iommu/ipmmu-vmsa: Return proper error if devm_request_irq fails
[linux-2.6/btrfs-unstable.git] / net / sctp / socket.c
blob634a2abb5f3a25f15ca99c4737190c1e5e17b27a
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, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
35 * email address(es):
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
59 #include <linux/ip.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
69 #include <net/ip.h>
70 #include <net/icmp.h>
71 #include <net/route.h>
72 #include <net/ipv6.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
78 #include <net/sock.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
86 size_t msg_len);
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
105 extern struct kmem_cache *sctp_bucket_cachep;
106 extern long sysctl_sctp_mem[3];
107 extern int sysctl_sctp_rmem[3];
108 extern int sysctl_sctp_wmem[3];
110 static int sctp_memory_pressure;
111 static atomic_long_t sctp_memory_allocated;
112 struct percpu_counter sctp_sockets_allocated;
114 static void sctp_enter_memory_pressure(struct sock *sk)
116 sctp_memory_pressure = 1;
120 /* Get the sndbuf space available at the time on the association. */
121 static inline int sctp_wspace(struct sctp_association *asoc)
123 int amt;
125 if (asoc->ep->sndbuf_policy)
126 amt = asoc->sndbuf_used;
127 else
128 amt = sk_wmem_alloc_get(asoc->base.sk);
130 if (amt >= asoc->base.sk->sk_sndbuf) {
131 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
132 amt = 0;
133 else {
134 amt = sk_stream_wspace(asoc->base.sk);
135 if (amt < 0)
136 amt = 0;
138 } else {
139 amt = asoc->base.sk->sk_sndbuf - amt;
141 return amt;
144 /* Increment the used sndbuf space count of the corresponding association by
145 * the size of the outgoing data chunk.
146 * Also, set the skb destructor for sndbuf accounting later.
148 * Since it is always 1-1 between chunk and skb, and also a new skb is always
149 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
150 * destructor in the data chunk skb for the purpose of the sndbuf space
151 * tracking.
153 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
155 struct sctp_association *asoc = chunk->asoc;
156 struct sock *sk = asoc->base.sk;
158 /* The sndbuf space is tracked per association. */
159 sctp_association_hold(asoc);
161 skb_set_owner_w(chunk->skb, sk);
163 chunk->skb->destructor = sctp_wfree;
164 /* Save the chunk pointer in skb for sctp_wfree to use later. */
165 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
167 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
168 sizeof(struct sk_buff) +
169 sizeof(struct sctp_chunk);
171 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
172 sk->sk_wmem_queued += chunk->skb->truesize;
173 sk_mem_charge(sk, chunk->skb->truesize);
176 /* Verify that this is a valid address. */
177 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
178 int len)
180 struct sctp_af *af;
182 /* Verify basic sockaddr. */
183 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
184 if (!af)
185 return -EINVAL;
187 /* Is this a valid SCTP address? */
188 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
189 return -EINVAL;
191 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
192 return -EINVAL;
194 return 0;
197 /* Look up the association by its id. If this is not a UDP-style
198 * socket, the ID field is always ignored.
200 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
202 struct sctp_association *asoc = NULL;
204 /* If this is not a UDP-style socket, assoc id should be ignored. */
205 if (!sctp_style(sk, UDP)) {
206 /* Return NULL if the socket state is not ESTABLISHED. It
207 * could be a TCP-style listening socket or a socket which
208 * hasn't yet called connect() to establish an association.
210 if (!sctp_sstate(sk, ESTABLISHED))
211 return NULL;
213 /* Get the first and the only association from the list. */
214 if (!list_empty(&sctp_sk(sk)->ep->asocs))
215 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
216 struct sctp_association, asocs);
217 return asoc;
220 /* Otherwise this is a UDP-style socket. */
221 if (!id || (id == (sctp_assoc_t)-1))
222 return NULL;
224 spin_lock_bh(&sctp_assocs_id_lock);
225 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
226 spin_unlock_bh(&sctp_assocs_id_lock);
228 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
229 return NULL;
231 return asoc;
234 /* Look up the transport from an address and an assoc id. If both address and
235 * id are specified, the associations matching the address and the id should be
236 * the same.
238 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
239 struct sockaddr_storage *addr,
240 sctp_assoc_t id)
242 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
243 struct sctp_transport *transport;
244 union sctp_addr *laddr = (union sctp_addr *)addr;
246 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
247 laddr,
248 &transport);
250 if (!addr_asoc)
251 return NULL;
253 id_asoc = sctp_id2assoc(sk, id);
254 if (id_asoc && (id_asoc != addr_asoc))
255 return NULL;
257 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
258 (union sctp_addr *)addr);
260 return transport;
263 /* API 3.1.2 bind() - UDP Style Syntax
264 * The syntax of bind() is,
266 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
268 * sd - the socket descriptor returned by socket().
269 * addr - the address structure (struct sockaddr_in or struct
270 * sockaddr_in6 [RFC 2553]),
271 * addr_len - the size of the address structure.
273 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
275 int retval = 0;
277 lock_sock(sk);
279 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
280 addr, addr_len);
282 /* Disallow binding twice. */
283 if (!sctp_sk(sk)->ep->base.bind_addr.port)
284 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
285 addr_len);
286 else
287 retval = -EINVAL;
289 release_sock(sk);
291 return retval;
294 static long sctp_get_port_local(struct sock *, union sctp_addr *);
296 /* Verify this is a valid sockaddr. */
297 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
298 union sctp_addr *addr, int len)
300 struct sctp_af *af;
302 /* Check minimum size. */
303 if (len < sizeof (struct sockaddr))
304 return NULL;
306 /* V4 mapped address are really of AF_INET family */
307 if (addr->sa.sa_family == AF_INET6 &&
308 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
309 if (!opt->pf->af_supported(AF_INET, opt))
310 return NULL;
311 } else {
312 /* Does this PF support this AF? */
313 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
314 return NULL;
317 /* If we get this far, af is valid. */
318 af = sctp_get_af_specific(addr->sa.sa_family);
320 if (len < af->sockaddr_len)
321 return NULL;
323 return af;
326 /* Bind a local address either to an endpoint or to an association. */
327 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
329 struct net *net = sock_net(sk);
330 struct sctp_sock *sp = sctp_sk(sk);
331 struct sctp_endpoint *ep = sp->ep;
332 struct sctp_bind_addr *bp = &ep->base.bind_addr;
333 struct sctp_af *af;
334 unsigned short snum;
335 int ret = 0;
337 /* Common sockaddr verification. */
338 af = sctp_sockaddr_af(sp, addr, len);
339 if (!af) {
340 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
341 __func__, sk, addr, len);
342 return -EINVAL;
345 snum = ntohs(addr->v4.sin_port);
347 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
348 __func__, sk, &addr->sa, bp->port, snum, len);
350 /* PF specific bind() address verification. */
351 if (!sp->pf->bind_verify(sp, addr))
352 return -EADDRNOTAVAIL;
354 /* We must either be unbound, or bind to the same port.
355 * It's OK to allow 0 ports if we are already bound.
356 * We'll just inhert an already bound port in this case
358 if (bp->port) {
359 if (!snum)
360 snum = bp->port;
361 else if (snum != bp->port) {
362 pr_debug("%s: new port %d doesn't match existing port "
363 "%d\n", __func__, snum, bp->port);
364 return -EINVAL;
368 if (snum && snum < PROT_SOCK &&
369 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
370 return -EACCES;
372 /* See if the address matches any of the addresses we may have
373 * already bound before checking against other endpoints.
375 if (sctp_bind_addr_match(bp, addr, sp))
376 return -EINVAL;
378 /* Make sure we are allowed to bind here.
379 * The function sctp_get_port_local() does duplicate address
380 * detection.
382 addr->v4.sin_port = htons(snum);
383 if ((ret = sctp_get_port_local(sk, addr))) {
384 return -EADDRINUSE;
387 /* Refresh ephemeral port. */
388 if (!bp->port)
389 bp->port = inet_sk(sk)->inet_num;
391 /* Add the address to the bind address list.
392 * Use GFP_ATOMIC since BHs will be disabled.
394 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
396 /* Copy back into socket for getsockname() use. */
397 if (!ret) {
398 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
399 sp->pf->to_sk_saddr(addr, sk);
402 return ret;
405 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
407 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
408 * at any one time. If a sender, after sending an ASCONF chunk, decides
409 * it needs to transfer another ASCONF Chunk, it MUST wait until the
410 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
411 * subsequent ASCONF. Note this restriction binds each side, so at any
412 * time two ASCONF may be in-transit on any given association (one sent
413 * from each endpoint).
415 static int sctp_send_asconf(struct sctp_association *asoc,
416 struct sctp_chunk *chunk)
418 struct net *net = sock_net(asoc->base.sk);
419 int retval = 0;
421 /* If there is an outstanding ASCONF chunk, queue it for later
422 * transmission.
424 if (asoc->addip_last_asconf) {
425 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
426 goto out;
429 /* Hold the chunk until an ASCONF_ACK is received. */
430 sctp_chunk_hold(chunk);
431 retval = sctp_primitive_ASCONF(net, asoc, chunk);
432 if (retval)
433 sctp_chunk_free(chunk);
434 else
435 asoc->addip_last_asconf = chunk;
437 out:
438 return retval;
441 /* Add a list of addresses as bind addresses to local endpoint or
442 * association.
444 * Basically run through each address specified in the addrs/addrcnt
445 * array/length pair, determine if it is IPv6 or IPv4 and call
446 * sctp_do_bind() on it.
448 * If any of them fails, then the operation will be reversed and the
449 * ones that were added will be removed.
451 * Only sctp_setsockopt_bindx() is supposed to call this function.
453 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
455 int cnt;
456 int retval = 0;
457 void *addr_buf;
458 struct sockaddr *sa_addr;
459 struct sctp_af *af;
461 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
462 addrs, addrcnt);
464 addr_buf = addrs;
465 for (cnt = 0; cnt < addrcnt; cnt++) {
466 /* The list may contain either IPv4 or IPv6 address;
467 * determine the address length for walking thru the list.
469 sa_addr = addr_buf;
470 af = sctp_get_af_specific(sa_addr->sa_family);
471 if (!af) {
472 retval = -EINVAL;
473 goto err_bindx_add;
476 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
477 af->sockaddr_len);
479 addr_buf += af->sockaddr_len;
481 err_bindx_add:
482 if (retval < 0) {
483 /* Failed. Cleanup the ones that have been added */
484 if (cnt > 0)
485 sctp_bindx_rem(sk, addrs, cnt);
486 return retval;
490 return retval;
493 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
494 * associations that are part of the endpoint indicating that a list of local
495 * addresses are added to the endpoint.
497 * If any of the addresses is already in the bind address list of the
498 * association, we do not send the chunk for that association. But it will not
499 * affect other associations.
501 * Only sctp_setsockopt_bindx() is supposed to call this function.
503 static int sctp_send_asconf_add_ip(struct sock *sk,
504 struct sockaddr *addrs,
505 int addrcnt)
507 struct net *net = sock_net(sk);
508 struct sctp_sock *sp;
509 struct sctp_endpoint *ep;
510 struct sctp_association *asoc;
511 struct sctp_bind_addr *bp;
512 struct sctp_chunk *chunk;
513 struct sctp_sockaddr_entry *laddr;
514 union sctp_addr *addr;
515 union sctp_addr saveaddr;
516 void *addr_buf;
517 struct sctp_af *af;
518 struct list_head *p;
519 int i;
520 int retval = 0;
522 if (!net->sctp.addip_enable)
523 return retval;
525 sp = sctp_sk(sk);
526 ep = sp->ep;
528 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
529 __func__, sk, addrs, addrcnt);
531 list_for_each_entry(asoc, &ep->asocs, asocs) {
532 if (!asoc->peer.asconf_capable)
533 continue;
535 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
536 continue;
538 if (!sctp_state(asoc, ESTABLISHED))
539 continue;
541 /* Check if any address in the packed array of addresses is
542 * in the bind address list of the association. If so,
543 * do not send the asconf chunk to its peer, but continue with
544 * other associations.
546 addr_buf = addrs;
547 for (i = 0; i < addrcnt; i++) {
548 addr = addr_buf;
549 af = sctp_get_af_specific(addr->v4.sin_family);
550 if (!af) {
551 retval = -EINVAL;
552 goto out;
555 if (sctp_assoc_lookup_laddr(asoc, addr))
556 break;
558 addr_buf += af->sockaddr_len;
560 if (i < addrcnt)
561 continue;
563 /* Use the first valid address in bind addr list of
564 * association as Address Parameter of ASCONF CHUNK.
566 bp = &asoc->base.bind_addr;
567 p = bp->address_list.next;
568 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
569 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
570 addrcnt, SCTP_PARAM_ADD_IP);
571 if (!chunk) {
572 retval = -ENOMEM;
573 goto out;
576 /* Add the new addresses to the bind address list with
577 * use_as_src set to 0.
579 addr_buf = addrs;
580 for (i = 0; i < addrcnt; i++) {
581 addr = addr_buf;
582 af = sctp_get_af_specific(addr->v4.sin_family);
583 memcpy(&saveaddr, addr, af->sockaddr_len);
584 retval = sctp_add_bind_addr(bp, &saveaddr,
585 SCTP_ADDR_NEW, GFP_ATOMIC);
586 addr_buf += af->sockaddr_len;
588 if (asoc->src_out_of_asoc_ok) {
589 struct sctp_transport *trans;
591 list_for_each_entry(trans,
592 &asoc->peer.transport_addr_list, transports) {
593 /* Clear the source and route cache */
594 dst_release(trans->dst);
595 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
596 2*asoc->pathmtu, 4380));
597 trans->ssthresh = asoc->peer.i.a_rwnd;
598 trans->rto = asoc->rto_initial;
599 sctp_max_rto(asoc, trans);
600 trans->rtt = trans->srtt = trans->rttvar = 0;
601 sctp_transport_route(trans, NULL,
602 sctp_sk(asoc->base.sk));
605 retval = sctp_send_asconf(asoc, chunk);
608 out:
609 return retval;
612 /* Remove a list of addresses from bind addresses list. Do not remove the
613 * last address.
615 * Basically run through each address specified in the addrs/addrcnt
616 * array/length pair, determine if it is IPv6 or IPv4 and call
617 * sctp_del_bind() on it.
619 * If any of them fails, then the operation will be reversed and the
620 * ones that were removed will be added back.
622 * At least one address has to be left; if only one address is
623 * available, the operation will return -EBUSY.
625 * Only sctp_setsockopt_bindx() is supposed to call this function.
627 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
629 struct sctp_sock *sp = sctp_sk(sk);
630 struct sctp_endpoint *ep = sp->ep;
631 int cnt;
632 struct sctp_bind_addr *bp = &ep->base.bind_addr;
633 int retval = 0;
634 void *addr_buf;
635 union sctp_addr *sa_addr;
636 struct sctp_af *af;
638 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
639 __func__, sk, addrs, addrcnt);
641 addr_buf = addrs;
642 for (cnt = 0; cnt < addrcnt; cnt++) {
643 /* If the bind address list is empty or if there is only one
644 * bind address, there is nothing more to be removed (we need
645 * at least one address here).
647 if (list_empty(&bp->address_list) ||
648 (sctp_list_single_entry(&bp->address_list))) {
649 retval = -EBUSY;
650 goto err_bindx_rem;
653 sa_addr = addr_buf;
654 af = sctp_get_af_specific(sa_addr->sa.sa_family);
655 if (!af) {
656 retval = -EINVAL;
657 goto err_bindx_rem;
660 if (!af->addr_valid(sa_addr, sp, NULL)) {
661 retval = -EADDRNOTAVAIL;
662 goto err_bindx_rem;
665 if (sa_addr->v4.sin_port &&
666 sa_addr->v4.sin_port != htons(bp->port)) {
667 retval = -EINVAL;
668 goto err_bindx_rem;
671 if (!sa_addr->v4.sin_port)
672 sa_addr->v4.sin_port = htons(bp->port);
674 /* FIXME - There is probably a need to check if sk->sk_saddr and
675 * sk->sk_rcv_addr are currently set to one of the addresses to
676 * be removed. This is something which needs to be looked into
677 * when we are fixing the outstanding issues with multi-homing
678 * socket routing and failover schemes. Refer to comments in
679 * sctp_do_bind(). -daisy
681 retval = sctp_del_bind_addr(bp, sa_addr);
683 addr_buf += af->sockaddr_len;
684 err_bindx_rem:
685 if (retval < 0) {
686 /* Failed. Add the ones that has been removed back */
687 if (cnt > 0)
688 sctp_bindx_add(sk, addrs, cnt);
689 return retval;
693 return retval;
696 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
697 * the associations that are part of the endpoint indicating that a list of
698 * local addresses are removed from the endpoint.
700 * If any of the addresses is already in the bind address list of the
701 * association, we do not send the chunk for that association. But it will not
702 * affect other associations.
704 * Only sctp_setsockopt_bindx() is supposed to call this function.
706 static int sctp_send_asconf_del_ip(struct sock *sk,
707 struct sockaddr *addrs,
708 int addrcnt)
710 struct net *net = sock_net(sk);
711 struct sctp_sock *sp;
712 struct sctp_endpoint *ep;
713 struct sctp_association *asoc;
714 struct sctp_transport *transport;
715 struct sctp_bind_addr *bp;
716 struct sctp_chunk *chunk;
717 union sctp_addr *laddr;
718 void *addr_buf;
719 struct sctp_af *af;
720 struct sctp_sockaddr_entry *saddr;
721 int i;
722 int retval = 0;
723 int stored = 0;
725 chunk = NULL;
726 if (!net->sctp.addip_enable)
727 return retval;
729 sp = sctp_sk(sk);
730 ep = sp->ep;
732 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
733 __func__, sk, addrs, addrcnt);
735 list_for_each_entry(asoc, &ep->asocs, asocs) {
737 if (!asoc->peer.asconf_capable)
738 continue;
740 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
741 continue;
743 if (!sctp_state(asoc, ESTABLISHED))
744 continue;
746 /* Check if any address in the packed array of addresses is
747 * not present in the bind address list of the association.
748 * If so, do not send the asconf chunk to its peer, but
749 * continue with other associations.
751 addr_buf = addrs;
752 for (i = 0; i < addrcnt; i++) {
753 laddr = addr_buf;
754 af = sctp_get_af_specific(laddr->v4.sin_family);
755 if (!af) {
756 retval = -EINVAL;
757 goto out;
760 if (!sctp_assoc_lookup_laddr(asoc, laddr))
761 break;
763 addr_buf += af->sockaddr_len;
765 if (i < addrcnt)
766 continue;
768 /* Find one address in the association's bind address list
769 * that is not in the packed array of addresses. This is to
770 * make sure that we do not delete all the addresses in the
771 * association.
773 bp = &asoc->base.bind_addr;
774 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
775 addrcnt, sp);
776 if ((laddr == NULL) && (addrcnt == 1)) {
777 if (asoc->asconf_addr_del_pending)
778 continue;
779 asoc->asconf_addr_del_pending =
780 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
781 if (asoc->asconf_addr_del_pending == NULL) {
782 retval = -ENOMEM;
783 goto out;
785 asoc->asconf_addr_del_pending->sa.sa_family =
786 addrs->sa_family;
787 asoc->asconf_addr_del_pending->v4.sin_port =
788 htons(bp->port);
789 if (addrs->sa_family == AF_INET) {
790 struct sockaddr_in *sin;
792 sin = (struct sockaddr_in *)addrs;
793 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
794 } else if (addrs->sa_family == AF_INET6) {
795 struct sockaddr_in6 *sin6;
797 sin6 = (struct sockaddr_in6 *)addrs;
798 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
801 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
802 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
803 asoc->asconf_addr_del_pending);
805 asoc->src_out_of_asoc_ok = 1;
806 stored = 1;
807 goto skip_mkasconf;
810 if (laddr == NULL)
811 return -EINVAL;
813 /* We do not need RCU protection throughout this loop
814 * because this is done under a socket lock from the
815 * setsockopt call.
817 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
818 SCTP_PARAM_DEL_IP);
819 if (!chunk) {
820 retval = -ENOMEM;
821 goto out;
824 skip_mkasconf:
825 /* Reset use_as_src flag for the addresses in the bind address
826 * list that are to be deleted.
828 addr_buf = addrs;
829 for (i = 0; i < addrcnt; i++) {
830 laddr = addr_buf;
831 af = sctp_get_af_specific(laddr->v4.sin_family);
832 list_for_each_entry(saddr, &bp->address_list, list) {
833 if (sctp_cmp_addr_exact(&saddr->a, laddr))
834 saddr->state = SCTP_ADDR_DEL;
836 addr_buf += af->sockaddr_len;
839 /* Update the route and saddr entries for all the transports
840 * as some of the addresses in the bind address list are
841 * about to be deleted and cannot be used as source addresses.
843 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
844 transports) {
845 dst_release(transport->dst);
846 sctp_transport_route(transport, NULL,
847 sctp_sk(asoc->base.sk));
850 if (stored)
851 /* We don't need to transmit ASCONF */
852 continue;
853 retval = sctp_send_asconf(asoc, chunk);
855 out:
856 return retval;
859 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
860 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
862 struct sock *sk = sctp_opt2sk(sp);
863 union sctp_addr *addr;
864 struct sctp_af *af;
866 /* It is safe to write port space in caller. */
867 addr = &addrw->a;
868 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
869 af = sctp_get_af_specific(addr->sa.sa_family);
870 if (!af)
871 return -EINVAL;
872 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
873 return -EINVAL;
875 if (addrw->state == SCTP_ADDR_NEW)
876 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
877 else
878 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
881 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
883 * API 8.1
884 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
885 * int flags);
887 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
888 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
889 * or IPv6 addresses.
891 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
892 * Section 3.1.2 for this usage.
894 * addrs is a pointer to an array of one or more socket addresses. Each
895 * address is contained in its appropriate structure (i.e. struct
896 * sockaddr_in or struct sockaddr_in6) the family of the address type
897 * must be used to distinguish the address length (note that this
898 * representation is termed a "packed array" of addresses). The caller
899 * specifies the number of addresses in the array with addrcnt.
901 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
902 * -1, and sets errno to the appropriate error code.
904 * For SCTP, the port given in each socket address must be the same, or
905 * sctp_bindx() will fail, setting errno to EINVAL.
907 * The flags parameter is formed from the bitwise OR of zero or more of
908 * the following currently defined flags:
910 * SCTP_BINDX_ADD_ADDR
912 * SCTP_BINDX_REM_ADDR
914 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
915 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
916 * addresses from the association. The two flags are mutually exclusive;
917 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
918 * not remove all addresses from an association; sctp_bindx() will
919 * reject such an attempt with EINVAL.
921 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
922 * additional addresses with an endpoint after calling bind(). Or use
923 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
924 * socket is associated with so that no new association accepted will be
925 * associated with those addresses. If the endpoint supports dynamic
926 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
927 * endpoint to send the appropriate message to the peer to change the
928 * peers address lists.
930 * Adding and removing addresses from a connected association is
931 * optional functionality. Implementations that do not support this
932 * functionality should return EOPNOTSUPP.
934 * Basically do nothing but copying the addresses from user to kernel
935 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
936 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
937 * from userspace.
939 * We don't use copy_from_user() for optimization: we first do the
940 * sanity checks (buffer size -fast- and access check-healthy
941 * pointer); if all of those succeed, then we can alloc the memory
942 * (expensive operation) needed to copy the data to kernel. Then we do
943 * the copying without checking the user space area
944 * (__copy_from_user()).
946 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
947 * it.
949 * sk The sk of the socket
950 * addrs The pointer to the addresses in user land
951 * addrssize Size of the addrs buffer
952 * op Operation to perform (add or remove, see the flags of
953 * sctp_bindx)
955 * Returns 0 if ok, <0 errno code on error.
957 static int sctp_setsockopt_bindx(struct sock *sk,
958 struct sockaddr __user *addrs,
959 int addrs_size, int op)
961 struct sockaddr *kaddrs;
962 int err;
963 int addrcnt = 0;
964 int walk_size = 0;
965 struct sockaddr *sa_addr;
966 void *addr_buf;
967 struct sctp_af *af;
969 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
970 __func__, sk, addrs, addrs_size, op);
972 if (unlikely(addrs_size <= 0))
973 return -EINVAL;
975 /* Check the user passed a healthy pointer. */
976 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
977 return -EFAULT;
979 /* Alloc space for the address array in kernel memory. */
980 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
981 if (unlikely(!kaddrs))
982 return -ENOMEM;
984 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
985 kfree(kaddrs);
986 return -EFAULT;
989 /* Walk through the addrs buffer and count the number of addresses. */
990 addr_buf = kaddrs;
991 while (walk_size < addrs_size) {
992 if (walk_size + sizeof(sa_family_t) > addrs_size) {
993 kfree(kaddrs);
994 return -EINVAL;
997 sa_addr = addr_buf;
998 af = sctp_get_af_specific(sa_addr->sa_family);
1000 /* If the address family is not supported or if this address
1001 * causes the address buffer to overflow return EINVAL.
1003 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1004 kfree(kaddrs);
1005 return -EINVAL;
1007 addrcnt++;
1008 addr_buf += af->sockaddr_len;
1009 walk_size += af->sockaddr_len;
1012 /* Do the work. */
1013 switch (op) {
1014 case SCTP_BINDX_ADD_ADDR:
1015 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1016 if (err)
1017 goto out;
1018 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1019 break;
1021 case SCTP_BINDX_REM_ADDR:
1022 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1023 if (err)
1024 goto out;
1025 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1026 break;
1028 default:
1029 err = -EINVAL;
1030 break;
1033 out:
1034 kfree(kaddrs);
1036 return err;
1039 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1041 * Common routine for handling connect() and sctp_connectx().
1042 * Connect will come in with just a single address.
1044 static int __sctp_connect(struct sock *sk,
1045 struct sockaddr *kaddrs,
1046 int addrs_size,
1047 sctp_assoc_t *assoc_id)
1049 struct net *net = sock_net(sk);
1050 struct sctp_sock *sp;
1051 struct sctp_endpoint *ep;
1052 struct sctp_association *asoc = NULL;
1053 struct sctp_association *asoc2;
1054 struct sctp_transport *transport;
1055 union sctp_addr to;
1056 sctp_scope_t scope;
1057 long timeo;
1058 int err = 0;
1059 int addrcnt = 0;
1060 int walk_size = 0;
1061 union sctp_addr *sa_addr = NULL;
1062 void *addr_buf;
1063 unsigned short port;
1064 unsigned int f_flags = 0;
1066 sp = sctp_sk(sk);
1067 ep = sp->ep;
1069 /* connect() cannot be done on a socket that is already in ESTABLISHED
1070 * state - UDP-style peeled off socket or a TCP-style socket that
1071 * is already connected.
1072 * It cannot be done even on a TCP-style listening socket.
1074 if (sctp_sstate(sk, ESTABLISHED) ||
1075 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1076 err = -EISCONN;
1077 goto out_free;
1080 /* Walk through the addrs buffer and count the number of addresses. */
1081 addr_buf = kaddrs;
1082 while (walk_size < addrs_size) {
1083 struct sctp_af *af;
1085 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1086 err = -EINVAL;
1087 goto out_free;
1090 sa_addr = addr_buf;
1091 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1093 /* If the address family is not supported or if this address
1094 * causes the address buffer to overflow return EINVAL.
1096 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1097 err = -EINVAL;
1098 goto out_free;
1101 port = ntohs(sa_addr->v4.sin_port);
1103 /* Save current address so we can work with it */
1104 memcpy(&to, sa_addr, af->sockaddr_len);
1106 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1107 if (err)
1108 goto out_free;
1110 /* Make sure the destination port is correctly set
1111 * in all addresses.
1113 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1114 err = -EINVAL;
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 !ns_capable(net->user_ns, 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(net, 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 sp->pf->to_sk_daddr(sa_addr, sk);
1210 sk->sk_err = 0;
1212 /* in-kernel sockets don't generally have a file allocated to them
1213 * if all they do is call sock_create_kern().
1215 if (sk->sk_socket->file)
1216 f_flags = sk->sk_socket->file->f_flags;
1218 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1220 err = sctp_wait_for_connect(asoc, &timeo);
1221 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1222 *assoc_id = asoc->assoc_id;
1224 /* Don't free association on exit. */
1225 asoc = NULL;
1227 out_free:
1228 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1229 __func__, asoc, kaddrs, err);
1231 if (asoc) {
1232 /* sctp_primitive_ASSOCIATE may have added this association
1233 * To the hash table, try to unhash it, just in case, its a noop
1234 * if it wasn't hashed so we're safe
1236 sctp_unhash_established(asoc);
1237 sctp_association_free(asoc);
1239 return err;
1242 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1244 * API 8.9
1245 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1246 * sctp_assoc_t *asoc);
1248 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1249 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1250 * or IPv6 addresses.
1252 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1253 * Section 3.1.2 for this usage.
1255 * addrs is a pointer to an array of one or more socket addresses. Each
1256 * address is contained in its appropriate structure (i.e. struct
1257 * sockaddr_in or struct sockaddr_in6) the family of the address type
1258 * must be used to distengish the address length (note that this
1259 * representation is termed a "packed array" of addresses). The caller
1260 * specifies the number of addresses in the array with addrcnt.
1262 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1263 * the association id of the new association. On failure, sctp_connectx()
1264 * returns -1, and sets errno to the appropriate error code. The assoc_id
1265 * is not touched by the kernel.
1267 * For SCTP, the port given in each socket address must be the same, or
1268 * sctp_connectx() will fail, setting errno to EINVAL.
1270 * An application can use sctp_connectx to initiate an association with
1271 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1272 * allows a caller to specify multiple addresses at which a peer can be
1273 * reached. The way the SCTP stack uses the list of addresses to set up
1274 * the association is implementation dependent. This function only
1275 * specifies that the stack will try to make use of all the addresses in
1276 * the list when needed.
1278 * Note that the list of addresses passed in is only used for setting up
1279 * the association. It does not necessarily equal the set of addresses
1280 * the peer uses for the resulting association. If the caller wants to
1281 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1282 * retrieve them after the association has been set up.
1284 * Basically do nothing but copying the addresses from user to kernel
1285 * land and invoking either sctp_connectx(). This is used for tunneling
1286 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1288 * We don't use copy_from_user() for optimization: we first do the
1289 * sanity checks (buffer size -fast- and access check-healthy
1290 * pointer); if all of those succeed, then we can alloc the memory
1291 * (expensive operation) needed to copy the data to kernel. Then we do
1292 * the copying without checking the user space area
1293 * (__copy_from_user()).
1295 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1296 * it.
1298 * sk The sk of the socket
1299 * addrs The pointer to the addresses in user land
1300 * addrssize Size of the addrs buffer
1302 * Returns >=0 if ok, <0 errno code on error.
1304 static int __sctp_setsockopt_connectx(struct sock *sk,
1305 struct sockaddr __user *addrs,
1306 int addrs_size,
1307 sctp_assoc_t *assoc_id)
1309 int err = 0;
1310 struct sockaddr *kaddrs;
1312 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1313 __func__, sk, addrs, addrs_size);
1315 if (unlikely(addrs_size <= 0))
1316 return -EINVAL;
1318 /* Check the user passed a healthy pointer. */
1319 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1320 return -EFAULT;
1322 /* Alloc space for the address array in kernel memory. */
1323 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1324 if (unlikely(!kaddrs))
1325 return -ENOMEM;
1327 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1328 err = -EFAULT;
1329 } else {
1330 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1333 kfree(kaddrs);
1335 return err;
1339 * This is an older interface. It's kept for backward compatibility
1340 * to the option that doesn't provide association id.
1342 static int sctp_setsockopt_connectx_old(struct sock *sk,
1343 struct sockaddr __user *addrs,
1344 int addrs_size)
1346 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1350 * New interface for the API. The since the API is done with a socket
1351 * option, to make it simple we feed back the association id is as a return
1352 * indication to the call. Error is always negative and association id is
1353 * always positive.
1355 static int sctp_setsockopt_connectx(struct sock *sk,
1356 struct sockaddr __user *addrs,
1357 int addrs_size)
1359 sctp_assoc_t assoc_id = 0;
1360 int err = 0;
1362 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1364 if (err)
1365 return err;
1366 else
1367 return assoc_id;
1371 * New (hopefully final) interface for the API.
1372 * We use the sctp_getaddrs_old structure so that use-space library
1373 * can avoid any unnecessary allocations. The only different part
1374 * is that we store the actual length of the address buffer into the
1375 * addrs_num structure member. That way we can re-use the existing
1376 * code.
1378 #ifdef CONFIG_COMPAT
1379 struct compat_sctp_getaddrs_old {
1380 sctp_assoc_t assoc_id;
1381 s32 addr_num;
1382 compat_uptr_t addrs; /* struct sockaddr * */
1384 #endif
1386 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1387 char __user *optval,
1388 int __user *optlen)
1390 struct sctp_getaddrs_old param;
1391 sctp_assoc_t assoc_id = 0;
1392 int err = 0;
1394 #ifdef CONFIG_COMPAT
1395 if (is_compat_task()) {
1396 struct compat_sctp_getaddrs_old param32;
1398 if (len < sizeof(param32))
1399 return -EINVAL;
1400 if (copy_from_user(&param32, optval, sizeof(param32)))
1401 return -EFAULT;
1403 param.assoc_id = param32.assoc_id;
1404 param.addr_num = param32.addr_num;
1405 param.addrs = compat_ptr(param32.addrs);
1406 } else
1407 #endif
1409 if (len < sizeof(param))
1410 return -EINVAL;
1411 if (copy_from_user(&param, optval, sizeof(param)))
1412 return -EFAULT;
1415 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1416 param.addrs, param.addr_num,
1417 &assoc_id);
1418 if (err == 0 || err == -EINPROGRESS) {
1419 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1420 return -EFAULT;
1421 if (put_user(sizeof(assoc_id), optlen))
1422 return -EFAULT;
1425 return err;
1428 /* API 3.1.4 close() - UDP Style Syntax
1429 * Applications use close() to perform graceful shutdown (as described in
1430 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1431 * by a UDP-style socket.
1433 * The syntax is
1435 * ret = close(int sd);
1437 * sd - the socket descriptor of the associations to be closed.
1439 * To gracefully shutdown a specific association represented by the
1440 * UDP-style socket, an application should use the sendmsg() call,
1441 * passing no user data, but including the appropriate flag in the
1442 * ancillary data (see Section xxxx).
1444 * If sd in the close() call is a branched-off socket representing only
1445 * one association, the shutdown is performed on that association only.
1447 * 4.1.6 close() - TCP Style Syntax
1449 * Applications use close() to gracefully close down an association.
1451 * The syntax is:
1453 * int close(int sd);
1455 * sd - the socket descriptor of the association to be closed.
1457 * After an application calls close() on a socket descriptor, no further
1458 * socket operations will succeed on that descriptor.
1460 * API 7.1.4 SO_LINGER
1462 * An application using the TCP-style socket can use this option to
1463 * perform the SCTP ABORT primitive. The linger option structure is:
1465 * struct linger {
1466 * int l_onoff; // option on/off
1467 * int l_linger; // linger time
1468 * };
1470 * To enable the option, set l_onoff to 1. If the l_linger value is set
1471 * to 0, calling close() is the same as the ABORT primitive. If the
1472 * value is set to a negative value, the setsockopt() call will return
1473 * an error. If the value is set to a positive value linger_time, the
1474 * close() can be blocked for at most linger_time ms. If the graceful
1475 * shutdown phase does not finish during this period, close() will
1476 * return but the graceful shutdown phase continues in the system.
1478 static void sctp_close(struct sock *sk, long timeout)
1480 struct net *net = sock_net(sk);
1481 struct sctp_endpoint *ep;
1482 struct sctp_association *asoc;
1483 struct list_head *pos, *temp;
1484 unsigned int data_was_unread;
1486 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1488 lock_sock(sk);
1489 sk->sk_shutdown = SHUTDOWN_MASK;
1490 sk->sk_state = SCTP_SS_CLOSING;
1492 ep = sctp_sk(sk)->ep;
1494 /* Clean up any skbs sitting on the receive queue. */
1495 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1496 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1498 /* Walk all associations on an endpoint. */
1499 list_for_each_safe(pos, temp, &ep->asocs) {
1500 asoc = list_entry(pos, struct sctp_association, asocs);
1502 if (sctp_style(sk, TCP)) {
1503 /* A closed association can still be in the list if
1504 * it belongs to a TCP-style listening socket that is
1505 * not yet accepted. If so, free it. If not, send an
1506 * ABORT or SHUTDOWN based on the linger options.
1508 if (sctp_state(asoc, CLOSED)) {
1509 sctp_unhash_established(asoc);
1510 sctp_association_free(asoc);
1511 continue;
1515 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1516 !skb_queue_empty(&asoc->ulpq.reasm) ||
1517 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1518 struct sctp_chunk *chunk;
1520 chunk = sctp_make_abort_user(asoc, NULL, 0);
1521 if (chunk)
1522 sctp_primitive_ABORT(net, asoc, chunk);
1523 } else
1524 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1527 /* On a TCP-style socket, block for at most linger_time if set. */
1528 if (sctp_style(sk, TCP) && timeout)
1529 sctp_wait_for_close(sk, timeout);
1531 /* This will run the backlog queue. */
1532 release_sock(sk);
1534 /* Supposedly, no process has access to the socket, but
1535 * the net layers still may.
1537 local_bh_disable();
1538 bh_lock_sock(sk);
1540 /* Hold the sock, since sk_common_release() will put sock_put()
1541 * and we have just a little more cleanup.
1543 sock_hold(sk);
1544 sk_common_release(sk);
1546 bh_unlock_sock(sk);
1547 local_bh_enable();
1549 sock_put(sk);
1551 SCTP_DBG_OBJCNT_DEC(sock);
1554 /* Handle EPIPE error. */
1555 static int sctp_error(struct sock *sk, int flags, int err)
1557 if (err == -EPIPE)
1558 err = sock_error(sk) ? : -EPIPE;
1559 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1560 send_sig(SIGPIPE, current, 0);
1561 return err;
1564 /* API 3.1.3 sendmsg() - UDP Style Syntax
1566 * An application uses sendmsg() and recvmsg() calls to transmit data to
1567 * and receive data from its peer.
1569 * ssize_t sendmsg(int socket, const struct msghdr *message,
1570 * int flags);
1572 * socket - the socket descriptor of the endpoint.
1573 * message - pointer to the msghdr structure which contains a single
1574 * user message and possibly some ancillary data.
1576 * See Section 5 for complete description of the data
1577 * structures.
1579 * flags - flags sent or received with the user message, see Section
1580 * 5 for complete description of the flags.
1582 * Note: This function could use a rewrite especially when explicit
1583 * connect support comes in.
1585 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1587 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1589 static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1590 struct msghdr *msg, size_t msg_len)
1592 struct net *net = sock_net(sk);
1593 struct sctp_sock *sp;
1594 struct sctp_endpoint *ep;
1595 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1596 struct sctp_transport *transport, *chunk_tp;
1597 struct sctp_chunk *chunk;
1598 union sctp_addr to;
1599 struct sockaddr *msg_name = NULL;
1600 struct sctp_sndrcvinfo default_sinfo;
1601 struct sctp_sndrcvinfo *sinfo;
1602 struct sctp_initmsg *sinit;
1603 sctp_assoc_t associd = 0;
1604 sctp_cmsgs_t cmsgs = { NULL };
1605 sctp_scope_t scope;
1606 bool fill_sinfo_ttl = false;
1607 struct sctp_datamsg *datamsg;
1608 int msg_flags = msg->msg_flags;
1609 __u16 sinfo_flags = 0;
1610 long timeo;
1611 int err;
1613 err = 0;
1614 sp = sctp_sk(sk);
1615 ep = sp->ep;
1617 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1618 msg, msg_len, ep);
1620 /* We cannot send a message over a TCP-style listening socket. */
1621 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1622 err = -EPIPE;
1623 goto out_nounlock;
1626 /* Parse out the SCTP CMSGs. */
1627 err = sctp_msghdr_parse(msg, &cmsgs);
1628 if (err) {
1629 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1630 goto out_nounlock;
1633 /* Fetch the destination address for this packet. This
1634 * address only selects the association--it is not necessarily
1635 * the address we will send to.
1636 * For a peeled-off socket, msg_name is ignored.
1638 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1639 int msg_namelen = msg->msg_namelen;
1641 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1642 msg_namelen);
1643 if (err)
1644 return err;
1646 if (msg_namelen > sizeof(to))
1647 msg_namelen = sizeof(to);
1648 memcpy(&to, msg->msg_name, msg_namelen);
1649 msg_name = msg->msg_name;
1652 sinit = cmsgs.init;
1653 if (cmsgs.sinfo != NULL) {
1654 memset(&default_sinfo, 0, sizeof(default_sinfo));
1655 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1656 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1657 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1658 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1659 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1661 sinfo = &default_sinfo;
1662 fill_sinfo_ttl = true;
1663 } else {
1664 sinfo = cmsgs.srinfo;
1666 /* Did the user specify SNDINFO/SNDRCVINFO? */
1667 if (sinfo) {
1668 sinfo_flags = sinfo->sinfo_flags;
1669 associd = sinfo->sinfo_assoc_id;
1672 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1673 msg_len, sinfo_flags);
1675 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1676 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1677 err = -EINVAL;
1678 goto out_nounlock;
1681 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1682 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1683 * If SCTP_ABORT is set, the message length could be non zero with
1684 * the msg_iov set to the user abort reason.
1686 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1687 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1688 err = -EINVAL;
1689 goto out_nounlock;
1692 /* If SCTP_ADDR_OVER is set, there must be an address
1693 * specified in msg_name.
1695 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1696 err = -EINVAL;
1697 goto out_nounlock;
1700 transport = NULL;
1702 pr_debug("%s: about to look up association\n", __func__);
1704 lock_sock(sk);
1706 /* If a msg_name has been specified, assume this is to be used. */
1707 if (msg_name) {
1708 /* Look for a matching association on the endpoint. */
1709 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1710 if (!asoc) {
1711 /* If we could not find a matching association on the
1712 * endpoint, make sure that it is not a TCP-style
1713 * socket that already has an association or there is
1714 * no peeled-off association on another socket.
1716 if ((sctp_style(sk, TCP) &&
1717 sctp_sstate(sk, ESTABLISHED)) ||
1718 sctp_endpoint_is_peeled_off(ep, &to)) {
1719 err = -EADDRNOTAVAIL;
1720 goto out_unlock;
1723 } else {
1724 asoc = sctp_id2assoc(sk, associd);
1725 if (!asoc) {
1726 err = -EPIPE;
1727 goto out_unlock;
1731 if (asoc) {
1732 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1734 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1735 * socket that has an association in CLOSED state. This can
1736 * happen when an accepted socket has an association that is
1737 * already CLOSED.
1739 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1740 err = -EPIPE;
1741 goto out_unlock;
1744 if (sinfo_flags & SCTP_EOF) {
1745 pr_debug("%s: shutting down association:%p\n",
1746 __func__, asoc);
1748 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1749 err = 0;
1750 goto out_unlock;
1752 if (sinfo_flags & SCTP_ABORT) {
1754 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1755 if (!chunk) {
1756 err = -ENOMEM;
1757 goto out_unlock;
1760 pr_debug("%s: aborting association:%p\n",
1761 __func__, asoc);
1763 sctp_primitive_ABORT(net, asoc, chunk);
1764 err = 0;
1765 goto out_unlock;
1769 /* Do we need to create the association? */
1770 if (!asoc) {
1771 pr_debug("%s: there is no association yet\n", __func__);
1773 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1774 err = -EINVAL;
1775 goto out_unlock;
1778 /* Check for invalid stream against the stream counts,
1779 * either the default or the user specified stream counts.
1781 if (sinfo) {
1782 if (!sinit || !sinit->sinit_num_ostreams) {
1783 /* Check against the defaults. */
1784 if (sinfo->sinfo_stream >=
1785 sp->initmsg.sinit_num_ostreams) {
1786 err = -EINVAL;
1787 goto out_unlock;
1789 } else {
1790 /* Check against the requested. */
1791 if (sinfo->sinfo_stream >=
1792 sinit->sinit_num_ostreams) {
1793 err = -EINVAL;
1794 goto out_unlock;
1800 * API 3.1.2 bind() - UDP Style Syntax
1801 * If a bind() or sctp_bindx() is not called prior to a
1802 * sendmsg() call that initiates a new association, the
1803 * system picks an ephemeral port and will choose an address
1804 * set equivalent to binding with a wildcard address.
1806 if (!ep->base.bind_addr.port) {
1807 if (sctp_autobind(sk)) {
1808 err = -EAGAIN;
1809 goto out_unlock;
1811 } else {
1813 * If an unprivileged user inherits a one-to-many
1814 * style socket with open associations on a privileged
1815 * port, it MAY be permitted to accept new associations,
1816 * but it SHOULD NOT be permitted to open new
1817 * associations.
1819 if (ep->base.bind_addr.port < PROT_SOCK &&
1820 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1821 err = -EACCES;
1822 goto out_unlock;
1826 scope = sctp_scope(&to);
1827 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1828 if (!new_asoc) {
1829 err = -ENOMEM;
1830 goto out_unlock;
1832 asoc = new_asoc;
1833 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1834 if (err < 0) {
1835 err = -ENOMEM;
1836 goto out_free;
1839 /* If the SCTP_INIT ancillary data is specified, set all
1840 * the association init values accordingly.
1842 if (sinit) {
1843 if (sinit->sinit_num_ostreams) {
1844 asoc->c.sinit_num_ostreams =
1845 sinit->sinit_num_ostreams;
1847 if (sinit->sinit_max_instreams) {
1848 asoc->c.sinit_max_instreams =
1849 sinit->sinit_max_instreams;
1851 if (sinit->sinit_max_attempts) {
1852 asoc->max_init_attempts
1853 = sinit->sinit_max_attempts;
1855 if (sinit->sinit_max_init_timeo) {
1856 asoc->max_init_timeo =
1857 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1861 /* Prime the peer's transport structures. */
1862 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1863 if (!transport) {
1864 err = -ENOMEM;
1865 goto out_free;
1869 /* ASSERT: we have a valid association at this point. */
1870 pr_debug("%s: we have a valid association\n", __func__);
1872 if (!sinfo) {
1873 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1874 * one with some defaults.
1876 memset(&default_sinfo, 0, sizeof(default_sinfo));
1877 default_sinfo.sinfo_stream = asoc->default_stream;
1878 default_sinfo.sinfo_flags = asoc->default_flags;
1879 default_sinfo.sinfo_ppid = asoc->default_ppid;
1880 default_sinfo.sinfo_context = asoc->default_context;
1881 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1882 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1884 sinfo = &default_sinfo;
1885 } else if (fill_sinfo_ttl) {
1886 /* In case SNDINFO was specified, we still need to fill
1887 * it with a default ttl from the assoc here.
1889 sinfo->sinfo_timetolive = asoc->default_timetolive;
1892 /* API 7.1.7, the sndbuf size per association bounds the
1893 * maximum size of data that can be sent in a single send call.
1895 if (msg_len > sk->sk_sndbuf) {
1896 err = -EMSGSIZE;
1897 goto out_free;
1900 if (asoc->pmtu_pending)
1901 sctp_assoc_pending_pmtu(sk, asoc);
1903 /* If fragmentation is disabled and the message length exceeds the
1904 * association fragmentation point, return EMSGSIZE. The I-D
1905 * does not specify what this error is, but this looks like
1906 * a great fit.
1908 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1909 err = -EMSGSIZE;
1910 goto out_free;
1913 /* Check for invalid stream. */
1914 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1915 err = -EINVAL;
1916 goto out_free;
1919 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1920 if (!sctp_wspace(asoc)) {
1921 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1922 if (err)
1923 goto out_free;
1926 /* If an address is passed with the sendto/sendmsg call, it is used
1927 * to override the primary destination address in the TCP model, or
1928 * when SCTP_ADDR_OVER flag is set in the UDP model.
1930 if ((sctp_style(sk, TCP) && msg_name) ||
1931 (sinfo_flags & SCTP_ADDR_OVER)) {
1932 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1933 if (!chunk_tp) {
1934 err = -EINVAL;
1935 goto out_free;
1937 } else
1938 chunk_tp = NULL;
1940 /* Auto-connect, if we aren't connected already. */
1941 if (sctp_state(asoc, CLOSED)) {
1942 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1943 if (err < 0)
1944 goto out_free;
1946 pr_debug("%s: we associated primitively\n", __func__);
1949 /* Break the message into multiple chunks of maximum size. */
1950 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1951 if (IS_ERR(datamsg)) {
1952 err = PTR_ERR(datamsg);
1953 goto out_free;
1956 /* Now send the (possibly) fragmented message. */
1957 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1958 sctp_chunk_hold(chunk);
1960 /* Do accounting for the write space. */
1961 sctp_set_owner_w(chunk);
1963 chunk->transport = chunk_tp;
1966 /* Send it to the lower layers. Note: all chunks
1967 * must either fail or succeed. The lower layer
1968 * works that way today. Keep it that way or this
1969 * breaks.
1971 err = sctp_primitive_SEND(net, asoc, datamsg);
1972 /* Did the lower layer accept the chunk? */
1973 if (err) {
1974 sctp_datamsg_free(datamsg);
1975 goto out_free;
1978 pr_debug("%s: we sent primitively\n", __func__);
1980 sctp_datamsg_put(datamsg);
1981 err = msg_len;
1983 /* If we are already past ASSOCIATE, the lower
1984 * layers are responsible for association cleanup.
1986 goto out_unlock;
1988 out_free:
1989 if (new_asoc) {
1990 sctp_unhash_established(asoc);
1991 sctp_association_free(asoc);
1993 out_unlock:
1994 release_sock(sk);
1996 out_nounlock:
1997 return sctp_error(sk, msg_flags, err);
1999 #if 0
2000 do_sock_err:
2001 if (msg_len)
2002 err = msg_len;
2003 else
2004 err = sock_error(sk);
2005 goto out;
2007 do_interrupted:
2008 if (msg_len)
2009 err = msg_len;
2010 goto out;
2011 #endif /* 0 */
2014 /* This is an extended version of skb_pull() that removes the data from the
2015 * start of a skb even when data is spread across the list of skb's in the
2016 * frag_list. len specifies the total amount of data that needs to be removed.
2017 * when 'len' bytes could be removed from the skb, it returns 0.
2018 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2019 * could not be removed.
2021 static int sctp_skb_pull(struct sk_buff *skb, int len)
2023 struct sk_buff *list;
2024 int skb_len = skb_headlen(skb);
2025 int rlen;
2027 if (len <= skb_len) {
2028 __skb_pull(skb, len);
2029 return 0;
2031 len -= skb_len;
2032 __skb_pull(skb, skb_len);
2034 skb_walk_frags(skb, list) {
2035 rlen = sctp_skb_pull(list, len);
2036 skb->len -= (len-rlen);
2037 skb->data_len -= (len-rlen);
2039 if (!rlen)
2040 return 0;
2042 len = rlen;
2045 return len;
2048 /* API 3.1.3 recvmsg() - UDP Style Syntax
2050 * ssize_t recvmsg(int socket, struct msghdr *message,
2051 * int flags);
2053 * socket - the socket descriptor of the endpoint.
2054 * message - pointer to the msghdr structure which contains a single
2055 * user message and possibly some ancillary data.
2057 * See Section 5 for complete description of the data
2058 * structures.
2060 * flags - flags sent or received with the user message, see Section
2061 * 5 for complete description of the flags.
2063 static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2064 struct msghdr *msg, size_t len, int noblock,
2065 int flags, int *addr_len)
2067 struct sctp_ulpevent *event = NULL;
2068 struct sctp_sock *sp = sctp_sk(sk);
2069 struct sk_buff *skb;
2070 int copied;
2071 int err = 0;
2072 int skb_len;
2074 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2075 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2076 addr_len);
2078 lock_sock(sk);
2080 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2081 err = -ENOTCONN;
2082 goto out;
2085 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2086 if (!skb)
2087 goto out;
2089 /* Get the total length of the skb including any skb's in the
2090 * frag_list.
2092 skb_len = skb->len;
2094 copied = skb_len;
2095 if (copied > len)
2096 copied = len;
2098 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2100 event = sctp_skb2event(skb);
2102 if (err)
2103 goto out_free;
2105 sock_recv_ts_and_drops(msg, sk, skb);
2106 if (sctp_ulpevent_is_notification(event)) {
2107 msg->msg_flags |= MSG_NOTIFICATION;
2108 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2109 } else {
2110 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2113 /* Check if we allow SCTP_NXTINFO. */
2114 if (sp->recvnxtinfo)
2115 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2116 /* Check if we allow SCTP_RCVINFO. */
2117 if (sp->recvrcvinfo)
2118 sctp_ulpevent_read_rcvinfo(event, msg);
2119 /* Check if we allow SCTP_SNDRCVINFO. */
2120 if (sp->subscribe.sctp_data_io_event)
2121 sctp_ulpevent_read_sndrcvinfo(event, msg);
2123 #if 0
2124 /* FIXME: we should be calling IP/IPv6 layers. */
2125 if (sk->sk_protinfo.af_inet.cmsg_flags)
2126 ip_cmsg_recv(msg, skb);
2127 #endif
2129 err = copied;
2131 /* If skb's length exceeds the user's buffer, update the skb and
2132 * push it back to the receive_queue so that the next call to
2133 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2135 if (skb_len > copied) {
2136 msg->msg_flags &= ~MSG_EOR;
2137 if (flags & MSG_PEEK)
2138 goto out_free;
2139 sctp_skb_pull(skb, copied);
2140 skb_queue_head(&sk->sk_receive_queue, skb);
2142 /* When only partial message is copied to the user, increase
2143 * rwnd by that amount. If all the data in the skb is read,
2144 * rwnd is updated when the event is freed.
2146 if (!sctp_ulpevent_is_notification(event))
2147 sctp_assoc_rwnd_increase(event->asoc, copied);
2148 goto out;
2149 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2150 (event->msg_flags & MSG_EOR))
2151 msg->msg_flags |= MSG_EOR;
2152 else
2153 msg->msg_flags &= ~MSG_EOR;
2155 out_free:
2156 if (flags & MSG_PEEK) {
2157 /* Release the skb reference acquired after peeking the skb in
2158 * sctp_skb_recv_datagram().
2160 kfree_skb(skb);
2161 } else {
2162 /* Free the event which includes releasing the reference to
2163 * the owner of the skb, freeing the skb and updating the
2164 * rwnd.
2166 sctp_ulpevent_free(event);
2168 out:
2169 release_sock(sk);
2170 return err;
2173 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2175 * This option is a on/off flag. If enabled no SCTP message
2176 * fragmentation will be performed. Instead if a message being sent
2177 * exceeds the current PMTU size, the message will NOT be sent and
2178 * instead a error will be indicated to the user.
2180 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2181 char __user *optval,
2182 unsigned int optlen)
2184 int val;
2186 if (optlen < sizeof(int))
2187 return -EINVAL;
2189 if (get_user(val, (int __user *)optval))
2190 return -EFAULT;
2192 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2194 return 0;
2197 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2198 unsigned int optlen)
2200 struct sctp_association *asoc;
2201 struct sctp_ulpevent *event;
2203 if (optlen > sizeof(struct sctp_event_subscribe))
2204 return -EINVAL;
2205 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2206 return -EFAULT;
2208 if (sctp_sk(sk)->subscribe.sctp_data_io_event)
2209 pr_warn_ratelimited(DEPRECATED "%s (pid %d) "
2210 "Requested SCTP_SNDRCVINFO event.\n"
2211 "Use SCTP_RCVINFO through SCTP_RECVRCVINFO option instead.\n",
2212 current->comm, task_pid_nr(current));
2214 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2215 * if there is no data to be sent or retransmit, the stack will
2216 * immediately send up this notification.
2218 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2219 &sctp_sk(sk)->subscribe)) {
2220 asoc = sctp_id2assoc(sk, 0);
2222 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2223 event = sctp_ulpevent_make_sender_dry_event(asoc,
2224 GFP_ATOMIC);
2225 if (!event)
2226 return -ENOMEM;
2228 sctp_ulpq_tail_event(&asoc->ulpq, event);
2232 return 0;
2235 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2237 * This socket option is applicable to the UDP-style socket only. When
2238 * set it will cause associations that are idle for more than the
2239 * specified number of seconds to automatically close. An association
2240 * being idle is defined an association that has NOT sent or received
2241 * user data. The special value of '0' indicates that no automatic
2242 * close of any associations should be performed. The option expects an
2243 * integer defining the number of seconds of idle time before an
2244 * association is closed.
2246 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2247 unsigned int optlen)
2249 struct sctp_sock *sp = sctp_sk(sk);
2250 struct net *net = sock_net(sk);
2252 /* Applicable to UDP-style socket only */
2253 if (sctp_style(sk, TCP))
2254 return -EOPNOTSUPP;
2255 if (optlen != sizeof(int))
2256 return -EINVAL;
2257 if (copy_from_user(&sp->autoclose, optval, optlen))
2258 return -EFAULT;
2260 if (sp->autoclose > net->sctp.max_autoclose)
2261 sp->autoclose = net->sctp.max_autoclose;
2263 return 0;
2266 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2268 * Applications can enable or disable heartbeats for any peer address of
2269 * an association, modify an address's heartbeat interval, force a
2270 * heartbeat to be sent immediately, and adjust the address's maximum
2271 * number of retransmissions sent before an address is considered
2272 * unreachable. The following structure is used to access and modify an
2273 * address's parameters:
2275 * struct sctp_paddrparams {
2276 * sctp_assoc_t spp_assoc_id;
2277 * struct sockaddr_storage spp_address;
2278 * uint32_t spp_hbinterval;
2279 * uint16_t spp_pathmaxrxt;
2280 * uint32_t spp_pathmtu;
2281 * uint32_t spp_sackdelay;
2282 * uint32_t spp_flags;
2283 * };
2285 * spp_assoc_id - (one-to-many style socket) This is filled in the
2286 * application, and identifies the association for
2287 * this query.
2288 * spp_address - This specifies which address is of interest.
2289 * spp_hbinterval - This contains the value of the heartbeat interval,
2290 * in milliseconds. If a value of zero
2291 * is present in this field then no changes are to
2292 * be made to this parameter.
2293 * spp_pathmaxrxt - This contains the maximum number of
2294 * retransmissions before this address shall be
2295 * considered unreachable. If a value of zero
2296 * is present in this field then no changes are to
2297 * be made to this parameter.
2298 * spp_pathmtu - When Path MTU discovery is disabled the value
2299 * specified here will be the "fixed" path mtu.
2300 * Note that if the spp_address field is empty
2301 * then all associations on this address will
2302 * have this fixed path mtu set upon them.
2304 * spp_sackdelay - When delayed sack is enabled, this value specifies
2305 * the number of milliseconds that sacks will be delayed
2306 * for. This value will apply to all addresses of an
2307 * association if the spp_address field is empty. Note
2308 * also, that if delayed sack is enabled and this
2309 * value is set to 0, no change is made to the last
2310 * recorded delayed sack timer value.
2312 * spp_flags - These flags are used to control various features
2313 * on an association. The flag field may contain
2314 * zero or more of the following options.
2316 * SPP_HB_ENABLE - Enable heartbeats on the
2317 * specified address. Note that if the address
2318 * field is empty all addresses for the association
2319 * have heartbeats enabled upon them.
2321 * SPP_HB_DISABLE - Disable heartbeats on the
2322 * speicifed address. Note that if the address
2323 * field is empty all addresses for the association
2324 * will have their heartbeats disabled. Note also
2325 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2326 * mutually exclusive, only one of these two should
2327 * be specified. Enabling both fields will have
2328 * undetermined results.
2330 * SPP_HB_DEMAND - Request a user initiated heartbeat
2331 * to be made immediately.
2333 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2334 * heartbeat delayis to be set to the value of 0
2335 * milliseconds.
2337 * SPP_PMTUD_ENABLE - This field will enable PMTU
2338 * discovery upon the specified address. Note that
2339 * if the address feild is empty then all addresses
2340 * on the association are effected.
2342 * SPP_PMTUD_DISABLE - This field will disable PMTU
2343 * discovery upon the specified address. Note that
2344 * if the address feild is empty then all addresses
2345 * on the association are effected. Not also that
2346 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2347 * exclusive. Enabling both will have undetermined
2348 * results.
2350 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2351 * on delayed sack. The time specified in spp_sackdelay
2352 * is used to specify the sack delay for this address. Note
2353 * that if spp_address is empty then all addresses will
2354 * enable delayed sack and take on the sack delay
2355 * value specified in spp_sackdelay.
2356 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2357 * off delayed sack. If the spp_address field is blank then
2358 * delayed sack is disabled for the entire association. Note
2359 * also that this field is mutually exclusive to
2360 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2361 * results.
2363 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2364 struct sctp_transport *trans,
2365 struct sctp_association *asoc,
2366 struct sctp_sock *sp,
2367 int hb_change,
2368 int pmtud_change,
2369 int sackdelay_change)
2371 int error;
2373 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2374 struct net *net = sock_net(trans->asoc->base.sk);
2376 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2377 if (error)
2378 return error;
2381 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2382 * this field is ignored. Note also that a value of zero indicates
2383 * the current setting should be left unchanged.
2385 if (params->spp_flags & SPP_HB_ENABLE) {
2387 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2388 * set. This lets us use 0 value when this flag
2389 * is set.
2391 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2392 params->spp_hbinterval = 0;
2394 if (params->spp_hbinterval ||
2395 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2396 if (trans) {
2397 trans->hbinterval =
2398 msecs_to_jiffies(params->spp_hbinterval);
2399 } else if (asoc) {
2400 asoc->hbinterval =
2401 msecs_to_jiffies(params->spp_hbinterval);
2402 } else {
2403 sp->hbinterval = params->spp_hbinterval;
2408 if (hb_change) {
2409 if (trans) {
2410 trans->param_flags =
2411 (trans->param_flags & ~SPP_HB) | hb_change;
2412 } else if (asoc) {
2413 asoc->param_flags =
2414 (asoc->param_flags & ~SPP_HB) | hb_change;
2415 } else {
2416 sp->param_flags =
2417 (sp->param_flags & ~SPP_HB) | hb_change;
2421 /* When Path MTU discovery is disabled the value specified here will
2422 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2423 * include the flag SPP_PMTUD_DISABLE for this field to have any
2424 * effect).
2426 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2427 if (trans) {
2428 trans->pathmtu = params->spp_pathmtu;
2429 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2430 } else if (asoc) {
2431 asoc->pathmtu = params->spp_pathmtu;
2432 sctp_frag_point(asoc, params->spp_pathmtu);
2433 } else {
2434 sp->pathmtu = params->spp_pathmtu;
2438 if (pmtud_change) {
2439 if (trans) {
2440 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2441 (params->spp_flags & SPP_PMTUD_ENABLE);
2442 trans->param_flags =
2443 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2444 if (update) {
2445 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2446 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2448 } else if (asoc) {
2449 asoc->param_flags =
2450 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2451 } else {
2452 sp->param_flags =
2453 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2457 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2458 * value of this field is ignored. Note also that a value of zero
2459 * indicates the current setting should be left unchanged.
2461 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2462 if (trans) {
2463 trans->sackdelay =
2464 msecs_to_jiffies(params->spp_sackdelay);
2465 } else if (asoc) {
2466 asoc->sackdelay =
2467 msecs_to_jiffies(params->spp_sackdelay);
2468 } else {
2469 sp->sackdelay = params->spp_sackdelay;
2473 if (sackdelay_change) {
2474 if (trans) {
2475 trans->param_flags =
2476 (trans->param_flags & ~SPP_SACKDELAY) |
2477 sackdelay_change;
2478 } else if (asoc) {
2479 asoc->param_flags =
2480 (asoc->param_flags & ~SPP_SACKDELAY) |
2481 sackdelay_change;
2482 } else {
2483 sp->param_flags =
2484 (sp->param_flags & ~SPP_SACKDELAY) |
2485 sackdelay_change;
2489 /* Note that a value of zero indicates the current setting should be
2490 left unchanged.
2492 if (params->spp_pathmaxrxt) {
2493 if (trans) {
2494 trans->pathmaxrxt = params->spp_pathmaxrxt;
2495 } else if (asoc) {
2496 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2497 } else {
2498 sp->pathmaxrxt = params->spp_pathmaxrxt;
2502 return 0;
2505 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2506 char __user *optval,
2507 unsigned int optlen)
2509 struct sctp_paddrparams params;
2510 struct sctp_transport *trans = NULL;
2511 struct sctp_association *asoc = NULL;
2512 struct sctp_sock *sp = sctp_sk(sk);
2513 int error;
2514 int hb_change, pmtud_change, sackdelay_change;
2516 if (optlen != sizeof(struct sctp_paddrparams))
2517 return -EINVAL;
2519 if (copy_from_user(&params, optval, optlen))
2520 return -EFAULT;
2522 /* Validate flags and value parameters. */
2523 hb_change = params.spp_flags & SPP_HB;
2524 pmtud_change = params.spp_flags & SPP_PMTUD;
2525 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2527 if (hb_change == SPP_HB ||
2528 pmtud_change == SPP_PMTUD ||
2529 sackdelay_change == SPP_SACKDELAY ||
2530 params.spp_sackdelay > 500 ||
2531 (params.spp_pathmtu &&
2532 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2533 return -EINVAL;
2535 /* If an address other than INADDR_ANY is specified, and
2536 * no transport is found, then the request is invalid.
2538 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2539 trans = sctp_addr_id2transport(sk, &params.spp_address,
2540 params.spp_assoc_id);
2541 if (!trans)
2542 return -EINVAL;
2545 /* Get association, if assoc_id != 0 and the socket is a one
2546 * to many style socket, and an association was not found, then
2547 * the id was invalid.
2549 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2550 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2551 return -EINVAL;
2553 /* Heartbeat demand can only be sent on a transport or
2554 * association, but not a socket.
2556 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2557 return -EINVAL;
2559 /* Process parameters. */
2560 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2561 hb_change, pmtud_change,
2562 sackdelay_change);
2564 if (error)
2565 return error;
2567 /* If changes are for association, also apply parameters to each
2568 * transport.
2570 if (!trans && asoc) {
2571 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2572 transports) {
2573 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2574 hb_change, pmtud_change,
2575 sackdelay_change);
2579 return 0;
2582 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2584 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2587 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2589 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2593 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2595 * This option will effect the way delayed acks are performed. This
2596 * option allows you to get or set the delayed ack time, in
2597 * milliseconds. It also allows changing the delayed ack frequency.
2598 * Changing the frequency to 1 disables the delayed sack algorithm. If
2599 * the assoc_id is 0, then this sets or gets the endpoints default
2600 * values. If the assoc_id field is non-zero, then the set or get
2601 * effects the specified association for the one to many model (the
2602 * assoc_id field is ignored by the one to one model). Note that if
2603 * sack_delay or sack_freq are 0 when setting this option, then the
2604 * current values will remain unchanged.
2606 * struct sctp_sack_info {
2607 * sctp_assoc_t sack_assoc_id;
2608 * uint32_t sack_delay;
2609 * uint32_t sack_freq;
2610 * };
2612 * sack_assoc_id - This parameter, indicates which association the user
2613 * is performing an action upon. Note that if this field's value is
2614 * zero then the endpoints default value is changed (effecting future
2615 * associations only).
2617 * sack_delay - This parameter contains the number of milliseconds that
2618 * the user is requesting the delayed ACK timer be set to. Note that
2619 * this value is defined in the standard to be between 200 and 500
2620 * milliseconds.
2622 * sack_freq - This parameter contains the number of packets that must
2623 * be received before a sack is sent without waiting for the delay
2624 * timer to expire. The default value for this is 2, setting this
2625 * value to 1 will disable the delayed sack algorithm.
2628 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2629 char __user *optval, unsigned int optlen)
2631 struct sctp_sack_info params;
2632 struct sctp_transport *trans = NULL;
2633 struct sctp_association *asoc = NULL;
2634 struct sctp_sock *sp = sctp_sk(sk);
2636 if (optlen == sizeof(struct sctp_sack_info)) {
2637 if (copy_from_user(&params, optval, optlen))
2638 return -EFAULT;
2640 if (params.sack_delay == 0 && params.sack_freq == 0)
2641 return 0;
2642 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2643 pr_warn_ratelimited(DEPRECATED
2644 "%s (pid %d) "
2645 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2646 "Use struct sctp_sack_info instead\n",
2647 current->comm, task_pid_nr(current));
2648 if (copy_from_user(&params, optval, optlen))
2649 return -EFAULT;
2651 if (params.sack_delay == 0)
2652 params.sack_freq = 1;
2653 else
2654 params.sack_freq = 0;
2655 } else
2656 return -EINVAL;
2658 /* Validate value parameter. */
2659 if (params.sack_delay > 500)
2660 return -EINVAL;
2662 /* Get association, if sack_assoc_id != 0 and the socket is a one
2663 * to many style socket, and an association was not found, then
2664 * the id was invalid.
2666 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2667 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2668 return -EINVAL;
2670 if (params.sack_delay) {
2671 if (asoc) {
2672 asoc->sackdelay =
2673 msecs_to_jiffies(params.sack_delay);
2674 asoc->param_flags =
2675 sctp_spp_sackdelay_enable(asoc->param_flags);
2676 } else {
2677 sp->sackdelay = params.sack_delay;
2678 sp->param_flags =
2679 sctp_spp_sackdelay_enable(sp->param_flags);
2683 if (params.sack_freq == 1) {
2684 if (asoc) {
2685 asoc->param_flags =
2686 sctp_spp_sackdelay_disable(asoc->param_flags);
2687 } else {
2688 sp->param_flags =
2689 sctp_spp_sackdelay_disable(sp->param_flags);
2691 } else if (params.sack_freq > 1) {
2692 if (asoc) {
2693 asoc->sackfreq = params.sack_freq;
2694 asoc->param_flags =
2695 sctp_spp_sackdelay_enable(asoc->param_flags);
2696 } else {
2697 sp->sackfreq = params.sack_freq;
2698 sp->param_flags =
2699 sctp_spp_sackdelay_enable(sp->param_flags);
2703 /* If change is for association, also apply to each transport. */
2704 if (asoc) {
2705 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2706 transports) {
2707 if (params.sack_delay) {
2708 trans->sackdelay =
2709 msecs_to_jiffies(params.sack_delay);
2710 trans->param_flags =
2711 sctp_spp_sackdelay_enable(trans->param_flags);
2713 if (params.sack_freq == 1) {
2714 trans->param_flags =
2715 sctp_spp_sackdelay_disable(trans->param_flags);
2716 } else if (params.sack_freq > 1) {
2717 trans->sackfreq = params.sack_freq;
2718 trans->param_flags =
2719 sctp_spp_sackdelay_enable(trans->param_flags);
2724 return 0;
2727 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2729 * Applications can specify protocol parameters for the default association
2730 * initialization. The option name argument to setsockopt() and getsockopt()
2731 * is SCTP_INITMSG.
2733 * Setting initialization parameters is effective only on an unconnected
2734 * socket (for UDP-style sockets only future associations are effected
2735 * by the change). With TCP-style sockets, this option is inherited by
2736 * sockets derived from a listener socket.
2738 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2740 struct sctp_initmsg sinit;
2741 struct sctp_sock *sp = sctp_sk(sk);
2743 if (optlen != sizeof(struct sctp_initmsg))
2744 return -EINVAL;
2745 if (copy_from_user(&sinit, optval, optlen))
2746 return -EFAULT;
2748 if (sinit.sinit_num_ostreams)
2749 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2750 if (sinit.sinit_max_instreams)
2751 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2752 if (sinit.sinit_max_attempts)
2753 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2754 if (sinit.sinit_max_init_timeo)
2755 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2757 return 0;
2761 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2763 * Applications that wish to use the sendto() system call may wish to
2764 * specify a default set of parameters that would normally be supplied
2765 * through the inclusion of ancillary data. This socket option allows
2766 * such an application to set the default sctp_sndrcvinfo structure.
2767 * The application that wishes to use this socket option simply passes
2768 * in to this call the sctp_sndrcvinfo structure defined in Section
2769 * 5.2.2) The input parameters accepted by this call include
2770 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2771 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2772 * to this call if the caller is using the UDP model.
2774 static int sctp_setsockopt_default_send_param(struct sock *sk,
2775 char __user *optval,
2776 unsigned int optlen)
2778 struct sctp_sock *sp = sctp_sk(sk);
2779 struct sctp_association *asoc;
2780 struct sctp_sndrcvinfo info;
2782 if (optlen != sizeof(info))
2783 return -EINVAL;
2784 if (copy_from_user(&info, optval, optlen))
2785 return -EFAULT;
2786 if (info.sinfo_flags &
2787 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2788 SCTP_ABORT | SCTP_EOF))
2789 return -EINVAL;
2791 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2792 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2793 return -EINVAL;
2794 if (asoc) {
2795 asoc->default_stream = info.sinfo_stream;
2796 asoc->default_flags = info.sinfo_flags;
2797 asoc->default_ppid = info.sinfo_ppid;
2798 asoc->default_context = info.sinfo_context;
2799 asoc->default_timetolive = info.sinfo_timetolive;
2800 } else {
2801 sp->default_stream = info.sinfo_stream;
2802 sp->default_flags = info.sinfo_flags;
2803 sp->default_ppid = info.sinfo_ppid;
2804 sp->default_context = info.sinfo_context;
2805 sp->default_timetolive = info.sinfo_timetolive;
2808 return 0;
2811 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2812 * (SCTP_DEFAULT_SNDINFO)
2814 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2815 char __user *optval,
2816 unsigned int optlen)
2818 struct sctp_sock *sp = sctp_sk(sk);
2819 struct sctp_association *asoc;
2820 struct sctp_sndinfo info;
2822 if (optlen != sizeof(info))
2823 return -EINVAL;
2824 if (copy_from_user(&info, optval, optlen))
2825 return -EFAULT;
2826 if (info.snd_flags &
2827 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2828 SCTP_ABORT | SCTP_EOF))
2829 return -EINVAL;
2831 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2832 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2833 return -EINVAL;
2834 if (asoc) {
2835 asoc->default_stream = info.snd_sid;
2836 asoc->default_flags = info.snd_flags;
2837 asoc->default_ppid = info.snd_ppid;
2838 asoc->default_context = info.snd_context;
2839 } else {
2840 sp->default_stream = info.snd_sid;
2841 sp->default_flags = info.snd_flags;
2842 sp->default_ppid = info.snd_ppid;
2843 sp->default_context = info.snd_context;
2846 return 0;
2849 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2851 * Requests that the local SCTP stack use the enclosed peer address as
2852 * the association primary. The enclosed address must be one of the
2853 * association peer's addresses.
2855 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2856 unsigned int optlen)
2858 struct sctp_prim prim;
2859 struct sctp_transport *trans;
2861 if (optlen != sizeof(struct sctp_prim))
2862 return -EINVAL;
2864 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2865 return -EFAULT;
2867 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2868 if (!trans)
2869 return -EINVAL;
2871 sctp_assoc_set_primary(trans->asoc, trans);
2873 return 0;
2877 * 7.1.5 SCTP_NODELAY
2879 * Turn on/off any Nagle-like algorithm. This means that packets are
2880 * generally sent as soon as possible and no unnecessary delays are
2881 * introduced, at the cost of more packets in the network. Expects an
2882 * integer boolean flag.
2884 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2885 unsigned int optlen)
2887 int val;
2889 if (optlen < sizeof(int))
2890 return -EINVAL;
2891 if (get_user(val, (int __user *)optval))
2892 return -EFAULT;
2894 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2895 return 0;
2900 * 7.1.1 SCTP_RTOINFO
2902 * The protocol parameters used to initialize and bound retransmission
2903 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2904 * and modify these parameters.
2905 * All parameters are time values, in milliseconds. A value of 0, when
2906 * modifying the parameters, indicates that the current value should not
2907 * be changed.
2910 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2912 struct sctp_rtoinfo rtoinfo;
2913 struct sctp_association *asoc;
2914 unsigned long rto_min, rto_max;
2915 struct sctp_sock *sp = sctp_sk(sk);
2917 if (optlen != sizeof (struct sctp_rtoinfo))
2918 return -EINVAL;
2920 if (copy_from_user(&rtoinfo, optval, optlen))
2921 return -EFAULT;
2923 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2925 /* Set the values to the specific association */
2926 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2927 return -EINVAL;
2929 rto_max = rtoinfo.srto_max;
2930 rto_min = rtoinfo.srto_min;
2932 if (rto_max)
2933 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2934 else
2935 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2937 if (rto_min)
2938 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2939 else
2940 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2942 if (rto_min > rto_max)
2943 return -EINVAL;
2945 if (asoc) {
2946 if (rtoinfo.srto_initial != 0)
2947 asoc->rto_initial =
2948 msecs_to_jiffies(rtoinfo.srto_initial);
2949 asoc->rto_max = rto_max;
2950 asoc->rto_min = rto_min;
2951 } else {
2952 /* If there is no association or the association-id = 0
2953 * set the values to the endpoint.
2955 if (rtoinfo.srto_initial != 0)
2956 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2957 sp->rtoinfo.srto_max = rto_max;
2958 sp->rtoinfo.srto_min = rto_min;
2961 return 0;
2966 * 7.1.2 SCTP_ASSOCINFO
2968 * This option is used to tune the maximum retransmission attempts
2969 * of the association.
2970 * Returns an error if the new association retransmission value is
2971 * greater than the sum of the retransmission value of the peer.
2972 * See [SCTP] for more information.
2975 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2978 struct sctp_assocparams assocparams;
2979 struct sctp_association *asoc;
2981 if (optlen != sizeof(struct sctp_assocparams))
2982 return -EINVAL;
2983 if (copy_from_user(&assocparams, optval, optlen))
2984 return -EFAULT;
2986 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2988 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2989 return -EINVAL;
2991 /* Set the values to the specific association */
2992 if (asoc) {
2993 if (assocparams.sasoc_asocmaxrxt != 0) {
2994 __u32 path_sum = 0;
2995 int paths = 0;
2996 struct sctp_transport *peer_addr;
2998 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2999 transports) {
3000 path_sum += peer_addr->pathmaxrxt;
3001 paths++;
3004 /* Only validate asocmaxrxt if we have more than
3005 * one path/transport. We do this because path
3006 * retransmissions are only counted when we have more
3007 * then one path.
3009 if (paths > 1 &&
3010 assocparams.sasoc_asocmaxrxt > path_sum)
3011 return -EINVAL;
3013 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3016 if (assocparams.sasoc_cookie_life != 0)
3017 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3018 } else {
3019 /* Set the values to the endpoint */
3020 struct sctp_sock *sp = sctp_sk(sk);
3022 if (assocparams.sasoc_asocmaxrxt != 0)
3023 sp->assocparams.sasoc_asocmaxrxt =
3024 assocparams.sasoc_asocmaxrxt;
3025 if (assocparams.sasoc_cookie_life != 0)
3026 sp->assocparams.sasoc_cookie_life =
3027 assocparams.sasoc_cookie_life;
3029 return 0;
3033 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3035 * This socket option is a boolean flag which turns on or off mapped V4
3036 * addresses. If this option is turned on and the socket is type
3037 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3038 * If this option is turned off, then no mapping will be done of V4
3039 * addresses and a user will receive both PF_INET6 and PF_INET type
3040 * addresses on the socket.
3042 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3044 int val;
3045 struct sctp_sock *sp = sctp_sk(sk);
3047 if (optlen < sizeof(int))
3048 return -EINVAL;
3049 if (get_user(val, (int __user *)optval))
3050 return -EFAULT;
3051 if (val)
3052 sp->v4mapped = 1;
3053 else
3054 sp->v4mapped = 0;
3056 return 0;
3060 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3061 * This option will get or set the maximum size to put in any outgoing
3062 * SCTP DATA chunk. If a message is larger than this size it will be
3063 * fragmented by SCTP into the specified size. Note that the underlying
3064 * SCTP implementation may fragment into smaller sized chunks when the
3065 * PMTU of the underlying association is smaller than the value set by
3066 * the user. The default value for this option is '0' which indicates
3067 * the user is NOT limiting fragmentation and only the PMTU will effect
3068 * SCTP's choice of DATA chunk size. Note also that values set larger
3069 * than the maximum size of an IP datagram will effectively let SCTP
3070 * control fragmentation (i.e. the same as setting this option to 0).
3072 * The following structure is used to access and modify this parameter:
3074 * struct sctp_assoc_value {
3075 * sctp_assoc_t assoc_id;
3076 * uint32_t assoc_value;
3077 * };
3079 * assoc_id: This parameter is ignored for one-to-one style sockets.
3080 * For one-to-many style sockets this parameter indicates which
3081 * association the user is performing an action upon. Note that if
3082 * this field's value is zero then the endpoints default value is
3083 * changed (effecting future associations only).
3084 * assoc_value: This parameter specifies the maximum size in bytes.
3086 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3088 struct sctp_assoc_value params;
3089 struct sctp_association *asoc;
3090 struct sctp_sock *sp = sctp_sk(sk);
3091 int val;
3093 if (optlen == sizeof(int)) {
3094 pr_warn_ratelimited(DEPRECATED
3095 "%s (pid %d) "
3096 "Use of int in maxseg socket option.\n"
3097 "Use struct sctp_assoc_value instead\n",
3098 current->comm, task_pid_nr(current));
3099 if (copy_from_user(&val, optval, optlen))
3100 return -EFAULT;
3101 params.assoc_id = 0;
3102 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3103 if (copy_from_user(&params, optval, optlen))
3104 return -EFAULT;
3105 val = params.assoc_value;
3106 } else
3107 return -EINVAL;
3109 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3110 return -EINVAL;
3112 asoc = sctp_id2assoc(sk, params.assoc_id);
3113 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3114 return -EINVAL;
3116 if (asoc) {
3117 if (val == 0) {
3118 val = asoc->pathmtu;
3119 val -= sp->pf->af->net_header_len;
3120 val -= sizeof(struct sctphdr) +
3121 sizeof(struct sctp_data_chunk);
3123 asoc->user_frag = val;
3124 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3125 } else {
3126 sp->user_frag = val;
3129 return 0;
3134 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3136 * Requests that the peer mark the enclosed address as the association
3137 * primary. The enclosed address must be one of the association's
3138 * locally bound addresses. The following structure is used to make a
3139 * set primary request:
3141 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3142 unsigned int optlen)
3144 struct net *net = sock_net(sk);
3145 struct sctp_sock *sp;
3146 struct sctp_association *asoc = NULL;
3147 struct sctp_setpeerprim prim;
3148 struct sctp_chunk *chunk;
3149 struct sctp_af *af;
3150 int err;
3152 sp = sctp_sk(sk);
3154 if (!net->sctp.addip_enable)
3155 return -EPERM;
3157 if (optlen != sizeof(struct sctp_setpeerprim))
3158 return -EINVAL;
3160 if (copy_from_user(&prim, optval, optlen))
3161 return -EFAULT;
3163 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3164 if (!asoc)
3165 return -EINVAL;
3167 if (!asoc->peer.asconf_capable)
3168 return -EPERM;
3170 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3171 return -EPERM;
3173 if (!sctp_state(asoc, ESTABLISHED))
3174 return -ENOTCONN;
3176 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3177 if (!af)
3178 return -EINVAL;
3180 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3181 return -EADDRNOTAVAIL;
3183 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3184 return -EADDRNOTAVAIL;
3186 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3187 chunk = sctp_make_asconf_set_prim(asoc,
3188 (union sctp_addr *)&prim.sspp_addr);
3189 if (!chunk)
3190 return -ENOMEM;
3192 err = sctp_send_asconf(asoc, chunk);
3194 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3196 return err;
3199 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3200 unsigned int optlen)
3202 struct sctp_setadaptation adaptation;
3204 if (optlen != sizeof(struct sctp_setadaptation))
3205 return -EINVAL;
3206 if (copy_from_user(&adaptation, optval, optlen))
3207 return -EFAULT;
3209 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3211 return 0;
3215 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3217 * The context field in the sctp_sndrcvinfo structure is normally only
3218 * used when a failed message is retrieved holding the value that was
3219 * sent down on the actual send call. This option allows the setting of
3220 * a default context on an association basis that will be received on
3221 * reading messages from the peer. This is especially helpful in the
3222 * one-2-many model for an application to keep some reference to an
3223 * internal state machine that is processing messages on the
3224 * association. Note that the setting of this value only effects
3225 * received messages from the peer and does not effect the value that is
3226 * saved with outbound messages.
3228 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3229 unsigned int optlen)
3231 struct sctp_assoc_value params;
3232 struct sctp_sock *sp;
3233 struct sctp_association *asoc;
3235 if (optlen != sizeof(struct sctp_assoc_value))
3236 return -EINVAL;
3237 if (copy_from_user(&params, optval, optlen))
3238 return -EFAULT;
3240 sp = sctp_sk(sk);
3242 if (params.assoc_id != 0) {
3243 asoc = sctp_id2assoc(sk, params.assoc_id);
3244 if (!asoc)
3245 return -EINVAL;
3246 asoc->default_rcv_context = params.assoc_value;
3247 } else {
3248 sp->default_rcv_context = params.assoc_value;
3251 return 0;
3255 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3257 * This options will at a minimum specify if the implementation is doing
3258 * fragmented interleave. Fragmented interleave, for a one to many
3259 * socket, is when subsequent calls to receive a message may return
3260 * parts of messages from different associations. Some implementations
3261 * may allow you to turn this value on or off. If so, when turned off,
3262 * no fragment interleave will occur (which will cause a head of line
3263 * blocking amongst multiple associations sharing the same one to many
3264 * socket). When this option is turned on, then each receive call may
3265 * come from a different association (thus the user must receive data
3266 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3267 * association each receive belongs to.
3269 * This option takes a boolean value. A non-zero value indicates that
3270 * fragmented interleave is on. A value of zero indicates that
3271 * fragmented interleave is off.
3273 * Note that it is important that an implementation that allows this
3274 * option to be turned on, have it off by default. Otherwise an unaware
3275 * application using the one to many model may become confused and act
3276 * incorrectly.
3278 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3279 char __user *optval,
3280 unsigned int optlen)
3282 int val;
3284 if (optlen != sizeof(int))
3285 return -EINVAL;
3286 if (get_user(val, (int __user *)optval))
3287 return -EFAULT;
3289 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3291 return 0;
3295 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3296 * (SCTP_PARTIAL_DELIVERY_POINT)
3298 * This option will set or get the SCTP partial delivery point. This
3299 * point is the size of a message where the partial delivery API will be
3300 * invoked to help free up rwnd space for the peer. Setting this to a
3301 * lower value will cause partial deliveries to happen more often. The
3302 * calls argument is an integer that sets or gets the partial delivery
3303 * point. Note also that the call will fail if the user attempts to set
3304 * this value larger than the socket receive buffer size.
3306 * Note that any single message having a length smaller than or equal to
3307 * the SCTP partial delivery point will be delivered in one single read
3308 * call as long as the user provided buffer is large enough to hold the
3309 * message.
3311 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3312 char __user *optval,
3313 unsigned int optlen)
3315 u32 val;
3317 if (optlen != sizeof(u32))
3318 return -EINVAL;
3319 if (get_user(val, (int __user *)optval))
3320 return -EFAULT;
3322 /* Note: We double the receive buffer from what the user sets
3323 * it to be, also initial rwnd is based on rcvbuf/2.
3325 if (val > (sk->sk_rcvbuf >> 1))
3326 return -EINVAL;
3328 sctp_sk(sk)->pd_point = val;
3330 return 0; /* is this the right error code? */
3334 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3336 * This option will allow a user to change the maximum burst of packets
3337 * that can be emitted by this association. Note that the default value
3338 * is 4, and some implementations may restrict this setting so that it
3339 * can only be lowered.
3341 * NOTE: This text doesn't seem right. Do this on a socket basis with
3342 * future associations inheriting the socket value.
3344 static int sctp_setsockopt_maxburst(struct sock *sk,
3345 char __user *optval,
3346 unsigned int optlen)
3348 struct sctp_assoc_value params;
3349 struct sctp_sock *sp;
3350 struct sctp_association *asoc;
3351 int val;
3352 int assoc_id = 0;
3354 if (optlen == sizeof(int)) {
3355 pr_warn_ratelimited(DEPRECATED
3356 "%s (pid %d) "
3357 "Use of int in max_burst socket option deprecated.\n"
3358 "Use struct sctp_assoc_value instead\n",
3359 current->comm, task_pid_nr(current));
3360 if (copy_from_user(&val, optval, optlen))
3361 return -EFAULT;
3362 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3363 if (copy_from_user(&params, optval, optlen))
3364 return -EFAULT;
3365 val = params.assoc_value;
3366 assoc_id = params.assoc_id;
3367 } else
3368 return -EINVAL;
3370 sp = sctp_sk(sk);
3372 if (assoc_id != 0) {
3373 asoc = sctp_id2assoc(sk, assoc_id);
3374 if (!asoc)
3375 return -EINVAL;
3376 asoc->max_burst = val;
3377 } else
3378 sp->max_burst = val;
3380 return 0;
3384 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3386 * This set option adds a chunk type that the user is requesting to be
3387 * received only in an authenticated way. Changes to the list of chunks
3388 * will only effect future associations on the socket.
3390 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3391 char __user *optval,
3392 unsigned int optlen)
3394 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3395 struct sctp_authchunk val;
3397 if (!ep->auth_enable)
3398 return -EACCES;
3400 if (optlen != sizeof(struct sctp_authchunk))
3401 return -EINVAL;
3402 if (copy_from_user(&val, optval, optlen))
3403 return -EFAULT;
3405 switch (val.sauth_chunk) {
3406 case SCTP_CID_INIT:
3407 case SCTP_CID_INIT_ACK:
3408 case SCTP_CID_SHUTDOWN_COMPLETE:
3409 case SCTP_CID_AUTH:
3410 return -EINVAL;
3413 /* add this chunk id to the endpoint */
3414 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3418 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3420 * This option gets or sets the list of HMAC algorithms that the local
3421 * endpoint requires the peer to use.
3423 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3424 char __user *optval,
3425 unsigned int optlen)
3427 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3428 struct sctp_hmacalgo *hmacs;
3429 u32 idents;
3430 int err;
3432 if (!ep->auth_enable)
3433 return -EACCES;
3435 if (optlen < sizeof(struct sctp_hmacalgo))
3436 return -EINVAL;
3438 hmacs = memdup_user(optval, optlen);
3439 if (IS_ERR(hmacs))
3440 return PTR_ERR(hmacs);
3442 idents = hmacs->shmac_num_idents;
3443 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3444 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3445 err = -EINVAL;
3446 goto out;
3449 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3450 out:
3451 kfree(hmacs);
3452 return err;
3456 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3458 * This option will set a shared secret key which is used to build an
3459 * association shared key.
3461 static int sctp_setsockopt_auth_key(struct sock *sk,
3462 char __user *optval,
3463 unsigned int optlen)
3465 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3466 struct sctp_authkey *authkey;
3467 struct sctp_association *asoc;
3468 int ret;
3470 if (!ep->auth_enable)
3471 return -EACCES;
3473 if (optlen <= sizeof(struct sctp_authkey))
3474 return -EINVAL;
3476 authkey = memdup_user(optval, optlen);
3477 if (IS_ERR(authkey))
3478 return PTR_ERR(authkey);
3480 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3481 ret = -EINVAL;
3482 goto out;
3485 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3486 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3487 ret = -EINVAL;
3488 goto out;
3491 ret = sctp_auth_set_key(ep, asoc, authkey);
3492 out:
3493 kzfree(authkey);
3494 return ret;
3498 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3500 * This option will get or set the active shared key to be used to build
3501 * the association shared key.
3503 static int sctp_setsockopt_active_key(struct sock *sk,
3504 char __user *optval,
3505 unsigned int optlen)
3507 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3508 struct sctp_authkeyid val;
3509 struct sctp_association *asoc;
3511 if (!ep->auth_enable)
3512 return -EACCES;
3514 if (optlen != sizeof(struct sctp_authkeyid))
3515 return -EINVAL;
3516 if (copy_from_user(&val, optval, optlen))
3517 return -EFAULT;
3519 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3520 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3521 return -EINVAL;
3523 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3527 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3529 * This set option will delete a shared secret key from use.
3531 static int sctp_setsockopt_del_key(struct sock *sk,
3532 char __user *optval,
3533 unsigned int optlen)
3535 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3536 struct sctp_authkeyid val;
3537 struct sctp_association *asoc;
3539 if (!ep->auth_enable)
3540 return -EACCES;
3542 if (optlen != sizeof(struct sctp_authkeyid))
3543 return -EINVAL;
3544 if (copy_from_user(&val, optval, optlen))
3545 return -EFAULT;
3547 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3548 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3549 return -EINVAL;
3551 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3556 * 8.1.23 SCTP_AUTO_ASCONF
3558 * This option will enable or disable the use of the automatic generation of
3559 * ASCONF chunks to add and delete addresses to an existing association. Note
3560 * that this option has two caveats namely: a) it only affects sockets that
3561 * are bound to all addresses available to the SCTP stack, and b) the system
3562 * administrator may have an overriding control that turns the ASCONF feature
3563 * off no matter what setting the socket option may have.
3564 * This option expects an integer boolean flag, where a non-zero value turns on
3565 * the option, and a zero value turns off the option.
3566 * Note. In this implementation, socket operation overrides default parameter
3567 * being set by sysctl as well as FreeBSD implementation
3569 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3570 unsigned int optlen)
3572 int val;
3573 struct sctp_sock *sp = sctp_sk(sk);
3575 if (optlen < sizeof(int))
3576 return -EINVAL;
3577 if (get_user(val, (int __user *)optval))
3578 return -EFAULT;
3579 if (!sctp_is_ep_boundall(sk) && val)
3580 return -EINVAL;
3581 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3582 return 0;
3584 if (val == 0 && sp->do_auto_asconf) {
3585 list_del(&sp->auto_asconf_list);
3586 sp->do_auto_asconf = 0;
3587 } else if (val && !sp->do_auto_asconf) {
3588 list_add_tail(&sp->auto_asconf_list,
3589 &sock_net(sk)->sctp.auto_asconf_splist);
3590 sp->do_auto_asconf = 1;
3592 return 0;
3596 * SCTP_PEER_ADDR_THLDS
3598 * This option allows us to alter the partially failed threshold for one or all
3599 * transports in an association. See Section 6.1 of:
3600 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3602 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3603 char __user *optval,
3604 unsigned int optlen)
3606 struct sctp_paddrthlds val;
3607 struct sctp_transport *trans;
3608 struct sctp_association *asoc;
3610 if (optlen < sizeof(struct sctp_paddrthlds))
3611 return -EINVAL;
3612 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3613 sizeof(struct sctp_paddrthlds)))
3614 return -EFAULT;
3617 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3618 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3619 if (!asoc)
3620 return -ENOENT;
3621 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3622 transports) {
3623 if (val.spt_pathmaxrxt)
3624 trans->pathmaxrxt = val.spt_pathmaxrxt;
3625 trans->pf_retrans = val.spt_pathpfthld;
3628 if (val.spt_pathmaxrxt)
3629 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3630 asoc->pf_retrans = val.spt_pathpfthld;
3631 } else {
3632 trans = sctp_addr_id2transport(sk, &val.spt_address,
3633 val.spt_assoc_id);
3634 if (!trans)
3635 return -ENOENT;
3637 if (val.spt_pathmaxrxt)
3638 trans->pathmaxrxt = val.spt_pathmaxrxt;
3639 trans->pf_retrans = val.spt_pathpfthld;
3642 return 0;
3645 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3646 char __user *optval,
3647 unsigned int optlen)
3649 int val;
3651 if (optlen < sizeof(int))
3652 return -EINVAL;
3653 if (get_user(val, (int __user *) optval))
3654 return -EFAULT;
3656 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3658 return 0;
3661 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3662 char __user *optval,
3663 unsigned int optlen)
3665 int val;
3667 if (optlen < sizeof(int))
3668 return -EINVAL;
3669 if (get_user(val, (int __user *) optval))
3670 return -EFAULT;
3672 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3674 return 0;
3677 /* API 6.2 setsockopt(), getsockopt()
3679 * Applications use setsockopt() and getsockopt() to set or retrieve
3680 * socket options. Socket options are used to change the default
3681 * behavior of sockets calls. They are described in Section 7.
3683 * The syntax is:
3685 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3686 * int __user *optlen);
3687 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3688 * int optlen);
3690 * sd - the socket descript.
3691 * level - set to IPPROTO_SCTP for all SCTP options.
3692 * optname - the option name.
3693 * optval - the buffer to store the value of the option.
3694 * optlen - the size of the buffer.
3696 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3697 char __user *optval, unsigned int optlen)
3699 int retval = 0;
3701 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3703 /* I can hardly begin to describe how wrong this is. This is
3704 * so broken as to be worse than useless. The API draft
3705 * REALLY is NOT helpful here... I am not convinced that the
3706 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3707 * are at all well-founded.
3709 if (level != SOL_SCTP) {
3710 struct sctp_af *af = sctp_sk(sk)->pf->af;
3711 retval = af->setsockopt(sk, level, optname, optval, optlen);
3712 goto out_nounlock;
3715 lock_sock(sk);
3717 switch (optname) {
3718 case SCTP_SOCKOPT_BINDX_ADD:
3719 /* 'optlen' is the size of the addresses buffer. */
3720 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3721 optlen, SCTP_BINDX_ADD_ADDR);
3722 break;
3724 case SCTP_SOCKOPT_BINDX_REM:
3725 /* 'optlen' is the size of the addresses buffer. */
3726 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3727 optlen, SCTP_BINDX_REM_ADDR);
3728 break;
3730 case SCTP_SOCKOPT_CONNECTX_OLD:
3731 /* 'optlen' is the size of the addresses buffer. */
3732 retval = sctp_setsockopt_connectx_old(sk,
3733 (struct sockaddr __user *)optval,
3734 optlen);
3735 break;
3737 case SCTP_SOCKOPT_CONNECTX:
3738 /* 'optlen' is the size of the addresses buffer. */
3739 retval = sctp_setsockopt_connectx(sk,
3740 (struct sockaddr __user *)optval,
3741 optlen);
3742 break;
3744 case SCTP_DISABLE_FRAGMENTS:
3745 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3746 break;
3748 case SCTP_EVENTS:
3749 retval = sctp_setsockopt_events(sk, optval, optlen);
3750 break;
3752 case SCTP_AUTOCLOSE:
3753 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3754 break;
3756 case SCTP_PEER_ADDR_PARAMS:
3757 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3758 break;
3760 case SCTP_DELAYED_SACK:
3761 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3762 break;
3763 case SCTP_PARTIAL_DELIVERY_POINT:
3764 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3765 break;
3767 case SCTP_INITMSG:
3768 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3769 break;
3770 case SCTP_DEFAULT_SEND_PARAM:
3771 retval = sctp_setsockopt_default_send_param(sk, optval,
3772 optlen);
3773 break;
3774 case SCTP_DEFAULT_SNDINFO:
3775 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3776 break;
3777 case SCTP_PRIMARY_ADDR:
3778 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3779 break;
3780 case SCTP_SET_PEER_PRIMARY_ADDR:
3781 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3782 break;
3783 case SCTP_NODELAY:
3784 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3785 break;
3786 case SCTP_RTOINFO:
3787 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3788 break;
3789 case SCTP_ASSOCINFO:
3790 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3791 break;
3792 case SCTP_I_WANT_MAPPED_V4_ADDR:
3793 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3794 break;
3795 case SCTP_MAXSEG:
3796 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3797 break;
3798 case SCTP_ADAPTATION_LAYER:
3799 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3800 break;
3801 case SCTP_CONTEXT:
3802 retval = sctp_setsockopt_context(sk, optval, optlen);
3803 break;
3804 case SCTP_FRAGMENT_INTERLEAVE:
3805 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3806 break;
3807 case SCTP_MAX_BURST:
3808 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3809 break;
3810 case SCTP_AUTH_CHUNK:
3811 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3812 break;
3813 case SCTP_HMAC_IDENT:
3814 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3815 break;
3816 case SCTP_AUTH_KEY:
3817 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3818 break;
3819 case SCTP_AUTH_ACTIVE_KEY:
3820 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3821 break;
3822 case SCTP_AUTH_DELETE_KEY:
3823 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3824 break;
3825 case SCTP_AUTO_ASCONF:
3826 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3827 break;
3828 case SCTP_PEER_ADDR_THLDS:
3829 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3830 break;
3831 case SCTP_RECVRCVINFO:
3832 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3833 break;
3834 case SCTP_RECVNXTINFO:
3835 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3836 break;
3837 default:
3838 retval = -ENOPROTOOPT;
3839 break;
3842 release_sock(sk);
3844 out_nounlock:
3845 return retval;
3848 /* API 3.1.6 connect() - UDP Style Syntax
3850 * An application may use the connect() call in the UDP model to initiate an
3851 * association without sending data.
3853 * The syntax is:
3855 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3857 * sd: the socket descriptor to have a new association added to.
3859 * nam: the address structure (either struct sockaddr_in or struct
3860 * sockaddr_in6 defined in RFC2553 [7]).
3862 * len: the size of the address.
3864 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3865 int addr_len)
3867 int err = 0;
3868 struct sctp_af *af;
3870 lock_sock(sk);
3872 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3873 addr, addr_len);
3875 /* Validate addr_len before calling common connect/connectx routine. */
3876 af = sctp_get_af_specific(addr->sa_family);
3877 if (!af || addr_len < af->sockaddr_len) {
3878 err = -EINVAL;
3879 } else {
3880 /* Pass correct addr len to common routine (so it knows there
3881 * is only one address being passed.
3883 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3886 release_sock(sk);
3887 return err;
3890 /* FIXME: Write comments. */
3891 static int sctp_disconnect(struct sock *sk, int flags)
3893 return -EOPNOTSUPP; /* STUB */
3896 /* 4.1.4 accept() - TCP Style Syntax
3898 * Applications use accept() call to remove an established SCTP
3899 * association from the accept queue of the endpoint. A new socket
3900 * descriptor will be returned from accept() to represent the newly
3901 * formed association.
3903 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3905 struct sctp_sock *sp;
3906 struct sctp_endpoint *ep;
3907 struct sock *newsk = NULL;
3908 struct sctp_association *asoc;
3909 long timeo;
3910 int error = 0;
3912 lock_sock(sk);
3914 sp = sctp_sk(sk);
3915 ep = sp->ep;
3917 if (!sctp_style(sk, TCP)) {
3918 error = -EOPNOTSUPP;
3919 goto out;
3922 if (!sctp_sstate(sk, LISTENING)) {
3923 error = -EINVAL;
3924 goto out;
3927 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3929 error = sctp_wait_for_accept(sk, timeo);
3930 if (error)
3931 goto out;
3933 /* We treat the list of associations on the endpoint as the accept
3934 * queue and pick the first association on the list.
3936 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3938 newsk = sp->pf->create_accept_sk(sk, asoc);
3939 if (!newsk) {
3940 error = -ENOMEM;
3941 goto out;
3944 /* Populate the fields of the newsk from the oldsk and migrate the
3945 * asoc to the newsk.
3947 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3949 out:
3950 release_sock(sk);
3951 *err = error;
3952 return newsk;
3955 /* The SCTP ioctl handler. */
3956 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3958 int rc = -ENOTCONN;
3960 lock_sock(sk);
3963 * SEQPACKET-style sockets in LISTENING state are valid, for
3964 * SCTP, so only discard TCP-style sockets in LISTENING state.
3966 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3967 goto out;
3969 switch (cmd) {
3970 case SIOCINQ: {
3971 struct sk_buff *skb;
3972 unsigned int amount = 0;
3974 skb = skb_peek(&sk->sk_receive_queue);
3975 if (skb != NULL) {
3977 * We will only return the amount of this packet since
3978 * that is all that will be read.
3980 amount = skb->len;
3982 rc = put_user(amount, (int __user *)arg);
3983 break;
3985 default:
3986 rc = -ENOIOCTLCMD;
3987 break;
3989 out:
3990 release_sock(sk);
3991 return rc;
3994 /* This is the function which gets called during socket creation to
3995 * initialized the SCTP-specific portion of the sock.
3996 * The sock structure should already be zero-filled memory.
3998 static int sctp_init_sock(struct sock *sk)
4000 struct net *net = sock_net(sk);
4001 struct sctp_sock *sp;
4003 pr_debug("%s: sk:%p\n", __func__, sk);
4005 sp = sctp_sk(sk);
4007 /* Initialize the SCTP per socket area. */
4008 switch (sk->sk_type) {
4009 case SOCK_SEQPACKET:
4010 sp->type = SCTP_SOCKET_UDP;
4011 break;
4012 case SOCK_STREAM:
4013 sp->type = SCTP_SOCKET_TCP;
4014 break;
4015 default:
4016 return -ESOCKTNOSUPPORT;
4019 /* Initialize default send parameters. These parameters can be
4020 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4022 sp->default_stream = 0;
4023 sp->default_ppid = 0;
4024 sp->default_flags = 0;
4025 sp->default_context = 0;
4026 sp->default_timetolive = 0;
4028 sp->default_rcv_context = 0;
4029 sp->max_burst = net->sctp.max_burst;
4031 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4033 /* Initialize default setup parameters. These parameters
4034 * can be modified with the SCTP_INITMSG socket option or
4035 * overridden by the SCTP_INIT CMSG.
4037 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4038 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4039 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4040 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4042 /* Initialize default RTO related parameters. These parameters can
4043 * be modified for with the SCTP_RTOINFO socket option.
4045 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4046 sp->rtoinfo.srto_max = net->sctp.rto_max;
4047 sp->rtoinfo.srto_min = net->sctp.rto_min;
4049 /* Initialize default association related parameters. These parameters
4050 * can be modified with the SCTP_ASSOCINFO socket option.
4052 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4053 sp->assocparams.sasoc_number_peer_destinations = 0;
4054 sp->assocparams.sasoc_peer_rwnd = 0;
4055 sp->assocparams.sasoc_local_rwnd = 0;
4056 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4058 /* Initialize default event subscriptions. By default, all the
4059 * options are off.
4061 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4063 /* Default Peer Address Parameters. These defaults can
4064 * be modified via SCTP_PEER_ADDR_PARAMS
4066 sp->hbinterval = net->sctp.hb_interval;
4067 sp->pathmaxrxt = net->sctp.max_retrans_path;
4068 sp->pathmtu = 0; /* allow default discovery */
4069 sp->sackdelay = net->sctp.sack_timeout;
4070 sp->sackfreq = 2;
4071 sp->param_flags = SPP_HB_ENABLE |
4072 SPP_PMTUD_ENABLE |
4073 SPP_SACKDELAY_ENABLE;
4075 /* If enabled no SCTP message fragmentation will be performed.
4076 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4078 sp->disable_fragments = 0;
4080 /* Enable Nagle algorithm by default. */
4081 sp->nodelay = 0;
4083 sp->recvrcvinfo = 0;
4084 sp->recvnxtinfo = 0;
4086 /* Enable by default. */
4087 sp->v4mapped = 1;
4089 /* Auto-close idle associations after the configured
4090 * number of seconds. A value of 0 disables this
4091 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4092 * for UDP-style sockets only.
4094 sp->autoclose = 0;
4096 /* User specified fragmentation limit. */
4097 sp->user_frag = 0;
4099 sp->adaptation_ind = 0;
4101 sp->pf = sctp_get_pf_specific(sk->sk_family);
4103 /* Control variables for partial data delivery. */
4104 atomic_set(&sp->pd_mode, 0);
4105 skb_queue_head_init(&sp->pd_lobby);
4106 sp->frag_interleave = 0;
4108 /* Create a per socket endpoint structure. Even if we
4109 * change the data structure relationships, this may still
4110 * be useful for storing pre-connect address information.
4112 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4113 if (!sp->ep)
4114 return -ENOMEM;
4116 sp->hmac = NULL;
4118 sk->sk_destruct = sctp_destruct_sock;
4120 SCTP_DBG_OBJCNT_INC(sock);
4122 local_bh_disable();
4123 percpu_counter_inc(&sctp_sockets_allocated);
4124 sock_prot_inuse_add(net, sk->sk_prot, 1);
4125 if (net->sctp.default_auto_asconf) {
4126 list_add_tail(&sp->auto_asconf_list,
4127 &net->sctp.auto_asconf_splist);
4128 sp->do_auto_asconf = 1;
4129 } else
4130 sp->do_auto_asconf = 0;
4131 local_bh_enable();
4133 return 0;
4136 /* Cleanup any SCTP per socket resources. */
4137 static void sctp_destroy_sock(struct sock *sk)
4139 struct sctp_sock *sp;
4141 pr_debug("%s: sk:%p\n", __func__, sk);
4143 /* Release our hold on the endpoint. */
4144 sp = sctp_sk(sk);
4145 /* This could happen during socket init, thus we bail out
4146 * early, since the rest of the below is not setup either.
4148 if (sp->ep == NULL)
4149 return;
4151 if (sp->do_auto_asconf) {
4152 sp->do_auto_asconf = 0;
4153 list_del(&sp->auto_asconf_list);
4155 sctp_endpoint_free(sp->ep);
4156 local_bh_disable();
4157 percpu_counter_dec(&sctp_sockets_allocated);
4158 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4159 local_bh_enable();
4162 /* Triggered when there are no references on the socket anymore */
4163 static void sctp_destruct_sock(struct sock *sk)
4165 struct sctp_sock *sp = sctp_sk(sk);
4167 /* Free up the HMAC transform. */
4168 crypto_free_hash(sp->hmac);
4170 inet_sock_destruct(sk);
4173 /* API 4.1.7 shutdown() - TCP Style Syntax
4174 * int shutdown(int socket, int how);
4176 * sd - the socket descriptor of the association to be closed.
4177 * how - Specifies the type of shutdown. The values are
4178 * as follows:
4179 * SHUT_RD
4180 * Disables further receive operations. No SCTP
4181 * protocol action is taken.
4182 * SHUT_WR
4183 * Disables further send operations, and initiates
4184 * the SCTP shutdown sequence.
4185 * SHUT_RDWR
4186 * Disables further send and receive operations
4187 * and initiates the SCTP shutdown sequence.
4189 static void sctp_shutdown(struct sock *sk, int how)
4191 struct net *net = sock_net(sk);
4192 struct sctp_endpoint *ep;
4193 struct sctp_association *asoc;
4195 if (!sctp_style(sk, TCP))
4196 return;
4198 if (how & SEND_SHUTDOWN) {
4199 ep = sctp_sk(sk)->ep;
4200 if (!list_empty(&ep->asocs)) {
4201 asoc = list_entry(ep->asocs.next,
4202 struct sctp_association, asocs);
4203 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4208 /* 7.2.1 Association Status (SCTP_STATUS)
4210 * Applications can retrieve current status information about an
4211 * association, including association state, peer receiver window size,
4212 * number of unacked data chunks, and number of data chunks pending
4213 * receipt. This information is read-only.
4215 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4216 char __user *optval,
4217 int __user *optlen)
4219 struct sctp_status status;
4220 struct sctp_association *asoc = NULL;
4221 struct sctp_transport *transport;
4222 sctp_assoc_t associd;
4223 int retval = 0;
4225 if (len < sizeof(status)) {
4226 retval = -EINVAL;
4227 goto out;
4230 len = sizeof(status);
4231 if (copy_from_user(&status, optval, len)) {
4232 retval = -EFAULT;
4233 goto out;
4236 associd = status.sstat_assoc_id;
4237 asoc = sctp_id2assoc(sk, associd);
4238 if (!asoc) {
4239 retval = -EINVAL;
4240 goto out;
4243 transport = asoc->peer.primary_path;
4245 status.sstat_assoc_id = sctp_assoc2id(asoc);
4246 status.sstat_state = sctp_assoc_to_state(asoc);
4247 status.sstat_rwnd = asoc->peer.rwnd;
4248 status.sstat_unackdata = asoc->unack_data;
4250 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4251 status.sstat_instrms = asoc->c.sinit_max_instreams;
4252 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4253 status.sstat_fragmentation_point = asoc->frag_point;
4254 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4255 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4256 transport->af_specific->sockaddr_len);
4257 /* Map ipv4 address into v4-mapped-on-v6 address. */
4258 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4259 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4260 status.sstat_primary.spinfo_state = transport->state;
4261 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4262 status.sstat_primary.spinfo_srtt = transport->srtt;
4263 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4264 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4266 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4267 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4269 if (put_user(len, optlen)) {
4270 retval = -EFAULT;
4271 goto out;
4274 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4275 __func__, len, status.sstat_state, status.sstat_rwnd,
4276 status.sstat_assoc_id);
4278 if (copy_to_user(optval, &status, len)) {
4279 retval = -EFAULT;
4280 goto out;
4283 out:
4284 return retval;
4288 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4290 * Applications can retrieve information about a specific peer address
4291 * of an association, including its reachability state, congestion
4292 * window, and retransmission timer values. This information is
4293 * read-only.
4295 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4296 char __user *optval,
4297 int __user *optlen)
4299 struct sctp_paddrinfo pinfo;
4300 struct sctp_transport *transport;
4301 int retval = 0;
4303 if (len < sizeof(pinfo)) {
4304 retval = -EINVAL;
4305 goto out;
4308 len = sizeof(pinfo);
4309 if (copy_from_user(&pinfo, optval, len)) {
4310 retval = -EFAULT;
4311 goto out;
4314 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4315 pinfo.spinfo_assoc_id);
4316 if (!transport)
4317 return -EINVAL;
4319 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4320 pinfo.spinfo_state = transport->state;
4321 pinfo.spinfo_cwnd = transport->cwnd;
4322 pinfo.spinfo_srtt = transport->srtt;
4323 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4324 pinfo.spinfo_mtu = transport->pathmtu;
4326 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4327 pinfo.spinfo_state = SCTP_ACTIVE;
4329 if (put_user(len, optlen)) {
4330 retval = -EFAULT;
4331 goto out;
4334 if (copy_to_user(optval, &pinfo, len)) {
4335 retval = -EFAULT;
4336 goto out;
4339 out:
4340 return retval;
4343 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4345 * This option is a on/off flag. If enabled no SCTP message
4346 * fragmentation will be performed. Instead if a message being sent
4347 * exceeds the current PMTU size, the message will NOT be sent and
4348 * instead a error will be indicated to the user.
4350 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4351 char __user *optval, int __user *optlen)
4353 int val;
4355 if (len < sizeof(int))
4356 return -EINVAL;
4358 len = sizeof(int);
4359 val = (sctp_sk(sk)->disable_fragments == 1);
4360 if (put_user(len, optlen))
4361 return -EFAULT;
4362 if (copy_to_user(optval, &val, len))
4363 return -EFAULT;
4364 return 0;
4367 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4369 * This socket option is used to specify various notifications and
4370 * ancillary data the user wishes to receive.
4372 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4373 int __user *optlen)
4375 if (len <= 0)
4376 return -EINVAL;
4377 if (len > sizeof(struct sctp_event_subscribe))
4378 len = sizeof(struct sctp_event_subscribe);
4379 if (put_user(len, optlen))
4380 return -EFAULT;
4381 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4382 return -EFAULT;
4383 return 0;
4386 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4388 * This socket option is applicable to the UDP-style socket only. When
4389 * set it will cause associations that are idle for more than the
4390 * specified number of seconds to automatically close. An association
4391 * being idle is defined an association that has NOT sent or received
4392 * user data. The special value of '0' indicates that no automatic
4393 * close of any associations should be performed. The option expects an
4394 * integer defining the number of seconds of idle time before an
4395 * association is closed.
4397 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4399 /* Applicable to UDP-style socket only */
4400 if (sctp_style(sk, TCP))
4401 return -EOPNOTSUPP;
4402 if (len < sizeof(int))
4403 return -EINVAL;
4404 len = sizeof(int);
4405 if (put_user(len, optlen))
4406 return -EFAULT;
4407 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4408 return -EFAULT;
4409 return 0;
4412 /* Helper routine to branch off an association to a new socket. */
4413 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4415 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4416 struct sctp_sock *sp = sctp_sk(sk);
4417 struct socket *sock;
4418 int err = 0;
4420 if (!asoc)
4421 return -EINVAL;
4423 /* An association cannot be branched off from an already peeled-off
4424 * socket, nor is this supported for tcp style sockets.
4426 if (!sctp_style(sk, UDP))
4427 return -EINVAL;
4429 /* Create a new socket. */
4430 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4431 if (err < 0)
4432 return err;
4434 sctp_copy_sock(sock->sk, sk, asoc);
4436 /* Make peeled-off sockets more like 1-1 accepted sockets.
4437 * Set the daddr and initialize id to something more random
4439 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4441 /* Populate the fields of the newsk from the oldsk and migrate the
4442 * asoc to the newsk.
4444 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4446 *sockp = sock;
4448 return err;
4450 EXPORT_SYMBOL(sctp_do_peeloff);
4452 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4454 sctp_peeloff_arg_t peeloff;
4455 struct socket *newsock;
4456 struct file *newfile;
4457 int retval = 0;
4459 if (len < sizeof(sctp_peeloff_arg_t))
4460 return -EINVAL;
4461 len = sizeof(sctp_peeloff_arg_t);
4462 if (copy_from_user(&peeloff, optval, len))
4463 return -EFAULT;
4465 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4466 if (retval < 0)
4467 goto out;
4469 /* Map the socket to an unused fd that can be returned to the user. */
4470 retval = get_unused_fd_flags(0);
4471 if (retval < 0) {
4472 sock_release(newsock);
4473 goto out;
4476 newfile = sock_alloc_file(newsock, 0, NULL);
4477 if (unlikely(IS_ERR(newfile))) {
4478 put_unused_fd(retval);
4479 sock_release(newsock);
4480 return PTR_ERR(newfile);
4483 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4484 retval);
4486 /* Return the fd mapped to the new socket. */
4487 if (put_user(len, optlen)) {
4488 fput(newfile);
4489 put_unused_fd(retval);
4490 return -EFAULT;
4492 peeloff.sd = retval;
4493 if (copy_to_user(optval, &peeloff, len)) {
4494 fput(newfile);
4495 put_unused_fd(retval);
4496 return -EFAULT;
4498 fd_install(retval, newfile);
4499 out:
4500 return retval;
4503 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4505 * Applications can enable or disable heartbeats for any peer address of
4506 * an association, modify an address's heartbeat interval, force a
4507 * heartbeat to be sent immediately, and adjust the address's maximum
4508 * number of retransmissions sent before an address is considered
4509 * unreachable. The following structure is used to access and modify an
4510 * address's parameters:
4512 * struct sctp_paddrparams {
4513 * sctp_assoc_t spp_assoc_id;
4514 * struct sockaddr_storage spp_address;
4515 * uint32_t spp_hbinterval;
4516 * uint16_t spp_pathmaxrxt;
4517 * uint32_t spp_pathmtu;
4518 * uint32_t spp_sackdelay;
4519 * uint32_t spp_flags;
4520 * };
4522 * spp_assoc_id - (one-to-many style socket) This is filled in the
4523 * application, and identifies the association for
4524 * this query.
4525 * spp_address - This specifies which address is of interest.
4526 * spp_hbinterval - This contains the value of the heartbeat interval,
4527 * in milliseconds. If a value of zero
4528 * is present in this field then no changes are to
4529 * be made to this parameter.
4530 * spp_pathmaxrxt - This contains the maximum number of
4531 * retransmissions before this address shall be
4532 * considered unreachable. If a value of zero
4533 * is present in this field then no changes are to
4534 * be made to this parameter.
4535 * spp_pathmtu - When Path MTU discovery is disabled the value
4536 * specified here will be the "fixed" path mtu.
4537 * Note that if the spp_address field is empty
4538 * then all associations on this address will
4539 * have this fixed path mtu set upon them.
4541 * spp_sackdelay - When delayed sack is enabled, this value specifies
4542 * the number of milliseconds that sacks will be delayed
4543 * for. This value will apply to all addresses of an
4544 * association if the spp_address field is empty. Note
4545 * also, that if delayed sack is enabled and this
4546 * value is set to 0, no change is made to the last
4547 * recorded delayed sack timer value.
4549 * spp_flags - These flags are used to control various features
4550 * on an association. The flag field may contain
4551 * zero or more of the following options.
4553 * SPP_HB_ENABLE - Enable heartbeats on the
4554 * specified address. Note that if the address
4555 * field is empty all addresses for the association
4556 * have heartbeats enabled upon them.
4558 * SPP_HB_DISABLE - Disable heartbeats on the
4559 * speicifed address. Note that if the address
4560 * field is empty all addresses for the association
4561 * will have their heartbeats disabled. Note also
4562 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4563 * mutually exclusive, only one of these two should
4564 * be specified. Enabling both fields will have
4565 * undetermined results.
4567 * SPP_HB_DEMAND - Request a user initiated heartbeat
4568 * to be made immediately.
4570 * SPP_PMTUD_ENABLE - This field will enable PMTU
4571 * discovery upon the specified address. Note that
4572 * if the address feild is empty then all addresses
4573 * on the association are effected.
4575 * SPP_PMTUD_DISABLE - This field will disable PMTU
4576 * discovery upon the specified address. Note that
4577 * if the address feild is empty then all addresses
4578 * on the association are effected. Not also that
4579 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4580 * exclusive. Enabling both will have undetermined
4581 * results.
4583 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4584 * on delayed sack. The time specified in spp_sackdelay
4585 * is used to specify the sack delay for this address. Note
4586 * that if spp_address is empty then all addresses will
4587 * enable delayed sack and take on the sack delay
4588 * value specified in spp_sackdelay.
4589 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4590 * off delayed sack. If the spp_address field is blank then
4591 * delayed sack is disabled for the entire association. Note
4592 * also that this field is mutually exclusive to
4593 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4594 * results.
4596 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4597 char __user *optval, int __user *optlen)
4599 struct sctp_paddrparams params;
4600 struct sctp_transport *trans = NULL;
4601 struct sctp_association *asoc = NULL;
4602 struct sctp_sock *sp = sctp_sk(sk);
4604 if (len < sizeof(struct sctp_paddrparams))
4605 return -EINVAL;
4606 len = sizeof(struct sctp_paddrparams);
4607 if (copy_from_user(&params, optval, len))
4608 return -EFAULT;
4610 /* If an address other than INADDR_ANY is specified, and
4611 * no transport is found, then the request is invalid.
4613 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4614 trans = sctp_addr_id2transport(sk, &params.spp_address,
4615 params.spp_assoc_id);
4616 if (!trans) {
4617 pr_debug("%s: failed no transport\n", __func__);
4618 return -EINVAL;
4622 /* Get association, if assoc_id != 0 and the socket is a one
4623 * to many style socket, and an association was not found, then
4624 * the id was invalid.
4626 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4627 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4628 pr_debug("%s: failed no association\n", __func__);
4629 return -EINVAL;
4632 if (trans) {
4633 /* Fetch transport values. */
4634 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4635 params.spp_pathmtu = trans->pathmtu;
4636 params.spp_pathmaxrxt = trans->pathmaxrxt;
4637 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4639 /*draft-11 doesn't say what to return in spp_flags*/
4640 params.spp_flags = trans->param_flags;
4641 } else if (asoc) {
4642 /* Fetch association values. */
4643 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4644 params.spp_pathmtu = asoc->pathmtu;
4645 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4646 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4648 /*draft-11 doesn't say what to return in spp_flags*/
4649 params.spp_flags = asoc->param_flags;
4650 } else {
4651 /* Fetch socket values. */
4652 params.spp_hbinterval = sp->hbinterval;
4653 params.spp_pathmtu = sp->pathmtu;
4654 params.spp_sackdelay = sp->sackdelay;
4655 params.spp_pathmaxrxt = sp->pathmaxrxt;
4657 /*draft-11 doesn't say what to return in spp_flags*/
4658 params.spp_flags = sp->param_flags;
4661 if (copy_to_user(optval, &params, len))
4662 return -EFAULT;
4664 if (put_user(len, optlen))
4665 return -EFAULT;
4667 return 0;
4671 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4673 * This option will effect the way delayed acks are performed. This
4674 * option allows you to get or set the delayed ack time, in
4675 * milliseconds. It also allows changing the delayed ack frequency.
4676 * Changing the frequency to 1 disables the delayed sack algorithm. If
4677 * the assoc_id is 0, then this sets or gets the endpoints default
4678 * values. If the assoc_id field is non-zero, then the set or get
4679 * effects the specified association for the one to many model (the
4680 * assoc_id field is ignored by the one to one model). Note that if
4681 * sack_delay or sack_freq are 0 when setting this option, then the
4682 * current values will remain unchanged.
4684 * struct sctp_sack_info {
4685 * sctp_assoc_t sack_assoc_id;
4686 * uint32_t sack_delay;
4687 * uint32_t sack_freq;
4688 * };
4690 * sack_assoc_id - This parameter, indicates which association the user
4691 * is performing an action upon. Note that if this field's value is
4692 * zero then the endpoints default value is changed (effecting future
4693 * associations only).
4695 * sack_delay - This parameter contains the number of milliseconds that
4696 * the user is requesting the delayed ACK timer be set to. Note that
4697 * this value is defined in the standard to be between 200 and 500
4698 * milliseconds.
4700 * sack_freq - This parameter contains the number of packets that must
4701 * be received before a sack is sent without waiting for the delay
4702 * timer to expire. The default value for this is 2, setting this
4703 * value to 1 will disable the delayed sack algorithm.
4705 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4706 char __user *optval,
4707 int __user *optlen)
4709 struct sctp_sack_info params;
4710 struct sctp_association *asoc = NULL;
4711 struct sctp_sock *sp = sctp_sk(sk);
4713 if (len >= sizeof(struct sctp_sack_info)) {
4714 len = sizeof(struct sctp_sack_info);
4716 if (copy_from_user(&params, optval, len))
4717 return -EFAULT;
4718 } else if (len == sizeof(struct sctp_assoc_value)) {
4719 pr_warn_ratelimited(DEPRECATED
4720 "%s (pid %d) "
4721 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4722 "Use struct sctp_sack_info instead\n",
4723 current->comm, task_pid_nr(current));
4724 if (copy_from_user(&params, optval, len))
4725 return -EFAULT;
4726 } else
4727 return -EINVAL;
4729 /* Get association, if sack_assoc_id != 0 and the socket is a one
4730 * to many style socket, and an association was not found, then
4731 * the id was invalid.
4733 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4734 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4735 return -EINVAL;
4737 if (asoc) {
4738 /* Fetch association values. */
4739 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4740 params.sack_delay = jiffies_to_msecs(
4741 asoc->sackdelay);
4742 params.sack_freq = asoc->sackfreq;
4744 } else {
4745 params.sack_delay = 0;
4746 params.sack_freq = 1;
4748 } else {
4749 /* Fetch socket values. */
4750 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4751 params.sack_delay = sp->sackdelay;
4752 params.sack_freq = sp->sackfreq;
4753 } else {
4754 params.sack_delay = 0;
4755 params.sack_freq = 1;
4759 if (copy_to_user(optval, &params, len))
4760 return -EFAULT;
4762 if (put_user(len, optlen))
4763 return -EFAULT;
4765 return 0;
4768 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4770 * Applications can specify protocol parameters for the default association
4771 * initialization. The option name argument to setsockopt() and getsockopt()
4772 * is SCTP_INITMSG.
4774 * Setting initialization parameters is effective only on an unconnected
4775 * socket (for UDP-style sockets only future associations are effected
4776 * by the change). With TCP-style sockets, this option is inherited by
4777 * sockets derived from a listener socket.
4779 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4781 if (len < sizeof(struct sctp_initmsg))
4782 return -EINVAL;
4783 len = sizeof(struct sctp_initmsg);
4784 if (put_user(len, optlen))
4785 return -EFAULT;
4786 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4787 return -EFAULT;
4788 return 0;
4792 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4793 char __user *optval, int __user *optlen)
4795 struct sctp_association *asoc;
4796 int cnt = 0;
4797 struct sctp_getaddrs getaddrs;
4798 struct sctp_transport *from;
4799 void __user *to;
4800 union sctp_addr temp;
4801 struct sctp_sock *sp = sctp_sk(sk);
4802 int addrlen;
4803 size_t space_left;
4804 int bytes_copied;
4806 if (len < sizeof(struct sctp_getaddrs))
4807 return -EINVAL;
4809 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4810 return -EFAULT;
4812 /* For UDP-style sockets, id specifies the association to query. */
4813 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4814 if (!asoc)
4815 return -EINVAL;
4817 to = optval + offsetof(struct sctp_getaddrs, addrs);
4818 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4820 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4821 transports) {
4822 memcpy(&temp, &from->ipaddr, sizeof(temp));
4823 addrlen = sctp_get_pf_specific(sk->sk_family)
4824 ->addr_to_user(sp, &temp);
4825 if (space_left < addrlen)
4826 return -ENOMEM;
4827 if (copy_to_user(to, &temp, addrlen))
4828 return -EFAULT;
4829 to += addrlen;
4830 cnt++;
4831 space_left -= addrlen;
4834 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4835 return -EFAULT;
4836 bytes_copied = ((char __user *)to) - optval;
4837 if (put_user(bytes_copied, optlen))
4838 return -EFAULT;
4840 return 0;
4843 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4844 size_t space_left, int *bytes_copied)
4846 struct sctp_sockaddr_entry *addr;
4847 union sctp_addr temp;
4848 int cnt = 0;
4849 int addrlen;
4850 struct net *net = sock_net(sk);
4852 rcu_read_lock();
4853 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4854 if (!addr->valid)
4855 continue;
4857 if ((PF_INET == sk->sk_family) &&
4858 (AF_INET6 == addr->a.sa.sa_family))
4859 continue;
4860 if ((PF_INET6 == sk->sk_family) &&
4861 inet_v6_ipv6only(sk) &&
4862 (AF_INET == addr->a.sa.sa_family))
4863 continue;
4864 memcpy(&temp, &addr->a, sizeof(temp));
4865 if (!temp.v4.sin_port)
4866 temp.v4.sin_port = htons(port);
4868 addrlen = sctp_get_pf_specific(sk->sk_family)
4869 ->addr_to_user(sctp_sk(sk), &temp);
4871 if (space_left < addrlen) {
4872 cnt = -ENOMEM;
4873 break;
4875 memcpy(to, &temp, addrlen);
4877 to += addrlen;
4878 cnt++;
4879 space_left -= addrlen;
4880 *bytes_copied += addrlen;
4882 rcu_read_unlock();
4884 return cnt;
4888 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4889 char __user *optval, int __user *optlen)
4891 struct sctp_bind_addr *bp;
4892 struct sctp_association *asoc;
4893 int cnt = 0;
4894 struct sctp_getaddrs getaddrs;
4895 struct sctp_sockaddr_entry *addr;
4896 void __user *to;
4897 union sctp_addr temp;
4898 struct sctp_sock *sp = sctp_sk(sk);
4899 int addrlen;
4900 int err = 0;
4901 size_t space_left;
4902 int bytes_copied = 0;
4903 void *addrs;
4904 void *buf;
4906 if (len < sizeof(struct sctp_getaddrs))
4907 return -EINVAL;
4909 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4910 return -EFAULT;
4913 * For UDP-style sockets, id specifies the association to query.
4914 * If the id field is set to the value '0' then the locally bound
4915 * addresses are returned without regard to any particular
4916 * association.
4918 if (0 == getaddrs.assoc_id) {
4919 bp = &sctp_sk(sk)->ep->base.bind_addr;
4920 } else {
4921 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4922 if (!asoc)
4923 return -EINVAL;
4924 bp = &asoc->base.bind_addr;
4927 to = optval + offsetof(struct sctp_getaddrs, addrs);
4928 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4930 addrs = kmalloc(space_left, GFP_KERNEL);
4931 if (!addrs)
4932 return -ENOMEM;
4934 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4935 * addresses from the global local address list.
4937 if (sctp_list_single_entry(&bp->address_list)) {
4938 addr = list_entry(bp->address_list.next,
4939 struct sctp_sockaddr_entry, list);
4940 if (sctp_is_any(sk, &addr->a)) {
4941 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4942 space_left, &bytes_copied);
4943 if (cnt < 0) {
4944 err = cnt;
4945 goto out;
4947 goto copy_getaddrs;
4951 buf = addrs;
4952 /* Protection on the bound address list is not needed since
4953 * in the socket option context we hold a socket lock and
4954 * thus the bound address list can't change.
4956 list_for_each_entry(addr, &bp->address_list, list) {
4957 memcpy(&temp, &addr->a, sizeof(temp));
4958 addrlen = sctp_get_pf_specific(sk->sk_family)
4959 ->addr_to_user(sp, &temp);
4960 if (space_left < addrlen) {
4961 err = -ENOMEM; /*fixme: right error?*/
4962 goto out;
4964 memcpy(buf, &temp, addrlen);
4965 buf += addrlen;
4966 bytes_copied += addrlen;
4967 cnt++;
4968 space_left -= addrlen;
4971 copy_getaddrs:
4972 if (copy_to_user(to, addrs, bytes_copied)) {
4973 err = -EFAULT;
4974 goto out;
4976 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4977 err = -EFAULT;
4978 goto out;
4980 if (put_user(bytes_copied, optlen))
4981 err = -EFAULT;
4982 out:
4983 kfree(addrs);
4984 return err;
4987 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4989 * Requests that the local SCTP stack use the enclosed peer address as
4990 * the association primary. The enclosed address must be one of the
4991 * association peer's addresses.
4993 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4994 char __user *optval, int __user *optlen)
4996 struct sctp_prim prim;
4997 struct sctp_association *asoc;
4998 struct sctp_sock *sp = sctp_sk(sk);
5000 if (len < sizeof(struct sctp_prim))
5001 return -EINVAL;
5003 len = sizeof(struct sctp_prim);
5005 if (copy_from_user(&prim, optval, len))
5006 return -EFAULT;
5008 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5009 if (!asoc)
5010 return -EINVAL;
5012 if (!asoc->peer.primary_path)
5013 return -ENOTCONN;
5015 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5016 asoc->peer.primary_path->af_specific->sockaddr_len);
5018 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5019 (union sctp_addr *)&prim.ssp_addr);
5021 if (put_user(len, optlen))
5022 return -EFAULT;
5023 if (copy_to_user(optval, &prim, len))
5024 return -EFAULT;
5026 return 0;
5030 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5032 * Requests that the local endpoint set the specified Adaptation Layer
5033 * Indication parameter for all future INIT and INIT-ACK exchanges.
5035 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5036 char __user *optval, int __user *optlen)
5038 struct sctp_setadaptation adaptation;
5040 if (len < sizeof(struct sctp_setadaptation))
5041 return -EINVAL;
5043 len = sizeof(struct sctp_setadaptation);
5045 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5047 if (put_user(len, optlen))
5048 return -EFAULT;
5049 if (copy_to_user(optval, &adaptation, len))
5050 return -EFAULT;
5052 return 0;
5057 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5059 * Applications that wish to use the sendto() system call may wish to
5060 * specify a default set of parameters that would normally be supplied
5061 * through the inclusion of ancillary data. This socket option allows
5062 * such an application to set the default sctp_sndrcvinfo structure.
5065 * The application that wishes to use this socket option simply passes
5066 * in to this call the sctp_sndrcvinfo structure defined in Section
5067 * 5.2.2) The input parameters accepted by this call include
5068 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5069 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5070 * to this call if the caller is using the UDP model.
5072 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5074 static int sctp_getsockopt_default_send_param(struct sock *sk,
5075 int len, char __user *optval,
5076 int __user *optlen)
5078 struct sctp_sock *sp = sctp_sk(sk);
5079 struct sctp_association *asoc;
5080 struct sctp_sndrcvinfo info;
5082 if (len < sizeof(info))
5083 return -EINVAL;
5085 len = sizeof(info);
5087 if (copy_from_user(&info, optval, len))
5088 return -EFAULT;
5090 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5091 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5092 return -EINVAL;
5093 if (asoc) {
5094 info.sinfo_stream = asoc->default_stream;
5095 info.sinfo_flags = asoc->default_flags;
5096 info.sinfo_ppid = asoc->default_ppid;
5097 info.sinfo_context = asoc->default_context;
5098 info.sinfo_timetolive = asoc->default_timetolive;
5099 } else {
5100 info.sinfo_stream = sp->default_stream;
5101 info.sinfo_flags = sp->default_flags;
5102 info.sinfo_ppid = sp->default_ppid;
5103 info.sinfo_context = sp->default_context;
5104 info.sinfo_timetolive = sp->default_timetolive;
5107 if (put_user(len, optlen))
5108 return -EFAULT;
5109 if (copy_to_user(optval, &info, len))
5110 return -EFAULT;
5112 return 0;
5115 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5116 * (SCTP_DEFAULT_SNDINFO)
5118 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5119 char __user *optval,
5120 int __user *optlen)
5122 struct sctp_sock *sp = sctp_sk(sk);
5123 struct sctp_association *asoc;
5124 struct sctp_sndinfo info;
5126 if (len < sizeof(info))
5127 return -EINVAL;
5129 len = sizeof(info);
5131 if (copy_from_user(&info, optval, len))
5132 return -EFAULT;
5134 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5135 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5136 return -EINVAL;
5137 if (asoc) {
5138 info.snd_sid = asoc->default_stream;
5139 info.snd_flags = asoc->default_flags;
5140 info.snd_ppid = asoc->default_ppid;
5141 info.snd_context = asoc->default_context;
5142 } else {
5143 info.snd_sid = sp->default_stream;
5144 info.snd_flags = sp->default_flags;
5145 info.snd_ppid = sp->default_ppid;
5146 info.snd_context = sp->default_context;
5149 if (put_user(len, optlen))
5150 return -EFAULT;
5151 if (copy_to_user(optval, &info, len))
5152 return -EFAULT;
5154 return 0;
5159 * 7.1.5 SCTP_NODELAY
5161 * Turn on/off any Nagle-like algorithm. This means that packets are
5162 * generally sent as soon as possible and no unnecessary delays are
5163 * introduced, at the cost of more packets in the network. Expects an
5164 * integer boolean flag.
5167 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5168 char __user *optval, int __user *optlen)
5170 int val;
5172 if (len < sizeof(int))
5173 return -EINVAL;
5175 len = sizeof(int);
5176 val = (sctp_sk(sk)->nodelay == 1);
5177 if (put_user(len, optlen))
5178 return -EFAULT;
5179 if (copy_to_user(optval, &val, len))
5180 return -EFAULT;
5181 return 0;
5186 * 7.1.1 SCTP_RTOINFO
5188 * The protocol parameters used to initialize and bound retransmission
5189 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5190 * and modify these parameters.
5191 * All parameters are time values, in milliseconds. A value of 0, when
5192 * modifying the parameters, indicates that the current value should not
5193 * be changed.
5196 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5197 char __user *optval,
5198 int __user *optlen) {
5199 struct sctp_rtoinfo rtoinfo;
5200 struct sctp_association *asoc;
5202 if (len < sizeof (struct sctp_rtoinfo))
5203 return -EINVAL;
5205 len = sizeof(struct sctp_rtoinfo);
5207 if (copy_from_user(&rtoinfo, optval, len))
5208 return -EFAULT;
5210 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5212 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5213 return -EINVAL;
5215 /* Values corresponding to the specific association. */
5216 if (asoc) {
5217 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5218 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5219 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5220 } else {
5221 /* Values corresponding to the endpoint. */
5222 struct sctp_sock *sp = sctp_sk(sk);
5224 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5225 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5226 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5229 if (put_user(len, optlen))
5230 return -EFAULT;
5232 if (copy_to_user(optval, &rtoinfo, len))
5233 return -EFAULT;
5235 return 0;
5240 * 7.1.2 SCTP_ASSOCINFO
5242 * This option is used to tune the maximum retransmission attempts
5243 * of the association.
5244 * Returns an error if the new association retransmission value is
5245 * greater than the sum of the retransmission value of the peer.
5246 * See [SCTP] for more information.
5249 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5250 char __user *optval,
5251 int __user *optlen)
5254 struct sctp_assocparams assocparams;
5255 struct sctp_association *asoc;
5256 struct list_head *pos;
5257 int cnt = 0;
5259 if (len < sizeof (struct sctp_assocparams))
5260 return -EINVAL;
5262 len = sizeof(struct sctp_assocparams);
5264 if (copy_from_user(&assocparams, optval, len))
5265 return -EFAULT;
5267 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5269 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5270 return -EINVAL;
5272 /* Values correspoinding to the specific association */
5273 if (asoc) {
5274 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5275 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5276 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5277 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5279 list_for_each(pos, &asoc->peer.transport_addr_list) {
5280 cnt++;
5283 assocparams.sasoc_number_peer_destinations = cnt;
5284 } else {
5285 /* Values corresponding to the endpoint */
5286 struct sctp_sock *sp = sctp_sk(sk);
5288 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5289 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5290 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5291 assocparams.sasoc_cookie_life =
5292 sp->assocparams.sasoc_cookie_life;
5293 assocparams.sasoc_number_peer_destinations =
5294 sp->assocparams.
5295 sasoc_number_peer_destinations;
5298 if (put_user(len, optlen))
5299 return -EFAULT;
5301 if (copy_to_user(optval, &assocparams, len))
5302 return -EFAULT;
5304 return 0;
5308 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5310 * This socket option is a boolean flag which turns on or off mapped V4
5311 * addresses. If this option is turned on and the socket is type
5312 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5313 * If this option is turned off, then no mapping will be done of V4
5314 * addresses and a user will receive both PF_INET6 and PF_INET type
5315 * addresses on the socket.
5317 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5318 char __user *optval, int __user *optlen)
5320 int val;
5321 struct sctp_sock *sp = sctp_sk(sk);
5323 if (len < sizeof(int))
5324 return -EINVAL;
5326 len = sizeof(int);
5327 val = sp->v4mapped;
5328 if (put_user(len, optlen))
5329 return -EFAULT;
5330 if (copy_to_user(optval, &val, len))
5331 return -EFAULT;
5333 return 0;
5337 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5338 * (chapter and verse is quoted at sctp_setsockopt_context())
5340 static int sctp_getsockopt_context(struct sock *sk, int len,
5341 char __user *optval, int __user *optlen)
5343 struct sctp_assoc_value params;
5344 struct sctp_sock *sp;
5345 struct sctp_association *asoc;
5347 if (len < sizeof(struct sctp_assoc_value))
5348 return -EINVAL;
5350 len = sizeof(struct sctp_assoc_value);
5352 if (copy_from_user(&params, optval, len))
5353 return -EFAULT;
5355 sp = sctp_sk(sk);
5357 if (params.assoc_id != 0) {
5358 asoc = sctp_id2assoc(sk, params.assoc_id);
5359 if (!asoc)
5360 return -EINVAL;
5361 params.assoc_value = asoc->default_rcv_context;
5362 } else {
5363 params.assoc_value = sp->default_rcv_context;
5366 if (put_user(len, optlen))
5367 return -EFAULT;
5368 if (copy_to_user(optval, &params, len))
5369 return -EFAULT;
5371 return 0;
5375 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5376 * This option will get or set the maximum size to put in any outgoing
5377 * SCTP DATA chunk. If a message is larger than this size it will be
5378 * fragmented by SCTP into the specified size. Note that the underlying
5379 * SCTP implementation may fragment into smaller sized chunks when the
5380 * PMTU of the underlying association is smaller than the value set by
5381 * the user. The default value for this option is '0' which indicates
5382 * the user is NOT limiting fragmentation and only the PMTU will effect
5383 * SCTP's choice of DATA chunk size. Note also that values set larger
5384 * than the maximum size of an IP datagram will effectively let SCTP
5385 * control fragmentation (i.e. the same as setting this option to 0).
5387 * The following structure is used to access and modify this parameter:
5389 * struct sctp_assoc_value {
5390 * sctp_assoc_t assoc_id;
5391 * uint32_t assoc_value;
5392 * };
5394 * assoc_id: This parameter is ignored for one-to-one style sockets.
5395 * For one-to-many style sockets this parameter indicates which
5396 * association the user is performing an action upon. Note that if
5397 * this field's value is zero then the endpoints default value is
5398 * changed (effecting future associations only).
5399 * assoc_value: This parameter specifies the maximum size in bytes.
5401 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5402 char __user *optval, int __user *optlen)
5404 struct sctp_assoc_value params;
5405 struct sctp_association *asoc;
5407 if (len == sizeof(int)) {
5408 pr_warn_ratelimited(DEPRECATED
5409 "%s (pid %d) "
5410 "Use of int in maxseg socket option.\n"
5411 "Use struct sctp_assoc_value instead\n",
5412 current->comm, task_pid_nr(current));
5413 params.assoc_id = 0;
5414 } else if (len >= sizeof(struct sctp_assoc_value)) {
5415 len = sizeof(struct sctp_assoc_value);
5416 if (copy_from_user(&params, optval, sizeof(params)))
5417 return -EFAULT;
5418 } else
5419 return -EINVAL;
5421 asoc = sctp_id2assoc(sk, params.assoc_id);
5422 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5423 return -EINVAL;
5425 if (asoc)
5426 params.assoc_value = asoc->frag_point;
5427 else
5428 params.assoc_value = sctp_sk(sk)->user_frag;
5430 if (put_user(len, optlen))
5431 return -EFAULT;
5432 if (len == sizeof(int)) {
5433 if (copy_to_user(optval, &params.assoc_value, len))
5434 return -EFAULT;
5435 } else {
5436 if (copy_to_user(optval, &params, len))
5437 return -EFAULT;
5440 return 0;
5444 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5445 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5447 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5448 char __user *optval, int __user *optlen)
5450 int val;
5452 if (len < sizeof(int))
5453 return -EINVAL;
5455 len = sizeof(int);
5457 val = sctp_sk(sk)->frag_interleave;
5458 if (put_user(len, optlen))
5459 return -EFAULT;
5460 if (copy_to_user(optval, &val, len))
5461 return -EFAULT;
5463 return 0;
5467 * 7.1.25. Set or Get the sctp partial delivery point
5468 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5470 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5471 char __user *optval,
5472 int __user *optlen)
5474 u32 val;
5476 if (len < sizeof(u32))
5477 return -EINVAL;
5479 len = sizeof(u32);
5481 val = sctp_sk(sk)->pd_point;
5482 if (put_user(len, optlen))
5483 return -EFAULT;
5484 if (copy_to_user(optval, &val, len))
5485 return -EFAULT;
5487 return 0;
5491 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5492 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5494 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5495 char __user *optval,
5496 int __user *optlen)
5498 struct sctp_assoc_value params;
5499 struct sctp_sock *sp;
5500 struct sctp_association *asoc;
5502 if (len == sizeof(int)) {
5503 pr_warn_ratelimited(DEPRECATED
5504 "%s (pid %d) "
5505 "Use of int in max_burst socket option.\n"
5506 "Use struct sctp_assoc_value instead\n",
5507 current->comm, task_pid_nr(current));
5508 params.assoc_id = 0;
5509 } else if (len >= sizeof(struct sctp_assoc_value)) {
5510 len = sizeof(struct sctp_assoc_value);
5511 if (copy_from_user(&params, optval, len))
5512 return -EFAULT;
5513 } else
5514 return -EINVAL;
5516 sp = sctp_sk(sk);
5518 if (params.assoc_id != 0) {
5519 asoc = sctp_id2assoc(sk, params.assoc_id);
5520 if (!asoc)
5521 return -EINVAL;
5522 params.assoc_value = asoc->max_burst;
5523 } else
5524 params.assoc_value = sp->max_burst;
5526 if (len == sizeof(int)) {
5527 if (copy_to_user(optval, &params.assoc_value, len))
5528 return -EFAULT;
5529 } else {
5530 if (copy_to_user(optval, &params, len))
5531 return -EFAULT;
5534 return 0;
5538 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5539 char __user *optval, int __user *optlen)
5541 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5542 struct sctp_hmacalgo __user *p = (void __user *)optval;
5543 struct sctp_hmac_algo_param *hmacs;
5544 __u16 data_len = 0;
5545 u32 num_idents;
5547 if (!ep->auth_enable)
5548 return -EACCES;
5550 hmacs = ep->auth_hmacs_list;
5551 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5553 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5554 return -EINVAL;
5556 len = sizeof(struct sctp_hmacalgo) + data_len;
5557 num_idents = data_len / sizeof(u16);
5559 if (put_user(len, optlen))
5560 return -EFAULT;
5561 if (put_user(num_idents, &p->shmac_num_idents))
5562 return -EFAULT;
5563 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5564 return -EFAULT;
5565 return 0;
5568 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5569 char __user *optval, int __user *optlen)
5571 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5572 struct sctp_authkeyid val;
5573 struct sctp_association *asoc;
5575 if (!ep->auth_enable)
5576 return -EACCES;
5578 if (len < sizeof(struct sctp_authkeyid))
5579 return -EINVAL;
5580 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5581 return -EFAULT;
5583 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5584 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5585 return -EINVAL;
5587 if (asoc)
5588 val.scact_keynumber = asoc->active_key_id;
5589 else
5590 val.scact_keynumber = ep->active_key_id;
5592 len = sizeof(struct sctp_authkeyid);
5593 if (put_user(len, optlen))
5594 return -EFAULT;
5595 if (copy_to_user(optval, &val, len))
5596 return -EFAULT;
5598 return 0;
5601 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5602 char __user *optval, int __user *optlen)
5604 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5605 struct sctp_authchunks __user *p = (void __user *)optval;
5606 struct sctp_authchunks val;
5607 struct sctp_association *asoc;
5608 struct sctp_chunks_param *ch;
5609 u32 num_chunks = 0;
5610 char __user *to;
5612 if (!ep->auth_enable)
5613 return -EACCES;
5615 if (len < sizeof(struct sctp_authchunks))
5616 return -EINVAL;
5618 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5619 return -EFAULT;
5621 to = p->gauth_chunks;
5622 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5623 if (!asoc)
5624 return -EINVAL;
5626 ch = asoc->peer.peer_chunks;
5627 if (!ch)
5628 goto num;
5630 /* See if the user provided enough room for all the data */
5631 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5632 if (len < num_chunks)
5633 return -EINVAL;
5635 if (copy_to_user(to, ch->chunks, num_chunks))
5636 return -EFAULT;
5637 num:
5638 len = sizeof(struct sctp_authchunks) + num_chunks;
5639 if (put_user(len, optlen))
5640 return -EFAULT;
5641 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5642 return -EFAULT;
5643 return 0;
5646 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5647 char __user *optval, int __user *optlen)
5649 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5650 struct sctp_authchunks __user *p = (void __user *)optval;
5651 struct sctp_authchunks val;
5652 struct sctp_association *asoc;
5653 struct sctp_chunks_param *ch;
5654 u32 num_chunks = 0;
5655 char __user *to;
5657 if (!ep->auth_enable)
5658 return -EACCES;
5660 if (len < sizeof(struct sctp_authchunks))
5661 return -EINVAL;
5663 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5664 return -EFAULT;
5666 to = p->gauth_chunks;
5667 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5668 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5669 return -EINVAL;
5671 if (asoc)
5672 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5673 else
5674 ch = ep->auth_chunk_list;
5676 if (!ch)
5677 goto num;
5679 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5680 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5681 return -EINVAL;
5683 if (copy_to_user(to, ch->chunks, num_chunks))
5684 return -EFAULT;
5685 num:
5686 len = sizeof(struct sctp_authchunks) + num_chunks;
5687 if (put_user(len, optlen))
5688 return -EFAULT;
5689 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5690 return -EFAULT;
5692 return 0;
5696 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5697 * This option gets the current number of associations that are attached
5698 * to a one-to-many style socket. The option value is an uint32_t.
5700 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5701 char __user *optval, int __user *optlen)
5703 struct sctp_sock *sp = sctp_sk(sk);
5704 struct sctp_association *asoc;
5705 u32 val = 0;
5707 if (sctp_style(sk, TCP))
5708 return -EOPNOTSUPP;
5710 if (len < sizeof(u32))
5711 return -EINVAL;
5713 len = sizeof(u32);
5715 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5716 val++;
5719 if (put_user(len, optlen))
5720 return -EFAULT;
5721 if (copy_to_user(optval, &val, len))
5722 return -EFAULT;
5724 return 0;
5728 * 8.1.23 SCTP_AUTO_ASCONF
5729 * See the corresponding setsockopt entry as description
5731 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5732 char __user *optval, int __user *optlen)
5734 int val = 0;
5736 if (len < sizeof(int))
5737 return -EINVAL;
5739 len = sizeof(int);
5740 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5741 val = 1;
5742 if (put_user(len, optlen))
5743 return -EFAULT;
5744 if (copy_to_user(optval, &val, len))
5745 return -EFAULT;
5746 return 0;
5750 * 8.2.6. Get the Current Identifiers of Associations
5751 * (SCTP_GET_ASSOC_ID_LIST)
5753 * This option gets the current list of SCTP association identifiers of
5754 * the SCTP associations handled by a one-to-many style socket.
5756 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5757 char __user *optval, int __user *optlen)
5759 struct sctp_sock *sp = sctp_sk(sk);
5760 struct sctp_association *asoc;
5761 struct sctp_assoc_ids *ids;
5762 u32 num = 0;
5764 if (sctp_style(sk, TCP))
5765 return -EOPNOTSUPP;
5767 if (len < sizeof(struct sctp_assoc_ids))
5768 return -EINVAL;
5770 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5771 num++;
5774 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5775 return -EINVAL;
5777 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5779 ids = kmalloc(len, GFP_KERNEL);
5780 if (unlikely(!ids))
5781 return -ENOMEM;
5783 ids->gaids_number_of_ids = num;
5784 num = 0;
5785 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5786 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5789 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5790 kfree(ids);
5791 return -EFAULT;
5794 kfree(ids);
5795 return 0;
5799 * SCTP_PEER_ADDR_THLDS
5801 * This option allows us to fetch the partially failed threshold for one or all
5802 * transports in an association. See Section 6.1 of:
5803 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5805 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5806 char __user *optval,
5807 int len,
5808 int __user *optlen)
5810 struct sctp_paddrthlds val;
5811 struct sctp_transport *trans;
5812 struct sctp_association *asoc;
5814 if (len < sizeof(struct sctp_paddrthlds))
5815 return -EINVAL;
5816 len = sizeof(struct sctp_paddrthlds);
5817 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5818 return -EFAULT;
5820 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5821 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5822 if (!asoc)
5823 return -ENOENT;
5825 val.spt_pathpfthld = asoc->pf_retrans;
5826 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5827 } else {
5828 trans = sctp_addr_id2transport(sk, &val.spt_address,
5829 val.spt_assoc_id);
5830 if (!trans)
5831 return -ENOENT;
5833 val.spt_pathmaxrxt = trans->pathmaxrxt;
5834 val.spt_pathpfthld = trans->pf_retrans;
5837 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5838 return -EFAULT;
5840 return 0;
5844 * SCTP_GET_ASSOC_STATS
5846 * This option retrieves local per endpoint statistics. It is modeled
5847 * after OpenSolaris' implementation
5849 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5850 char __user *optval,
5851 int __user *optlen)
5853 struct sctp_assoc_stats sas;
5854 struct sctp_association *asoc = NULL;
5856 /* User must provide at least the assoc id */
5857 if (len < sizeof(sctp_assoc_t))
5858 return -EINVAL;
5860 /* Allow the struct to grow and fill in as much as possible */
5861 len = min_t(size_t, len, sizeof(sas));
5863 if (copy_from_user(&sas, optval, len))
5864 return -EFAULT;
5866 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5867 if (!asoc)
5868 return -EINVAL;
5870 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5871 sas.sas_gapcnt = asoc->stats.gapcnt;
5872 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5873 sas.sas_osacks = asoc->stats.osacks;
5874 sas.sas_isacks = asoc->stats.isacks;
5875 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5876 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5877 sas.sas_oodchunks = asoc->stats.oodchunks;
5878 sas.sas_iodchunks = asoc->stats.iodchunks;
5879 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5880 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5881 sas.sas_idupchunks = asoc->stats.idupchunks;
5882 sas.sas_opackets = asoc->stats.opackets;
5883 sas.sas_ipackets = asoc->stats.ipackets;
5885 /* New high max rto observed, will return 0 if not a single
5886 * RTO update took place. obs_rto_ipaddr will be bogus
5887 * in such a case
5889 sas.sas_maxrto = asoc->stats.max_obs_rto;
5890 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5891 sizeof(struct sockaddr_storage));
5893 /* Mark beginning of a new observation period */
5894 asoc->stats.max_obs_rto = asoc->rto_min;
5896 if (put_user(len, optlen))
5897 return -EFAULT;
5899 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5901 if (copy_to_user(optval, &sas, len))
5902 return -EFAULT;
5904 return 0;
5907 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
5908 char __user *optval,
5909 int __user *optlen)
5911 int val = 0;
5913 if (len < sizeof(int))
5914 return -EINVAL;
5916 len = sizeof(int);
5917 if (sctp_sk(sk)->recvrcvinfo)
5918 val = 1;
5919 if (put_user(len, optlen))
5920 return -EFAULT;
5921 if (copy_to_user(optval, &val, len))
5922 return -EFAULT;
5924 return 0;
5927 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
5928 char __user *optval,
5929 int __user *optlen)
5931 int val = 0;
5933 if (len < sizeof(int))
5934 return -EINVAL;
5936 len = sizeof(int);
5937 if (sctp_sk(sk)->recvnxtinfo)
5938 val = 1;
5939 if (put_user(len, optlen))
5940 return -EFAULT;
5941 if (copy_to_user(optval, &val, len))
5942 return -EFAULT;
5944 return 0;
5947 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5948 char __user *optval, int __user *optlen)
5950 int retval = 0;
5951 int len;
5953 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5955 /* I can hardly begin to describe how wrong this is. This is
5956 * so broken as to be worse than useless. The API draft
5957 * REALLY is NOT helpful here... I am not convinced that the
5958 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5959 * are at all well-founded.
5961 if (level != SOL_SCTP) {
5962 struct sctp_af *af = sctp_sk(sk)->pf->af;
5964 retval = af->getsockopt(sk, level, optname, optval, optlen);
5965 return retval;
5968 if (get_user(len, optlen))
5969 return -EFAULT;
5971 lock_sock(sk);
5973 switch (optname) {
5974 case SCTP_STATUS:
5975 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5976 break;
5977 case SCTP_DISABLE_FRAGMENTS:
5978 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5979 optlen);
5980 break;
5981 case SCTP_EVENTS:
5982 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5983 break;
5984 case SCTP_AUTOCLOSE:
5985 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5986 break;
5987 case SCTP_SOCKOPT_PEELOFF:
5988 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5989 break;
5990 case SCTP_PEER_ADDR_PARAMS:
5991 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5992 optlen);
5993 break;
5994 case SCTP_DELAYED_SACK:
5995 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5996 optlen);
5997 break;
5998 case SCTP_INITMSG:
5999 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6000 break;
6001 case SCTP_GET_PEER_ADDRS:
6002 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6003 optlen);
6004 break;
6005 case SCTP_GET_LOCAL_ADDRS:
6006 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6007 optlen);
6008 break;
6009 case SCTP_SOCKOPT_CONNECTX3:
6010 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6011 break;
6012 case SCTP_DEFAULT_SEND_PARAM:
6013 retval = sctp_getsockopt_default_send_param(sk, len,
6014 optval, optlen);
6015 break;
6016 case SCTP_DEFAULT_SNDINFO:
6017 retval = sctp_getsockopt_default_sndinfo(sk, len,
6018 optval, optlen);
6019 break;
6020 case SCTP_PRIMARY_ADDR:
6021 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6022 break;
6023 case SCTP_NODELAY:
6024 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6025 break;
6026 case SCTP_RTOINFO:
6027 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6028 break;
6029 case SCTP_ASSOCINFO:
6030 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6031 break;
6032 case SCTP_I_WANT_MAPPED_V4_ADDR:
6033 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6034 break;
6035 case SCTP_MAXSEG:
6036 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6037 break;
6038 case SCTP_GET_PEER_ADDR_INFO:
6039 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6040 optlen);
6041 break;
6042 case SCTP_ADAPTATION_LAYER:
6043 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6044 optlen);
6045 break;
6046 case SCTP_CONTEXT:
6047 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6048 break;
6049 case SCTP_FRAGMENT_INTERLEAVE:
6050 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6051 optlen);
6052 break;
6053 case SCTP_PARTIAL_DELIVERY_POINT:
6054 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6055 optlen);
6056 break;
6057 case SCTP_MAX_BURST:
6058 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6059 break;
6060 case SCTP_AUTH_KEY:
6061 case SCTP_AUTH_CHUNK:
6062 case SCTP_AUTH_DELETE_KEY:
6063 retval = -EOPNOTSUPP;
6064 break;
6065 case SCTP_HMAC_IDENT:
6066 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6067 break;
6068 case SCTP_AUTH_ACTIVE_KEY:
6069 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6070 break;
6071 case SCTP_PEER_AUTH_CHUNKS:
6072 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6073 optlen);
6074 break;
6075 case SCTP_LOCAL_AUTH_CHUNKS:
6076 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6077 optlen);
6078 break;
6079 case SCTP_GET_ASSOC_NUMBER:
6080 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6081 break;
6082 case SCTP_GET_ASSOC_ID_LIST:
6083 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6084 break;
6085 case SCTP_AUTO_ASCONF:
6086 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6087 break;
6088 case SCTP_PEER_ADDR_THLDS:
6089 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6090 break;
6091 case SCTP_GET_ASSOC_STATS:
6092 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6093 break;
6094 case SCTP_RECVRCVINFO:
6095 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6096 break;
6097 case SCTP_RECVNXTINFO:
6098 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6099 break;
6100 default:
6101 retval = -ENOPROTOOPT;
6102 break;
6105 release_sock(sk);
6106 return retval;
6109 static void sctp_hash(struct sock *sk)
6111 /* STUB */
6114 static void sctp_unhash(struct sock *sk)
6116 /* STUB */
6119 /* Check if port is acceptable. Possibly find first available port.
6121 * The port hash table (contained in the 'global' SCTP protocol storage
6122 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6123 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6124 * list (the list number is the port number hashed out, so as you
6125 * would expect from a hash function, all the ports in a given list have
6126 * such a number that hashes out to the same list number; you were
6127 * expecting that, right?); so each list has a set of ports, with a
6128 * link to the socket (struct sock) that uses it, the port number and
6129 * a fastreuse flag (FIXME: NPI ipg).
6131 static struct sctp_bind_bucket *sctp_bucket_create(
6132 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6134 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6136 struct sctp_bind_hashbucket *head; /* hash list */
6137 struct sctp_bind_bucket *pp;
6138 unsigned short snum;
6139 int ret;
6141 snum = ntohs(addr->v4.sin_port);
6143 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6145 local_bh_disable();
6147 if (snum == 0) {
6148 /* Search for an available port. */
6149 int low, high, remaining, index;
6150 unsigned int rover;
6151 struct net *net = sock_net(sk);
6153 inet_get_local_port_range(net, &low, &high);
6154 remaining = (high - low) + 1;
6155 rover = prandom_u32() % remaining + low;
6157 do {
6158 rover++;
6159 if ((rover < low) || (rover > high))
6160 rover = low;
6161 if (inet_is_local_reserved_port(net, rover))
6162 continue;
6163 index = sctp_phashfn(sock_net(sk), rover);
6164 head = &sctp_port_hashtable[index];
6165 spin_lock(&head->lock);
6166 sctp_for_each_hentry(pp, &head->chain)
6167 if ((pp->port == rover) &&
6168 net_eq(sock_net(sk), pp->net))
6169 goto next;
6170 break;
6171 next:
6172 spin_unlock(&head->lock);
6173 } while (--remaining > 0);
6175 /* Exhausted local port range during search? */
6176 ret = 1;
6177 if (remaining <= 0)
6178 goto fail;
6180 /* OK, here is the one we will use. HEAD (the port
6181 * hash table list entry) is non-NULL and we hold it's
6182 * mutex.
6184 snum = rover;
6185 } else {
6186 /* We are given an specific port number; we verify
6187 * that it is not being used. If it is used, we will
6188 * exahust the search in the hash list corresponding
6189 * to the port number (snum) - we detect that with the
6190 * port iterator, pp being NULL.
6192 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6193 spin_lock(&head->lock);
6194 sctp_for_each_hentry(pp, &head->chain) {
6195 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6196 goto pp_found;
6199 pp = NULL;
6200 goto pp_not_found;
6201 pp_found:
6202 if (!hlist_empty(&pp->owner)) {
6203 /* We had a port hash table hit - there is an
6204 * available port (pp != NULL) and it is being
6205 * used by other socket (pp->owner not empty); that other
6206 * socket is going to be sk2.
6208 int reuse = sk->sk_reuse;
6209 struct sock *sk2;
6211 pr_debug("%s: found a possible match\n", __func__);
6213 if (pp->fastreuse && sk->sk_reuse &&
6214 sk->sk_state != SCTP_SS_LISTENING)
6215 goto success;
6217 /* Run through the list of sockets bound to the port
6218 * (pp->port) [via the pointers bind_next and
6219 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6220 * we get the endpoint they describe and run through
6221 * the endpoint's list of IP (v4 or v6) addresses,
6222 * comparing each of the addresses with the address of
6223 * the socket sk. If we find a match, then that means
6224 * that this port/socket (sk) combination are already
6225 * in an endpoint.
6227 sk_for_each_bound(sk2, &pp->owner) {
6228 struct sctp_endpoint *ep2;
6229 ep2 = sctp_sk(sk2)->ep;
6231 if (sk == sk2 ||
6232 (reuse && sk2->sk_reuse &&
6233 sk2->sk_state != SCTP_SS_LISTENING))
6234 continue;
6236 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6237 sctp_sk(sk2), sctp_sk(sk))) {
6238 ret = (long)sk2;
6239 goto fail_unlock;
6243 pr_debug("%s: found a match\n", __func__);
6245 pp_not_found:
6246 /* If there was a hash table miss, create a new port. */
6247 ret = 1;
6248 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6249 goto fail_unlock;
6251 /* In either case (hit or miss), make sure fastreuse is 1 only
6252 * if sk->sk_reuse is too (that is, if the caller requested
6253 * SO_REUSEADDR on this socket -sk-).
6255 if (hlist_empty(&pp->owner)) {
6256 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6257 pp->fastreuse = 1;
6258 else
6259 pp->fastreuse = 0;
6260 } else if (pp->fastreuse &&
6261 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6262 pp->fastreuse = 0;
6264 /* We are set, so fill up all the data in the hash table
6265 * entry, tie the socket list information with the rest of the
6266 * sockets FIXME: Blurry, NPI (ipg).
6268 success:
6269 if (!sctp_sk(sk)->bind_hash) {
6270 inet_sk(sk)->inet_num = snum;
6271 sk_add_bind_node(sk, &pp->owner);
6272 sctp_sk(sk)->bind_hash = pp;
6274 ret = 0;
6276 fail_unlock:
6277 spin_unlock(&head->lock);
6279 fail:
6280 local_bh_enable();
6281 return ret;
6284 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6285 * port is requested.
6287 static int sctp_get_port(struct sock *sk, unsigned short snum)
6289 union sctp_addr addr;
6290 struct sctp_af *af = sctp_sk(sk)->pf->af;
6292 /* Set up a dummy address struct from the sk. */
6293 af->from_sk(&addr, sk);
6294 addr.v4.sin_port = htons(snum);
6296 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6297 return !!sctp_get_port_local(sk, &addr);
6301 * Move a socket to LISTENING state.
6303 static int sctp_listen_start(struct sock *sk, int backlog)
6305 struct sctp_sock *sp = sctp_sk(sk);
6306 struct sctp_endpoint *ep = sp->ep;
6307 struct crypto_hash *tfm = NULL;
6308 char alg[32];
6310 /* Allocate HMAC for generating cookie. */
6311 if (!sp->hmac && sp->sctp_hmac_alg) {
6312 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6313 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6314 if (IS_ERR(tfm)) {
6315 net_info_ratelimited("failed to load transform for %s: %ld\n",
6316 sp->sctp_hmac_alg, PTR_ERR(tfm));
6317 return -ENOSYS;
6319 sctp_sk(sk)->hmac = tfm;
6323 * If a bind() or sctp_bindx() is not called prior to a listen()
6324 * call that allows new associations to be accepted, the system
6325 * picks an ephemeral port and will choose an address set equivalent
6326 * to binding with a wildcard address.
6328 * This is not currently spelled out in the SCTP sockets
6329 * extensions draft, but follows the practice as seen in TCP
6330 * sockets.
6333 sk->sk_state = SCTP_SS_LISTENING;
6334 if (!ep->base.bind_addr.port) {
6335 if (sctp_autobind(sk))
6336 return -EAGAIN;
6337 } else {
6338 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6339 sk->sk_state = SCTP_SS_CLOSED;
6340 return -EADDRINUSE;
6344 sk->sk_max_ack_backlog = backlog;
6345 sctp_hash_endpoint(ep);
6346 return 0;
6350 * 4.1.3 / 5.1.3 listen()
6352 * By default, new associations are not accepted for UDP style sockets.
6353 * An application uses listen() to mark a socket as being able to
6354 * accept new associations.
6356 * On TCP style sockets, applications use listen() to ready the SCTP
6357 * endpoint for accepting inbound associations.
6359 * On both types of endpoints a backlog of '0' disables listening.
6361 * Move a socket to LISTENING state.
6363 int sctp_inet_listen(struct socket *sock, int backlog)
6365 struct sock *sk = sock->sk;
6366 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6367 int err = -EINVAL;
6369 if (unlikely(backlog < 0))
6370 return err;
6372 lock_sock(sk);
6374 /* Peeled-off sockets are not allowed to listen(). */
6375 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6376 goto out;
6378 if (sock->state != SS_UNCONNECTED)
6379 goto out;
6381 /* If backlog is zero, disable listening. */
6382 if (!backlog) {
6383 if (sctp_sstate(sk, CLOSED))
6384 goto out;
6386 err = 0;
6387 sctp_unhash_endpoint(ep);
6388 sk->sk_state = SCTP_SS_CLOSED;
6389 if (sk->sk_reuse)
6390 sctp_sk(sk)->bind_hash->fastreuse = 1;
6391 goto out;
6394 /* If we are already listening, just update the backlog */
6395 if (sctp_sstate(sk, LISTENING))
6396 sk->sk_max_ack_backlog = backlog;
6397 else {
6398 err = sctp_listen_start(sk, backlog);
6399 if (err)
6400 goto out;
6403 err = 0;
6404 out:
6405 release_sock(sk);
6406 return err;
6410 * This function is done by modeling the current datagram_poll() and the
6411 * tcp_poll(). Note that, based on these implementations, we don't
6412 * lock the socket in this function, even though it seems that,
6413 * ideally, locking or some other mechanisms can be used to ensure
6414 * the integrity of the counters (sndbuf and wmem_alloc) used
6415 * in this place. We assume that we don't need locks either until proven
6416 * otherwise.
6418 * Another thing to note is that we include the Async I/O support
6419 * here, again, by modeling the current TCP/UDP code. We don't have
6420 * a good way to test with it yet.
6422 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6424 struct sock *sk = sock->sk;
6425 struct sctp_sock *sp = sctp_sk(sk);
6426 unsigned int mask;
6428 poll_wait(file, sk_sleep(sk), wait);
6430 /* A TCP-style listening socket becomes readable when the accept queue
6431 * is not empty.
6433 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6434 return (!list_empty(&sp->ep->asocs)) ?
6435 (POLLIN | POLLRDNORM) : 0;
6437 mask = 0;
6439 /* Is there any exceptional events? */
6440 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6441 mask |= POLLERR |
6442 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6443 if (sk->sk_shutdown & RCV_SHUTDOWN)
6444 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6445 if (sk->sk_shutdown == SHUTDOWN_MASK)
6446 mask |= POLLHUP;
6448 /* Is it readable? Reconsider this code with TCP-style support. */
6449 if (!skb_queue_empty(&sk->sk_receive_queue))
6450 mask |= POLLIN | POLLRDNORM;
6452 /* The association is either gone or not ready. */
6453 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6454 return mask;
6456 /* Is it writable? */
6457 if (sctp_writeable(sk)) {
6458 mask |= POLLOUT | POLLWRNORM;
6459 } else {
6460 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6462 * Since the socket is not locked, the buffer
6463 * might be made available after the writeable check and
6464 * before the bit is set. This could cause a lost I/O
6465 * signal. tcp_poll() has a race breaker for this race
6466 * condition. Based on their implementation, we put
6467 * in the following code to cover it as well.
6469 if (sctp_writeable(sk))
6470 mask |= POLLOUT | POLLWRNORM;
6472 return mask;
6475 /********************************************************************
6476 * 2nd Level Abstractions
6477 ********************************************************************/
6479 static struct sctp_bind_bucket *sctp_bucket_create(
6480 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6482 struct sctp_bind_bucket *pp;
6484 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6485 if (pp) {
6486 SCTP_DBG_OBJCNT_INC(bind_bucket);
6487 pp->port = snum;
6488 pp->fastreuse = 0;
6489 INIT_HLIST_HEAD(&pp->owner);
6490 pp->net = net;
6491 hlist_add_head(&pp->node, &head->chain);
6493 return pp;
6496 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6497 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6499 if (pp && hlist_empty(&pp->owner)) {
6500 __hlist_del(&pp->node);
6501 kmem_cache_free(sctp_bucket_cachep, pp);
6502 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6506 /* Release this socket's reference to a local port. */
6507 static inline void __sctp_put_port(struct sock *sk)
6509 struct sctp_bind_hashbucket *head =
6510 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6511 inet_sk(sk)->inet_num)];
6512 struct sctp_bind_bucket *pp;
6514 spin_lock(&head->lock);
6515 pp = sctp_sk(sk)->bind_hash;
6516 __sk_del_bind_node(sk);
6517 sctp_sk(sk)->bind_hash = NULL;
6518 inet_sk(sk)->inet_num = 0;
6519 sctp_bucket_destroy(pp);
6520 spin_unlock(&head->lock);
6523 void sctp_put_port(struct sock *sk)
6525 local_bh_disable();
6526 __sctp_put_port(sk);
6527 local_bh_enable();
6531 * The system picks an ephemeral port and choose an address set equivalent
6532 * to binding with a wildcard address.
6533 * One of those addresses will be the primary address for the association.
6534 * This automatically enables the multihoming capability of SCTP.
6536 static int sctp_autobind(struct sock *sk)
6538 union sctp_addr autoaddr;
6539 struct sctp_af *af;
6540 __be16 port;
6542 /* Initialize a local sockaddr structure to INADDR_ANY. */
6543 af = sctp_sk(sk)->pf->af;
6545 port = htons(inet_sk(sk)->inet_num);
6546 af->inaddr_any(&autoaddr, port);
6548 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6551 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6553 * From RFC 2292
6554 * 4.2 The cmsghdr Structure *
6556 * When ancillary data is sent or received, any number of ancillary data
6557 * objects can be specified by the msg_control and msg_controllen members of
6558 * the msghdr structure, because each object is preceded by
6559 * a cmsghdr structure defining the object's length (the cmsg_len member).
6560 * Historically Berkeley-derived implementations have passed only one object
6561 * at a time, but this API allows multiple objects to be
6562 * passed in a single call to sendmsg() or recvmsg(). The following example
6563 * shows two ancillary data objects in a control buffer.
6565 * |<--------------------------- msg_controllen -------------------------->|
6566 * | |
6568 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6570 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6571 * | | |
6573 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6575 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6576 * | | | | |
6578 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6579 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6581 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6583 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6587 * msg_control
6588 * points here
6590 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6592 struct cmsghdr *cmsg;
6593 struct msghdr *my_msg = (struct msghdr *)msg;
6595 for (cmsg = CMSG_FIRSTHDR(msg); cmsg != NULL;
6596 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6597 if (!CMSG_OK(my_msg, cmsg))
6598 return -EINVAL;
6600 /* Should we parse this header or ignore? */
6601 if (cmsg->cmsg_level != IPPROTO_SCTP)
6602 continue;
6604 /* Strictly check lengths following example in SCM code. */
6605 switch (cmsg->cmsg_type) {
6606 case SCTP_INIT:
6607 /* SCTP Socket API Extension
6608 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6610 * This cmsghdr structure provides information for
6611 * initializing new SCTP associations with sendmsg().
6612 * The SCTP_INITMSG socket option uses this same data
6613 * structure. This structure is not used for
6614 * recvmsg().
6616 * cmsg_level cmsg_type cmsg_data[]
6617 * ------------ ------------ ----------------------
6618 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6620 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
6621 return -EINVAL;
6623 cmsgs->init = CMSG_DATA(cmsg);
6624 break;
6626 case SCTP_SNDRCV:
6627 /* SCTP Socket API Extension
6628 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6630 * This cmsghdr structure specifies SCTP options for
6631 * sendmsg() and describes SCTP header information
6632 * about a received message through recvmsg().
6634 * cmsg_level cmsg_type cmsg_data[]
6635 * ------------ ------------ ----------------------
6636 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6638 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6639 return -EINVAL;
6641 cmsgs->srinfo = CMSG_DATA(cmsg);
6643 if (cmsgs->srinfo->sinfo_flags &
6644 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6645 SCTP_ABORT | SCTP_EOF))
6646 return -EINVAL;
6647 break;
6649 case SCTP_SNDINFO:
6650 /* SCTP Socket API Extension
6651 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6653 * This cmsghdr structure specifies SCTP options for
6654 * sendmsg(). This structure and SCTP_RCVINFO replaces
6655 * SCTP_SNDRCV which has been deprecated.
6657 * cmsg_level cmsg_type cmsg_data[]
6658 * ------------ ------------ ---------------------
6659 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6661 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
6662 return -EINVAL;
6664 cmsgs->sinfo = CMSG_DATA(cmsg);
6666 if (cmsgs->sinfo->snd_flags &
6667 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6668 SCTP_ABORT | SCTP_EOF))
6669 return -EINVAL;
6670 break;
6671 default:
6672 return -EINVAL;
6676 return 0;
6680 * Wait for a packet..
6681 * Note: This function is the same function as in core/datagram.c
6682 * with a few modifications to make lksctp work.
6684 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6686 int error;
6687 DEFINE_WAIT(wait);
6689 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6691 /* Socket errors? */
6692 error = sock_error(sk);
6693 if (error)
6694 goto out;
6696 if (!skb_queue_empty(&sk->sk_receive_queue))
6697 goto ready;
6699 /* Socket shut down? */
6700 if (sk->sk_shutdown & RCV_SHUTDOWN)
6701 goto out;
6703 /* Sequenced packets can come disconnected. If so we report the
6704 * problem.
6706 error = -ENOTCONN;
6708 /* Is there a good reason to think that we may receive some data? */
6709 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6710 goto out;
6712 /* Handle signals. */
6713 if (signal_pending(current))
6714 goto interrupted;
6716 /* Let another process have a go. Since we are going to sleep
6717 * anyway. Note: This may cause odd behaviors if the message
6718 * does not fit in the user's buffer, but this seems to be the
6719 * only way to honor MSG_DONTWAIT realistically.
6721 release_sock(sk);
6722 *timeo_p = schedule_timeout(*timeo_p);
6723 lock_sock(sk);
6725 ready:
6726 finish_wait(sk_sleep(sk), &wait);
6727 return 0;
6729 interrupted:
6730 error = sock_intr_errno(*timeo_p);
6732 out:
6733 finish_wait(sk_sleep(sk), &wait);
6734 *err = error;
6735 return error;
6738 /* Receive a datagram.
6739 * Note: This is pretty much the same routine as in core/datagram.c
6740 * with a few changes to make lksctp work.
6742 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6743 int noblock, int *err)
6745 int error;
6746 struct sk_buff *skb;
6747 long timeo;
6749 timeo = sock_rcvtimeo(sk, noblock);
6751 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6752 MAX_SCHEDULE_TIMEOUT);
6754 do {
6755 /* Again only user level code calls this function,
6756 * so nothing interrupt level
6757 * will suddenly eat the receive_queue.
6759 * Look at current nfs client by the way...
6760 * However, this function was correct in any case. 8)
6762 if (flags & MSG_PEEK) {
6763 spin_lock_bh(&sk->sk_receive_queue.lock);
6764 skb = skb_peek(&sk->sk_receive_queue);
6765 if (skb)
6766 atomic_inc(&skb->users);
6767 spin_unlock_bh(&sk->sk_receive_queue.lock);
6768 } else {
6769 skb = skb_dequeue(&sk->sk_receive_queue);
6772 if (skb)
6773 return skb;
6775 /* Caller is allowed not to check sk->sk_err before calling. */
6776 error = sock_error(sk);
6777 if (error)
6778 goto no_packet;
6780 if (sk->sk_shutdown & RCV_SHUTDOWN)
6781 break;
6783 if (sk_can_busy_loop(sk) &&
6784 sk_busy_loop(sk, noblock))
6785 continue;
6787 /* User doesn't want to wait. */
6788 error = -EAGAIN;
6789 if (!timeo)
6790 goto no_packet;
6791 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6793 return NULL;
6795 no_packet:
6796 *err = error;
6797 return NULL;
6800 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6801 static void __sctp_write_space(struct sctp_association *asoc)
6803 struct sock *sk = asoc->base.sk;
6804 struct socket *sock = sk->sk_socket;
6806 if ((sctp_wspace(asoc) > 0) && sock) {
6807 if (waitqueue_active(&asoc->wait))
6808 wake_up_interruptible(&asoc->wait);
6810 if (sctp_writeable(sk)) {
6811 wait_queue_head_t *wq = sk_sleep(sk);
6813 if (wq && waitqueue_active(wq))
6814 wake_up_interruptible(wq);
6816 /* Note that we try to include the Async I/O support
6817 * here by modeling from the current TCP/UDP code.
6818 * We have not tested with it yet.
6820 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6821 sock_wake_async(sock,
6822 SOCK_WAKE_SPACE, POLL_OUT);
6827 static void sctp_wake_up_waiters(struct sock *sk,
6828 struct sctp_association *asoc)
6830 struct sctp_association *tmp = asoc;
6832 /* We do accounting for the sndbuf space per association,
6833 * so we only need to wake our own association.
6835 if (asoc->ep->sndbuf_policy)
6836 return __sctp_write_space(asoc);
6838 /* If association goes down and is just flushing its
6839 * outq, then just normally notify others.
6841 if (asoc->base.dead)
6842 return sctp_write_space(sk);
6844 /* Accounting for the sndbuf space is per socket, so we
6845 * need to wake up others, try to be fair and in case of
6846 * other associations, let them have a go first instead
6847 * of just doing a sctp_write_space() call.
6849 * Note that we reach sctp_wake_up_waiters() only when
6850 * associations free up queued chunks, thus we are under
6851 * lock and the list of associations on a socket is
6852 * guaranteed not to change.
6854 for (tmp = list_next_entry(tmp, asocs); 1;
6855 tmp = list_next_entry(tmp, asocs)) {
6856 /* Manually skip the head element. */
6857 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6858 continue;
6859 /* Wake up association. */
6860 __sctp_write_space(tmp);
6861 /* We've reached the end. */
6862 if (tmp == asoc)
6863 break;
6867 /* Do accounting for the sndbuf space.
6868 * Decrement the used sndbuf space of the corresponding association by the
6869 * data size which was just transmitted(freed).
6871 static void sctp_wfree(struct sk_buff *skb)
6873 struct sctp_association *asoc;
6874 struct sctp_chunk *chunk;
6875 struct sock *sk;
6877 /* Get the saved chunk pointer. */
6878 chunk = *((struct sctp_chunk **)(skb->cb));
6879 asoc = chunk->asoc;
6880 sk = asoc->base.sk;
6881 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6882 sizeof(struct sk_buff) +
6883 sizeof(struct sctp_chunk);
6885 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6888 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6890 sk->sk_wmem_queued -= skb->truesize;
6891 sk_mem_uncharge(sk, skb->truesize);
6893 sock_wfree(skb);
6894 sctp_wake_up_waiters(sk, asoc);
6896 sctp_association_put(asoc);
6899 /* Do accounting for the receive space on the socket.
6900 * Accounting for the association is done in ulpevent.c
6901 * We set this as a destructor for the cloned data skbs so that
6902 * accounting is done at the correct time.
6904 void sctp_sock_rfree(struct sk_buff *skb)
6906 struct sock *sk = skb->sk;
6907 struct sctp_ulpevent *event = sctp_skb2event(skb);
6909 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6912 * Mimic the behavior of sock_rfree
6914 sk_mem_uncharge(sk, event->rmem_len);
6918 /* Helper function to wait for space in the sndbuf. */
6919 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6920 size_t msg_len)
6922 struct sock *sk = asoc->base.sk;
6923 int err = 0;
6924 long current_timeo = *timeo_p;
6925 DEFINE_WAIT(wait);
6927 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6928 *timeo_p, msg_len);
6930 /* Increment the association's refcnt. */
6931 sctp_association_hold(asoc);
6933 /* Wait on the association specific sndbuf space. */
6934 for (;;) {
6935 prepare_to_wait_exclusive(&asoc->wait, &wait,
6936 TASK_INTERRUPTIBLE);
6937 if (!*timeo_p)
6938 goto do_nonblock;
6939 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6940 asoc->base.dead)
6941 goto do_error;
6942 if (signal_pending(current))
6943 goto do_interrupted;
6944 if (msg_len <= sctp_wspace(asoc))
6945 break;
6947 /* Let another process have a go. Since we are going
6948 * to sleep anyway.
6950 release_sock(sk);
6951 current_timeo = schedule_timeout(current_timeo);
6952 BUG_ON(sk != asoc->base.sk);
6953 lock_sock(sk);
6955 *timeo_p = current_timeo;
6958 out:
6959 finish_wait(&asoc->wait, &wait);
6961 /* Release the association's refcnt. */
6962 sctp_association_put(asoc);
6964 return err;
6966 do_error:
6967 err = -EPIPE;
6968 goto out;
6970 do_interrupted:
6971 err = sock_intr_errno(*timeo_p);
6972 goto out;
6974 do_nonblock:
6975 err = -EAGAIN;
6976 goto out;
6979 void sctp_data_ready(struct sock *sk)
6981 struct socket_wq *wq;
6983 rcu_read_lock();
6984 wq = rcu_dereference(sk->sk_wq);
6985 if (wq_has_sleeper(wq))
6986 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6987 POLLRDNORM | POLLRDBAND);
6988 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6989 rcu_read_unlock();
6992 /* If socket sndbuf has changed, wake up all per association waiters. */
6993 void sctp_write_space(struct sock *sk)
6995 struct sctp_association *asoc;
6997 /* Wake up the tasks in each wait queue. */
6998 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6999 __sctp_write_space(asoc);
7003 /* Is there any sndbuf space available on the socket?
7005 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7006 * associations on the same socket. For a UDP-style socket with
7007 * multiple associations, it is possible for it to be "unwriteable"
7008 * prematurely. I assume that this is acceptable because
7009 * a premature "unwriteable" is better than an accidental "writeable" which
7010 * would cause an unwanted block under certain circumstances. For the 1-1
7011 * UDP-style sockets or TCP-style sockets, this code should work.
7012 * - Daisy
7014 static int sctp_writeable(struct sock *sk)
7016 int amt = 0;
7018 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7019 if (amt < 0)
7020 amt = 0;
7021 return amt;
7024 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7025 * returns immediately with EINPROGRESS.
7027 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7029 struct sock *sk = asoc->base.sk;
7030 int err = 0;
7031 long current_timeo = *timeo_p;
7032 DEFINE_WAIT(wait);
7034 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7036 /* Increment the association's refcnt. */
7037 sctp_association_hold(asoc);
7039 for (;;) {
7040 prepare_to_wait_exclusive(&asoc->wait, &wait,
7041 TASK_INTERRUPTIBLE);
7042 if (!*timeo_p)
7043 goto do_nonblock;
7044 if (sk->sk_shutdown & RCV_SHUTDOWN)
7045 break;
7046 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7047 asoc->base.dead)
7048 goto do_error;
7049 if (signal_pending(current))
7050 goto do_interrupted;
7052 if (sctp_state(asoc, ESTABLISHED))
7053 break;
7055 /* Let another process have a go. Since we are going
7056 * to sleep anyway.
7058 release_sock(sk);
7059 current_timeo = schedule_timeout(current_timeo);
7060 lock_sock(sk);
7062 *timeo_p = current_timeo;
7065 out:
7066 finish_wait(&asoc->wait, &wait);
7068 /* Release the association's refcnt. */
7069 sctp_association_put(asoc);
7071 return err;
7073 do_error:
7074 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7075 err = -ETIMEDOUT;
7076 else
7077 err = -ECONNREFUSED;
7078 goto out;
7080 do_interrupted:
7081 err = sock_intr_errno(*timeo_p);
7082 goto out;
7084 do_nonblock:
7085 err = -EINPROGRESS;
7086 goto out;
7089 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7091 struct sctp_endpoint *ep;
7092 int err = 0;
7093 DEFINE_WAIT(wait);
7095 ep = sctp_sk(sk)->ep;
7098 for (;;) {
7099 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7100 TASK_INTERRUPTIBLE);
7102 if (list_empty(&ep->asocs)) {
7103 release_sock(sk);
7104 timeo = schedule_timeout(timeo);
7105 lock_sock(sk);
7108 err = -EINVAL;
7109 if (!sctp_sstate(sk, LISTENING))
7110 break;
7112 err = 0;
7113 if (!list_empty(&ep->asocs))
7114 break;
7116 err = sock_intr_errno(timeo);
7117 if (signal_pending(current))
7118 break;
7120 err = -EAGAIN;
7121 if (!timeo)
7122 break;
7125 finish_wait(sk_sleep(sk), &wait);
7127 return err;
7130 static void sctp_wait_for_close(struct sock *sk, long timeout)
7132 DEFINE_WAIT(wait);
7134 do {
7135 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7136 if (list_empty(&sctp_sk(sk)->ep->asocs))
7137 break;
7138 release_sock(sk);
7139 timeout = schedule_timeout(timeout);
7140 lock_sock(sk);
7141 } while (!signal_pending(current) && timeout);
7143 finish_wait(sk_sleep(sk), &wait);
7146 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7148 struct sk_buff *frag;
7150 if (!skb->data_len)
7151 goto done;
7153 /* Don't forget the fragments. */
7154 skb_walk_frags(skb, frag)
7155 sctp_skb_set_owner_r_frag(frag, sk);
7157 done:
7158 sctp_skb_set_owner_r(skb, sk);
7161 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7162 struct sctp_association *asoc)
7164 struct inet_sock *inet = inet_sk(sk);
7165 struct inet_sock *newinet;
7167 newsk->sk_type = sk->sk_type;
7168 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7169 newsk->sk_flags = sk->sk_flags;
7170 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7171 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7172 newsk->sk_reuse = sk->sk_reuse;
7174 newsk->sk_shutdown = sk->sk_shutdown;
7175 newsk->sk_destruct = sctp_destruct_sock;
7176 newsk->sk_family = sk->sk_family;
7177 newsk->sk_protocol = IPPROTO_SCTP;
7178 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7179 newsk->sk_sndbuf = sk->sk_sndbuf;
7180 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7181 newsk->sk_lingertime = sk->sk_lingertime;
7182 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7183 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7185 newinet = inet_sk(newsk);
7187 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7188 * getsockname() and getpeername()
7190 newinet->inet_sport = inet->inet_sport;
7191 newinet->inet_saddr = inet->inet_saddr;
7192 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7193 newinet->inet_dport = htons(asoc->peer.port);
7194 newinet->pmtudisc = inet->pmtudisc;
7195 newinet->inet_id = asoc->next_tsn ^ jiffies;
7197 newinet->uc_ttl = inet->uc_ttl;
7198 newinet->mc_loop = 1;
7199 newinet->mc_ttl = 1;
7200 newinet->mc_index = 0;
7201 newinet->mc_list = NULL;
7204 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7205 * and its messages to the newsk.
7207 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7208 struct sctp_association *assoc,
7209 sctp_socket_type_t type)
7211 struct sctp_sock *oldsp = sctp_sk(oldsk);
7212 struct sctp_sock *newsp = sctp_sk(newsk);
7213 struct sctp_bind_bucket *pp; /* hash list port iterator */
7214 struct sctp_endpoint *newep = newsp->ep;
7215 struct sk_buff *skb, *tmp;
7216 struct sctp_ulpevent *event;
7217 struct sctp_bind_hashbucket *head;
7218 struct list_head tmplist;
7220 /* Migrate socket buffer sizes and all the socket level options to the
7221 * new socket.
7223 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7224 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7225 /* Brute force copy old sctp opt. */
7226 if (oldsp->do_auto_asconf) {
7227 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
7228 inet_sk_copy_descendant(newsk, oldsk);
7229 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
7230 } else
7231 inet_sk_copy_descendant(newsk, oldsk);
7233 /* Restore the ep value that was overwritten with the above structure
7234 * copy.
7236 newsp->ep = newep;
7237 newsp->hmac = NULL;
7239 /* Hook this new socket in to the bind_hash list. */
7240 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7241 inet_sk(oldsk)->inet_num)];
7242 local_bh_disable();
7243 spin_lock(&head->lock);
7244 pp = sctp_sk(oldsk)->bind_hash;
7245 sk_add_bind_node(newsk, &pp->owner);
7246 sctp_sk(newsk)->bind_hash = pp;
7247 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7248 spin_unlock(&head->lock);
7249 local_bh_enable();
7251 /* Copy the bind_addr list from the original endpoint to the new
7252 * endpoint so that we can handle restarts properly
7254 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7255 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7257 /* Move any messages in the old socket's receive queue that are for the
7258 * peeled off association to the new socket's receive queue.
7260 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7261 event = sctp_skb2event(skb);
7262 if (event->asoc == assoc) {
7263 __skb_unlink(skb, &oldsk->sk_receive_queue);
7264 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7265 sctp_skb_set_owner_r_frag(skb, newsk);
7269 /* Clean up any messages pending delivery due to partial
7270 * delivery. Three cases:
7271 * 1) No partial deliver; no work.
7272 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7273 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7275 skb_queue_head_init(&newsp->pd_lobby);
7276 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7278 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7279 struct sk_buff_head *queue;
7281 /* Decide which queue to move pd_lobby skbs to. */
7282 if (assoc->ulpq.pd_mode) {
7283 queue = &newsp->pd_lobby;
7284 } else
7285 queue = &newsk->sk_receive_queue;
7287 /* Walk through the pd_lobby, looking for skbs that
7288 * need moved to the new socket.
7290 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7291 event = sctp_skb2event(skb);
7292 if (event->asoc == assoc) {
7293 __skb_unlink(skb, &oldsp->pd_lobby);
7294 __skb_queue_tail(queue, skb);
7295 sctp_skb_set_owner_r_frag(skb, newsk);
7299 /* Clear up any skbs waiting for the partial
7300 * delivery to finish.
7302 if (assoc->ulpq.pd_mode)
7303 sctp_clear_pd(oldsk, NULL);
7307 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7308 sctp_skb_set_owner_r_frag(skb, newsk);
7310 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7311 sctp_skb_set_owner_r_frag(skb, newsk);
7313 /* Set the type of socket to indicate that it is peeled off from the
7314 * original UDP-style socket or created with the accept() call on a
7315 * TCP-style socket..
7317 newsp->type = type;
7319 /* Mark the new socket "in-use" by the user so that any packets
7320 * that may arrive on the association after we've moved it are
7321 * queued to the backlog. This prevents a potential race between
7322 * backlog processing on the old socket and new-packet processing
7323 * on the new socket.
7325 * The caller has just allocated newsk so we can guarantee that other
7326 * paths won't try to lock it and then oldsk.
7328 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7329 sctp_assoc_migrate(assoc, newsk);
7331 /* If the association on the newsk is already closed before accept()
7332 * is called, set RCV_SHUTDOWN flag.
7334 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7335 newsk->sk_shutdown |= RCV_SHUTDOWN;
7337 newsk->sk_state = SCTP_SS_ESTABLISHED;
7338 release_sock(newsk);
7342 /* This proto struct describes the ULP interface for SCTP. */
7343 struct proto sctp_prot = {
7344 .name = "SCTP",
7345 .owner = THIS_MODULE,
7346 .close = sctp_close,
7347 .connect = sctp_connect,
7348 .disconnect = sctp_disconnect,
7349 .accept = sctp_accept,
7350 .ioctl = sctp_ioctl,
7351 .init = sctp_init_sock,
7352 .destroy = sctp_destroy_sock,
7353 .shutdown = sctp_shutdown,
7354 .setsockopt = sctp_setsockopt,
7355 .getsockopt = sctp_getsockopt,
7356 .sendmsg = sctp_sendmsg,
7357 .recvmsg = sctp_recvmsg,
7358 .bind = sctp_bind,
7359 .backlog_rcv = sctp_backlog_rcv,
7360 .hash = sctp_hash,
7361 .unhash = sctp_unhash,
7362 .get_port = sctp_get_port,
7363 .obj_size = sizeof(struct sctp_sock),
7364 .sysctl_mem = sysctl_sctp_mem,
7365 .sysctl_rmem = sysctl_sctp_rmem,
7366 .sysctl_wmem = sysctl_sctp_wmem,
7367 .memory_pressure = &sctp_memory_pressure,
7368 .enter_memory_pressure = sctp_enter_memory_pressure,
7369 .memory_allocated = &sctp_memory_allocated,
7370 .sockets_allocated = &sctp_sockets_allocated,
7373 #if IS_ENABLED(CONFIG_IPV6)
7375 struct proto sctpv6_prot = {
7376 .name = "SCTPv6",
7377 .owner = THIS_MODULE,
7378 .close = sctp_close,
7379 .connect = sctp_connect,
7380 .disconnect = sctp_disconnect,
7381 .accept = sctp_accept,
7382 .ioctl = sctp_ioctl,
7383 .init = sctp_init_sock,
7384 .destroy = sctp_destroy_sock,
7385 .shutdown = sctp_shutdown,
7386 .setsockopt = sctp_setsockopt,
7387 .getsockopt = sctp_getsockopt,
7388 .sendmsg = sctp_sendmsg,
7389 .recvmsg = sctp_recvmsg,
7390 .bind = sctp_bind,
7391 .backlog_rcv = sctp_backlog_rcv,
7392 .hash = sctp_hash,
7393 .unhash = sctp_unhash,
7394 .get_port = sctp_get_port,
7395 .obj_size = sizeof(struct sctp6_sock),
7396 .sysctl_mem = sysctl_sctp_mem,
7397 .sysctl_rmem = sysctl_sctp_rmem,
7398 .sysctl_wmem = sysctl_sctp_wmem,
7399 .memory_pressure = &sctp_memory_pressure,
7400 .enter_memory_pressure = sctp_enter_memory_pressure,
7401 .memory_allocated = &sctp_memory_allocated,
7402 .sockets_allocated = &sctp_sockets_allocated,
7404 #endif /* IS_ENABLED(CONFIG_IPV6) */