drm/radeon/kms: allocate vram scratch page on 6xx+
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / sctp / socket.c
blob836aa63ee121977d820e4f648135e65e3dd2f5d8
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 = 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 = 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 /* Add the new addresses to the bind address list with
587 * use_as_src set to 0.
589 addr_buf = addrs;
590 for (i = 0; i < addrcnt; i++) {
591 addr = addr_buf;
592 af = sctp_get_af_specific(addr->v4.sin_family);
593 memcpy(&saveaddr, addr, af->sockaddr_len);
594 retval = sctp_add_bind_addr(bp, &saveaddr,
595 SCTP_ADDR_NEW, GFP_ATOMIC);
596 addr_buf += af->sockaddr_len;
598 if (asoc->src_out_of_asoc_ok) {
599 struct sctp_transport *trans;
601 list_for_each_entry(trans,
602 &asoc->peer.transport_addr_list, transports) {
603 /* Clear the source and route cache */
604 dst_release(trans->dst);
605 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
606 2*asoc->pathmtu, 4380));
607 trans->ssthresh = asoc->peer.i.a_rwnd;
608 trans->rto = asoc->rto_initial;
609 trans->rtt = trans->srtt = trans->rttvar = 0;
610 sctp_transport_route(trans, NULL,
611 sctp_sk(asoc->base.sk));
614 retval = sctp_send_asconf(asoc, chunk);
617 out:
618 return retval;
621 /* Remove a list of addresses from bind addresses list. Do not remove the
622 * last address.
624 * Basically run through each address specified in the addrs/addrcnt
625 * array/length pair, determine if it is IPv6 or IPv4 and call
626 * sctp_del_bind() on it.
628 * If any of them fails, then the operation will be reversed and the
629 * ones that were removed will be added back.
631 * At least one address has to be left; if only one address is
632 * available, the operation will return -EBUSY.
634 * Only sctp_setsockopt_bindx() is supposed to call this function.
636 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
638 struct sctp_sock *sp = sctp_sk(sk);
639 struct sctp_endpoint *ep = sp->ep;
640 int cnt;
641 struct sctp_bind_addr *bp = &ep->base.bind_addr;
642 int retval = 0;
643 void *addr_buf;
644 union sctp_addr *sa_addr;
645 struct sctp_af *af;
647 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
648 sk, addrs, addrcnt);
650 addr_buf = addrs;
651 for (cnt = 0; cnt < addrcnt; cnt++) {
652 /* If the bind address list is empty or if there is only one
653 * bind address, there is nothing more to be removed (we need
654 * at least one address here).
656 if (list_empty(&bp->address_list) ||
657 (sctp_list_single_entry(&bp->address_list))) {
658 retval = -EBUSY;
659 goto err_bindx_rem;
662 sa_addr = addr_buf;
663 af = sctp_get_af_specific(sa_addr->sa.sa_family);
664 if (!af) {
665 retval = -EINVAL;
666 goto err_bindx_rem;
669 if (!af->addr_valid(sa_addr, sp, NULL)) {
670 retval = -EADDRNOTAVAIL;
671 goto err_bindx_rem;
674 if (sa_addr->v4.sin_port &&
675 sa_addr->v4.sin_port != htons(bp->port)) {
676 retval = -EINVAL;
677 goto err_bindx_rem;
680 if (!sa_addr->v4.sin_port)
681 sa_addr->v4.sin_port = htons(bp->port);
683 /* FIXME - There is probably a need to check if sk->sk_saddr and
684 * sk->sk_rcv_addr are currently set to one of the addresses to
685 * be removed. This is something which needs to be looked into
686 * when we are fixing the outstanding issues with multi-homing
687 * socket routing and failover schemes. Refer to comments in
688 * sctp_do_bind(). -daisy
690 retval = sctp_del_bind_addr(bp, sa_addr);
692 addr_buf += af->sockaddr_len;
693 err_bindx_rem:
694 if (retval < 0) {
695 /* Failed. Add the ones that has been removed back */
696 if (cnt > 0)
697 sctp_bindx_add(sk, addrs, cnt);
698 return retval;
702 return retval;
705 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
706 * the associations that are part of the endpoint indicating that a list of
707 * local addresses are removed from the endpoint.
709 * If any of the addresses is already in the bind address list of the
710 * association, we do not send the chunk for that association. But it will not
711 * affect other associations.
713 * Only sctp_setsockopt_bindx() is supposed to call this function.
715 static int sctp_send_asconf_del_ip(struct sock *sk,
716 struct sockaddr *addrs,
717 int addrcnt)
719 struct sctp_sock *sp;
720 struct sctp_endpoint *ep;
721 struct sctp_association *asoc;
722 struct sctp_transport *transport;
723 struct sctp_bind_addr *bp;
724 struct sctp_chunk *chunk;
725 union sctp_addr *laddr;
726 void *addr_buf;
727 struct sctp_af *af;
728 struct sctp_sockaddr_entry *saddr;
729 int i;
730 int retval = 0;
731 int stored = 0;
733 chunk = NULL;
734 if (!sctp_addip_enable)
735 return retval;
737 sp = sctp_sk(sk);
738 ep = sp->ep;
740 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
741 __func__, sk, addrs, addrcnt);
743 list_for_each_entry(asoc, &ep->asocs, asocs) {
745 if (!asoc->peer.asconf_capable)
746 continue;
748 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
749 continue;
751 if (!sctp_state(asoc, ESTABLISHED))
752 continue;
754 /* Check if any address in the packed array of addresses is
755 * not present in the bind address list of the association.
756 * If so, do not send the asconf chunk to its peer, but
757 * continue with other associations.
759 addr_buf = addrs;
760 for (i = 0; i < addrcnt; i++) {
761 laddr = addr_buf;
762 af = sctp_get_af_specific(laddr->v4.sin_family);
763 if (!af) {
764 retval = -EINVAL;
765 goto out;
768 if (!sctp_assoc_lookup_laddr(asoc, laddr))
769 break;
771 addr_buf += af->sockaddr_len;
773 if (i < addrcnt)
774 continue;
776 /* Find one address in the association's bind address list
777 * that is not in the packed array of addresses. This is to
778 * make sure that we do not delete all the addresses in the
779 * association.
781 bp = &asoc->base.bind_addr;
782 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
783 addrcnt, sp);
784 if ((laddr == NULL) && (addrcnt == 1)) {
785 if (asoc->asconf_addr_del_pending)
786 continue;
787 asoc->asconf_addr_del_pending =
788 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
789 if (asoc->asconf_addr_del_pending == NULL) {
790 retval = -ENOMEM;
791 goto out;
793 asoc->asconf_addr_del_pending->sa.sa_family =
794 addrs->sa_family;
795 asoc->asconf_addr_del_pending->v4.sin_port =
796 htons(bp->port);
797 if (addrs->sa_family == AF_INET) {
798 struct sockaddr_in *sin;
800 sin = (struct sockaddr_in *)addrs;
801 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
802 } else if (addrs->sa_family == AF_INET6) {
803 struct sockaddr_in6 *sin6;
805 sin6 = (struct sockaddr_in6 *)addrs;
806 ipv6_addr_copy(&asoc->asconf_addr_del_pending->v6.sin6_addr, &sin6->sin6_addr);
808 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
809 " at %p\n", asoc, asoc->asconf_addr_del_pending,
810 asoc->asconf_addr_del_pending);
811 asoc->src_out_of_asoc_ok = 1;
812 stored = 1;
813 goto skip_mkasconf;
816 /* We do not need RCU protection throughout this loop
817 * because this is done under a socket lock from the
818 * setsockopt call.
820 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
821 SCTP_PARAM_DEL_IP);
822 if (!chunk) {
823 retval = -ENOMEM;
824 goto out;
827 skip_mkasconf:
828 /* Reset use_as_src flag for the addresses in the bind address
829 * list that are to be deleted.
831 addr_buf = addrs;
832 for (i = 0; i < addrcnt; i++) {
833 laddr = addr_buf;
834 af = sctp_get_af_specific(laddr->v4.sin_family);
835 list_for_each_entry(saddr, &bp->address_list, list) {
836 if (sctp_cmp_addr_exact(&saddr->a, laddr))
837 saddr->state = SCTP_ADDR_DEL;
839 addr_buf += af->sockaddr_len;
842 /* Update the route and saddr entries for all the transports
843 * as some of the addresses in the bind address list are
844 * about to be deleted and cannot be used as source addresses.
846 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
847 transports) {
848 dst_release(transport->dst);
849 sctp_transport_route(transport, NULL,
850 sctp_sk(asoc->base.sk));
853 if (stored)
854 /* We don't need to transmit ASCONF */
855 continue;
856 retval = sctp_send_asconf(asoc, chunk);
858 out:
859 return retval;
862 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
863 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
865 struct sock *sk = sctp_opt2sk(sp);
866 union sctp_addr *addr;
867 struct sctp_af *af;
869 /* It is safe to write port space in caller. */
870 addr = &addrw->a;
871 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
872 af = sctp_get_af_specific(addr->sa.sa_family);
873 if (!af)
874 return -EINVAL;
875 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
876 return -EINVAL;
878 if (addrw->state == SCTP_ADDR_NEW)
879 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
880 else
881 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
884 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
886 * API 8.1
887 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
888 * int flags);
890 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
891 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
892 * or IPv6 addresses.
894 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
895 * Section 3.1.2 for this usage.
897 * addrs is a pointer to an array of one or more socket addresses. Each
898 * address is contained in its appropriate structure (i.e. struct
899 * sockaddr_in or struct sockaddr_in6) the family of the address type
900 * must be used to distinguish the address length (note that this
901 * representation is termed a "packed array" of addresses). The caller
902 * specifies the number of addresses in the array with addrcnt.
904 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
905 * -1, and sets errno to the appropriate error code.
907 * For SCTP, the port given in each socket address must be the same, or
908 * sctp_bindx() will fail, setting errno to EINVAL.
910 * The flags parameter is formed from the bitwise OR of zero or more of
911 * the following currently defined flags:
913 * SCTP_BINDX_ADD_ADDR
915 * SCTP_BINDX_REM_ADDR
917 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
918 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
919 * addresses from the association. The two flags are mutually exclusive;
920 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
921 * not remove all addresses from an association; sctp_bindx() will
922 * reject such an attempt with EINVAL.
924 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
925 * additional addresses with an endpoint after calling bind(). Or use
926 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
927 * socket is associated with so that no new association accepted will be
928 * associated with those addresses. If the endpoint supports dynamic
929 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
930 * endpoint to send the appropriate message to the peer to change the
931 * peers address lists.
933 * Adding and removing addresses from a connected association is
934 * optional functionality. Implementations that do not support this
935 * functionality should return EOPNOTSUPP.
937 * Basically do nothing but copying the addresses from user to kernel
938 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
939 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
940 * from userspace.
942 * We don't use copy_from_user() for optimization: we first do the
943 * sanity checks (buffer size -fast- and access check-healthy
944 * pointer); if all of those succeed, then we can alloc the memory
945 * (expensive operation) needed to copy the data to kernel. Then we do
946 * the copying without checking the user space area
947 * (__copy_from_user()).
949 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
950 * it.
952 * sk The sk of the socket
953 * addrs The pointer to the addresses in user land
954 * addrssize Size of the addrs buffer
955 * op Operation to perform (add or remove, see the flags of
956 * sctp_bindx)
958 * Returns 0 if ok, <0 errno code on error.
960 SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
961 struct sockaddr __user *addrs,
962 int addrs_size, int op)
964 struct sockaddr *kaddrs;
965 int err;
966 int addrcnt = 0;
967 int walk_size = 0;
968 struct sockaddr *sa_addr;
969 void *addr_buf;
970 struct sctp_af *af;
972 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
973 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
975 if (unlikely(addrs_size <= 0))
976 return -EINVAL;
978 /* Check the user passed a healthy pointer. */
979 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
980 return -EFAULT;
982 /* Alloc space for the address array in kernel memory. */
983 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
984 if (unlikely(!kaddrs))
985 return -ENOMEM;
987 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
988 kfree(kaddrs);
989 return -EFAULT;
992 /* Walk through the addrs buffer and count the number of addresses. */
993 addr_buf = kaddrs;
994 while (walk_size < addrs_size) {
995 if (walk_size + sizeof(sa_family_t) > addrs_size) {
996 kfree(kaddrs);
997 return -EINVAL;
1000 sa_addr = addr_buf;
1001 af = sctp_get_af_specific(sa_addr->sa_family);
1003 /* If the address family is not supported or if this address
1004 * causes the address buffer to overflow return EINVAL.
1006 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1007 kfree(kaddrs);
1008 return -EINVAL;
1010 addrcnt++;
1011 addr_buf += af->sockaddr_len;
1012 walk_size += af->sockaddr_len;
1015 /* Do the work. */
1016 switch (op) {
1017 case SCTP_BINDX_ADD_ADDR:
1018 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1019 if (err)
1020 goto out;
1021 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1022 break;
1024 case SCTP_BINDX_REM_ADDR:
1025 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1026 if (err)
1027 goto out;
1028 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1029 break;
1031 default:
1032 err = -EINVAL;
1033 break;
1036 out:
1037 kfree(kaddrs);
1039 return err;
1042 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1044 * Common routine for handling connect() and sctp_connectx().
1045 * Connect will come in with just a single address.
1047 static int __sctp_connect(struct sock* sk,
1048 struct sockaddr *kaddrs,
1049 int addrs_size,
1050 sctp_assoc_t *assoc_id)
1052 struct sctp_sock *sp;
1053 struct sctp_endpoint *ep;
1054 struct sctp_association *asoc = NULL;
1055 struct sctp_association *asoc2;
1056 struct sctp_transport *transport;
1057 union sctp_addr to;
1058 struct sctp_af *af;
1059 sctp_scope_t scope;
1060 long timeo;
1061 int err = 0;
1062 int addrcnt = 0;
1063 int walk_size = 0;
1064 union sctp_addr *sa_addr = NULL;
1065 void *addr_buf;
1066 unsigned short port;
1067 unsigned int f_flags = 0;
1069 sp = sctp_sk(sk);
1070 ep = sp->ep;
1072 /* connect() cannot be done on a socket that is already in ESTABLISHED
1073 * state - UDP-style peeled off socket or a TCP-style socket that
1074 * is already connected.
1075 * It cannot be done even on a TCP-style listening socket.
1077 if (sctp_sstate(sk, ESTABLISHED) ||
1078 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1079 err = -EISCONN;
1080 goto out_free;
1083 /* Walk through the addrs buffer and count the number of addresses. */
1084 addr_buf = kaddrs;
1085 while (walk_size < addrs_size) {
1086 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1087 err = -EINVAL;
1088 goto out_free;
1091 sa_addr = addr_buf;
1092 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1094 /* If the address family is not supported or if this address
1095 * causes the address buffer to overflow return EINVAL.
1097 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1098 err = -EINVAL;
1099 goto out_free;
1102 port = ntohs(sa_addr->v4.sin_port);
1104 /* Save current address so we can work with it */
1105 memcpy(&to, sa_addr, af->sockaddr_len);
1107 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1108 if (err)
1109 goto out_free;
1111 /* Make sure the destination port is correctly set
1112 * in all addresses.
1114 if (asoc && asoc->peer.port && asoc->peer.port != port)
1115 goto out_free;
1118 /* Check if there already is a matching association on the
1119 * endpoint (other than the one created here).
1121 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1122 if (asoc2 && asoc2 != asoc) {
1123 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1124 err = -EISCONN;
1125 else
1126 err = -EALREADY;
1127 goto out_free;
1130 /* If we could not find a matching association on the endpoint,
1131 * make sure that there is no peeled-off association matching
1132 * the peer address even on another socket.
1134 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1135 err = -EADDRNOTAVAIL;
1136 goto out_free;
1139 if (!asoc) {
1140 /* If a bind() or sctp_bindx() is not called prior to
1141 * an sctp_connectx() call, the system picks an
1142 * ephemeral port and will choose an address set
1143 * equivalent to binding with a wildcard address.
1145 if (!ep->base.bind_addr.port) {
1146 if (sctp_autobind(sk)) {
1147 err = -EAGAIN;
1148 goto out_free;
1150 } else {
1152 * If an unprivileged user inherits a 1-many
1153 * style socket with open associations on a
1154 * privileged port, it MAY be permitted to
1155 * accept new associations, but it SHOULD NOT
1156 * be permitted to open new associations.
1158 if (ep->base.bind_addr.port < PROT_SOCK &&
1159 !capable(CAP_NET_BIND_SERVICE)) {
1160 err = -EACCES;
1161 goto out_free;
1165 scope = sctp_scope(&to);
1166 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1167 if (!asoc) {
1168 err = -ENOMEM;
1169 goto out_free;
1172 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1173 GFP_KERNEL);
1174 if (err < 0) {
1175 goto out_free;
1180 /* Prime the peer's transport structures. */
1181 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1182 SCTP_UNKNOWN);
1183 if (!transport) {
1184 err = -ENOMEM;
1185 goto out_free;
1188 addrcnt++;
1189 addr_buf += af->sockaddr_len;
1190 walk_size += af->sockaddr_len;
1193 /* In case the user of sctp_connectx() wants an association
1194 * id back, assign one now.
1196 if (assoc_id) {
1197 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1198 if (err < 0)
1199 goto out_free;
1202 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1203 if (err < 0) {
1204 goto out_free;
1207 /* Initialize sk's dport and daddr for getpeername() */
1208 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1209 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1210 af->to_sk_daddr(sa_addr, sk);
1211 sk->sk_err = 0;
1213 /* in-kernel sockets don't generally have a file allocated to them
1214 * if all they do is call sock_create_kern().
1216 if (sk->sk_socket->file)
1217 f_flags = sk->sk_socket->file->f_flags;
1219 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1221 err = sctp_wait_for_connect(asoc, &timeo);
1222 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1223 *assoc_id = asoc->assoc_id;
1225 /* Don't free association on exit. */
1226 asoc = NULL;
1228 out_free:
1230 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1231 " kaddrs: %p err: %d\n",
1232 asoc, kaddrs, err);
1233 if (asoc)
1234 sctp_association_free(asoc);
1235 return err;
1238 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1240 * API 8.9
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1292 * it.
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1298 * Returns >=0 if ok, <0 errno code on error.
1300 SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1301 struct sockaddr __user *addrs,
1302 int addrs_size,
1303 sctp_assoc_t *assoc_id)
1305 int err = 0;
1306 struct sockaddr *kaddrs;
1308 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1309 __func__, sk, addrs, addrs_size);
1311 if (unlikely(addrs_size <= 0))
1312 return -EINVAL;
1314 /* Check the user passed a healthy pointer. */
1315 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1316 return -EFAULT;
1318 /* Alloc space for the address array in kernel memory. */
1319 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1320 if (unlikely(!kaddrs))
1321 return -ENOMEM;
1323 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1324 err = -EFAULT;
1325 } else {
1326 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1329 kfree(kaddrs);
1331 return err;
1335 * This is an older interface. It's kept for backward compatibility
1336 * to the option that doesn't provide association id.
1338 SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1339 struct sockaddr __user *addrs,
1340 int addrs_size)
1342 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1346 * New interface for the API. The since the API is done with a socket
1347 * option, to make it simple we feed back the association id is as a return
1348 * indication to the call. Error is always negative and association id is
1349 * always positive.
1351 SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1352 struct sockaddr __user *addrs,
1353 int addrs_size)
1355 sctp_assoc_t assoc_id = 0;
1356 int err = 0;
1358 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1360 if (err)
1361 return err;
1362 else
1363 return assoc_id;
1367 * New (hopefully final) interface for the API.
1368 * We use the sctp_getaddrs_old structure so that use-space library
1369 * can avoid any unnecessary allocations. The only defferent part
1370 * is that we store the actual length of the address buffer into the
1371 * addrs_num structure member. That way we can re-use the existing
1372 * code.
1374 SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1375 char __user *optval,
1376 int __user *optlen)
1378 struct sctp_getaddrs_old param;
1379 sctp_assoc_t assoc_id = 0;
1380 int err = 0;
1382 if (len < sizeof(param))
1383 return -EINVAL;
1385 if (copy_from_user(&param, optval, sizeof(param)))
1386 return -EFAULT;
1388 err = __sctp_setsockopt_connectx(sk,
1389 (struct sockaddr __user *)param.addrs,
1390 param.addr_num, &assoc_id);
1392 if (err == 0 || err == -EINPROGRESS) {
1393 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1394 return -EFAULT;
1395 if (put_user(sizeof(assoc_id), optlen))
1396 return -EFAULT;
1399 return err;
1402 /* API 3.1.4 close() - UDP Style Syntax
1403 * Applications use close() to perform graceful shutdown (as described in
1404 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1405 * by a UDP-style socket.
1407 * The syntax is
1409 * ret = close(int sd);
1411 * sd - the socket descriptor of the associations to be closed.
1413 * To gracefully shutdown a specific association represented by the
1414 * UDP-style socket, an application should use the sendmsg() call,
1415 * passing no user data, but including the appropriate flag in the
1416 * ancillary data (see Section xxxx).
1418 * If sd in the close() call is a branched-off socket representing only
1419 * one association, the shutdown is performed on that association only.
1421 * 4.1.6 close() - TCP Style Syntax
1423 * Applications use close() to gracefully close down an association.
1425 * The syntax is:
1427 * int close(int sd);
1429 * sd - the socket descriptor of the association to be closed.
1431 * After an application calls close() on a socket descriptor, no further
1432 * socket operations will succeed on that descriptor.
1434 * API 7.1.4 SO_LINGER
1436 * An application using the TCP-style socket can use this option to
1437 * perform the SCTP ABORT primitive. The linger option structure is:
1439 * struct linger {
1440 * int l_onoff; // option on/off
1441 * int l_linger; // linger time
1442 * };
1444 * To enable the option, set l_onoff to 1. If the l_linger value is set
1445 * to 0, calling close() is the same as the ABORT primitive. If the
1446 * value is set to a negative value, the setsockopt() call will return
1447 * an error. If the value is set to a positive value linger_time, the
1448 * close() can be blocked for at most linger_time ms. If the graceful
1449 * shutdown phase does not finish during this period, close() will
1450 * return but the graceful shutdown phase continues in the system.
1452 SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1454 struct sctp_endpoint *ep;
1455 struct sctp_association *asoc;
1456 struct list_head *pos, *temp;
1457 unsigned int data_was_unread;
1459 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1461 sctp_lock_sock(sk);
1462 sk->sk_shutdown = SHUTDOWN_MASK;
1463 sk->sk_state = SCTP_SS_CLOSING;
1465 ep = sctp_sk(sk)->ep;
1467 /* Clean up any skbs sitting on the receive queue. */
1468 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1469 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1471 /* Walk all associations on an endpoint. */
1472 list_for_each_safe(pos, temp, &ep->asocs) {
1473 asoc = list_entry(pos, struct sctp_association, asocs);
1475 if (sctp_style(sk, TCP)) {
1476 /* A closed association can still be in the list if
1477 * it belongs to a TCP-style listening socket that is
1478 * not yet accepted. If so, free it. If not, send an
1479 * ABORT or SHUTDOWN based on the linger options.
1481 if (sctp_state(asoc, CLOSED)) {
1482 sctp_unhash_established(asoc);
1483 sctp_association_free(asoc);
1484 continue;
1488 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1489 !skb_queue_empty(&asoc->ulpq.reasm) ||
1490 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1491 struct sctp_chunk *chunk;
1493 chunk = sctp_make_abort_user(asoc, NULL, 0);
1494 if (chunk)
1495 sctp_primitive_ABORT(asoc, chunk);
1496 } else
1497 sctp_primitive_SHUTDOWN(asoc, NULL);
1500 /* On a TCP-style socket, block for at most linger_time if set. */
1501 if (sctp_style(sk, TCP) && timeout)
1502 sctp_wait_for_close(sk, timeout);
1504 /* This will run the backlog queue. */
1505 sctp_release_sock(sk);
1507 /* Supposedly, no process has access to the socket, but
1508 * the net layers still may.
1510 sctp_local_bh_disable();
1511 sctp_bh_lock_sock(sk);
1513 /* Hold the sock, since sk_common_release() will put sock_put()
1514 * and we have just a little more cleanup.
1516 sock_hold(sk);
1517 sk_common_release(sk);
1519 sctp_bh_unlock_sock(sk);
1520 sctp_local_bh_enable();
1522 sock_put(sk);
1524 SCTP_DBG_OBJCNT_DEC(sock);
1527 /* Handle EPIPE error. */
1528 static int sctp_error(struct sock *sk, int flags, int err)
1530 if (err == -EPIPE)
1531 err = sock_error(sk) ? : -EPIPE;
1532 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1533 send_sig(SIGPIPE, current, 0);
1534 return err;
1537 /* API 3.1.3 sendmsg() - UDP Style Syntax
1539 * An application uses sendmsg() and recvmsg() calls to transmit data to
1540 * and receive data from its peer.
1542 * ssize_t sendmsg(int socket, const struct msghdr *message,
1543 * int flags);
1545 * socket - the socket descriptor of the endpoint.
1546 * message - pointer to the msghdr structure which contains a single
1547 * user message and possibly some ancillary data.
1549 * See Section 5 for complete description of the data
1550 * structures.
1552 * flags - flags sent or received with the user message, see Section
1553 * 5 for complete description of the flags.
1555 * Note: This function could use a rewrite especially when explicit
1556 * connect support comes in.
1558 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1560 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1562 SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1563 struct msghdr *msg, size_t msg_len)
1565 struct sctp_sock *sp;
1566 struct sctp_endpoint *ep;
1567 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1568 struct sctp_transport *transport, *chunk_tp;
1569 struct sctp_chunk *chunk;
1570 union sctp_addr to;
1571 struct sockaddr *msg_name = NULL;
1572 struct sctp_sndrcvinfo default_sinfo;
1573 struct sctp_sndrcvinfo *sinfo;
1574 struct sctp_initmsg *sinit;
1575 sctp_assoc_t associd = 0;
1576 sctp_cmsgs_t cmsgs = { NULL };
1577 int err;
1578 sctp_scope_t scope;
1579 long timeo;
1580 __u16 sinfo_flags = 0;
1581 struct sctp_datamsg *datamsg;
1582 int msg_flags = msg->msg_flags;
1584 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1585 sk, msg, msg_len);
1587 err = 0;
1588 sp = sctp_sk(sk);
1589 ep = sp->ep;
1591 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1593 /* We cannot send a message over a TCP-style listening socket. */
1594 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1595 err = -EPIPE;
1596 goto out_nounlock;
1599 /* Parse out the SCTP CMSGs. */
1600 err = sctp_msghdr_parse(msg, &cmsgs);
1602 if (err) {
1603 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1604 goto out_nounlock;
1607 /* Fetch the destination address for this packet. This
1608 * address only selects the association--it is not necessarily
1609 * the address we will send to.
1610 * For a peeled-off socket, msg_name is ignored.
1612 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1613 int msg_namelen = msg->msg_namelen;
1615 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1616 msg_namelen);
1617 if (err)
1618 return err;
1620 if (msg_namelen > sizeof(to))
1621 msg_namelen = sizeof(to);
1622 memcpy(&to, msg->msg_name, msg_namelen);
1623 msg_name = msg->msg_name;
1626 sinfo = cmsgs.info;
1627 sinit = cmsgs.init;
1629 /* Did the user specify SNDRCVINFO? */
1630 if (sinfo) {
1631 sinfo_flags = sinfo->sinfo_flags;
1632 associd = sinfo->sinfo_assoc_id;
1635 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1636 msg_len, sinfo_flags);
1638 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1639 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1640 err = -EINVAL;
1641 goto out_nounlock;
1644 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1645 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1646 * If SCTP_ABORT is set, the message length could be non zero with
1647 * the msg_iov set to the user abort reason.
1649 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1650 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1651 err = -EINVAL;
1652 goto out_nounlock;
1655 /* If SCTP_ADDR_OVER is set, there must be an address
1656 * specified in msg_name.
1658 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1659 err = -EINVAL;
1660 goto out_nounlock;
1663 transport = NULL;
1665 SCTP_DEBUG_PRINTK("About to look up association.\n");
1667 sctp_lock_sock(sk);
1669 /* If a msg_name has been specified, assume this is to be used. */
1670 if (msg_name) {
1671 /* Look for a matching association on the endpoint. */
1672 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1673 if (!asoc) {
1674 /* If we could not find a matching association on the
1675 * endpoint, make sure that it is not a TCP-style
1676 * socket that already has an association or there is
1677 * no peeled-off association on another socket.
1679 if ((sctp_style(sk, TCP) &&
1680 sctp_sstate(sk, ESTABLISHED)) ||
1681 sctp_endpoint_is_peeled_off(ep, &to)) {
1682 err = -EADDRNOTAVAIL;
1683 goto out_unlock;
1686 } else {
1687 asoc = sctp_id2assoc(sk, associd);
1688 if (!asoc) {
1689 err = -EPIPE;
1690 goto out_unlock;
1694 if (asoc) {
1695 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1697 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1698 * socket that has an association in CLOSED state. This can
1699 * happen when an accepted socket has an association that is
1700 * already CLOSED.
1702 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1703 err = -EPIPE;
1704 goto out_unlock;
1707 if (sinfo_flags & SCTP_EOF) {
1708 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1709 asoc);
1710 sctp_primitive_SHUTDOWN(asoc, NULL);
1711 err = 0;
1712 goto out_unlock;
1714 if (sinfo_flags & SCTP_ABORT) {
1716 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1717 if (!chunk) {
1718 err = -ENOMEM;
1719 goto out_unlock;
1722 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1723 sctp_primitive_ABORT(asoc, chunk);
1724 err = 0;
1725 goto out_unlock;
1729 /* Do we need to create the association? */
1730 if (!asoc) {
1731 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1733 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1734 err = -EINVAL;
1735 goto out_unlock;
1738 /* Check for invalid stream against the stream counts,
1739 * either the default or the user specified stream counts.
1741 if (sinfo) {
1742 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1743 /* Check against the defaults. */
1744 if (sinfo->sinfo_stream >=
1745 sp->initmsg.sinit_num_ostreams) {
1746 err = -EINVAL;
1747 goto out_unlock;
1749 } else {
1750 /* Check against the requested. */
1751 if (sinfo->sinfo_stream >=
1752 sinit->sinit_num_ostreams) {
1753 err = -EINVAL;
1754 goto out_unlock;
1760 * API 3.1.2 bind() - UDP Style Syntax
1761 * If a bind() or sctp_bindx() is not called prior to a
1762 * sendmsg() call that initiates a new association, the
1763 * system picks an ephemeral port and will choose an address
1764 * set equivalent to binding with a wildcard address.
1766 if (!ep->base.bind_addr.port) {
1767 if (sctp_autobind(sk)) {
1768 err = -EAGAIN;
1769 goto out_unlock;
1771 } else {
1773 * If an unprivileged user inherits a one-to-many
1774 * style socket with open associations on a privileged
1775 * port, it MAY be permitted to accept new associations,
1776 * but it SHOULD NOT be permitted to open new
1777 * associations.
1779 if (ep->base.bind_addr.port < PROT_SOCK &&
1780 !capable(CAP_NET_BIND_SERVICE)) {
1781 err = -EACCES;
1782 goto out_unlock;
1786 scope = sctp_scope(&to);
1787 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1788 if (!new_asoc) {
1789 err = -ENOMEM;
1790 goto out_unlock;
1792 asoc = new_asoc;
1793 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1794 if (err < 0) {
1795 err = -ENOMEM;
1796 goto out_free;
1799 /* If the SCTP_INIT ancillary data is specified, set all
1800 * the association init values accordingly.
1802 if (sinit) {
1803 if (sinit->sinit_num_ostreams) {
1804 asoc->c.sinit_num_ostreams =
1805 sinit->sinit_num_ostreams;
1807 if (sinit->sinit_max_instreams) {
1808 asoc->c.sinit_max_instreams =
1809 sinit->sinit_max_instreams;
1811 if (sinit->sinit_max_attempts) {
1812 asoc->max_init_attempts
1813 = sinit->sinit_max_attempts;
1815 if (sinit->sinit_max_init_timeo) {
1816 asoc->max_init_timeo =
1817 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1821 /* Prime the peer's transport structures. */
1822 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1823 if (!transport) {
1824 err = -ENOMEM;
1825 goto out_free;
1829 /* ASSERT: we have a valid association at this point. */
1830 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1832 if (!sinfo) {
1833 /* If the user didn't specify SNDRCVINFO, make up one with
1834 * some defaults.
1836 memset(&default_sinfo, 0, sizeof(default_sinfo));
1837 default_sinfo.sinfo_stream = asoc->default_stream;
1838 default_sinfo.sinfo_flags = asoc->default_flags;
1839 default_sinfo.sinfo_ppid = asoc->default_ppid;
1840 default_sinfo.sinfo_context = asoc->default_context;
1841 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1842 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1843 sinfo = &default_sinfo;
1846 /* API 7.1.7, the sndbuf size per association bounds the
1847 * maximum size of data that can be sent in a single send call.
1849 if (msg_len > sk->sk_sndbuf) {
1850 err = -EMSGSIZE;
1851 goto out_free;
1854 if (asoc->pmtu_pending)
1855 sctp_assoc_pending_pmtu(asoc);
1857 /* If fragmentation is disabled and the message length exceeds the
1858 * association fragmentation point, return EMSGSIZE. The I-D
1859 * does not specify what this error is, but this looks like
1860 * a great fit.
1862 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1863 err = -EMSGSIZE;
1864 goto out_free;
1867 /* Check for invalid stream. */
1868 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1869 err = -EINVAL;
1870 goto out_free;
1873 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1874 if (!sctp_wspace(asoc)) {
1875 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1876 if (err)
1877 goto out_free;
1880 /* If an address is passed with the sendto/sendmsg call, it is used
1881 * to override the primary destination address in the TCP model, or
1882 * when SCTP_ADDR_OVER flag is set in the UDP model.
1884 if ((sctp_style(sk, TCP) && msg_name) ||
1885 (sinfo_flags & SCTP_ADDR_OVER)) {
1886 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1887 if (!chunk_tp) {
1888 err = -EINVAL;
1889 goto out_free;
1891 } else
1892 chunk_tp = NULL;
1894 /* Auto-connect, if we aren't connected already. */
1895 if (sctp_state(asoc, CLOSED)) {
1896 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1897 if (err < 0)
1898 goto out_free;
1899 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1902 /* Break the message into multiple chunks of maximum size. */
1903 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1904 if (!datamsg) {
1905 err = -ENOMEM;
1906 goto out_free;
1909 /* Now send the (possibly) fragmented message. */
1910 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1911 sctp_chunk_hold(chunk);
1913 /* Do accounting for the write space. */
1914 sctp_set_owner_w(chunk);
1916 chunk->transport = chunk_tp;
1919 /* Send it to the lower layers. Note: all chunks
1920 * must either fail or succeed. The lower layer
1921 * works that way today. Keep it that way or this
1922 * breaks.
1924 err = sctp_primitive_SEND(asoc, datamsg);
1925 /* Did the lower layer accept the chunk? */
1926 if (err)
1927 sctp_datamsg_free(datamsg);
1928 else
1929 sctp_datamsg_put(datamsg);
1931 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1933 if (err)
1934 goto out_free;
1935 else
1936 err = msg_len;
1938 /* If we are already past ASSOCIATE, the lower
1939 * layers are responsible for association cleanup.
1941 goto out_unlock;
1943 out_free:
1944 if (new_asoc)
1945 sctp_association_free(asoc);
1946 out_unlock:
1947 sctp_release_sock(sk);
1949 out_nounlock:
1950 return sctp_error(sk, msg_flags, err);
1952 #if 0
1953 do_sock_err:
1954 if (msg_len)
1955 err = msg_len;
1956 else
1957 err = sock_error(sk);
1958 goto out;
1960 do_interrupted:
1961 if (msg_len)
1962 err = msg_len;
1963 goto out;
1964 #endif /* 0 */
1967 /* This is an extended version of skb_pull() that removes the data from the
1968 * start of a skb even when data is spread across the list of skb's in the
1969 * frag_list. len specifies the total amount of data that needs to be removed.
1970 * when 'len' bytes could be removed from the skb, it returns 0.
1971 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1972 * could not be removed.
1974 static int sctp_skb_pull(struct sk_buff *skb, int len)
1976 struct sk_buff *list;
1977 int skb_len = skb_headlen(skb);
1978 int rlen;
1980 if (len <= skb_len) {
1981 __skb_pull(skb, len);
1982 return 0;
1984 len -= skb_len;
1985 __skb_pull(skb, skb_len);
1987 skb_walk_frags(skb, list) {
1988 rlen = sctp_skb_pull(list, len);
1989 skb->len -= (len-rlen);
1990 skb->data_len -= (len-rlen);
1992 if (!rlen)
1993 return 0;
1995 len = rlen;
1998 return len;
2001 /* API 3.1.3 recvmsg() - UDP Style Syntax
2003 * ssize_t recvmsg(int socket, struct msghdr *message,
2004 * int flags);
2006 * socket - the socket descriptor of the endpoint.
2007 * message - pointer to the msghdr structure which contains a single
2008 * user message and possibly some ancillary data.
2010 * See Section 5 for complete description of the data
2011 * structures.
2013 * flags - flags sent or received with the user message, see Section
2014 * 5 for complete description of the flags.
2016 static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2018 SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2019 struct msghdr *msg, size_t len, int noblock,
2020 int flags, int *addr_len)
2022 struct sctp_ulpevent *event = NULL;
2023 struct sctp_sock *sp = sctp_sk(sk);
2024 struct sk_buff *skb;
2025 int copied;
2026 int err = 0;
2027 int skb_len;
2029 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2030 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2031 "len", len, "knoblauch", noblock,
2032 "flags", flags, "addr_len", addr_len);
2034 sctp_lock_sock(sk);
2036 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2037 err = -ENOTCONN;
2038 goto out;
2041 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2042 if (!skb)
2043 goto out;
2045 /* Get the total length of the skb including any skb's in the
2046 * frag_list.
2048 skb_len = skb->len;
2050 copied = skb_len;
2051 if (copied > len)
2052 copied = len;
2054 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2056 event = sctp_skb2event(skb);
2058 if (err)
2059 goto out_free;
2061 sock_recv_ts_and_drops(msg, sk, skb);
2062 if (sctp_ulpevent_is_notification(event)) {
2063 msg->msg_flags |= MSG_NOTIFICATION;
2064 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2065 } else {
2066 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2069 /* Check if we allow SCTP_SNDRCVINFO. */
2070 if (sp->subscribe.sctp_data_io_event)
2071 sctp_ulpevent_read_sndrcvinfo(event, msg);
2072 #if 0
2073 /* FIXME: we should be calling IP/IPv6 layers. */
2074 if (sk->sk_protinfo.af_inet.cmsg_flags)
2075 ip_cmsg_recv(msg, skb);
2076 #endif
2078 err = copied;
2080 /* If skb's length exceeds the user's buffer, update the skb and
2081 * push it back to the receive_queue so that the next call to
2082 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2084 if (skb_len > copied) {
2085 msg->msg_flags &= ~MSG_EOR;
2086 if (flags & MSG_PEEK)
2087 goto out_free;
2088 sctp_skb_pull(skb, copied);
2089 skb_queue_head(&sk->sk_receive_queue, skb);
2091 /* When only partial message is copied to the user, increase
2092 * rwnd by that amount. If all the data in the skb is read,
2093 * rwnd is updated when the event is freed.
2095 if (!sctp_ulpevent_is_notification(event))
2096 sctp_assoc_rwnd_increase(event->asoc, copied);
2097 goto out;
2098 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2099 (event->msg_flags & MSG_EOR))
2100 msg->msg_flags |= MSG_EOR;
2101 else
2102 msg->msg_flags &= ~MSG_EOR;
2104 out_free:
2105 if (flags & MSG_PEEK) {
2106 /* Release the skb reference acquired after peeking the skb in
2107 * sctp_skb_recv_datagram().
2109 kfree_skb(skb);
2110 } else {
2111 /* Free the event which includes releasing the reference to
2112 * the owner of the skb, freeing the skb and updating the
2113 * rwnd.
2115 sctp_ulpevent_free(event);
2117 out:
2118 sctp_release_sock(sk);
2119 return err;
2122 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2124 * This option is a on/off flag. If enabled no SCTP message
2125 * fragmentation will be performed. Instead if a message being sent
2126 * exceeds the current PMTU size, the message will NOT be sent and
2127 * instead a error will be indicated to the user.
2129 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2130 char __user *optval,
2131 unsigned int optlen)
2133 int val;
2135 if (optlen < sizeof(int))
2136 return -EINVAL;
2138 if (get_user(val, (int __user *)optval))
2139 return -EFAULT;
2141 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2143 return 0;
2146 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2147 unsigned int optlen)
2149 struct sctp_association *asoc;
2150 struct sctp_ulpevent *event;
2152 if (optlen > sizeof(struct sctp_event_subscribe))
2153 return -EINVAL;
2154 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2155 return -EFAULT;
2158 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2159 * if there is no data to be sent or retransmit, the stack will
2160 * immediately send up this notification.
2162 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2163 &sctp_sk(sk)->subscribe)) {
2164 asoc = sctp_id2assoc(sk, 0);
2166 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2167 event = sctp_ulpevent_make_sender_dry_event(asoc,
2168 GFP_ATOMIC);
2169 if (!event)
2170 return -ENOMEM;
2172 sctp_ulpq_tail_event(&asoc->ulpq, event);
2176 return 0;
2179 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2181 * This socket option is applicable to the UDP-style socket only. When
2182 * set it will cause associations that are idle for more than the
2183 * specified number of seconds to automatically close. An association
2184 * being idle is defined an association that has NOT sent or received
2185 * user data. The special value of '0' indicates that no automatic
2186 * close of any associations should be performed. The option expects an
2187 * integer defining the number of seconds of idle time before an
2188 * association is closed.
2190 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2191 unsigned int optlen)
2193 struct sctp_sock *sp = sctp_sk(sk);
2195 /* Applicable to UDP-style socket only */
2196 if (sctp_style(sk, TCP))
2197 return -EOPNOTSUPP;
2198 if (optlen != sizeof(int))
2199 return -EINVAL;
2200 if (copy_from_user(&sp->autoclose, optval, optlen))
2201 return -EFAULT;
2202 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2203 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2205 return 0;
2208 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2210 * Applications can enable or disable heartbeats for any peer address of
2211 * an association, modify an address's heartbeat interval, force a
2212 * heartbeat to be sent immediately, and adjust the address's maximum
2213 * number of retransmissions sent before an address is considered
2214 * unreachable. The following structure is used to access and modify an
2215 * address's parameters:
2217 * struct sctp_paddrparams {
2218 * sctp_assoc_t spp_assoc_id;
2219 * struct sockaddr_storage spp_address;
2220 * uint32_t spp_hbinterval;
2221 * uint16_t spp_pathmaxrxt;
2222 * uint32_t spp_pathmtu;
2223 * uint32_t spp_sackdelay;
2224 * uint32_t spp_flags;
2225 * };
2227 * spp_assoc_id - (one-to-many style socket) This is filled in the
2228 * application, and identifies the association for
2229 * this query.
2230 * spp_address - This specifies which address is of interest.
2231 * spp_hbinterval - This contains the value of the heartbeat interval,
2232 * in milliseconds. If a value of zero
2233 * is present in this field then no changes are to
2234 * be made to this parameter.
2235 * spp_pathmaxrxt - This contains the maximum number of
2236 * retransmissions before this address shall be
2237 * considered unreachable. If a value of zero
2238 * is present in this field then no changes are to
2239 * be made to this parameter.
2240 * spp_pathmtu - When Path MTU discovery is disabled the value
2241 * specified here will be the "fixed" path mtu.
2242 * Note that if the spp_address field is empty
2243 * then all associations on this address will
2244 * have this fixed path mtu set upon them.
2246 * spp_sackdelay - When delayed sack is enabled, this value specifies
2247 * the number of milliseconds that sacks will be delayed
2248 * for. This value will apply to all addresses of an
2249 * association if the spp_address field is empty. Note
2250 * also, that if delayed sack is enabled and this
2251 * value is set to 0, no change is made to the last
2252 * recorded delayed sack timer value.
2254 * spp_flags - These flags are used to control various features
2255 * on an association. The flag field may contain
2256 * zero or more of the following options.
2258 * SPP_HB_ENABLE - Enable heartbeats on the
2259 * specified address. Note that if the address
2260 * field is empty all addresses for the association
2261 * have heartbeats enabled upon them.
2263 * SPP_HB_DISABLE - Disable heartbeats on the
2264 * speicifed address. Note that if the address
2265 * field is empty all addresses for the association
2266 * will have their heartbeats disabled. Note also
2267 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2268 * mutually exclusive, only one of these two should
2269 * be specified. Enabling both fields will have
2270 * undetermined results.
2272 * SPP_HB_DEMAND - Request a user initiated heartbeat
2273 * to be made immediately.
2275 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2276 * heartbeat delayis to be set to the value of 0
2277 * milliseconds.
2279 * SPP_PMTUD_ENABLE - This field will enable PMTU
2280 * discovery upon the specified address. Note that
2281 * if the address feild is empty then all addresses
2282 * on the association are effected.
2284 * SPP_PMTUD_DISABLE - This field will disable PMTU
2285 * discovery upon the specified address. Note that
2286 * if the address feild is empty then all addresses
2287 * on the association are effected. Not also that
2288 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2289 * exclusive. Enabling both will have undetermined
2290 * results.
2292 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2293 * on delayed sack. The time specified in spp_sackdelay
2294 * is used to specify the sack delay for this address. Note
2295 * that if spp_address is empty then all addresses will
2296 * enable delayed sack and take on the sack delay
2297 * value specified in spp_sackdelay.
2298 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2299 * off delayed sack. If the spp_address field is blank then
2300 * delayed sack is disabled for the entire association. Note
2301 * also that this field is mutually exclusive to
2302 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2303 * results.
2305 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2306 struct sctp_transport *trans,
2307 struct sctp_association *asoc,
2308 struct sctp_sock *sp,
2309 int hb_change,
2310 int pmtud_change,
2311 int sackdelay_change)
2313 int error;
2315 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2316 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2317 if (error)
2318 return error;
2321 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2322 * this field is ignored. Note also that a value of zero indicates
2323 * the current setting should be left unchanged.
2325 if (params->spp_flags & SPP_HB_ENABLE) {
2327 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2328 * set. This lets us use 0 value when this flag
2329 * is set.
2331 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2332 params->spp_hbinterval = 0;
2334 if (params->spp_hbinterval ||
2335 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2336 if (trans) {
2337 trans->hbinterval =
2338 msecs_to_jiffies(params->spp_hbinterval);
2339 } else if (asoc) {
2340 asoc->hbinterval =
2341 msecs_to_jiffies(params->spp_hbinterval);
2342 } else {
2343 sp->hbinterval = params->spp_hbinterval;
2348 if (hb_change) {
2349 if (trans) {
2350 trans->param_flags =
2351 (trans->param_flags & ~SPP_HB) | hb_change;
2352 } else if (asoc) {
2353 asoc->param_flags =
2354 (asoc->param_flags & ~SPP_HB) | hb_change;
2355 } else {
2356 sp->param_flags =
2357 (sp->param_flags & ~SPP_HB) | hb_change;
2361 /* When Path MTU discovery is disabled the value specified here will
2362 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2363 * include the flag SPP_PMTUD_DISABLE for this field to have any
2364 * effect).
2366 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2367 if (trans) {
2368 trans->pathmtu = params->spp_pathmtu;
2369 sctp_assoc_sync_pmtu(asoc);
2370 } else if (asoc) {
2371 asoc->pathmtu = params->spp_pathmtu;
2372 sctp_frag_point(asoc, params->spp_pathmtu);
2373 } else {
2374 sp->pathmtu = params->spp_pathmtu;
2378 if (pmtud_change) {
2379 if (trans) {
2380 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2381 (params->spp_flags & SPP_PMTUD_ENABLE);
2382 trans->param_flags =
2383 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2384 if (update) {
2385 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2386 sctp_assoc_sync_pmtu(asoc);
2388 } else if (asoc) {
2389 asoc->param_flags =
2390 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2391 } else {
2392 sp->param_flags =
2393 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2397 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2398 * value of this field is ignored. Note also that a value of zero
2399 * indicates the current setting should be left unchanged.
2401 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2402 if (trans) {
2403 trans->sackdelay =
2404 msecs_to_jiffies(params->spp_sackdelay);
2405 } else if (asoc) {
2406 asoc->sackdelay =
2407 msecs_to_jiffies(params->spp_sackdelay);
2408 } else {
2409 sp->sackdelay = params->spp_sackdelay;
2413 if (sackdelay_change) {
2414 if (trans) {
2415 trans->param_flags =
2416 (trans->param_flags & ~SPP_SACKDELAY) |
2417 sackdelay_change;
2418 } else if (asoc) {
2419 asoc->param_flags =
2420 (asoc->param_flags & ~SPP_SACKDELAY) |
2421 sackdelay_change;
2422 } else {
2423 sp->param_flags =
2424 (sp->param_flags & ~SPP_SACKDELAY) |
2425 sackdelay_change;
2429 /* Note that a value of zero indicates the current setting should be
2430 left unchanged.
2432 if (params->spp_pathmaxrxt) {
2433 if (trans) {
2434 trans->pathmaxrxt = params->spp_pathmaxrxt;
2435 } else if (asoc) {
2436 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2437 } else {
2438 sp->pathmaxrxt = params->spp_pathmaxrxt;
2442 return 0;
2445 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2446 char __user *optval,
2447 unsigned int optlen)
2449 struct sctp_paddrparams params;
2450 struct sctp_transport *trans = NULL;
2451 struct sctp_association *asoc = NULL;
2452 struct sctp_sock *sp = sctp_sk(sk);
2453 int error;
2454 int hb_change, pmtud_change, sackdelay_change;
2456 if (optlen != sizeof(struct sctp_paddrparams))
2457 return - EINVAL;
2459 if (copy_from_user(&params, optval, optlen))
2460 return -EFAULT;
2462 /* Validate flags and value parameters. */
2463 hb_change = params.spp_flags & SPP_HB;
2464 pmtud_change = params.spp_flags & SPP_PMTUD;
2465 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2467 if (hb_change == SPP_HB ||
2468 pmtud_change == SPP_PMTUD ||
2469 sackdelay_change == SPP_SACKDELAY ||
2470 params.spp_sackdelay > 500 ||
2471 (params.spp_pathmtu &&
2472 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2473 return -EINVAL;
2475 /* If an address other than INADDR_ANY is specified, and
2476 * no transport is found, then the request is invalid.
2478 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
2479 trans = sctp_addr_id2transport(sk, &params.spp_address,
2480 params.spp_assoc_id);
2481 if (!trans)
2482 return -EINVAL;
2485 /* Get association, if assoc_id != 0 and the socket is a one
2486 * to many style socket, and an association was not found, then
2487 * the id was invalid.
2489 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2490 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2491 return -EINVAL;
2493 /* Heartbeat demand can only be sent on a transport or
2494 * association, but not a socket.
2496 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2497 return -EINVAL;
2499 /* Process parameters. */
2500 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2501 hb_change, pmtud_change,
2502 sackdelay_change);
2504 if (error)
2505 return error;
2507 /* If changes are for association, also apply parameters to each
2508 * transport.
2510 if (!trans && asoc) {
2511 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2512 transports) {
2513 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2514 hb_change, pmtud_change,
2515 sackdelay_change);
2519 return 0;
2523 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2525 * This option will effect the way delayed acks are performed. This
2526 * option allows you to get or set the delayed ack time, in
2527 * milliseconds. It also allows changing the delayed ack frequency.
2528 * Changing the frequency to 1 disables the delayed sack algorithm. If
2529 * the assoc_id is 0, then this sets or gets the endpoints default
2530 * values. If the assoc_id field is non-zero, then the set or get
2531 * effects the specified association for the one to many model (the
2532 * assoc_id field is ignored by the one to one model). Note that if
2533 * sack_delay or sack_freq are 0 when setting this option, then the
2534 * current values will remain unchanged.
2536 * struct sctp_sack_info {
2537 * sctp_assoc_t sack_assoc_id;
2538 * uint32_t sack_delay;
2539 * uint32_t sack_freq;
2540 * };
2542 * sack_assoc_id - This parameter, indicates which association the user
2543 * is performing an action upon. Note that if this field's value is
2544 * zero then the endpoints default value is changed (effecting future
2545 * associations only).
2547 * sack_delay - This parameter contains the number of milliseconds that
2548 * the user is requesting the delayed ACK timer be set to. Note that
2549 * this value is defined in the standard to be between 200 and 500
2550 * milliseconds.
2552 * sack_freq - This parameter contains the number of packets that must
2553 * be received before a sack is sent without waiting for the delay
2554 * timer to expire. The default value for this is 2, setting this
2555 * value to 1 will disable the delayed sack algorithm.
2558 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2559 char __user *optval, unsigned int optlen)
2561 struct sctp_sack_info params;
2562 struct sctp_transport *trans = NULL;
2563 struct sctp_association *asoc = NULL;
2564 struct sctp_sock *sp = sctp_sk(sk);
2566 if (optlen == sizeof(struct sctp_sack_info)) {
2567 if (copy_from_user(&params, optval, optlen))
2568 return -EFAULT;
2570 if (params.sack_delay == 0 && params.sack_freq == 0)
2571 return 0;
2572 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2573 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2574 pr_warn("Use struct sctp_sack_info instead\n");
2575 if (copy_from_user(&params, optval, optlen))
2576 return -EFAULT;
2578 if (params.sack_delay == 0)
2579 params.sack_freq = 1;
2580 else
2581 params.sack_freq = 0;
2582 } else
2583 return - EINVAL;
2585 /* Validate value parameter. */
2586 if (params.sack_delay > 500)
2587 return -EINVAL;
2589 /* Get association, if sack_assoc_id != 0 and the socket is a one
2590 * to many style socket, and an association was not found, then
2591 * the id was invalid.
2593 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2594 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2595 return -EINVAL;
2597 if (params.sack_delay) {
2598 if (asoc) {
2599 asoc->sackdelay =
2600 msecs_to_jiffies(params.sack_delay);
2601 asoc->param_flags =
2602 (asoc->param_flags & ~SPP_SACKDELAY) |
2603 SPP_SACKDELAY_ENABLE;
2604 } else {
2605 sp->sackdelay = params.sack_delay;
2606 sp->param_flags =
2607 (sp->param_flags & ~SPP_SACKDELAY) |
2608 SPP_SACKDELAY_ENABLE;
2612 if (params.sack_freq == 1) {
2613 if (asoc) {
2614 asoc->param_flags =
2615 (asoc->param_flags & ~SPP_SACKDELAY) |
2616 SPP_SACKDELAY_DISABLE;
2617 } else {
2618 sp->param_flags =
2619 (sp->param_flags & ~SPP_SACKDELAY) |
2620 SPP_SACKDELAY_DISABLE;
2622 } else if (params.sack_freq > 1) {
2623 if (asoc) {
2624 asoc->sackfreq = params.sack_freq;
2625 asoc->param_flags =
2626 (asoc->param_flags & ~SPP_SACKDELAY) |
2627 SPP_SACKDELAY_ENABLE;
2628 } else {
2629 sp->sackfreq = params.sack_freq;
2630 sp->param_flags =
2631 (sp->param_flags & ~SPP_SACKDELAY) |
2632 SPP_SACKDELAY_ENABLE;
2636 /* If change is for association, also apply to each transport. */
2637 if (asoc) {
2638 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2639 transports) {
2640 if (params.sack_delay) {
2641 trans->sackdelay =
2642 msecs_to_jiffies(params.sack_delay);
2643 trans->param_flags =
2644 (trans->param_flags & ~SPP_SACKDELAY) |
2645 SPP_SACKDELAY_ENABLE;
2647 if (params.sack_freq == 1) {
2648 trans->param_flags =
2649 (trans->param_flags & ~SPP_SACKDELAY) |
2650 SPP_SACKDELAY_DISABLE;
2651 } else if (params.sack_freq > 1) {
2652 trans->sackfreq = params.sack_freq;
2653 trans->param_flags =
2654 (trans->param_flags & ~SPP_SACKDELAY) |
2655 SPP_SACKDELAY_ENABLE;
2660 return 0;
2663 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2665 * Applications can specify protocol parameters for the default association
2666 * initialization. The option name argument to setsockopt() and getsockopt()
2667 * is SCTP_INITMSG.
2669 * Setting initialization parameters is effective only on an unconnected
2670 * socket (for UDP-style sockets only future associations are effected
2671 * by the change). With TCP-style sockets, this option is inherited by
2672 * sockets derived from a listener socket.
2674 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2676 struct sctp_initmsg sinit;
2677 struct sctp_sock *sp = sctp_sk(sk);
2679 if (optlen != sizeof(struct sctp_initmsg))
2680 return -EINVAL;
2681 if (copy_from_user(&sinit, optval, optlen))
2682 return -EFAULT;
2684 if (sinit.sinit_num_ostreams)
2685 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2686 if (sinit.sinit_max_instreams)
2687 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2688 if (sinit.sinit_max_attempts)
2689 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2690 if (sinit.sinit_max_init_timeo)
2691 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2693 return 0;
2697 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2699 * Applications that wish to use the sendto() system call may wish to
2700 * specify a default set of parameters that would normally be supplied
2701 * through the inclusion of ancillary data. This socket option allows
2702 * such an application to set the default sctp_sndrcvinfo structure.
2703 * The application that wishes to use this socket option simply passes
2704 * in to this call the sctp_sndrcvinfo structure defined in Section
2705 * 5.2.2) The input parameters accepted by this call include
2706 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2707 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2708 * to this call if the caller is using the UDP model.
2710 static int sctp_setsockopt_default_send_param(struct sock *sk,
2711 char __user *optval,
2712 unsigned int optlen)
2714 struct sctp_sndrcvinfo info;
2715 struct sctp_association *asoc;
2716 struct sctp_sock *sp = sctp_sk(sk);
2718 if (optlen != sizeof(struct sctp_sndrcvinfo))
2719 return -EINVAL;
2720 if (copy_from_user(&info, optval, optlen))
2721 return -EFAULT;
2723 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2724 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2725 return -EINVAL;
2727 if (asoc) {
2728 asoc->default_stream = info.sinfo_stream;
2729 asoc->default_flags = info.sinfo_flags;
2730 asoc->default_ppid = info.sinfo_ppid;
2731 asoc->default_context = info.sinfo_context;
2732 asoc->default_timetolive = info.sinfo_timetolive;
2733 } else {
2734 sp->default_stream = info.sinfo_stream;
2735 sp->default_flags = info.sinfo_flags;
2736 sp->default_ppid = info.sinfo_ppid;
2737 sp->default_context = info.sinfo_context;
2738 sp->default_timetolive = info.sinfo_timetolive;
2741 return 0;
2744 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2746 * Requests that the local SCTP stack use the enclosed peer address as
2747 * the association primary. The enclosed address must be one of the
2748 * association peer's addresses.
2750 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2751 unsigned int optlen)
2753 struct sctp_prim prim;
2754 struct sctp_transport *trans;
2756 if (optlen != sizeof(struct sctp_prim))
2757 return -EINVAL;
2759 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2760 return -EFAULT;
2762 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2763 if (!trans)
2764 return -EINVAL;
2766 sctp_assoc_set_primary(trans->asoc, trans);
2768 return 0;
2772 * 7.1.5 SCTP_NODELAY
2774 * Turn on/off any Nagle-like algorithm. This means that packets are
2775 * generally sent as soon as possible and no unnecessary delays are
2776 * introduced, at the cost of more packets in the network. Expects an
2777 * integer boolean flag.
2779 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2780 unsigned int optlen)
2782 int val;
2784 if (optlen < sizeof(int))
2785 return -EINVAL;
2786 if (get_user(val, (int __user *)optval))
2787 return -EFAULT;
2789 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2790 return 0;
2795 * 7.1.1 SCTP_RTOINFO
2797 * The protocol parameters used to initialize and bound retransmission
2798 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2799 * and modify these parameters.
2800 * All parameters are time values, in milliseconds. A value of 0, when
2801 * modifying the parameters, indicates that the current value should not
2802 * be changed.
2805 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2807 struct sctp_rtoinfo rtoinfo;
2808 struct sctp_association *asoc;
2810 if (optlen != sizeof (struct sctp_rtoinfo))
2811 return -EINVAL;
2813 if (copy_from_user(&rtoinfo, optval, optlen))
2814 return -EFAULT;
2816 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2818 /* Set the values to the specific association */
2819 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2820 return -EINVAL;
2822 if (asoc) {
2823 if (rtoinfo.srto_initial != 0)
2824 asoc->rto_initial =
2825 msecs_to_jiffies(rtoinfo.srto_initial);
2826 if (rtoinfo.srto_max != 0)
2827 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2828 if (rtoinfo.srto_min != 0)
2829 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2830 } else {
2831 /* If there is no association or the association-id = 0
2832 * set the values to the endpoint.
2834 struct sctp_sock *sp = sctp_sk(sk);
2836 if (rtoinfo.srto_initial != 0)
2837 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2838 if (rtoinfo.srto_max != 0)
2839 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2840 if (rtoinfo.srto_min != 0)
2841 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2844 return 0;
2849 * 7.1.2 SCTP_ASSOCINFO
2851 * This option is used to tune the maximum retransmission attempts
2852 * of the association.
2853 * Returns an error if the new association retransmission value is
2854 * greater than the sum of the retransmission value of the peer.
2855 * See [SCTP] for more information.
2858 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2861 struct sctp_assocparams assocparams;
2862 struct sctp_association *asoc;
2864 if (optlen != sizeof(struct sctp_assocparams))
2865 return -EINVAL;
2866 if (copy_from_user(&assocparams, optval, optlen))
2867 return -EFAULT;
2869 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2871 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2872 return -EINVAL;
2874 /* Set the values to the specific association */
2875 if (asoc) {
2876 if (assocparams.sasoc_asocmaxrxt != 0) {
2877 __u32 path_sum = 0;
2878 int paths = 0;
2879 struct sctp_transport *peer_addr;
2881 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2882 transports) {
2883 path_sum += peer_addr->pathmaxrxt;
2884 paths++;
2887 /* Only validate asocmaxrxt if we have more than
2888 * one path/transport. We do this because path
2889 * retransmissions are only counted when we have more
2890 * then one path.
2892 if (paths > 1 &&
2893 assocparams.sasoc_asocmaxrxt > path_sum)
2894 return -EINVAL;
2896 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2899 if (assocparams.sasoc_cookie_life != 0) {
2900 asoc->cookie_life.tv_sec =
2901 assocparams.sasoc_cookie_life / 1000;
2902 asoc->cookie_life.tv_usec =
2903 (assocparams.sasoc_cookie_life % 1000)
2904 * 1000;
2906 } else {
2907 /* Set the values to the endpoint */
2908 struct sctp_sock *sp = sctp_sk(sk);
2910 if (assocparams.sasoc_asocmaxrxt != 0)
2911 sp->assocparams.sasoc_asocmaxrxt =
2912 assocparams.sasoc_asocmaxrxt;
2913 if (assocparams.sasoc_cookie_life != 0)
2914 sp->assocparams.sasoc_cookie_life =
2915 assocparams.sasoc_cookie_life;
2917 return 0;
2921 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2923 * This socket option is a boolean flag which turns on or off mapped V4
2924 * addresses. If this option is turned on and the socket is type
2925 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2926 * If this option is turned off, then no mapping will be done of V4
2927 * addresses and a user will receive both PF_INET6 and PF_INET type
2928 * addresses on the socket.
2930 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2932 int val;
2933 struct sctp_sock *sp = sctp_sk(sk);
2935 if (optlen < sizeof(int))
2936 return -EINVAL;
2937 if (get_user(val, (int __user *)optval))
2938 return -EFAULT;
2939 if (val)
2940 sp->v4mapped = 1;
2941 else
2942 sp->v4mapped = 0;
2944 return 0;
2948 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2949 * This option will get or set the maximum size to put in any outgoing
2950 * SCTP DATA chunk. If a message is larger than this size it will be
2951 * fragmented by SCTP into the specified size. Note that the underlying
2952 * SCTP implementation may fragment into smaller sized chunks when the
2953 * PMTU of the underlying association is smaller than the value set by
2954 * the user. The default value for this option is '0' which indicates
2955 * the user is NOT limiting fragmentation and only the PMTU will effect
2956 * SCTP's choice of DATA chunk size. Note also that values set larger
2957 * than the maximum size of an IP datagram will effectively let SCTP
2958 * control fragmentation (i.e. the same as setting this option to 0).
2960 * The following structure is used to access and modify this parameter:
2962 * struct sctp_assoc_value {
2963 * sctp_assoc_t assoc_id;
2964 * uint32_t assoc_value;
2965 * };
2967 * assoc_id: This parameter is ignored for one-to-one style sockets.
2968 * For one-to-many style sockets this parameter indicates which
2969 * association the user is performing an action upon. Note that if
2970 * this field's value is zero then the endpoints default value is
2971 * changed (effecting future associations only).
2972 * assoc_value: This parameter specifies the maximum size in bytes.
2974 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2976 struct sctp_assoc_value params;
2977 struct sctp_association *asoc;
2978 struct sctp_sock *sp = sctp_sk(sk);
2979 int val;
2981 if (optlen == sizeof(int)) {
2982 pr_warn("Use of int in maxseg socket option deprecated\n");
2983 pr_warn("Use struct sctp_assoc_value instead\n");
2984 if (copy_from_user(&val, optval, optlen))
2985 return -EFAULT;
2986 params.assoc_id = 0;
2987 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2988 if (copy_from_user(&params, optval, optlen))
2989 return -EFAULT;
2990 val = params.assoc_value;
2991 } else
2992 return -EINVAL;
2994 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2995 return -EINVAL;
2997 asoc = sctp_id2assoc(sk, params.assoc_id);
2998 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2999 return -EINVAL;
3001 if (asoc) {
3002 if (val == 0) {
3003 val = asoc->pathmtu;
3004 val -= sp->pf->af->net_header_len;
3005 val -= sizeof(struct sctphdr) +
3006 sizeof(struct sctp_data_chunk);
3008 asoc->user_frag = val;
3009 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3010 } else {
3011 sp->user_frag = val;
3014 return 0;
3019 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3021 * Requests that the peer mark the enclosed address as the association
3022 * primary. The enclosed address must be one of the association's
3023 * locally bound addresses. The following structure is used to make a
3024 * set primary request:
3026 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3027 unsigned int optlen)
3029 struct sctp_sock *sp;
3030 struct sctp_association *asoc = NULL;
3031 struct sctp_setpeerprim prim;
3032 struct sctp_chunk *chunk;
3033 struct sctp_af *af;
3034 int err;
3036 sp = sctp_sk(sk);
3038 if (!sctp_addip_enable)
3039 return -EPERM;
3041 if (optlen != sizeof(struct sctp_setpeerprim))
3042 return -EINVAL;
3044 if (copy_from_user(&prim, optval, optlen))
3045 return -EFAULT;
3047 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3048 if (!asoc)
3049 return -EINVAL;
3051 if (!asoc->peer.asconf_capable)
3052 return -EPERM;
3054 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3055 return -EPERM;
3057 if (!sctp_state(asoc, ESTABLISHED))
3058 return -ENOTCONN;
3060 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3061 if (!af)
3062 return -EINVAL;
3064 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3065 return -EADDRNOTAVAIL;
3067 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3068 return -EADDRNOTAVAIL;
3070 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3071 chunk = sctp_make_asconf_set_prim(asoc,
3072 (union sctp_addr *)&prim.sspp_addr);
3073 if (!chunk)
3074 return -ENOMEM;
3076 err = sctp_send_asconf(asoc, chunk);
3078 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3080 return err;
3083 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3084 unsigned int optlen)
3086 struct sctp_setadaptation adaptation;
3088 if (optlen != sizeof(struct sctp_setadaptation))
3089 return -EINVAL;
3090 if (copy_from_user(&adaptation, optval, optlen))
3091 return -EFAULT;
3093 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3095 return 0;
3099 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3101 * The context field in the sctp_sndrcvinfo structure is normally only
3102 * used when a failed message is retrieved holding the value that was
3103 * sent down on the actual send call. This option allows the setting of
3104 * a default context on an association basis that will be received on
3105 * reading messages from the peer. This is especially helpful in the
3106 * one-2-many model for an application to keep some reference to an
3107 * internal state machine that is processing messages on the
3108 * association. Note that the setting of this value only effects
3109 * received messages from the peer and does not effect the value that is
3110 * saved with outbound messages.
3112 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3113 unsigned int optlen)
3115 struct sctp_assoc_value params;
3116 struct sctp_sock *sp;
3117 struct sctp_association *asoc;
3119 if (optlen != sizeof(struct sctp_assoc_value))
3120 return -EINVAL;
3121 if (copy_from_user(&params, optval, optlen))
3122 return -EFAULT;
3124 sp = sctp_sk(sk);
3126 if (params.assoc_id != 0) {
3127 asoc = sctp_id2assoc(sk, params.assoc_id);
3128 if (!asoc)
3129 return -EINVAL;
3130 asoc->default_rcv_context = params.assoc_value;
3131 } else {
3132 sp->default_rcv_context = params.assoc_value;
3135 return 0;
3139 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3141 * This options will at a minimum specify if the implementation is doing
3142 * fragmented interleave. Fragmented interleave, for a one to many
3143 * socket, is when subsequent calls to receive a message may return
3144 * parts of messages from different associations. Some implementations
3145 * may allow you to turn this value on or off. If so, when turned off,
3146 * no fragment interleave will occur (which will cause a head of line
3147 * blocking amongst multiple associations sharing the same one to many
3148 * socket). When this option is turned on, then each receive call may
3149 * come from a different association (thus the user must receive data
3150 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3151 * association each receive belongs to.
3153 * This option takes a boolean value. A non-zero value indicates that
3154 * fragmented interleave is on. A value of zero indicates that
3155 * fragmented interleave is off.
3157 * Note that it is important that an implementation that allows this
3158 * option to be turned on, have it off by default. Otherwise an unaware
3159 * application using the one to many model may become confused and act
3160 * incorrectly.
3162 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3163 char __user *optval,
3164 unsigned int optlen)
3166 int val;
3168 if (optlen != sizeof(int))
3169 return -EINVAL;
3170 if (get_user(val, (int __user *)optval))
3171 return -EFAULT;
3173 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3175 return 0;
3179 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3180 * (SCTP_PARTIAL_DELIVERY_POINT)
3182 * This option will set or get the SCTP partial delivery point. This
3183 * point is the size of a message where the partial delivery API will be
3184 * invoked to help free up rwnd space for the peer. Setting this to a
3185 * lower value will cause partial deliveries to happen more often. The
3186 * calls argument is an integer that sets or gets the partial delivery
3187 * point. Note also that the call will fail if the user attempts to set
3188 * this value larger than the socket receive buffer size.
3190 * Note that any single message having a length smaller than or equal to
3191 * the SCTP partial delivery point will be delivered in one single read
3192 * call as long as the user provided buffer is large enough to hold the
3193 * message.
3195 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3196 char __user *optval,
3197 unsigned int optlen)
3199 u32 val;
3201 if (optlen != sizeof(u32))
3202 return -EINVAL;
3203 if (get_user(val, (int __user *)optval))
3204 return -EFAULT;
3206 /* Note: We double the receive buffer from what the user sets
3207 * it to be, also initial rwnd is based on rcvbuf/2.
3209 if (val > (sk->sk_rcvbuf >> 1))
3210 return -EINVAL;
3212 sctp_sk(sk)->pd_point = val;
3214 return 0; /* is this the right error code? */
3218 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3220 * This option will allow a user to change the maximum burst of packets
3221 * that can be emitted by this association. Note that the default value
3222 * is 4, and some implementations may restrict this setting so that it
3223 * can only be lowered.
3225 * NOTE: This text doesn't seem right. Do this on a socket basis with
3226 * future associations inheriting the socket value.
3228 static int sctp_setsockopt_maxburst(struct sock *sk,
3229 char __user *optval,
3230 unsigned int optlen)
3232 struct sctp_assoc_value params;
3233 struct sctp_sock *sp;
3234 struct sctp_association *asoc;
3235 int val;
3236 int assoc_id = 0;
3238 if (optlen == sizeof(int)) {
3239 pr_warn("Use of int in max_burst socket option deprecated\n");
3240 pr_warn("Use struct sctp_assoc_value instead\n");
3241 if (copy_from_user(&val, optval, optlen))
3242 return -EFAULT;
3243 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3244 if (copy_from_user(&params, optval, optlen))
3245 return -EFAULT;
3246 val = params.assoc_value;
3247 assoc_id = params.assoc_id;
3248 } else
3249 return -EINVAL;
3251 sp = sctp_sk(sk);
3253 if (assoc_id != 0) {
3254 asoc = sctp_id2assoc(sk, assoc_id);
3255 if (!asoc)
3256 return -EINVAL;
3257 asoc->max_burst = val;
3258 } else
3259 sp->max_burst = val;
3261 return 0;
3265 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3267 * This set option adds a chunk type that the user is requesting to be
3268 * received only in an authenticated way. Changes to the list of chunks
3269 * will only effect future associations on the socket.
3271 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3272 char __user *optval,
3273 unsigned int optlen)
3275 struct sctp_authchunk val;
3277 if (!sctp_auth_enable)
3278 return -EACCES;
3280 if (optlen != sizeof(struct sctp_authchunk))
3281 return -EINVAL;
3282 if (copy_from_user(&val, optval, optlen))
3283 return -EFAULT;
3285 switch (val.sauth_chunk) {
3286 case SCTP_CID_INIT:
3287 case SCTP_CID_INIT_ACK:
3288 case SCTP_CID_SHUTDOWN_COMPLETE:
3289 case SCTP_CID_AUTH:
3290 return -EINVAL;
3293 /* add this chunk id to the endpoint */
3294 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3298 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3300 * This option gets or sets the list of HMAC algorithms that the local
3301 * endpoint requires the peer to use.
3303 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3304 char __user *optval,
3305 unsigned int optlen)
3307 struct sctp_hmacalgo *hmacs;
3308 u32 idents;
3309 int err;
3311 if (!sctp_auth_enable)
3312 return -EACCES;
3314 if (optlen < sizeof(struct sctp_hmacalgo))
3315 return -EINVAL;
3317 hmacs= memdup_user(optval, optlen);
3318 if (IS_ERR(hmacs))
3319 return PTR_ERR(hmacs);
3321 idents = hmacs->shmac_num_idents;
3322 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3323 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3324 err = -EINVAL;
3325 goto out;
3328 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3329 out:
3330 kfree(hmacs);
3331 return err;
3335 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3337 * This option will set a shared secret key which is used to build an
3338 * association shared key.
3340 static int sctp_setsockopt_auth_key(struct sock *sk,
3341 char __user *optval,
3342 unsigned int optlen)
3344 struct sctp_authkey *authkey;
3345 struct sctp_association *asoc;
3346 int ret;
3348 if (!sctp_auth_enable)
3349 return -EACCES;
3351 if (optlen <= sizeof(struct sctp_authkey))
3352 return -EINVAL;
3354 authkey= memdup_user(optval, optlen);
3355 if (IS_ERR(authkey))
3356 return PTR_ERR(authkey);
3358 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3359 ret = -EINVAL;
3360 goto out;
3363 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3364 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3365 ret = -EINVAL;
3366 goto out;
3369 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3370 out:
3371 kfree(authkey);
3372 return ret;
3376 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3378 * This option will get or set the active shared key to be used to build
3379 * the association shared key.
3381 static int sctp_setsockopt_active_key(struct sock *sk,
3382 char __user *optval,
3383 unsigned int optlen)
3385 struct sctp_authkeyid val;
3386 struct sctp_association *asoc;
3388 if (!sctp_auth_enable)
3389 return -EACCES;
3391 if (optlen != sizeof(struct sctp_authkeyid))
3392 return -EINVAL;
3393 if (copy_from_user(&val, optval, optlen))
3394 return -EFAULT;
3396 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3397 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3398 return -EINVAL;
3400 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3401 val.scact_keynumber);
3405 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3407 * This set option will delete a shared secret key from use.
3409 static int sctp_setsockopt_del_key(struct sock *sk,
3410 char __user *optval,
3411 unsigned int optlen)
3413 struct sctp_authkeyid val;
3414 struct sctp_association *asoc;
3416 if (!sctp_auth_enable)
3417 return -EACCES;
3419 if (optlen != sizeof(struct sctp_authkeyid))
3420 return -EINVAL;
3421 if (copy_from_user(&val, optval, optlen))
3422 return -EFAULT;
3424 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3425 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3426 return -EINVAL;
3428 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3429 val.scact_keynumber);
3434 * 8.1.23 SCTP_AUTO_ASCONF
3436 * This option will enable or disable the use of the automatic generation of
3437 * ASCONF chunks to add and delete addresses to an existing association. Note
3438 * that this option has two caveats namely: a) it only affects sockets that
3439 * are bound to all addresses available to the SCTP stack, and b) the system
3440 * administrator may have an overriding control that turns the ASCONF feature
3441 * off no matter what setting the socket option may have.
3442 * This option expects an integer boolean flag, where a non-zero value turns on
3443 * the option, and a zero value turns off the option.
3444 * Note. In this implementation, socket operation overrides default parameter
3445 * being set by sysctl as well as FreeBSD implementation
3447 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3448 unsigned int optlen)
3450 int val;
3451 struct sctp_sock *sp = sctp_sk(sk);
3453 if (optlen < sizeof(int))
3454 return -EINVAL;
3455 if (get_user(val, (int __user *)optval))
3456 return -EFAULT;
3457 if (!sctp_is_ep_boundall(sk) && val)
3458 return -EINVAL;
3459 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3460 return 0;
3462 if (val == 0 && sp->do_auto_asconf) {
3463 list_del(&sp->auto_asconf_list);
3464 sp->do_auto_asconf = 0;
3465 } else if (val && !sp->do_auto_asconf) {
3466 list_add_tail(&sp->auto_asconf_list,
3467 &sctp_auto_asconf_splist);
3468 sp->do_auto_asconf = 1;
3470 return 0;
3474 /* API 6.2 setsockopt(), getsockopt()
3476 * Applications use setsockopt() and getsockopt() to set or retrieve
3477 * socket options. Socket options are used to change the default
3478 * behavior of sockets calls. They are described in Section 7.
3480 * The syntax is:
3482 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3483 * int __user *optlen);
3484 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3485 * int optlen);
3487 * sd - the socket descript.
3488 * level - set to IPPROTO_SCTP for all SCTP options.
3489 * optname - the option name.
3490 * optval - the buffer to store the value of the option.
3491 * optlen - the size of the buffer.
3493 SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3494 char __user *optval, unsigned int optlen)
3496 int retval = 0;
3498 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3499 sk, optname);
3501 /* I can hardly begin to describe how wrong this is. This is
3502 * so broken as to be worse than useless. The API draft
3503 * REALLY is NOT helpful here... I am not convinced that the
3504 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3505 * are at all well-founded.
3507 if (level != SOL_SCTP) {
3508 struct sctp_af *af = sctp_sk(sk)->pf->af;
3509 retval = af->setsockopt(sk, level, optname, optval, optlen);
3510 goto out_nounlock;
3513 sctp_lock_sock(sk);
3515 switch (optname) {
3516 case SCTP_SOCKOPT_BINDX_ADD:
3517 /* 'optlen' is the size of the addresses buffer. */
3518 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3519 optlen, SCTP_BINDX_ADD_ADDR);
3520 break;
3522 case SCTP_SOCKOPT_BINDX_REM:
3523 /* 'optlen' is the size of the addresses buffer. */
3524 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3525 optlen, SCTP_BINDX_REM_ADDR);
3526 break;
3528 case SCTP_SOCKOPT_CONNECTX_OLD:
3529 /* 'optlen' is the size of the addresses buffer. */
3530 retval = sctp_setsockopt_connectx_old(sk,
3531 (struct sockaddr __user *)optval,
3532 optlen);
3533 break;
3535 case SCTP_SOCKOPT_CONNECTX:
3536 /* 'optlen' is the size of the addresses buffer. */
3537 retval = sctp_setsockopt_connectx(sk,
3538 (struct sockaddr __user *)optval,
3539 optlen);
3540 break;
3542 case SCTP_DISABLE_FRAGMENTS:
3543 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3544 break;
3546 case SCTP_EVENTS:
3547 retval = sctp_setsockopt_events(sk, optval, optlen);
3548 break;
3550 case SCTP_AUTOCLOSE:
3551 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3552 break;
3554 case SCTP_PEER_ADDR_PARAMS:
3555 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3556 break;
3558 case SCTP_DELAYED_SACK:
3559 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3560 break;
3561 case SCTP_PARTIAL_DELIVERY_POINT:
3562 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3563 break;
3565 case SCTP_INITMSG:
3566 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3567 break;
3568 case SCTP_DEFAULT_SEND_PARAM:
3569 retval = sctp_setsockopt_default_send_param(sk, optval,
3570 optlen);
3571 break;
3572 case SCTP_PRIMARY_ADDR:
3573 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3574 break;
3575 case SCTP_SET_PEER_PRIMARY_ADDR:
3576 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3577 break;
3578 case SCTP_NODELAY:
3579 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3580 break;
3581 case SCTP_RTOINFO:
3582 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3583 break;
3584 case SCTP_ASSOCINFO:
3585 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3586 break;
3587 case SCTP_I_WANT_MAPPED_V4_ADDR:
3588 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3589 break;
3590 case SCTP_MAXSEG:
3591 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3592 break;
3593 case SCTP_ADAPTATION_LAYER:
3594 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3595 break;
3596 case SCTP_CONTEXT:
3597 retval = sctp_setsockopt_context(sk, optval, optlen);
3598 break;
3599 case SCTP_FRAGMENT_INTERLEAVE:
3600 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3601 break;
3602 case SCTP_MAX_BURST:
3603 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3604 break;
3605 case SCTP_AUTH_CHUNK:
3606 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3607 break;
3608 case SCTP_HMAC_IDENT:
3609 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3610 break;
3611 case SCTP_AUTH_KEY:
3612 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3613 break;
3614 case SCTP_AUTH_ACTIVE_KEY:
3615 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3616 break;
3617 case SCTP_AUTH_DELETE_KEY:
3618 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3619 break;
3620 case SCTP_AUTO_ASCONF:
3621 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3622 break;
3623 default:
3624 retval = -ENOPROTOOPT;
3625 break;
3628 sctp_release_sock(sk);
3630 out_nounlock:
3631 return retval;
3634 /* API 3.1.6 connect() - UDP Style Syntax
3636 * An application may use the connect() call in the UDP model to initiate an
3637 * association without sending data.
3639 * The syntax is:
3641 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3643 * sd: the socket descriptor to have a new association added to.
3645 * nam: the address structure (either struct sockaddr_in or struct
3646 * sockaddr_in6 defined in RFC2553 [7]).
3648 * len: the size of the address.
3650 SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3651 int addr_len)
3653 int err = 0;
3654 struct sctp_af *af;
3656 sctp_lock_sock(sk);
3658 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3659 __func__, sk, addr, addr_len);
3661 /* Validate addr_len before calling common connect/connectx routine. */
3662 af = sctp_get_af_specific(addr->sa_family);
3663 if (!af || addr_len < af->sockaddr_len) {
3664 err = -EINVAL;
3665 } else {
3666 /* Pass correct addr len to common routine (so it knows there
3667 * is only one address being passed.
3669 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3672 sctp_release_sock(sk);
3673 return err;
3676 /* FIXME: Write comments. */
3677 SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3679 return -EOPNOTSUPP; /* STUB */
3682 /* 4.1.4 accept() - TCP Style Syntax
3684 * Applications use accept() call to remove an established SCTP
3685 * association from the accept queue of the endpoint. A new socket
3686 * descriptor will be returned from accept() to represent the newly
3687 * formed association.
3689 SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3691 struct sctp_sock *sp;
3692 struct sctp_endpoint *ep;
3693 struct sock *newsk = NULL;
3694 struct sctp_association *asoc;
3695 long timeo;
3696 int error = 0;
3698 sctp_lock_sock(sk);
3700 sp = sctp_sk(sk);
3701 ep = sp->ep;
3703 if (!sctp_style(sk, TCP)) {
3704 error = -EOPNOTSUPP;
3705 goto out;
3708 if (!sctp_sstate(sk, LISTENING)) {
3709 error = -EINVAL;
3710 goto out;
3713 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3715 error = sctp_wait_for_accept(sk, timeo);
3716 if (error)
3717 goto out;
3719 /* We treat the list of associations on the endpoint as the accept
3720 * queue and pick the first association on the list.
3722 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3724 newsk = sp->pf->create_accept_sk(sk, asoc);
3725 if (!newsk) {
3726 error = -ENOMEM;
3727 goto out;
3730 /* Populate the fields of the newsk from the oldsk and migrate the
3731 * asoc to the newsk.
3733 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3735 out:
3736 sctp_release_sock(sk);
3737 *err = error;
3738 return newsk;
3741 /* The SCTP ioctl handler. */
3742 SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3744 int rc = -ENOTCONN;
3746 sctp_lock_sock(sk);
3749 * SEQPACKET-style sockets in LISTENING state are valid, for
3750 * SCTP, so only discard TCP-style sockets in LISTENING state.
3752 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3753 goto out;
3755 switch (cmd) {
3756 case SIOCINQ: {
3757 struct sk_buff *skb;
3758 unsigned int amount = 0;
3760 skb = skb_peek(&sk->sk_receive_queue);
3761 if (skb != NULL) {
3763 * We will only return the amount of this packet since
3764 * that is all that will be read.
3766 amount = skb->len;
3768 rc = put_user(amount, (int __user *)arg);
3769 break;
3771 default:
3772 rc = -ENOIOCTLCMD;
3773 break;
3775 out:
3776 sctp_release_sock(sk);
3777 return rc;
3780 /* This is the function which gets called during socket creation to
3781 * initialized the SCTP-specific portion of the sock.
3782 * The sock structure should already be zero-filled memory.
3784 SCTP_STATIC int sctp_init_sock(struct sock *sk)
3786 struct sctp_endpoint *ep;
3787 struct sctp_sock *sp;
3789 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3791 sp = sctp_sk(sk);
3793 /* Initialize the SCTP per socket area. */
3794 switch (sk->sk_type) {
3795 case SOCK_SEQPACKET:
3796 sp->type = SCTP_SOCKET_UDP;
3797 break;
3798 case SOCK_STREAM:
3799 sp->type = SCTP_SOCKET_TCP;
3800 break;
3801 default:
3802 return -ESOCKTNOSUPPORT;
3805 /* Initialize default send parameters. These parameters can be
3806 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3808 sp->default_stream = 0;
3809 sp->default_ppid = 0;
3810 sp->default_flags = 0;
3811 sp->default_context = 0;
3812 sp->default_timetolive = 0;
3814 sp->default_rcv_context = 0;
3815 sp->max_burst = sctp_max_burst;
3817 /* Initialize default setup parameters. These parameters
3818 * can be modified with the SCTP_INITMSG socket option or
3819 * overridden by the SCTP_INIT CMSG.
3821 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3822 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3823 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3824 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3826 /* Initialize default RTO related parameters. These parameters can
3827 * be modified for with the SCTP_RTOINFO socket option.
3829 sp->rtoinfo.srto_initial = sctp_rto_initial;
3830 sp->rtoinfo.srto_max = sctp_rto_max;
3831 sp->rtoinfo.srto_min = sctp_rto_min;
3833 /* Initialize default association related parameters. These parameters
3834 * can be modified with the SCTP_ASSOCINFO socket option.
3836 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3837 sp->assocparams.sasoc_number_peer_destinations = 0;
3838 sp->assocparams.sasoc_peer_rwnd = 0;
3839 sp->assocparams.sasoc_local_rwnd = 0;
3840 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3842 /* Initialize default event subscriptions. By default, all the
3843 * options are off.
3845 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3847 /* Default Peer Address Parameters. These defaults can
3848 * be modified via SCTP_PEER_ADDR_PARAMS
3850 sp->hbinterval = sctp_hb_interval;
3851 sp->pathmaxrxt = sctp_max_retrans_path;
3852 sp->pathmtu = 0; // allow default discovery
3853 sp->sackdelay = sctp_sack_timeout;
3854 sp->sackfreq = 2;
3855 sp->param_flags = SPP_HB_ENABLE |
3856 SPP_PMTUD_ENABLE |
3857 SPP_SACKDELAY_ENABLE;
3859 /* If enabled no SCTP message fragmentation will be performed.
3860 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3862 sp->disable_fragments = 0;
3864 /* Enable Nagle algorithm by default. */
3865 sp->nodelay = 0;
3867 /* Enable by default. */
3868 sp->v4mapped = 1;
3870 /* Auto-close idle associations after the configured
3871 * number of seconds. A value of 0 disables this
3872 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3873 * for UDP-style sockets only.
3875 sp->autoclose = 0;
3877 /* User specified fragmentation limit. */
3878 sp->user_frag = 0;
3880 sp->adaptation_ind = 0;
3882 sp->pf = sctp_get_pf_specific(sk->sk_family);
3884 /* Control variables for partial data delivery. */
3885 atomic_set(&sp->pd_mode, 0);
3886 skb_queue_head_init(&sp->pd_lobby);
3887 sp->frag_interleave = 0;
3889 /* Create a per socket endpoint structure. Even if we
3890 * change the data structure relationships, this may still
3891 * be useful for storing pre-connect address information.
3893 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3894 if (!ep)
3895 return -ENOMEM;
3897 sp->ep = ep;
3898 sp->hmac = NULL;
3900 SCTP_DBG_OBJCNT_INC(sock);
3902 local_bh_disable();
3903 percpu_counter_inc(&sctp_sockets_allocated);
3904 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3905 if (sctp_default_auto_asconf) {
3906 list_add_tail(&sp->auto_asconf_list,
3907 &sctp_auto_asconf_splist);
3908 sp->do_auto_asconf = 1;
3909 } else
3910 sp->do_auto_asconf = 0;
3911 local_bh_enable();
3913 return 0;
3916 /* Cleanup any SCTP per socket resources. */
3917 SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3919 struct sctp_sock *sp;
3921 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3923 /* Release our hold on the endpoint. */
3924 sp = sctp_sk(sk);
3925 if (sp->do_auto_asconf) {
3926 sp->do_auto_asconf = 0;
3927 list_del(&sp->auto_asconf_list);
3929 sctp_endpoint_free(sp->ep);
3930 local_bh_disable();
3931 percpu_counter_dec(&sctp_sockets_allocated);
3932 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3933 local_bh_enable();
3936 /* API 4.1.7 shutdown() - TCP Style Syntax
3937 * int shutdown(int socket, int how);
3939 * sd - the socket descriptor of the association to be closed.
3940 * how - Specifies the type of shutdown. The values are
3941 * as follows:
3942 * SHUT_RD
3943 * Disables further receive operations. No SCTP
3944 * protocol action is taken.
3945 * SHUT_WR
3946 * Disables further send operations, and initiates
3947 * the SCTP shutdown sequence.
3948 * SHUT_RDWR
3949 * Disables further send and receive operations
3950 * and initiates the SCTP shutdown sequence.
3952 SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3954 struct sctp_endpoint *ep;
3955 struct sctp_association *asoc;
3957 if (!sctp_style(sk, TCP))
3958 return;
3960 if (how & SEND_SHUTDOWN) {
3961 ep = sctp_sk(sk)->ep;
3962 if (!list_empty(&ep->asocs)) {
3963 asoc = list_entry(ep->asocs.next,
3964 struct sctp_association, asocs);
3965 sctp_primitive_SHUTDOWN(asoc, NULL);
3970 /* 7.2.1 Association Status (SCTP_STATUS)
3972 * Applications can retrieve current status information about an
3973 * association, including association state, peer receiver window size,
3974 * number of unacked data chunks, and number of data chunks pending
3975 * receipt. This information is read-only.
3977 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3978 char __user *optval,
3979 int __user *optlen)
3981 struct sctp_status status;
3982 struct sctp_association *asoc = NULL;
3983 struct sctp_transport *transport;
3984 sctp_assoc_t associd;
3985 int retval = 0;
3987 if (len < sizeof(status)) {
3988 retval = -EINVAL;
3989 goto out;
3992 len = sizeof(status);
3993 if (copy_from_user(&status, optval, len)) {
3994 retval = -EFAULT;
3995 goto out;
3998 associd = status.sstat_assoc_id;
3999 asoc = sctp_id2assoc(sk, associd);
4000 if (!asoc) {
4001 retval = -EINVAL;
4002 goto out;
4005 transport = asoc->peer.primary_path;
4007 status.sstat_assoc_id = sctp_assoc2id(asoc);
4008 status.sstat_state = asoc->state;
4009 status.sstat_rwnd = asoc->peer.rwnd;
4010 status.sstat_unackdata = asoc->unack_data;
4012 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4013 status.sstat_instrms = asoc->c.sinit_max_instreams;
4014 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4015 status.sstat_fragmentation_point = asoc->frag_point;
4016 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4017 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4018 transport->af_specific->sockaddr_len);
4019 /* Map ipv4 address into v4-mapped-on-v6 address. */
4020 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4021 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4022 status.sstat_primary.spinfo_state = transport->state;
4023 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4024 status.sstat_primary.spinfo_srtt = transport->srtt;
4025 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4026 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4028 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4029 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4031 if (put_user(len, optlen)) {
4032 retval = -EFAULT;
4033 goto out;
4036 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4037 len, status.sstat_state, status.sstat_rwnd,
4038 status.sstat_assoc_id);
4040 if (copy_to_user(optval, &status, len)) {
4041 retval = -EFAULT;
4042 goto out;
4045 out:
4046 return retval;
4050 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4052 * Applications can retrieve information about a specific peer address
4053 * of an association, including its reachability state, congestion
4054 * window, and retransmission timer values. This information is
4055 * read-only.
4057 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4058 char __user *optval,
4059 int __user *optlen)
4061 struct sctp_paddrinfo pinfo;
4062 struct sctp_transport *transport;
4063 int retval = 0;
4065 if (len < sizeof(pinfo)) {
4066 retval = -EINVAL;
4067 goto out;
4070 len = sizeof(pinfo);
4071 if (copy_from_user(&pinfo, optval, len)) {
4072 retval = -EFAULT;
4073 goto out;
4076 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4077 pinfo.spinfo_assoc_id);
4078 if (!transport)
4079 return -EINVAL;
4081 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4082 pinfo.spinfo_state = transport->state;
4083 pinfo.spinfo_cwnd = transport->cwnd;
4084 pinfo.spinfo_srtt = transport->srtt;
4085 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4086 pinfo.spinfo_mtu = transport->pathmtu;
4088 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4089 pinfo.spinfo_state = SCTP_ACTIVE;
4091 if (put_user(len, optlen)) {
4092 retval = -EFAULT;
4093 goto out;
4096 if (copy_to_user(optval, &pinfo, len)) {
4097 retval = -EFAULT;
4098 goto out;
4101 out:
4102 return retval;
4105 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4107 * This option is a on/off flag. If enabled no SCTP message
4108 * fragmentation will be performed. Instead if a message being sent
4109 * exceeds the current PMTU size, the message will NOT be sent and
4110 * instead a error will be indicated to the user.
4112 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4113 char __user *optval, int __user *optlen)
4115 int val;
4117 if (len < sizeof(int))
4118 return -EINVAL;
4120 len = sizeof(int);
4121 val = (sctp_sk(sk)->disable_fragments == 1);
4122 if (put_user(len, optlen))
4123 return -EFAULT;
4124 if (copy_to_user(optval, &val, len))
4125 return -EFAULT;
4126 return 0;
4129 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4131 * This socket option is used to specify various notifications and
4132 * ancillary data the user wishes to receive.
4134 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4135 int __user *optlen)
4137 if (len < sizeof(struct sctp_event_subscribe))
4138 return -EINVAL;
4139 len = sizeof(struct sctp_event_subscribe);
4140 if (put_user(len, optlen))
4141 return -EFAULT;
4142 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4143 return -EFAULT;
4144 return 0;
4147 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4149 * This socket option is applicable to the UDP-style socket only. When
4150 * set it will cause associations that are idle for more than the
4151 * specified number of seconds to automatically close. An association
4152 * being idle is defined an association that has NOT sent or received
4153 * user data. The special value of '0' indicates that no automatic
4154 * close of any associations should be performed. The option expects an
4155 * integer defining the number of seconds of idle time before an
4156 * association is closed.
4158 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4160 /* Applicable to UDP-style socket only */
4161 if (sctp_style(sk, TCP))
4162 return -EOPNOTSUPP;
4163 if (len < sizeof(int))
4164 return -EINVAL;
4165 len = sizeof(int);
4166 if (put_user(len, optlen))
4167 return -EFAULT;
4168 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4169 return -EFAULT;
4170 return 0;
4173 /* Helper routine to branch off an association to a new socket. */
4174 SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4175 struct socket **sockp)
4177 struct sock *sk = asoc->base.sk;
4178 struct socket *sock;
4179 struct sctp_af *af;
4180 int err = 0;
4182 /* An association cannot be branched off from an already peeled-off
4183 * socket, nor is this supported for tcp style sockets.
4185 if (!sctp_style(sk, UDP))
4186 return -EINVAL;
4188 /* Create a new socket. */
4189 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4190 if (err < 0)
4191 return err;
4193 sctp_copy_sock(sock->sk, sk, asoc);
4195 /* Make peeled-off sockets more like 1-1 accepted sockets.
4196 * Set the daddr and initialize id to something more random
4198 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4199 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4201 /* Populate the fields of the newsk from the oldsk and migrate the
4202 * asoc to the newsk.
4204 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4206 *sockp = sock;
4208 return err;
4211 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4213 sctp_peeloff_arg_t peeloff;
4214 struct socket *newsock;
4215 int retval = 0;
4216 struct sctp_association *asoc;
4218 if (len < sizeof(sctp_peeloff_arg_t))
4219 return -EINVAL;
4220 len = sizeof(sctp_peeloff_arg_t);
4221 if (copy_from_user(&peeloff, optval, len))
4222 return -EFAULT;
4224 asoc = sctp_id2assoc(sk, peeloff.associd);
4225 if (!asoc) {
4226 retval = -EINVAL;
4227 goto out;
4230 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4232 retval = sctp_do_peeloff(asoc, &newsock);
4233 if (retval < 0)
4234 goto out;
4236 /* Map the socket to an unused fd that can be returned to the user. */
4237 retval = sock_map_fd(newsock, 0);
4238 if (retval < 0) {
4239 sock_release(newsock);
4240 goto out;
4243 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4244 __func__, sk, asoc, newsock->sk, retval);
4246 /* Return the fd mapped to the new socket. */
4247 peeloff.sd = retval;
4248 if (put_user(len, optlen))
4249 return -EFAULT;
4250 if (copy_to_user(optval, &peeloff, len))
4251 retval = -EFAULT;
4253 out:
4254 return retval;
4257 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4259 * Applications can enable or disable heartbeats for any peer address of
4260 * an association, modify an address's heartbeat interval, force a
4261 * heartbeat to be sent immediately, and adjust the address's maximum
4262 * number of retransmissions sent before an address is considered
4263 * unreachable. The following structure is used to access and modify an
4264 * address's parameters:
4266 * struct sctp_paddrparams {
4267 * sctp_assoc_t spp_assoc_id;
4268 * struct sockaddr_storage spp_address;
4269 * uint32_t spp_hbinterval;
4270 * uint16_t spp_pathmaxrxt;
4271 * uint32_t spp_pathmtu;
4272 * uint32_t spp_sackdelay;
4273 * uint32_t spp_flags;
4274 * };
4276 * spp_assoc_id - (one-to-many style socket) This is filled in the
4277 * application, and identifies the association for
4278 * this query.
4279 * spp_address - This specifies which address is of interest.
4280 * spp_hbinterval - This contains the value of the heartbeat interval,
4281 * in milliseconds. If a value of zero
4282 * is present in this field then no changes are to
4283 * be made to this parameter.
4284 * spp_pathmaxrxt - This contains the maximum number of
4285 * retransmissions before this address shall be
4286 * considered unreachable. If a value of zero
4287 * is present in this field then no changes are to
4288 * be made to this parameter.
4289 * spp_pathmtu - When Path MTU discovery is disabled the value
4290 * specified here will be the "fixed" path mtu.
4291 * Note that if the spp_address field is empty
4292 * then all associations on this address will
4293 * have this fixed path mtu set upon them.
4295 * spp_sackdelay - When delayed sack is enabled, this value specifies
4296 * the number of milliseconds that sacks will be delayed
4297 * for. This value will apply to all addresses of an
4298 * association if the spp_address field is empty. Note
4299 * also, that if delayed sack is enabled and this
4300 * value is set to 0, no change is made to the last
4301 * recorded delayed sack timer value.
4303 * spp_flags - These flags are used to control various features
4304 * on an association. The flag field may contain
4305 * zero or more of the following options.
4307 * SPP_HB_ENABLE - Enable heartbeats on the
4308 * specified address. Note that if the address
4309 * field is empty all addresses for the association
4310 * have heartbeats enabled upon them.
4312 * SPP_HB_DISABLE - Disable heartbeats on the
4313 * speicifed address. Note that if the address
4314 * field is empty all addresses for the association
4315 * will have their heartbeats disabled. Note also
4316 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4317 * mutually exclusive, only one of these two should
4318 * be specified. Enabling both fields will have
4319 * undetermined results.
4321 * SPP_HB_DEMAND - Request a user initiated heartbeat
4322 * to be made immediately.
4324 * SPP_PMTUD_ENABLE - This field will enable PMTU
4325 * discovery upon the specified address. Note that
4326 * if the address feild is empty then all addresses
4327 * on the association are effected.
4329 * SPP_PMTUD_DISABLE - This field will disable PMTU
4330 * discovery upon the specified address. Note that
4331 * if the address feild is empty then all addresses
4332 * on the association are effected. Not also that
4333 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4334 * exclusive. Enabling both will have undetermined
4335 * results.
4337 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4338 * on delayed sack. The time specified in spp_sackdelay
4339 * is used to specify the sack delay for this address. Note
4340 * that if spp_address is empty then all addresses will
4341 * enable delayed sack and take on the sack delay
4342 * value specified in spp_sackdelay.
4343 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4344 * off delayed sack. If the spp_address field is blank then
4345 * delayed sack is disabled for the entire association. Note
4346 * also that this field is mutually exclusive to
4347 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4348 * results.
4350 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4351 char __user *optval, int __user *optlen)
4353 struct sctp_paddrparams params;
4354 struct sctp_transport *trans = NULL;
4355 struct sctp_association *asoc = NULL;
4356 struct sctp_sock *sp = sctp_sk(sk);
4358 if (len < sizeof(struct sctp_paddrparams))
4359 return -EINVAL;
4360 len = sizeof(struct sctp_paddrparams);
4361 if (copy_from_user(&params, optval, len))
4362 return -EFAULT;
4364 /* If an address other than INADDR_ANY is specified, and
4365 * no transport is found, then the request is invalid.
4367 if (!sctp_is_any(sk, ( union sctp_addr *)&params.spp_address)) {
4368 trans = sctp_addr_id2transport(sk, &params.spp_address,
4369 params.spp_assoc_id);
4370 if (!trans) {
4371 SCTP_DEBUG_PRINTK("Failed no transport\n");
4372 return -EINVAL;
4376 /* Get association, if assoc_id != 0 and the socket is a one
4377 * to many style socket, and an association was not found, then
4378 * the id was invalid.
4380 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4381 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4382 SCTP_DEBUG_PRINTK("Failed no association\n");
4383 return -EINVAL;
4386 if (trans) {
4387 /* Fetch transport values. */
4388 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4389 params.spp_pathmtu = trans->pathmtu;
4390 params.spp_pathmaxrxt = trans->pathmaxrxt;
4391 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4393 /*draft-11 doesn't say what to return in spp_flags*/
4394 params.spp_flags = trans->param_flags;
4395 } else if (asoc) {
4396 /* Fetch association values. */
4397 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4398 params.spp_pathmtu = asoc->pathmtu;
4399 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4400 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4402 /*draft-11 doesn't say what to return in spp_flags*/
4403 params.spp_flags = asoc->param_flags;
4404 } else {
4405 /* Fetch socket values. */
4406 params.spp_hbinterval = sp->hbinterval;
4407 params.spp_pathmtu = sp->pathmtu;
4408 params.spp_sackdelay = sp->sackdelay;
4409 params.spp_pathmaxrxt = sp->pathmaxrxt;
4411 /*draft-11 doesn't say what to return in spp_flags*/
4412 params.spp_flags = sp->param_flags;
4415 if (copy_to_user(optval, &params, len))
4416 return -EFAULT;
4418 if (put_user(len, optlen))
4419 return -EFAULT;
4421 return 0;
4425 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4427 * This option will effect the way delayed acks are performed. This
4428 * option allows you to get or set the delayed ack time, in
4429 * milliseconds. It also allows changing the delayed ack frequency.
4430 * Changing the frequency to 1 disables the delayed sack algorithm. If
4431 * the assoc_id is 0, then this sets or gets the endpoints default
4432 * values. If the assoc_id field is non-zero, then the set or get
4433 * effects the specified association for the one to many model (the
4434 * assoc_id field is ignored by the one to one model). Note that if
4435 * sack_delay or sack_freq are 0 when setting this option, then the
4436 * current values will remain unchanged.
4438 * struct sctp_sack_info {
4439 * sctp_assoc_t sack_assoc_id;
4440 * uint32_t sack_delay;
4441 * uint32_t sack_freq;
4442 * };
4444 * sack_assoc_id - This parameter, indicates which association the user
4445 * is performing an action upon. Note that if this field's value is
4446 * zero then the endpoints default value is changed (effecting future
4447 * associations only).
4449 * sack_delay - This parameter contains the number of milliseconds that
4450 * the user is requesting the delayed ACK timer be set to. Note that
4451 * this value is defined in the standard to be between 200 and 500
4452 * milliseconds.
4454 * sack_freq - This parameter contains the number of packets that must
4455 * be received before a sack is sent without waiting for the delay
4456 * timer to expire. The default value for this is 2, setting this
4457 * value to 1 will disable the delayed sack algorithm.
4459 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4460 char __user *optval,
4461 int __user *optlen)
4463 struct sctp_sack_info params;
4464 struct sctp_association *asoc = NULL;
4465 struct sctp_sock *sp = sctp_sk(sk);
4467 if (len >= sizeof(struct sctp_sack_info)) {
4468 len = sizeof(struct sctp_sack_info);
4470 if (copy_from_user(&params, optval, len))
4471 return -EFAULT;
4472 } else if (len == sizeof(struct sctp_assoc_value)) {
4473 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4474 pr_warn("Use struct sctp_sack_info instead\n");
4475 if (copy_from_user(&params, optval, len))
4476 return -EFAULT;
4477 } else
4478 return - EINVAL;
4480 /* Get association, if sack_assoc_id != 0 and the socket is a one
4481 * to many style socket, and an association was not found, then
4482 * the id was invalid.
4484 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4485 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4486 return -EINVAL;
4488 if (asoc) {
4489 /* Fetch association values. */
4490 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4491 params.sack_delay = jiffies_to_msecs(
4492 asoc->sackdelay);
4493 params.sack_freq = asoc->sackfreq;
4495 } else {
4496 params.sack_delay = 0;
4497 params.sack_freq = 1;
4499 } else {
4500 /* Fetch socket values. */
4501 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4502 params.sack_delay = sp->sackdelay;
4503 params.sack_freq = sp->sackfreq;
4504 } else {
4505 params.sack_delay = 0;
4506 params.sack_freq = 1;
4510 if (copy_to_user(optval, &params, len))
4511 return -EFAULT;
4513 if (put_user(len, optlen))
4514 return -EFAULT;
4516 return 0;
4519 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4521 * Applications can specify protocol parameters for the default association
4522 * initialization. The option name argument to setsockopt() and getsockopt()
4523 * is SCTP_INITMSG.
4525 * Setting initialization parameters is effective only on an unconnected
4526 * socket (for UDP-style sockets only future associations are effected
4527 * by the change). With TCP-style sockets, this option is inherited by
4528 * sockets derived from a listener socket.
4530 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4532 if (len < sizeof(struct sctp_initmsg))
4533 return -EINVAL;
4534 len = sizeof(struct sctp_initmsg);
4535 if (put_user(len, optlen))
4536 return -EFAULT;
4537 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4538 return -EFAULT;
4539 return 0;
4543 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4544 char __user *optval, int __user *optlen)
4546 struct sctp_association *asoc;
4547 int cnt = 0;
4548 struct sctp_getaddrs getaddrs;
4549 struct sctp_transport *from;
4550 void __user *to;
4551 union sctp_addr temp;
4552 struct sctp_sock *sp = sctp_sk(sk);
4553 int addrlen;
4554 size_t space_left;
4555 int bytes_copied;
4557 if (len < sizeof(struct sctp_getaddrs))
4558 return -EINVAL;
4560 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4561 return -EFAULT;
4563 /* For UDP-style sockets, id specifies the association to query. */
4564 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4565 if (!asoc)
4566 return -EINVAL;
4568 to = optval + offsetof(struct sctp_getaddrs,addrs);
4569 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4571 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4572 transports) {
4573 memcpy(&temp, &from->ipaddr, sizeof(temp));
4574 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4575 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4576 if (space_left < addrlen)
4577 return -ENOMEM;
4578 if (copy_to_user(to, &temp, addrlen))
4579 return -EFAULT;
4580 to += addrlen;
4581 cnt++;
4582 space_left -= addrlen;
4585 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4586 return -EFAULT;
4587 bytes_copied = ((char __user *)to) - optval;
4588 if (put_user(bytes_copied, optlen))
4589 return -EFAULT;
4591 return 0;
4594 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4595 size_t space_left, int *bytes_copied)
4597 struct sctp_sockaddr_entry *addr;
4598 union sctp_addr temp;
4599 int cnt = 0;
4600 int addrlen;
4602 rcu_read_lock();
4603 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4604 if (!addr->valid)
4605 continue;
4607 if ((PF_INET == sk->sk_family) &&
4608 (AF_INET6 == addr->a.sa.sa_family))
4609 continue;
4610 if ((PF_INET6 == sk->sk_family) &&
4611 inet_v6_ipv6only(sk) &&
4612 (AF_INET == addr->a.sa.sa_family))
4613 continue;
4614 memcpy(&temp, &addr->a, sizeof(temp));
4615 if (!temp.v4.sin_port)
4616 temp.v4.sin_port = htons(port);
4618 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4619 &temp);
4620 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4621 if (space_left < addrlen) {
4622 cnt = -ENOMEM;
4623 break;
4625 memcpy(to, &temp, addrlen);
4627 to += addrlen;
4628 cnt ++;
4629 space_left -= addrlen;
4630 *bytes_copied += addrlen;
4632 rcu_read_unlock();
4634 return cnt;
4638 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4639 char __user *optval, int __user *optlen)
4641 struct sctp_bind_addr *bp;
4642 struct sctp_association *asoc;
4643 int cnt = 0;
4644 struct sctp_getaddrs getaddrs;
4645 struct sctp_sockaddr_entry *addr;
4646 void __user *to;
4647 union sctp_addr temp;
4648 struct sctp_sock *sp = sctp_sk(sk);
4649 int addrlen;
4650 int err = 0;
4651 size_t space_left;
4652 int bytes_copied = 0;
4653 void *addrs;
4654 void *buf;
4656 if (len < sizeof(struct sctp_getaddrs))
4657 return -EINVAL;
4659 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4660 return -EFAULT;
4663 * For UDP-style sockets, id specifies the association to query.
4664 * If the id field is set to the value '0' then the locally bound
4665 * addresses are returned without regard to any particular
4666 * association.
4668 if (0 == getaddrs.assoc_id) {
4669 bp = &sctp_sk(sk)->ep->base.bind_addr;
4670 } else {
4671 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4672 if (!asoc)
4673 return -EINVAL;
4674 bp = &asoc->base.bind_addr;
4677 to = optval + offsetof(struct sctp_getaddrs,addrs);
4678 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4680 addrs = kmalloc(space_left, GFP_KERNEL);
4681 if (!addrs)
4682 return -ENOMEM;
4684 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4685 * addresses from the global local address list.
4687 if (sctp_list_single_entry(&bp->address_list)) {
4688 addr = list_entry(bp->address_list.next,
4689 struct sctp_sockaddr_entry, list);
4690 if (sctp_is_any(sk, &addr->a)) {
4691 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4692 space_left, &bytes_copied);
4693 if (cnt < 0) {
4694 err = cnt;
4695 goto out;
4697 goto copy_getaddrs;
4701 buf = addrs;
4702 /* Protection on the bound address list is not needed since
4703 * in the socket option context we hold a socket lock and
4704 * thus the bound address list can't change.
4706 list_for_each_entry(addr, &bp->address_list, list) {
4707 memcpy(&temp, &addr->a, sizeof(temp));
4708 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4709 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4710 if (space_left < addrlen) {
4711 err = -ENOMEM; /*fixme: right error?*/
4712 goto out;
4714 memcpy(buf, &temp, addrlen);
4715 buf += addrlen;
4716 bytes_copied += addrlen;
4717 cnt ++;
4718 space_left -= addrlen;
4721 copy_getaddrs:
4722 if (copy_to_user(to, addrs, bytes_copied)) {
4723 err = -EFAULT;
4724 goto out;
4726 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4727 err = -EFAULT;
4728 goto out;
4730 if (put_user(bytes_copied, optlen))
4731 err = -EFAULT;
4732 out:
4733 kfree(addrs);
4734 return err;
4737 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4739 * Requests that the local SCTP stack use the enclosed peer address as
4740 * the association primary. The enclosed address must be one of the
4741 * association peer's addresses.
4743 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4744 char __user *optval, int __user *optlen)
4746 struct sctp_prim prim;
4747 struct sctp_association *asoc;
4748 struct sctp_sock *sp = sctp_sk(sk);
4750 if (len < sizeof(struct sctp_prim))
4751 return -EINVAL;
4753 len = sizeof(struct sctp_prim);
4755 if (copy_from_user(&prim, optval, len))
4756 return -EFAULT;
4758 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4759 if (!asoc)
4760 return -EINVAL;
4762 if (!asoc->peer.primary_path)
4763 return -ENOTCONN;
4765 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4766 asoc->peer.primary_path->af_specific->sockaddr_len);
4768 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4769 (union sctp_addr *)&prim.ssp_addr);
4771 if (put_user(len, optlen))
4772 return -EFAULT;
4773 if (copy_to_user(optval, &prim, len))
4774 return -EFAULT;
4776 return 0;
4780 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4782 * Requests that the local endpoint set the specified Adaptation Layer
4783 * Indication parameter for all future INIT and INIT-ACK exchanges.
4785 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4786 char __user *optval, int __user *optlen)
4788 struct sctp_setadaptation adaptation;
4790 if (len < sizeof(struct sctp_setadaptation))
4791 return -EINVAL;
4793 len = sizeof(struct sctp_setadaptation);
4795 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4797 if (put_user(len, optlen))
4798 return -EFAULT;
4799 if (copy_to_user(optval, &adaptation, len))
4800 return -EFAULT;
4802 return 0;
4807 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4809 * Applications that wish to use the sendto() system call may wish to
4810 * specify a default set of parameters that would normally be supplied
4811 * through the inclusion of ancillary data. This socket option allows
4812 * such an application to set the default sctp_sndrcvinfo structure.
4815 * The application that wishes to use this socket option simply passes
4816 * in to this call the sctp_sndrcvinfo structure defined in Section
4817 * 5.2.2) The input parameters accepted by this call include
4818 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4819 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4820 * to this call if the caller is using the UDP model.
4822 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4824 static int sctp_getsockopt_default_send_param(struct sock *sk,
4825 int len, char __user *optval,
4826 int __user *optlen)
4828 struct sctp_sndrcvinfo info;
4829 struct sctp_association *asoc;
4830 struct sctp_sock *sp = sctp_sk(sk);
4832 if (len < sizeof(struct sctp_sndrcvinfo))
4833 return -EINVAL;
4835 len = sizeof(struct sctp_sndrcvinfo);
4837 if (copy_from_user(&info, optval, len))
4838 return -EFAULT;
4840 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4841 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4842 return -EINVAL;
4844 if (asoc) {
4845 info.sinfo_stream = asoc->default_stream;
4846 info.sinfo_flags = asoc->default_flags;
4847 info.sinfo_ppid = asoc->default_ppid;
4848 info.sinfo_context = asoc->default_context;
4849 info.sinfo_timetolive = asoc->default_timetolive;
4850 } else {
4851 info.sinfo_stream = sp->default_stream;
4852 info.sinfo_flags = sp->default_flags;
4853 info.sinfo_ppid = sp->default_ppid;
4854 info.sinfo_context = sp->default_context;
4855 info.sinfo_timetolive = sp->default_timetolive;
4858 if (put_user(len, optlen))
4859 return -EFAULT;
4860 if (copy_to_user(optval, &info, len))
4861 return -EFAULT;
4863 return 0;
4868 * 7.1.5 SCTP_NODELAY
4870 * Turn on/off any Nagle-like algorithm. This means that packets are
4871 * generally sent as soon as possible and no unnecessary delays are
4872 * introduced, at the cost of more packets in the network. Expects an
4873 * integer boolean flag.
4876 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4877 char __user *optval, int __user *optlen)
4879 int val;
4881 if (len < sizeof(int))
4882 return -EINVAL;
4884 len = sizeof(int);
4885 val = (sctp_sk(sk)->nodelay == 1);
4886 if (put_user(len, optlen))
4887 return -EFAULT;
4888 if (copy_to_user(optval, &val, len))
4889 return -EFAULT;
4890 return 0;
4895 * 7.1.1 SCTP_RTOINFO
4897 * The protocol parameters used to initialize and bound retransmission
4898 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4899 * and modify these parameters.
4900 * All parameters are time values, in milliseconds. A value of 0, when
4901 * modifying the parameters, indicates that the current value should not
4902 * be changed.
4905 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4906 char __user *optval,
4907 int __user *optlen) {
4908 struct sctp_rtoinfo rtoinfo;
4909 struct sctp_association *asoc;
4911 if (len < sizeof (struct sctp_rtoinfo))
4912 return -EINVAL;
4914 len = sizeof(struct sctp_rtoinfo);
4916 if (copy_from_user(&rtoinfo, optval, len))
4917 return -EFAULT;
4919 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4921 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4922 return -EINVAL;
4924 /* Values corresponding to the specific association. */
4925 if (asoc) {
4926 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4927 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4928 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4929 } else {
4930 /* Values corresponding to the endpoint. */
4931 struct sctp_sock *sp = sctp_sk(sk);
4933 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4934 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4935 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4938 if (put_user(len, optlen))
4939 return -EFAULT;
4941 if (copy_to_user(optval, &rtoinfo, len))
4942 return -EFAULT;
4944 return 0;
4949 * 7.1.2 SCTP_ASSOCINFO
4951 * This option is used to tune the maximum retransmission attempts
4952 * of the association.
4953 * Returns an error if the new association retransmission value is
4954 * greater than the sum of the retransmission value of the peer.
4955 * See [SCTP] for more information.
4958 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4959 char __user *optval,
4960 int __user *optlen)
4963 struct sctp_assocparams assocparams;
4964 struct sctp_association *asoc;
4965 struct list_head *pos;
4966 int cnt = 0;
4968 if (len < sizeof (struct sctp_assocparams))
4969 return -EINVAL;
4971 len = sizeof(struct sctp_assocparams);
4973 if (copy_from_user(&assocparams, optval, len))
4974 return -EFAULT;
4976 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4978 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4979 return -EINVAL;
4981 /* Values correspoinding to the specific association */
4982 if (asoc) {
4983 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4984 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4985 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4986 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4987 * 1000) +
4988 (asoc->cookie_life.tv_usec
4989 / 1000);
4991 list_for_each(pos, &asoc->peer.transport_addr_list) {
4992 cnt ++;
4995 assocparams.sasoc_number_peer_destinations = cnt;
4996 } else {
4997 /* Values corresponding to the endpoint */
4998 struct sctp_sock *sp = sctp_sk(sk);
5000 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5001 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5002 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5003 assocparams.sasoc_cookie_life =
5004 sp->assocparams.sasoc_cookie_life;
5005 assocparams.sasoc_number_peer_destinations =
5006 sp->assocparams.
5007 sasoc_number_peer_destinations;
5010 if (put_user(len, optlen))
5011 return -EFAULT;
5013 if (copy_to_user(optval, &assocparams, len))
5014 return -EFAULT;
5016 return 0;
5020 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5022 * This socket option is a boolean flag which turns on or off mapped V4
5023 * addresses. If this option is turned on and the socket is type
5024 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5025 * If this option is turned off, then no mapping will be done of V4
5026 * addresses and a user will receive both PF_INET6 and PF_INET type
5027 * addresses on the socket.
5029 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5030 char __user *optval, int __user *optlen)
5032 int val;
5033 struct sctp_sock *sp = sctp_sk(sk);
5035 if (len < sizeof(int))
5036 return -EINVAL;
5038 len = sizeof(int);
5039 val = sp->v4mapped;
5040 if (put_user(len, optlen))
5041 return -EFAULT;
5042 if (copy_to_user(optval, &val, len))
5043 return -EFAULT;
5045 return 0;
5049 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5050 * (chapter and verse is quoted at sctp_setsockopt_context())
5052 static int sctp_getsockopt_context(struct sock *sk, int len,
5053 char __user *optval, int __user *optlen)
5055 struct sctp_assoc_value params;
5056 struct sctp_sock *sp;
5057 struct sctp_association *asoc;
5059 if (len < sizeof(struct sctp_assoc_value))
5060 return -EINVAL;
5062 len = sizeof(struct sctp_assoc_value);
5064 if (copy_from_user(&params, optval, len))
5065 return -EFAULT;
5067 sp = sctp_sk(sk);
5069 if (params.assoc_id != 0) {
5070 asoc = sctp_id2assoc(sk, params.assoc_id);
5071 if (!asoc)
5072 return -EINVAL;
5073 params.assoc_value = asoc->default_rcv_context;
5074 } else {
5075 params.assoc_value = sp->default_rcv_context;
5078 if (put_user(len, optlen))
5079 return -EFAULT;
5080 if (copy_to_user(optval, &params, len))
5081 return -EFAULT;
5083 return 0;
5087 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5088 * This option will get or set the maximum size to put in any outgoing
5089 * SCTP DATA chunk. If a message is larger than this size it will be
5090 * fragmented by SCTP into the specified size. Note that the underlying
5091 * SCTP implementation may fragment into smaller sized chunks when the
5092 * PMTU of the underlying association is smaller than the value set by
5093 * the user. The default value for this option is '0' which indicates
5094 * the user is NOT limiting fragmentation and only the PMTU will effect
5095 * SCTP's choice of DATA chunk size. Note also that values set larger
5096 * than the maximum size of an IP datagram will effectively let SCTP
5097 * control fragmentation (i.e. the same as setting this option to 0).
5099 * The following structure is used to access and modify this parameter:
5101 * struct sctp_assoc_value {
5102 * sctp_assoc_t assoc_id;
5103 * uint32_t assoc_value;
5104 * };
5106 * assoc_id: This parameter is ignored for one-to-one style sockets.
5107 * For one-to-many style sockets this parameter indicates which
5108 * association the user is performing an action upon. Note that if
5109 * this field's value is zero then the endpoints default value is
5110 * changed (effecting future associations only).
5111 * assoc_value: This parameter specifies the maximum size in bytes.
5113 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5114 char __user *optval, int __user *optlen)
5116 struct sctp_assoc_value params;
5117 struct sctp_association *asoc;
5119 if (len == sizeof(int)) {
5120 pr_warn("Use of int in maxseg socket option deprecated\n");
5121 pr_warn("Use struct sctp_assoc_value instead\n");
5122 params.assoc_id = 0;
5123 } else if (len >= sizeof(struct sctp_assoc_value)) {
5124 len = sizeof(struct sctp_assoc_value);
5125 if (copy_from_user(&params, optval, sizeof(params)))
5126 return -EFAULT;
5127 } else
5128 return -EINVAL;
5130 asoc = sctp_id2assoc(sk, params.assoc_id);
5131 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5132 return -EINVAL;
5134 if (asoc)
5135 params.assoc_value = asoc->frag_point;
5136 else
5137 params.assoc_value = sctp_sk(sk)->user_frag;
5139 if (put_user(len, optlen))
5140 return -EFAULT;
5141 if (len == sizeof(int)) {
5142 if (copy_to_user(optval, &params.assoc_value, len))
5143 return -EFAULT;
5144 } else {
5145 if (copy_to_user(optval, &params, len))
5146 return -EFAULT;
5149 return 0;
5153 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5154 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5156 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5157 char __user *optval, int __user *optlen)
5159 int val;
5161 if (len < sizeof(int))
5162 return -EINVAL;
5164 len = sizeof(int);
5166 val = sctp_sk(sk)->frag_interleave;
5167 if (put_user(len, optlen))
5168 return -EFAULT;
5169 if (copy_to_user(optval, &val, len))
5170 return -EFAULT;
5172 return 0;
5176 * 7.1.25. Set or Get the sctp partial delivery point
5177 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5179 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5180 char __user *optval,
5181 int __user *optlen)
5183 u32 val;
5185 if (len < sizeof(u32))
5186 return -EINVAL;
5188 len = sizeof(u32);
5190 val = sctp_sk(sk)->pd_point;
5191 if (put_user(len, optlen))
5192 return -EFAULT;
5193 if (copy_to_user(optval, &val, len))
5194 return -EFAULT;
5196 return 0;
5200 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5201 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5203 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5204 char __user *optval,
5205 int __user *optlen)
5207 struct sctp_assoc_value params;
5208 struct sctp_sock *sp;
5209 struct sctp_association *asoc;
5211 if (len == sizeof(int)) {
5212 pr_warn("Use of int in max_burst socket option deprecated\n");
5213 pr_warn("Use struct sctp_assoc_value instead\n");
5214 params.assoc_id = 0;
5215 } else if (len >= sizeof(struct sctp_assoc_value)) {
5216 len = sizeof(struct sctp_assoc_value);
5217 if (copy_from_user(&params, optval, len))
5218 return -EFAULT;
5219 } else
5220 return -EINVAL;
5222 sp = sctp_sk(sk);
5224 if (params.assoc_id != 0) {
5225 asoc = sctp_id2assoc(sk, params.assoc_id);
5226 if (!asoc)
5227 return -EINVAL;
5228 params.assoc_value = asoc->max_burst;
5229 } else
5230 params.assoc_value = sp->max_burst;
5232 if (len == sizeof(int)) {
5233 if (copy_to_user(optval, &params.assoc_value, len))
5234 return -EFAULT;
5235 } else {
5236 if (copy_to_user(optval, &params, len))
5237 return -EFAULT;
5240 return 0;
5244 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5245 char __user *optval, int __user *optlen)
5247 struct sctp_hmacalgo __user *p = (void __user *)optval;
5248 struct sctp_hmac_algo_param *hmacs;
5249 __u16 data_len = 0;
5250 u32 num_idents;
5252 if (!sctp_auth_enable)
5253 return -EACCES;
5255 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5256 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5258 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5259 return -EINVAL;
5261 len = sizeof(struct sctp_hmacalgo) + data_len;
5262 num_idents = data_len / sizeof(u16);
5264 if (put_user(len, optlen))
5265 return -EFAULT;
5266 if (put_user(num_idents, &p->shmac_num_idents))
5267 return -EFAULT;
5268 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5269 return -EFAULT;
5270 return 0;
5273 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5274 char __user *optval, int __user *optlen)
5276 struct sctp_authkeyid val;
5277 struct sctp_association *asoc;
5279 if (!sctp_auth_enable)
5280 return -EACCES;
5282 if (len < sizeof(struct sctp_authkeyid))
5283 return -EINVAL;
5284 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5285 return -EFAULT;
5287 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5288 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5289 return -EINVAL;
5291 if (asoc)
5292 val.scact_keynumber = asoc->active_key_id;
5293 else
5294 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5296 len = sizeof(struct sctp_authkeyid);
5297 if (put_user(len, optlen))
5298 return -EFAULT;
5299 if (copy_to_user(optval, &val, len))
5300 return -EFAULT;
5302 return 0;
5305 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5306 char __user *optval, int __user *optlen)
5308 struct sctp_authchunks __user *p = (void __user *)optval;
5309 struct sctp_authchunks val;
5310 struct sctp_association *asoc;
5311 struct sctp_chunks_param *ch;
5312 u32 num_chunks = 0;
5313 char __user *to;
5315 if (!sctp_auth_enable)
5316 return -EACCES;
5318 if (len < sizeof(struct sctp_authchunks))
5319 return -EINVAL;
5321 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5322 return -EFAULT;
5324 to = p->gauth_chunks;
5325 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5326 if (!asoc)
5327 return -EINVAL;
5329 ch = asoc->peer.peer_chunks;
5330 if (!ch)
5331 goto num;
5333 /* See if the user provided enough room for all the data */
5334 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5335 if (len < num_chunks)
5336 return -EINVAL;
5338 if (copy_to_user(to, ch->chunks, num_chunks))
5339 return -EFAULT;
5340 num:
5341 len = sizeof(struct sctp_authchunks) + num_chunks;
5342 if (put_user(len, optlen)) return -EFAULT;
5343 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5344 return -EFAULT;
5345 return 0;
5348 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5349 char __user *optval, int __user *optlen)
5351 struct sctp_authchunks __user *p = (void __user *)optval;
5352 struct sctp_authchunks val;
5353 struct sctp_association *asoc;
5354 struct sctp_chunks_param *ch;
5355 u32 num_chunks = 0;
5356 char __user *to;
5358 if (!sctp_auth_enable)
5359 return -EACCES;
5361 if (len < sizeof(struct sctp_authchunks))
5362 return -EINVAL;
5364 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5365 return -EFAULT;
5367 to = p->gauth_chunks;
5368 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5369 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5370 return -EINVAL;
5372 if (asoc)
5373 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5374 else
5375 ch = sctp_sk(sk)->ep->auth_chunk_list;
5377 if (!ch)
5378 goto num;
5380 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5381 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5382 return -EINVAL;
5384 if (copy_to_user(to, ch->chunks, num_chunks))
5385 return -EFAULT;
5386 num:
5387 len = sizeof(struct sctp_authchunks) + num_chunks;
5388 if (put_user(len, optlen))
5389 return -EFAULT;
5390 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5391 return -EFAULT;
5393 return 0;
5397 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5398 * This option gets the current number of associations that are attached
5399 * to a one-to-many style socket. The option value is an uint32_t.
5401 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5402 char __user *optval, int __user *optlen)
5404 struct sctp_sock *sp = sctp_sk(sk);
5405 struct sctp_association *asoc;
5406 u32 val = 0;
5408 if (sctp_style(sk, TCP))
5409 return -EOPNOTSUPP;
5411 if (len < sizeof(u32))
5412 return -EINVAL;
5414 len = sizeof(u32);
5416 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5417 val++;
5420 if (put_user(len, optlen))
5421 return -EFAULT;
5422 if (copy_to_user(optval, &val, len))
5423 return -EFAULT;
5425 return 0;
5429 * 8.1.23 SCTP_AUTO_ASCONF
5430 * See the corresponding setsockopt entry as description
5432 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5433 char __user *optval, int __user *optlen)
5435 int val = 0;
5437 if (len < sizeof(int))
5438 return -EINVAL;
5440 len = sizeof(int);
5441 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5442 val = 1;
5443 if (put_user(len, optlen))
5444 return -EFAULT;
5445 if (copy_to_user(optval, &val, len))
5446 return -EFAULT;
5447 return 0;
5451 * 8.2.6. Get the Current Identifiers of Associations
5452 * (SCTP_GET_ASSOC_ID_LIST)
5454 * This option gets the current list of SCTP association identifiers of
5455 * the SCTP associations handled by a one-to-many style socket.
5457 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5458 char __user *optval, int __user *optlen)
5460 struct sctp_sock *sp = sctp_sk(sk);
5461 struct sctp_association *asoc;
5462 struct sctp_assoc_ids *ids;
5463 u32 num = 0;
5465 if (sctp_style(sk, TCP))
5466 return -EOPNOTSUPP;
5468 if (len < sizeof(struct sctp_assoc_ids))
5469 return -EINVAL;
5471 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5472 num++;
5475 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5476 return -EINVAL;
5478 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5480 ids = kmalloc(len, GFP_KERNEL);
5481 if (unlikely(!ids))
5482 return -ENOMEM;
5484 ids->gaids_number_of_ids = num;
5485 num = 0;
5486 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5487 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5490 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5491 kfree(ids);
5492 return -EFAULT;
5495 kfree(ids);
5496 return 0;
5499 SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5500 char __user *optval, int __user *optlen)
5502 int retval = 0;
5503 int len;
5505 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5506 sk, optname);
5508 /* I can hardly begin to describe how wrong this is. This is
5509 * so broken as to be worse than useless. The API draft
5510 * REALLY is NOT helpful here... I am not convinced that the
5511 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5512 * are at all well-founded.
5514 if (level != SOL_SCTP) {
5515 struct sctp_af *af = sctp_sk(sk)->pf->af;
5517 retval = af->getsockopt(sk, level, optname, optval, optlen);
5518 return retval;
5521 if (get_user(len, optlen))
5522 return -EFAULT;
5524 sctp_lock_sock(sk);
5526 switch (optname) {
5527 case SCTP_STATUS:
5528 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5529 break;
5530 case SCTP_DISABLE_FRAGMENTS:
5531 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5532 optlen);
5533 break;
5534 case SCTP_EVENTS:
5535 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5536 break;
5537 case SCTP_AUTOCLOSE:
5538 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5539 break;
5540 case SCTP_SOCKOPT_PEELOFF:
5541 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5542 break;
5543 case SCTP_PEER_ADDR_PARAMS:
5544 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5545 optlen);
5546 break;
5547 case SCTP_DELAYED_SACK:
5548 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5549 optlen);
5550 break;
5551 case SCTP_INITMSG:
5552 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5553 break;
5554 case SCTP_GET_PEER_ADDRS:
5555 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5556 optlen);
5557 break;
5558 case SCTP_GET_LOCAL_ADDRS:
5559 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5560 optlen);
5561 break;
5562 case SCTP_SOCKOPT_CONNECTX3:
5563 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5564 break;
5565 case SCTP_DEFAULT_SEND_PARAM:
5566 retval = sctp_getsockopt_default_send_param(sk, len,
5567 optval, optlen);
5568 break;
5569 case SCTP_PRIMARY_ADDR:
5570 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5571 break;
5572 case SCTP_NODELAY:
5573 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5574 break;
5575 case SCTP_RTOINFO:
5576 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5577 break;
5578 case SCTP_ASSOCINFO:
5579 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5580 break;
5581 case SCTP_I_WANT_MAPPED_V4_ADDR:
5582 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5583 break;
5584 case SCTP_MAXSEG:
5585 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5586 break;
5587 case SCTP_GET_PEER_ADDR_INFO:
5588 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5589 optlen);
5590 break;
5591 case SCTP_ADAPTATION_LAYER:
5592 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5593 optlen);
5594 break;
5595 case SCTP_CONTEXT:
5596 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5597 break;
5598 case SCTP_FRAGMENT_INTERLEAVE:
5599 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5600 optlen);
5601 break;
5602 case SCTP_PARTIAL_DELIVERY_POINT:
5603 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5604 optlen);
5605 break;
5606 case SCTP_MAX_BURST:
5607 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5608 break;
5609 case SCTP_AUTH_KEY:
5610 case SCTP_AUTH_CHUNK:
5611 case SCTP_AUTH_DELETE_KEY:
5612 retval = -EOPNOTSUPP;
5613 break;
5614 case SCTP_HMAC_IDENT:
5615 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5616 break;
5617 case SCTP_AUTH_ACTIVE_KEY:
5618 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5619 break;
5620 case SCTP_PEER_AUTH_CHUNKS:
5621 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5622 optlen);
5623 break;
5624 case SCTP_LOCAL_AUTH_CHUNKS:
5625 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5626 optlen);
5627 break;
5628 case SCTP_GET_ASSOC_NUMBER:
5629 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5630 break;
5631 case SCTP_GET_ASSOC_ID_LIST:
5632 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5633 break;
5634 case SCTP_AUTO_ASCONF:
5635 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5636 break;
5637 default:
5638 retval = -ENOPROTOOPT;
5639 break;
5642 sctp_release_sock(sk);
5643 return retval;
5646 static void sctp_hash(struct sock *sk)
5648 /* STUB */
5651 static void sctp_unhash(struct sock *sk)
5653 /* STUB */
5656 /* Check if port is acceptable. Possibly find first available port.
5658 * The port hash table (contained in the 'global' SCTP protocol storage
5659 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5660 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5661 * list (the list number is the port number hashed out, so as you
5662 * would expect from a hash function, all the ports in a given list have
5663 * such a number that hashes out to the same list number; you were
5664 * expecting that, right?); so each list has a set of ports, with a
5665 * link to the socket (struct sock) that uses it, the port number and
5666 * a fastreuse flag (FIXME: NPI ipg).
5668 static struct sctp_bind_bucket *sctp_bucket_create(
5669 struct sctp_bind_hashbucket *head, unsigned short snum);
5671 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5673 struct sctp_bind_hashbucket *head; /* hash list */
5674 struct sctp_bind_bucket *pp; /* hash list port iterator */
5675 struct hlist_node *node;
5676 unsigned short snum;
5677 int ret;
5679 snum = ntohs(addr->v4.sin_port);
5681 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5682 sctp_local_bh_disable();
5684 if (snum == 0) {
5685 /* Search for an available port. */
5686 int low, high, remaining, index;
5687 unsigned int rover;
5689 inet_get_local_port_range(&low, &high);
5690 remaining = (high - low) + 1;
5691 rover = net_random() % remaining + low;
5693 do {
5694 rover++;
5695 if ((rover < low) || (rover > high))
5696 rover = low;
5697 if (inet_is_reserved_local_port(rover))
5698 continue;
5699 index = sctp_phashfn(rover);
5700 head = &sctp_port_hashtable[index];
5701 sctp_spin_lock(&head->lock);
5702 sctp_for_each_hentry(pp, node, &head->chain)
5703 if (pp->port == rover)
5704 goto next;
5705 break;
5706 next:
5707 sctp_spin_unlock(&head->lock);
5708 } while (--remaining > 0);
5710 /* Exhausted local port range during search? */
5711 ret = 1;
5712 if (remaining <= 0)
5713 goto fail;
5715 /* OK, here is the one we will use. HEAD (the port
5716 * hash table list entry) is non-NULL and we hold it's
5717 * mutex.
5719 snum = rover;
5720 } else {
5721 /* We are given an specific port number; we verify
5722 * that it is not being used. If it is used, we will
5723 * exahust the search in the hash list corresponding
5724 * to the port number (snum) - we detect that with the
5725 * port iterator, pp being NULL.
5727 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5728 sctp_spin_lock(&head->lock);
5729 sctp_for_each_hentry(pp, node, &head->chain) {
5730 if (pp->port == snum)
5731 goto pp_found;
5734 pp = NULL;
5735 goto pp_not_found;
5736 pp_found:
5737 if (!hlist_empty(&pp->owner)) {
5738 /* We had a port hash table hit - there is an
5739 * available port (pp != NULL) and it is being
5740 * used by other socket (pp->owner not empty); that other
5741 * socket is going to be sk2.
5743 int reuse = sk->sk_reuse;
5744 struct sock *sk2;
5746 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5747 if (pp->fastreuse && sk->sk_reuse &&
5748 sk->sk_state != SCTP_SS_LISTENING)
5749 goto success;
5751 /* Run through the list of sockets bound to the port
5752 * (pp->port) [via the pointers bind_next and
5753 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5754 * we get the endpoint they describe and run through
5755 * the endpoint's list of IP (v4 or v6) addresses,
5756 * comparing each of the addresses with the address of
5757 * the socket sk. If we find a match, then that means
5758 * that this port/socket (sk) combination are already
5759 * in an endpoint.
5761 sk_for_each_bound(sk2, node, &pp->owner) {
5762 struct sctp_endpoint *ep2;
5763 ep2 = sctp_sk(sk2)->ep;
5765 if (sk == sk2 ||
5766 (reuse && sk2->sk_reuse &&
5767 sk2->sk_state != SCTP_SS_LISTENING))
5768 continue;
5770 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5771 sctp_sk(sk2), sctp_sk(sk))) {
5772 ret = (long)sk2;
5773 goto fail_unlock;
5776 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5778 pp_not_found:
5779 /* If there was a hash table miss, create a new port. */
5780 ret = 1;
5781 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5782 goto fail_unlock;
5784 /* In either case (hit or miss), make sure fastreuse is 1 only
5785 * if sk->sk_reuse is too (that is, if the caller requested
5786 * SO_REUSEADDR on this socket -sk-).
5788 if (hlist_empty(&pp->owner)) {
5789 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5790 pp->fastreuse = 1;
5791 else
5792 pp->fastreuse = 0;
5793 } else if (pp->fastreuse &&
5794 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5795 pp->fastreuse = 0;
5797 /* We are set, so fill up all the data in the hash table
5798 * entry, tie the socket list information with the rest of the
5799 * sockets FIXME: Blurry, NPI (ipg).
5801 success:
5802 if (!sctp_sk(sk)->bind_hash) {
5803 inet_sk(sk)->inet_num = snum;
5804 sk_add_bind_node(sk, &pp->owner);
5805 sctp_sk(sk)->bind_hash = pp;
5807 ret = 0;
5809 fail_unlock:
5810 sctp_spin_unlock(&head->lock);
5812 fail:
5813 sctp_local_bh_enable();
5814 return ret;
5817 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5818 * port is requested.
5820 static int sctp_get_port(struct sock *sk, unsigned short snum)
5822 long ret;
5823 union sctp_addr addr;
5824 struct sctp_af *af = sctp_sk(sk)->pf->af;
5826 /* Set up a dummy address struct from the sk. */
5827 af->from_sk(&addr, sk);
5828 addr.v4.sin_port = htons(snum);
5830 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5831 ret = sctp_get_port_local(sk, &addr);
5833 return ret ? 1 : 0;
5837 * Move a socket to LISTENING state.
5839 SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5841 struct sctp_sock *sp = sctp_sk(sk);
5842 struct sctp_endpoint *ep = sp->ep;
5843 struct crypto_hash *tfm = NULL;
5845 /* Allocate HMAC for generating cookie. */
5846 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5847 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5848 if (IS_ERR(tfm)) {
5849 if (net_ratelimit()) {
5850 pr_info("failed to load transform for %s: %ld\n",
5851 sctp_hmac_alg, PTR_ERR(tfm));
5853 return -ENOSYS;
5855 sctp_sk(sk)->hmac = tfm;
5859 * If a bind() or sctp_bindx() is not called prior to a listen()
5860 * call that allows new associations to be accepted, the system
5861 * picks an ephemeral port and will choose an address set equivalent
5862 * to binding with a wildcard address.
5864 * This is not currently spelled out in the SCTP sockets
5865 * extensions draft, but follows the practice as seen in TCP
5866 * sockets.
5869 sk->sk_state = SCTP_SS_LISTENING;
5870 if (!ep->base.bind_addr.port) {
5871 if (sctp_autobind(sk))
5872 return -EAGAIN;
5873 } else {
5874 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5875 sk->sk_state = SCTP_SS_CLOSED;
5876 return -EADDRINUSE;
5880 sk->sk_max_ack_backlog = backlog;
5881 sctp_hash_endpoint(ep);
5882 return 0;
5886 * 4.1.3 / 5.1.3 listen()
5888 * By default, new associations are not accepted for UDP style sockets.
5889 * An application uses listen() to mark a socket as being able to
5890 * accept new associations.
5892 * On TCP style sockets, applications use listen() to ready the SCTP
5893 * endpoint for accepting inbound associations.
5895 * On both types of endpoints a backlog of '0' disables listening.
5897 * Move a socket to LISTENING state.
5899 int sctp_inet_listen(struct socket *sock, int backlog)
5901 struct sock *sk = sock->sk;
5902 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5903 int err = -EINVAL;
5905 if (unlikely(backlog < 0))
5906 return err;
5908 sctp_lock_sock(sk);
5910 /* Peeled-off sockets are not allowed to listen(). */
5911 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5912 goto out;
5914 if (sock->state != SS_UNCONNECTED)
5915 goto out;
5917 /* If backlog is zero, disable listening. */
5918 if (!backlog) {
5919 if (sctp_sstate(sk, CLOSED))
5920 goto out;
5922 err = 0;
5923 sctp_unhash_endpoint(ep);
5924 sk->sk_state = SCTP_SS_CLOSED;
5925 if (sk->sk_reuse)
5926 sctp_sk(sk)->bind_hash->fastreuse = 1;
5927 goto out;
5930 /* If we are already listening, just update the backlog */
5931 if (sctp_sstate(sk, LISTENING))
5932 sk->sk_max_ack_backlog = backlog;
5933 else {
5934 err = sctp_listen_start(sk, backlog);
5935 if (err)
5936 goto out;
5939 err = 0;
5940 out:
5941 sctp_release_sock(sk);
5942 return err;
5946 * This function is done by modeling the current datagram_poll() and the
5947 * tcp_poll(). Note that, based on these implementations, we don't
5948 * lock the socket in this function, even though it seems that,
5949 * ideally, locking or some other mechanisms can be used to ensure
5950 * the integrity of the counters (sndbuf and wmem_alloc) used
5951 * in this place. We assume that we don't need locks either until proven
5952 * otherwise.
5954 * Another thing to note is that we include the Async I/O support
5955 * here, again, by modeling the current TCP/UDP code. We don't have
5956 * a good way to test with it yet.
5958 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5960 struct sock *sk = sock->sk;
5961 struct sctp_sock *sp = sctp_sk(sk);
5962 unsigned int mask;
5964 poll_wait(file, sk_sleep(sk), wait);
5966 /* A TCP-style listening socket becomes readable when the accept queue
5967 * is not empty.
5969 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5970 return (!list_empty(&sp->ep->asocs)) ?
5971 (POLLIN | POLLRDNORM) : 0;
5973 mask = 0;
5975 /* Is there any exceptional events? */
5976 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5977 mask |= POLLERR;
5978 if (sk->sk_shutdown & RCV_SHUTDOWN)
5979 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5980 if (sk->sk_shutdown == SHUTDOWN_MASK)
5981 mask |= POLLHUP;
5983 /* Is it readable? Reconsider this code with TCP-style support. */
5984 if (!skb_queue_empty(&sk->sk_receive_queue))
5985 mask |= POLLIN | POLLRDNORM;
5987 /* The association is either gone or not ready. */
5988 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5989 return mask;
5991 /* Is it writable? */
5992 if (sctp_writeable(sk)) {
5993 mask |= POLLOUT | POLLWRNORM;
5994 } else {
5995 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5997 * Since the socket is not locked, the buffer
5998 * might be made available after the writeable check and
5999 * before the bit is set. This could cause a lost I/O
6000 * signal. tcp_poll() has a race breaker for this race
6001 * condition. Based on their implementation, we put
6002 * in the following code to cover it as well.
6004 if (sctp_writeable(sk))
6005 mask |= POLLOUT | POLLWRNORM;
6007 return mask;
6010 /********************************************************************
6011 * 2nd Level Abstractions
6012 ********************************************************************/
6014 static struct sctp_bind_bucket *sctp_bucket_create(
6015 struct sctp_bind_hashbucket *head, unsigned short snum)
6017 struct sctp_bind_bucket *pp;
6019 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6020 if (pp) {
6021 SCTP_DBG_OBJCNT_INC(bind_bucket);
6022 pp->port = snum;
6023 pp->fastreuse = 0;
6024 INIT_HLIST_HEAD(&pp->owner);
6025 hlist_add_head(&pp->node, &head->chain);
6027 return pp;
6030 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6031 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6033 if (pp && hlist_empty(&pp->owner)) {
6034 __hlist_del(&pp->node);
6035 kmem_cache_free(sctp_bucket_cachep, pp);
6036 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6040 /* Release this socket's reference to a local port. */
6041 static inline void __sctp_put_port(struct sock *sk)
6043 struct sctp_bind_hashbucket *head =
6044 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6045 struct sctp_bind_bucket *pp;
6047 sctp_spin_lock(&head->lock);
6048 pp = sctp_sk(sk)->bind_hash;
6049 __sk_del_bind_node(sk);
6050 sctp_sk(sk)->bind_hash = NULL;
6051 inet_sk(sk)->inet_num = 0;
6052 sctp_bucket_destroy(pp);
6053 sctp_spin_unlock(&head->lock);
6056 void sctp_put_port(struct sock *sk)
6058 sctp_local_bh_disable();
6059 __sctp_put_port(sk);
6060 sctp_local_bh_enable();
6064 * The system picks an ephemeral port and choose an address set equivalent
6065 * to binding with a wildcard address.
6066 * One of those addresses will be the primary address for the association.
6067 * This automatically enables the multihoming capability of SCTP.
6069 static int sctp_autobind(struct sock *sk)
6071 union sctp_addr autoaddr;
6072 struct sctp_af *af;
6073 __be16 port;
6075 /* Initialize a local sockaddr structure to INADDR_ANY. */
6076 af = sctp_sk(sk)->pf->af;
6078 port = htons(inet_sk(sk)->inet_num);
6079 af->inaddr_any(&autoaddr, port);
6081 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6084 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6086 * From RFC 2292
6087 * 4.2 The cmsghdr Structure *
6089 * When ancillary data is sent or received, any number of ancillary data
6090 * objects can be specified by the msg_control and msg_controllen members of
6091 * the msghdr structure, because each object is preceded by
6092 * a cmsghdr structure defining the object's length (the cmsg_len member).
6093 * Historically Berkeley-derived implementations have passed only one object
6094 * at a time, but this API allows multiple objects to be
6095 * passed in a single call to sendmsg() or recvmsg(). The following example
6096 * shows two ancillary data objects in a control buffer.
6098 * |<--------------------------- msg_controllen -------------------------->|
6099 * | |
6101 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6103 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6104 * | | |
6106 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6108 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6109 * | | | | |
6111 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6112 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6114 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6116 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6120 * msg_control
6121 * points here
6123 SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6124 sctp_cmsgs_t *cmsgs)
6126 struct cmsghdr *cmsg;
6127 struct msghdr *my_msg = (struct msghdr *)msg;
6129 for (cmsg = CMSG_FIRSTHDR(msg);
6130 cmsg != NULL;
6131 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6132 if (!CMSG_OK(my_msg, cmsg))
6133 return -EINVAL;
6135 /* Should we parse this header or ignore? */
6136 if (cmsg->cmsg_level != IPPROTO_SCTP)
6137 continue;
6139 /* Strictly check lengths following example in SCM code. */
6140 switch (cmsg->cmsg_type) {
6141 case SCTP_INIT:
6142 /* SCTP Socket API Extension
6143 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6145 * This cmsghdr structure provides information for
6146 * initializing new SCTP associations with sendmsg().
6147 * The SCTP_INITMSG socket option uses this same data
6148 * structure. This structure is not used for
6149 * recvmsg().
6151 * cmsg_level cmsg_type cmsg_data[]
6152 * ------------ ------------ ----------------------
6153 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6155 if (cmsg->cmsg_len !=
6156 CMSG_LEN(sizeof(struct sctp_initmsg)))
6157 return -EINVAL;
6158 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6159 break;
6161 case SCTP_SNDRCV:
6162 /* SCTP Socket API Extension
6163 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6165 * This cmsghdr structure specifies SCTP options for
6166 * sendmsg() and describes SCTP header information
6167 * about a received message through recvmsg().
6169 * cmsg_level cmsg_type cmsg_data[]
6170 * ------------ ------------ ----------------------
6171 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6173 if (cmsg->cmsg_len !=
6174 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6175 return -EINVAL;
6177 cmsgs->info =
6178 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6180 /* Minimally, validate the sinfo_flags. */
6181 if (cmsgs->info->sinfo_flags &
6182 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6183 SCTP_ABORT | SCTP_EOF))
6184 return -EINVAL;
6185 break;
6187 default:
6188 return -EINVAL;
6191 return 0;
6195 * Wait for a packet..
6196 * Note: This function is the same function as in core/datagram.c
6197 * with a few modifications to make lksctp work.
6199 static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6201 int error;
6202 DEFINE_WAIT(wait);
6204 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6206 /* Socket errors? */
6207 error = sock_error(sk);
6208 if (error)
6209 goto out;
6211 if (!skb_queue_empty(&sk->sk_receive_queue))
6212 goto ready;
6214 /* Socket shut down? */
6215 if (sk->sk_shutdown & RCV_SHUTDOWN)
6216 goto out;
6218 /* Sequenced packets can come disconnected. If so we report the
6219 * problem.
6221 error = -ENOTCONN;
6223 /* Is there a good reason to think that we may receive some data? */
6224 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6225 goto out;
6227 /* Handle signals. */
6228 if (signal_pending(current))
6229 goto interrupted;
6231 /* Let another process have a go. Since we are going to sleep
6232 * anyway. Note: This may cause odd behaviors if the message
6233 * does not fit in the user's buffer, but this seems to be the
6234 * only way to honor MSG_DONTWAIT realistically.
6236 sctp_release_sock(sk);
6237 *timeo_p = schedule_timeout(*timeo_p);
6238 sctp_lock_sock(sk);
6240 ready:
6241 finish_wait(sk_sleep(sk), &wait);
6242 return 0;
6244 interrupted:
6245 error = sock_intr_errno(*timeo_p);
6247 out:
6248 finish_wait(sk_sleep(sk), &wait);
6249 *err = error;
6250 return error;
6253 /* Receive a datagram.
6254 * Note: This is pretty much the same routine as in core/datagram.c
6255 * with a few changes to make lksctp work.
6257 static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6258 int noblock, int *err)
6260 int error;
6261 struct sk_buff *skb;
6262 long timeo;
6264 timeo = sock_rcvtimeo(sk, noblock);
6266 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6267 timeo, MAX_SCHEDULE_TIMEOUT);
6269 do {
6270 /* Again only user level code calls this function,
6271 * so nothing interrupt level
6272 * will suddenly eat the receive_queue.
6274 * Look at current nfs client by the way...
6275 * However, this function was correct in any case. 8)
6277 if (flags & MSG_PEEK) {
6278 spin_lock_bh(&sk->sk_receive_queue.lock);
6279 skb = skb_peek(&sk->sk_receive_queue);
6280 if (skb)
6281 atomic_inc(&skb->users);
6282 spin_unlock_bh(&sk->sk_receive_queue.lock);
6283 } else {
6284 skb = skb_dequeue(&sk->sk_receive_queue);
6287 if (skb)
6288 return skb;
6290 /* Caller is allowed not to check sk->sk_err before calling. */
6291 error = sock_error(sk);
6292 if (error)
6293 goto no_packet;
6295 if (sk->sk_shutdown & RCV_SHUTDOWN)
6296 break;
6298 /* User doesn't want to wait. */
6299 error = -EAGAIN;
6300 if (!timeo)
6301 goto no_packet;
6302 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6304 return NULL;
6306 no_packet:
6307 *err = error;
6308 return NULL;
6311 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6312 static void __sctp_write_space(struct sctp_association *asoc)
6314 struct sock *sk = asoc->base.sk;
6315 struct socket *sock = sk->sk_socket;
6317 if ((sctp_wspace(asoc) > 0) && sock) {
6318 if (waitqueue_active(&asoc->wait))
6319 wake_up_interruptible(&asoc->wait);
6321 if (sctp_writeable(sk)) {
6322 wait_queue_head_t *wq = sk_sleep(sk);
6324 if (wq && waitqueue_active(wq))
6325 wake_up_interruptible(wq);
6327 /* Note that we try to include the Async I/O support
6328 * here by modeling from the current TCP/UDP code.
6329 * We have not tested with it yet.
6331 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6332 sock_wake_async(sock,
6333 SOCK_WAKE_SPACE, POLL_OUT);
6338 /* Do accounting for the sndbuf space.
6339 * Decrement the used sndbuf space of the corresponding association by the
6340 * data size which was just transmitted(freed).
6342 static void sctp_wfree(struct sk_buff *skb)
6344 struct sctp_association *asoc;
6345 struct sctp_chunk *chunk;
6346 struct sock *sk;
6348 /* Get the saved chunk pointer. */
6349 chunk = *((struct sctp_chunk **)(skb->cb));
6350 asoc = chunk->asoc;
6351 sk = asoc->base.sk;
6352 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6353 sizeof(struct sk_buff) +
6354 sizeof(struct sctp_chunk);
6356 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6359 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6361 sk->sk_wmem_queued -= skb->truesize;
6362 sk_mem_uncharge(sk, skb->truesize);
6364 sock_wfree(skb);
6365 __sctp_write_space(asoc);
6367 sctp_association_put(asoc);
6370 /* Do accounting for the receive space on the socket.
6371 * Accounting for the association is done in ulpevent.c
6372 * We set this as a destructor for the cloned data skbs so that
6373 * accounting is done at the correct time.
6375 void sctp_sock_rfree(struct sk_buff *skb)
6377 struct sock *sk = skb->sk;
6378 struct sctp_ulpevent *event = sctp_skb2event(skb);
6380 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6383 * Mimic the behavior of sock_rfree
6385 sk_mem_uncharge(sk, event->rmem_len);
6389 /* Helper function to wait for space in the sndbuf. */
6390 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6391 size_t msg_len)
6393 struct sock *sk = asoc->base.sk;
6394 int err = 0;
6395 long current_timeo = *timeo_p;
6396 DEFINE_WAIT(wait);
6398 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6399 asoc, (long)(*timeo_p), msg_len);
6401 /* Increment the association's refcnt. */
6402 sctp_association_hold(asoc);
6404 /* Wait on the association specific sndbuf space. */
6405 for (;;) {
6406 prepare_to_wait_exclusive(&asoc->wait, &wait,
6407 TASK_INTERRUPTIBLE);
6408 if (!*timeo_p)
6409 goto do_nonblock;
6410 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6411 asoc->base.dead)
6412 goto do_error;
6413 if (signal_pending(current))
6414 goto do_interrupted;
6415 if (msg_len <= sctp_wspace(asoc))
6416 break;
6418 /* Let another process have a go. Since we are going
6419 * to sleep anyway.
6421 sctp_release_sock(sk);
6422 current_timeo = schedule_timeout(current_timeo);
6423 BUG_ON(sk != asoc->base.sk);
6424 sctp_lock_sock(sk);
6426 *timeo_p = current_timeo;
6429 out:
6430 finish_wait(&asoc->wait, &wait);
6432 /* Release the association's refcnt. */
6433 sctp_association_put(asoc);
6435 return err;
6437 do_error:
6438 err = -EPIPE;
6439 goto out;
6441 do_interrupted:
6442 err = sock_intr_errno(*timeo_p);
6443 goto out;
6445 do_nonblock:
6446 err = -EAGAIN;
6447 goto out;
6450 void sctp_data_ready(struct sock *sk, int len)
6452 struct socket_wq *wq;
6454 rcu_read_lock();
6455 wq = rcu_dereference(sk->sk_wq);
6456 if (wq_has_sleeper(wq))
6457 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6458 POLLRDNORM | POLLRDBAND);
6459 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6460 rcu_read_unlock();
6463 /* If socket sndbuf has changed, wake up all per association waiters. */
6464 void sctp_write_space(struct sock *sk)
6466 struct sctp_association *asoc;
6468 /* Wake up the tasks in each wait queue. */
6469 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6470 __sctp_write_space(asoc);
6474 /* Is there any sndbuf space available on the socket?
6476 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6477 * associations on the same socket. For a UDP-style socket with
6478 * multiple associations, it is possible for it to be "unwriteable"
6479 * prematurely. I assume that this is acceptable because
6480 * a premature "unwriteable" is better than an accidental "writeable" which
6481 * would cause an unwanted block under certain circumstances. For the 1-1
6482 * UDP-style sockets or TCP-style sockets, this code should work.
6483 * - Daisy
6485 static int sctp_writeable(struct sock *sk)
6487 int amt = 0;
6489 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6490 if (amt < 0)
6491 amt = 0;
6492 return amt;
6495 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6496 * returns immediately with EINPROGRESS.
6498 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6500 struct sock *sk = asoc->base.sk;
6501 int err = 0;
6502 long current_timeo = *timeo_p;
6503 DEFINE_WAIT(wait);
6505 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6506 (long)(*timeo_p));
6508 /* Increment the association's refcnt. */
6509 sctp_association_hold(asoc);
6511 for (;;) {
6512 prepare_to_wait_exclusive(&asoc->wait, &wait,
6513 TASK_INTERRUPTIBLE);
6514 if (!*timeo_p)
6515 goto do_nonblock;
6516 if (sk->sk_shutdown & RCV_SHUTDOWN)
6517 break;
6518 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6519 asoc->base.dead)
6520 goto do_error;
6521 if (signal_pending(current))
6522 goto do_interrupted;
6524 if (sctp_state(asoc, ESTABLISHED))
6525 break;
6527 /* Let another process have a go. Since we are going
6528 * to sleep anyway.
6530 sctp_release_sock(sk);
6531 current_timeo = schedule_timeout(current_timeo);
6532 sctp_lock_sock(sk);
6534 *timeo_p = current_timeo;
6537 out:
6538 finish_wait(&asoc->wait, &wait);
6540 /* Release the association's refcnt. */
6541 sctp_association_put(asoc);
6543 return err;
6545 do_error:
6546 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6547 err = -ETIMEDOUT;
6548 else
6549 err = -ECONNREFUSED;
6550 goto out;
6552 do_interrupted:
6553 err = sock_intr_errno(*timeo_p);
6554 goto out;
6556 do_nonblock:
6557 err = -EINPROGRESS;
6558 goto out;
6561 static int sctp_wait_for_accept(struct sock *sk, long timeo)
6563 struct sctp_endpoint *ep;
6564 int err = 0;
6565 DEFINE_WAIT(wait);
6567 ep = sctp_sk(sk)->ep;
6570 for (;;) {
6571 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6572 TASK_INTERRUPTIBLE);
6574 if (list_empty(&ep->asocs)) {
6575 sctp_release_sock(sk);
6576 timeo = schedule_timeout(timeo);
6577 sctp_lock_sock(sk);
6580 err = -EINVAL;
6581 if (!sctp_sstate(sk, LISTENING))
6582 break;
6584 err = 0;
6585 if (!list_empty(&ep->asocs))
6586 break;
6588 err = sock_intr_errno(timeo);
6589 if (signal_pending(current))
6590 break;
6592 err = -EAGAIN;
6593 if (!timeo)
6594 break;
6597 finish_wait(sk_sleep(sk), &wait);
6599 return err;
6602 static void sctp_wait_for_close(struct sock *sk, long timeout)
6604 DEFINE_WAIT(wait);
6606 do {
6607 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6608 if (list_empty(&sctp_sk(sk)->ep->asocs))
6609 break;
6610 sctp_release_sock(sk);
6611 timeout = schedule_timeout(timeout);
6612 sctp_lock_sock(sk);
6613 } while (!signal_pending(current) && timeout);
6615 finish_wait(sk_sleep(sk), &wait);
6618 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6620 struct sk_buff *frag;
6622 if (!skb->data_len)
6623 goto done;
6625 /* Don't forget the fragments. */
6626 skb_walk_frags(skb, frag)
6627 sctp_skb_set_owner_r_frag(frag, sk);
6629 done:
6630 sctp_skb_set_owner_r(skb, sk);
6633 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6634 struct sctp_association *asoc)
6636 struct inet_sock *inet = inet_sk(sk);
6637 struct inet_sock *newinet;
6639 newsk->sk_type = sk->sk_type;
6640 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6641 newsk->sk_flags = sk->sk_flags;
6642 newsk->sk_no_check = sk->sk_no_check;
6643 newsk->sk_reuse = sk->sk_reuse;
6645 newsk->sk_shutdown = sk->sk_shutdown;
6646 newsk->sk_destruct = inet_sock_destruct;
6647 newsk->sk_family = sk->sk_family;
6648 newsk->sk_protocol = IPPROTO_SCTP;
6649 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6650 newsk->sk_sndbuf = sk->sk_sndbuf;
6651 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6652 newsk->sk_lingertime = sk->sk_lingertime;
6653 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6654 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6656 newinet = inet_sk(newsk);
6658 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6659 * getsockname() and getpeername()
6661 newinet->inet_sport = inet->inet_sport;
6662 newinet->inet_saddr = inet->inet_saddr;
6663 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6664 newinet->inet_dport = htons(asoc->peer.port);
6665 newinet->pmtudisc = inet->pmtudisc;
6666 newinet->inet_id = asoc->next_tsn ^ jiffies;
6668 newinet->uc_ttl = inet->uc_ttl;
6669 newinet->mc_loop = 1;
6670 newinet->mc_ttl = 1;
6671 newinet->mc_index = 0;
6672 newinet->mc_list = NULL;
6675 /* Populate the fields of the newsk from the oldsk and migrate the assoc
6676 * and its messages to the newsk.
6678 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6679 struct sctp_association *assoc,
6680 sctp_socket_type_t type)
6682 struct sctp_sock *oldsp = sctp_sk(oldsk);
6683 struct sctp_sock *newsp = sctp_sk(newsk);
6684 struct sctp_bind_bucket *pp; /* hash list port iterator */
6685 struct sctp_endpoint *newep = newsp->ep;
6686 struct sk_buff *skb, *tmp;
6687 struct sctp_ulpevent *event;
6688 struct sctp_bind_hashbucket *head;
6689 struct list_head tmplist;
6691 /* Migrate socket buffer sizes and all the socket level options to the
6692 * new socket.
6694 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6695 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6696 /* Brute force copy old sctp opt. */
6697 if (oldsp->do_auto_asconf) {
6698 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6699 inet_sk_copy_descendant(newsk, oldsk);
6700 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6701 } else
6702 inet_sk_copy_descendant(newsk, oldsk);
6704 /* Restore the ep value that was overwritten with the above structure
6705 * copy.
6707 newsp->ep = newep;
6708 newsp->hmac = NULL;
6710 /* Hook this new socket in to the bind_hash list. */
6711 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6712 sctp_local_bh_disable();
6713 sctp_spin_lock(&head->lock);
6714 pp = sctp_sk(oldsk)->bind_hash;
6715 sk_add_bind_node(newsk, &pp->owner);
6716 sctp_sk(newsk)->bind_hash = pp;
6717 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6718 sctp_spin_unlock(&head->lock);
6719 sctp_local_bh_enable();
6721 /* Copy the bind_addr list from the original endpoint to the new
6722 * endpoint so that we can handle restarts properly
6724 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6725 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6727 /* Move any messages in the old socket's receive queue that are for the
6728 * peeled off association to the new socket's receive queue.
6730 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6731 event = sctp_skb2event(skb);
6732 if (event->asoc == assoc) {
6733 __skb_unlink(skb, &oldsk->sk_receive_queue);
6734 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6735 sctp_skb_set_owner_r_frag(skb, newsk);
6739 /* Clean up any messages pending delivery due to partial
6740 * delivery. Three cases:
6741 * 1) No partial deliver; no work.
6742 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6743 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6745 skb_queue_head_init(&newsp->pd_lobby);
6746 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6748 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6749 struct sk_buff_head *queue;
6751 /* Decide which queue to move pd_lobby skbs to. */
6752 if (assoc->ulpq.pd_mode) {
6753 queue = &newsp->pd_lobby;
6754 } else
6755 queue = &newsk->sk_receive_queue;
6757 /* Walk through the pd_lobby, looking for skbs that
6758 * need moved to the new socket.
6760 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6761 event = sctp_skb2event(skb);
6762 if (event->asoc == assoc) {
6763 __skb_unlink(skb, &oldsp->pd_lobby);
6764 __skb_queue_tail(queue, skb);
6765 sctp_skb_set_owner_r_frag(skb, newsk);
6769 /* Clear up any skbs waiting for the partial
6770 * delivery to finish.
6772 if (assoc->ulpq.pd_mode)
6773 sctp_clear_pd(oldsk, NULL);
6777 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6778 sctp_skb_set_owner_r_frag(skb, newsk);
6780 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6781 sctp_skb_set_owner_r_frag(skb, newsk);
6783 /* Set the type of socket to indicate that it is peeled off from the
6784 * original UDP-style socket or created with the accept() call on a
6785 * TCP-style socket..
6787 newsp->type = type;
6789 /* Mark the new socket "in-use" by the user so that any packets
6790 * that may arrive on the association after we've moved it are
6791 * queued to the backlog. This prevents a potential race between
6792 * backlog processing on the old socket and new-packet processing
6793 * on the new socket.
6795 * The caller has just allocated newsk so we can guarantee that other
6796 * paths won't try to lock it and then oldsk.
6798 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6799 sctp_assoc_migrate(assoc, newsk);
6801 /* If the association on the newsk is already closed before accept()
6802 * is called, set RCV_SHUTDOWN flag.
6804 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6805 newsk->sk_shutdown |= RCV_SHUTDOWN;
6807 newsk->sk_state = SCTP_SS_ESTABLISHED;
6808 sctp_release_sock(newsk);
6812 /* This proto struct describes the ULP interface for SCTP. */
6813 struct proto sctp_prot = {
6814 .name = "SCTP",
6815 .owner = THIS_MODULE,
6816 .close = sctp_close,
6817 .connect = sctp_connect,
6818 .disconnect = sctp_disconnect,
6819 .accept = sctp_accept,
6820 .ioctl = sctp_ioctl,
6821 .init = sctp_init_sock,
6822 .destroy = sctp_destroy_sock,
6823 .shutdown = sctp_shutdown,
6824 .setsockopt = sctp_setsockopt,
6825 .getsockopt = sctp_getsockopt,
6826 .sendmsg = sctp_sendmsg,
6827 .recvmsg = sctp_recvmsg,
6828 .bind = sctp_bind,
6829 .backlog_rcv = sctp_backlog_rcv,
6830 .hash = sctp_hash,
6831 .unhash = sctp_unhash,
6832 .get_port = sctp_get_port,
6833 .obj_size = sizeof(struct sctp_sock),
6834 .sysctl_mem = sysctl_sctp_mem,
6835 .sysctl_rmem = sysctl_sctp_rmem,
6836 .sysctl_wmem = sysctl_sctp_wmem,
6837 .memory_pressure = &sctp_memory_pressure,
6838 .enter_memory_pressure = sctp_enter_memory_pressure,
6839 .memory_allocated = &sctp_memory_allocated,
6840 .sockets_allocated = &sctp_sockets_allocated,
6843 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6845 struct proto sctpv6_prot = {
6846 .name = "SCTPv6",
6847 .owner = THIS_MODULE,
6848 .close = sctp_close,
6849 .connect = sctp_connect,
6850 .disconnect = sctp_disconnect,
6851 .accept = sctp_accept,
6852 .ioctl = sctp_ioctl,
6853 .init = sctp_init_sock,
6854 .destroy = sctp_destroy_sock,
6855 .shutdown = sctp_shutdown,
6856 .setsockopt = sctp_setsockopt,
6857 .getsockopt = sctp_getsockopt,
6858 .sendmsg = sctp_sendmsg,
6859 .recvmsg = sctp_recvmsg,
6860 .bind = sctp_bind,
6861 .backlog_rcv = sctp_backlog_rcv,
6862 .hash = sctp_hash,
6863 .unhash = sctp_unhash,
6864 .get_port = sctp_get_port,
6865 .obj_size = sizeof(struct sctp6_sock),
6866 .sysctl_mem = sysctl_sctp_mem,
6867 .sysctl_rmem = sysctl_sctp_rmem,
6868 .sysctl_wmem = sysctl_sctp_wmem,
6869 .memory_pressure = &sctp_memory_pressure,
6870 .enter_memory_pressure = sctp_enter_memory_pressure,
6871 .memory_allocated = &sctp_memory_allocated,
6872 .sockets_allocated = &sctp_sockets_allocated,
6874 #endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */