_release_sock(sk);
6312 current_timeo = schedule_timeout(current_timeo);
6313 sctp_lock_sock(sk);
6315 *timeo_p = current_timeo;
6318 out:
6319 finish_wait(&asoc->wait, &wait);
6321 /* Release the association's refcnt. */
6322 sctp_association_put(asoc);
6324 return err;
6326 do_error:
6327 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6328 err = -ETIMEDOUT;
6329 else
6330 err = -ECONNREFUSED;
6331 goto out;
6333 do_interrupted:
6334 err = sock_intr_errno(*timeo_p);
6335 goto out;
6337 do_nonblock:
6338 err = -EINPROGRESS;
6339 goto out;
6342 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6344 struct sctp_endpoint *ep;
6345 int err = 0;
6346 DEFINE_WAIT(wait);
6348 ep = sctp_sk(sk)->ep;
6351 for (;;) {
6352 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6353 TASK_INTERRUPTIBLE);
6355 if (list_empty(&ep->asocs)) {
6356 sctp_release_sock(sk);
6357 timeo = schedule_timeout(timeo);
6358 sctp_lock_sock(sk);
6361 err = -EINVAL;
6362 if (!sctp_sstate(sk, LISTENING))
6363 break;
6365 err = 0;
6366 if (!list_empty(&ep->asocs))
6367 break;
6369 err = sock_intr_errno(timeo);
6370 if (signal_pending(current))
6371 break;
6373 err = -EAGAIN;
6374 if (!timeo)
6375 break;
6378 finish_wait(sk_sleep(sk), &wait);
6380 return err;
6383 static void sctp_wait_for_close(struct sock *sk, long timeout)
6385 DEFINE_WAIT(wait);
6387 do {
6388 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6389 if (list_empty(&sctp_sk(sk)->ep->asocs))
6390 break;
6391 sctp_release_sock(sk);
6392 timeout = schedule_timeout(timeout);
6393 sctp_lock_sock(sk);
6394 } while (!signal_pending(current) && timeout);
6396 finish_wait(sk_sleep(sk), &wait);
6399 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6401 struct sk_buff *frag;
6403 if (!skb->data_len)
6404 goto done;
6406 /* Don't forget the fragments. */
6407 skb_walk_frags(skb, frag)
6408 sctp_skb_set_owner_r_frag(frag, sk);
6410 done:
6411 sctp_skb_set_owner_r(skb, sk);
6414 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6415 struct sctp_association *asoc)
6417 struct inet_sock *inet = inet_sk(sk);
6418 struct inet_sock *newinet;
6420 newsk->sk_type = sk->sk_type;
6421 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6422 newsk->sk_flags = sk->sk_flags;
6423 newsk->sk_no_check = sk->sk_no_check;
6424 newsk->sk_reuse = sk->sk_reuse;
6426 newsk->sk_shutdown = sk->sk_shutdown;
6427 newsk->sk_destruct = inet_sock_destruct;
6428 newsk->sk_family = sk->sk_family;
6429 newsk->sk_protocol = IPPROTO_SCTP;
6430 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6431 newsk->sk_sndbuf = sk->sk_sndbuf;
6432 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6433 newsk->sk_lingertime = sk->sk_lingertime;
6434 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6435 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6437 newinet = inet_sk(newsk);
6439 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6440 * getsockname() and getpeername()
6442 newinet->inet_sport = inet->inet_sport;
6443 newinet->inet_saddr = inet->inet_saddr;
6444 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6445 newinet->inet_dport = htons(asoc->peer.port);
6446 newinet->pmtudisc = inet->pmtudisc;
6447 newinet->inet_id = asoc->next_tsn ^ jiffies;
6449 newinet->uc_ttl = inet->uc_ttl;
6450 newinet->mc_loop = 1;
6451 newinet->mc_ttl = 1;
6452 newinet->mc_index = 0;
6453 newinet->mc_list = NULL;
6456 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6457 * and its messages to the newsk.
6459 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6460 struct sctp_association *assoc,
6461 sctp_socket_type_t type)
6463 struct sctp_sock *oldsp = sctp_sk(oldsk);
6464 struct sctp_sock *newsp = sctp_sk(newsk);
6465 struct sctp_bind_bucket *pp; /* hash list port iterator */
6466 struct sctp_endpoint *newep = newsp->ep;
6467 struct sk_buff *skb, *tmp;
6468 struct sctp_ulpevent *event;
6469 struct sctp_bind_hashbucket *head;
6471 /* Migrate socket buffer sizes and all the socket level options to the
6472 * new socket.
6474 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6475 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6476 /* Brute force copy old sctp opt. */
6477 inet_sk_copy_descendant(newsk, oldsk);
6479 /* Restore the ep value that was overwritten with the above structure
6480 * copy.
6482 newsp->ep = newep;
6483 newsp->hmac = NULL;
6485 /* Hook this new socket in to the bind_hash list. */
6486 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6487 sctp_local_bh_disable();
6488 sctp_spin_lock(&head->lock);
6489 pp = sctp_sk(oldsk)->bind_hash;
6490 sk_add_bind_node(newsk, &pp->owner);
6491 sctp_sk(newsk)->bind_hash = pp;
6492 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6493 sctp_spin_unlock(&head->lock);
6494 sctp_local_bh_enable();
6496 /* Copy the bind_addr list from the original endpoint to the new
6497 * endpoint so that we can handle restarts properly
6499 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6500 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6502 /* Move any messages in the old socket's receive queue that are for the
6503 * peeled off association to the new socket's receive queue.
6505 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6506 event = sctp_skb2event(skb);
6507 if (event->asoc == assoc) {
6508 __skb_unlink(skb, &oldsk->sk_receive_queue);
6509 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6510 sctp_skb_set_owner_r_frag(skb, newsk);
6514 /* Clean up any messages pending delivery due to partial
6515 * delivery. Three cases:
6516 * 1) No partial deliver; no work.
6517 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6518 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6520 skb_queue_head_init(&newsp->pd_lobby);
6521 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6523 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6524 struct sk_buff_head *queue;
6526 /* Decide which queue to move pd_lobby skbs to. */
6527 if (assoc->ulpq.pd_mode) {
6528 queue = &newsp->pd_lobby;
6529 } else
6530 queue = &newsk->sk_receive_queue;
6532 /* Walk through the pd_lobby, looking for skbs that
6533 * need moved to the new socket.
6535 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6536 event = sctp_skb2event(skb);
6537 if (event->asoc == assoc) {
6538 __skb_unlink(skb, &oldsp->pd_lobby);
6539 __skb_queue_tail(queue, skb);
6540 sctp_skb_set_owner_r_frag(skb, newsk);
6544 /* Clear up any skbs waiting for the partial
6545 * delivery to finish.
6547 if (assoc->ulpq.pd_mode)
6548 sctp_clear_pd(oldsk, NULL);
6552 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6553 sctp_skb_set_owner_r_frag(skb, newsk);
6555 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6556 sctp_skb_set_owner_r_frag(skb, newsk);
6558 /* Set the type of socket to indicate that it is peeled off from the
6559 * original UDP-style socket or created with the accept() call on a
6560 * TCP-style socket..
6562 newsp->type = type;
6564 /* Mark the new socket "in-use" by the user so that any packets
6565 * that may arrive on the association after we've moved it are
6566 * queued to the backlog. This prevents a potential race between
6567 * backlog processing on the old socket and new-packet processing
6568 * on the new socket.
6570 * The caller has just allocated newsk so we can guarantee that other
6571 * paths won't try to lock it and then oldsk.
6573 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6574 sctp_assoc_migrate(assoc, newsk);
6576 /* If the association on the newsk is already closed before accept()
6577 * is called, set RCV_SHUTDOWN flag.
6579 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6580 newsk->sk_shutdown |= RCV_SHUTDOWN;
6582 newsk->sk_state = SCTP_SS_ESTABLISHED;
6583 sctp_release_sock(newsk);
6587 /* This proto struct describes the ULP interface for SCTP. */
6588 struct proto sctp_prot = {
6589 .name = "SCTP",
6590 .owner = THIS_MODULE,
6591 .close = sctp_close,
6592 .connect = sctp_connect,
6593 .disconnect = sctp_disconnect,
6594 .accept = sctp_accept,
6595 .ioctl = sctp_ioctl,
6596 .init = sctp_init_sock,
6597 .destroy = sctp_destroy_sock,
6598 .shutdown = sctp_shutdown,
6599 .setsockopt = sctp_setsockopt,
6600 .getsockopt = sctp_getsockopt,
6601 .sendmsg = sctp_sendmsg,
6602 .recvmsg = sctp_recvmsg,
6603 .bind = sctp_bind,
6604 .backlog_rcv = sctp_backlog_rcv,
6605 .hash = sctp_hash,
6606 .unhash = sctp_unhash,
6607 .get_port = sctp_get_port,
6608 .obj_size = sizeof(struct sctp_sock),
6609 .sysctl_mem = sysctl_sctp_mem,
6610 .sysctl_rmem = sysctl_sctp_rmem,
6611 .sysctl_wmem = sysctl_sctp_wmem,
6612 .memory_pressure = &sctp_memory_pressure,
6613 .enter_memory_pressure = sctp_enter_memory_pressure,
6614 .memory_allocated = &sctp_memory_allocated,
6615 .sockets_allocated = &sctp_sockets_allocated,
6618 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6620 struct proto sctpv6_prot = {
6621 .name = "SCTPv6",
6622 .owner = THIS_MODULE,
6623 .close = sctp_close,
6624 .connect = sctp_connect,
6625 .disconnect = sctp_disconnect,
6626 .accept = sctp_accept,
6627 .ioctl = sctp_ioctl,
6628 .init = sctp_init_sock,
6629 .destroy = sctp_destroy_sock,
6630 .shutdown = sctp_shutdown,
6631 .setsockopt = sctp_setsockopt,
6632 .getsockopt = sctp_getsockopt,
6633 .sendmsg = sctp_sendmsg,
6634 .recvmsg = sctp_recvmsg,
6635 .bind = sctp_bind,
6636 .backlog_rcv = sctp_backlog_rcv,
6637 .hash = sctp_hash,
6638 .unhash = sctp_unhash,
6639 .get_port = sctp_get_port,
6640 .obj_size = sizeof(struct sctp6_sock),
6641 .sysctl_mem = sysctl_sctp_mem,
6642 .sysctl_rmem = sysctl_sctp_rmem,
6643 .sysctl_wmem = sysctl_sctp_wmem,
6644 .memory_pressure = &sctp_memory_pressure,
6645 .enter_memory_pressure = sctp_enter_memory_pressure,
6646 .memory_allocated = &sctp_memory_allocated,
6647 .sockets_allocated = &sctp_sockets_allocated,
6649 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */