sysfs/cpu: Add vulnerability folder
[linux/fpc-iii.git] / net / sctp / socket.c
bloba870d27ca7780e2c31583df2eb112d3029012460
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 *asoc, long *timeo_p,
86 size_t msg_len, struct sock **orig_sk);
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 static int sctp_memory_pressure;
106 static atomic_long_t sctp_memory_allocated;
107 struct percpu_counter sctp_sockets_allocated;
109 static void sctp_enter_memory_pressure(struct sock *sk)
111 sctp_memory_pressure = 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association *asoc)
118 int amt;
120 if (asoc->ep->sndbuf_policy)
121 amt = asoc->sndbuf_used;
122 else
123 amt = sk_wmem_alloc_get(asoc->base.sk);
125 if (amt >= asoc->base.sk->sk_sndbuf) {
126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
127 amt = 0;
128 else {
129 amt = sk_stream_wspace(asoc->base.sk);
130 if (amt < 0)
131 amt = 0;
133 } else {
134 amt = asoc->base.sk->sk_sndbuf - amt;
136 return amt;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
146 * tracking.
148 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
150 struct sctp_association *asoc = chunk->asoc;
151 struct sock *sk = asoc->base.sk;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc);
156 skb_set_owner_w(chunk->skb, sk);
158 chunk->skb->destructor = sctp_wfree;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk->skb)->destructor_arg = chunk;
162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163 sizeof(struct sk_buff) +
164 sizeof(struct sctp_chunk);
166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
167 sk->sk_wmem_queued += chunk->skb->truesize;
168 sk_mem_charge(sk, chunk->skb->truesize);
171 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
173 skb_orphan(chunk->skb);
176 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
177 void (*cb)(struct sctp_chunk *))
180 struct sctp_outq *q = &asoc->outqueue;
181 struct sctp_transport *t;
182 struct sctp_chunk *chunk;
184 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
185 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
186 cb(chunk);
188 list_for_each_entry(chunk, &q->retransmit, list)
189 cb(chunk);
191 list_for_each_entry(chunk, &q->sacked, list)
192 cb(chunk);
194 list_for_each_entry(chunk, &q->abandoned, list)
195 cb(chunk);
197 list_for_each_entry(chunk, &q->out_chunk_list, list)
198 cb(chunk);
201 /* Verify that this is a valid address. */
202 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
203 int len)
205 struct sctp_af *af;
207 /* Verify basic sockaddr. */
208 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
209 if (!af)
210 return -EINVAL;
212 /* Is this a valid SCTP address? */
213 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
214 return -EINVAL;
216 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
217 return -EINVAL;
219 return 0;
222 /* Look up the association by its id. If this is not a UDP-style
223 * socket, the ID field is always ignored.
225 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
227 struct sctp_association *asoc = NULL;
229 /* If this is not a UDP-style socket, assoc id should be ignored. */
230 if (!sctp_style(sk, UDP)) {
231 /* Return NULL if the socket state is not ESTABLISHED. It
232 * could be a TCP-style listening socket or a socket which
233 * hasn't yet called connect() to establish an association.
235 if (!sctp_sstate(sk, ESTABLISHED))
236 return NULL;
238 /* Get the first and the only association from the list. */
239 if (!list_empty(&sctp_sk(sk)->ep->asocs))
240 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
241 struct sctp_association, asocs);
242 return asoc;
245 /* Otherwise this is a UDP-style socket. */
246 if (!id || (id == (sctp_assoc_t)-1))
247 return NULL;
249 spin_lock_bh(&sctp_assocs_id_lock);
250 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
251 spin_unlock_bh(&sctp_assocs_id_lock);
253 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
254 return NULL;
256 return asoc;
259 /* Look up the transport from an address and an assoc id. If both address and
260 * id are specified, the associations matching the address and the id should be
261 * the same.
263 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
264 struct sockaddr_storage *addr,
265 sctp_assoc_t id)
267 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
268 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
269 union sctp_addr *laddr = (union sctp_addr *)addr;
270 struct sctp_transport *transport;
272 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
273 return NULL;
275 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
276 laddr,
277 &transport);
279 if (!addr_asoc)
280 return NULL;
282 id_asoc = sctp_id2assoc(sk, id);
283 if (id_asoc && (id_asoc != addr_asoc))
284 return NULL;
286 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
287 (union sctp_addr *)addr);
289 return transport;
292 /* API 3.1.2 bind() - UDP Style Syntax
293 * The syntax of bind() is,
295 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
297 * sd - the socket descriptor returned by socket().
298 * addr - the address structure (struct sockaddr_in or struct
299 * sockaddr_in6 [RFC 2553]),
300 * addr_len - the size of the address structure.
302 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
304 int retval = 0;
306 lock_sock(sk);
308 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
309 addr, addr_len);
311 /* Disallow binding twice. */
312 if (!sctp_sk(sk)->ep->base.bind_addr.port)
313 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
314 addr_len);
315 else
316 retval = -EINVAL;
318 release_sock(sk);
320 return retval;
323 static long sctp_get_port_local(struct sock *, union sctp_addr *);
325 /* Verify this is a valid sockaddr. */
326 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
327 union sctp_addr *addr, int len)
329 struct sctp_af *af;
331 /* Check minimum size. */
332 if (len < sizeof (struct sockaddr))
333 return NULL;
335 /* V4 mapped address are really of AF_INET family */
336 if (addr->sa.sa_family == AF_INET6 &&
337 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
338 if (!opt->pf->af_supported(AF_INET, opt))
339 return NULL;
340 } else {
341 /* Does this PF support this AF? */
342 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
343 return NULL;
346 /* If we get this far, af is valid. */
347 af = sctp_get_af_specific(addr->sa.sa_family);
349 if (len < af->sockaddr_len)
350 return NULL;
352 return af;
355 /* Bind a local address either to an endpoint or to an association. */
356 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
358 struct net *net = sock_net(sk);
359 struct sctp_sock *sp = sctp_sk(sk);
360 struct sctp_endpoint *ep = sp->ep;
361 struct sctp_bind_addr *bp = &ep->base.bind_addr;
362 struct sctp_af *af;
363 unsigned short snum;
364 int ret = 0;
366 /* Common sockaddr verification. */
367 af = sctp_sockaddr_af(sp, addr, len);
368 if (!af) {
369 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
370 __func__, sk, addr, len);
371 return -EINVAL;
374 snum = ntohs(addr->v4.sin_port);
376 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
377 __func__, sk, &addr->sa, bp->port, snum, len);
379 /* PF specific bind() address verification. */
380 if (!sp->pf->bind_verify(sp, addr))
381 return -EADDRNOTAVAIL;
383 /* We must either be unbound, or bind to the same port.
384 * It's OK to allow 0 ports if we are already bound.
385 * We'll just inhert an already bound port in this case
387 if (bp->port) {
388 if (!snum)
389 snum = bp->port;
390 else if (snum != bp->port) {
391 pr_debug("%s: new port %d doesn't match existing port "
392 "%d\n", __func__, snum, bp->port);
393 return -EINVAL;
397 if (snum && snum < PROT_SOCK &&
398 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
399 return -EACCES;
401 /* See if the address matches any of the addresses we may have
402 * already bound before checking against other endpoints.
404 if (sctp_bind_addr_match(bp, addr, sp))
405 return -EINVAL;
407 /* Make sure we are allowed to bind here.
408 * The function sctp_get_port_local() does duplicate address
409 * detection.
411 addr->v4.sin_port = htons(snum);
412 if ((ret = sctp_get_port_local(sk, addr))) {
413 return -EADDRINUSE;
416 /* Refresh ephemeral port. */
417 if (!bp->port)
418 bp->port = inet_sk(sk)->inet_num;
420 /* Add the address to the bind address list.
421 * Use GFP_ATOMIC since BHs will be disabled.
423 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
425 /* Copy back into socket for getsockname() use. */
426 if (!ret) {
427 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
428 sp->pf->to_sk_saddr(addr, sk);
431 return ret;
434 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
436 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
437 * at any one time. If a sender, after sending an ASCONF chunk, decides
438 * it needs to transfer another ASCONF Chunk, it MUST wait until the
439 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
440 * subsequent ASCONF. Note this restriction binds each side, so at any
441 * time two ASCONF may be in-transit on any given association (one sent
442 * from each endpoint).
444 static int sctp_send_asconf(struct sctp_association *asoc,
445 struct sctp_chunk *chunk)
447 struct net *net = sock_net(asoc->base.sk);
448 int retval = 0;
450 /* If there is an outstanding ASCONF chunk, queue it for later
451 * transmission.
453 if (asoc->addip_last_asconf) {
454 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
455 goto out;
458 /* Hold the chunk until an ASCONF_ACK is received. */
459 sctp_chunk_hold(chunk);
460 retval = sctp_primitive_ASCONF(net, asoc, chunk);
461 if (retval)
462 sctp_chunk_free(chunk);
463 else
464 asoc->addip_last_asconf = chunk;
466 out:
467 return retval;
470 /* Add a list of addresses as bind addresses to local endpoint or
471 * association.
473 * Basically run through each address specified in the addrs/addrcnt
474 * array/length pair, determine if it is IPv6 or IPv4 and call
475 * sctp_do_bind() on it.
477 * If any of them fails, then the operation will be reversed and the
478 * ones that were added will be removed.
480 * Only sctp_setsockopt_bindx() is supposed to call this function.
482 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
484 int cnt;
485 int retval = 0;
486 void *addr_buf;
487 struct sockaddr *sa_addr;
488 struct sctp_af *af;
490 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
491 addrs, addrcnt);
493 addr_buf = addrs;
494 for (cnt = 0; cnt < addrcnt; cnt++) {
495 /* The list may contain either IPv4 or IPv6 address;
496 * determine the address length for walking thru the list.
498 sa_addr = addr_buf;
499 af = sctp_get_af_specific(sa_addr->sa_family);
500 if (!af) {
501 retval = -EINVAL;
502 goto err_bindx_add;
505 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
506 af->sockaddr_len);
508 addr_buf += af->sockaddr_len;
510 err_bindx_add:
511 if (retval < 0) {
512 /* Failed. Cleanup the ones that have been added */
513 if (cnt > 0)
514 sctp_bindx_rem(sk, addrs, cnt);
515 return retval;
519 return retval;
522 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
523 * associations that are part of the endpoint indicating that a list of local
524 * addresses are added to the endpoint.
526 * If any of the addresses is already in the bind address list of the
527 * association, we do not send the chunk for that association. But it will not
528 * affect other associations.
530 * Only sctp_setsockopt_bindx() is supposed to call this function.
532 static int sctp_send_asconf_add_ip(struct sock *sk,
533 struct sockaddr *addrs,
534 int addrcnt)
536 struct net *net = sock_net(sk);
537 struct sctp_sock *sp;
538 struct sctp_endpoint *ep;
539 struct sctp_association *asoc;
540 struct sctp_bind_addr *bp;
541 struct sctp_chunk *chunk;
542 struct sctp_sockaddr_entry *laddr;
543 union sctp_addr *addr;
544 union sctp_addr saveaddr;
545 void *addr_buf;
546 struct sctp_af *af;
547 struct list_head *p;
548 int i;
549 int retval = 0;
551 if (!net->sctp.addip_enable)
552 return retval;
554 sp = sctp_sk(sk);
555 ep = sp->ep;
557 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
558 __func__, sk, addrs, addrcnt);
560 list_for_each_entry(asoc, &ep->asocs, asocs) {
561 if (!asoc->peer.asconf_capable)
562 continue;
564 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
565 continue;
567 if (!sctp_state(asoc, ESTABLISHED))
568 continue;
570 /* Check if any address in the packed array of addresses is
571 * in the bind address list of the association. If so,
572 * do not send the asconf chunk to its peer, but continue with
573 * other associations.
575 addr_buf = addrs;
576 for (i = 0; i < addrcnt; i++) {
577 addr = addr_buf;
578 af = sctp_get_af_specific(addr->v4.sin_family);
579 if (!af) {
580 retval = -EINVAL;
581 goto out;
584 if (sctp_assoc_lookup_laddr(asoc, addr))
585 break;
587 addr_buf += af->sockaddr_len;
589 if (i < addrcnt)
590 continue;
592 /* Use the first valid address in bind addr list of
593 * association as Address Parameter of ASCONF CHUNK.
595 bp = &asoc->base.bind_addr;
596 p = bp->address_list.next;
597 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
598 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
599 addrcnt, SCTP_PARAM_ADD_IP);
600 if (!chunk) {
601 retval = -ENOMEM;
602 goto out;
605 /* Add the new addresses to the bind address list with
606 * use_as_src set to 0.
608 addr_buf = addrs;
609 for (i = 0; i < addrcnt; i++) {
610 addr = addr_buf;
611 af = sctp_get_af_specific(addr->v4.sin_family);
612 memcpy(&saveaddr, addr, af->sockaddr_len);
613 retval = sctp_add_bind_addr(bp, &saveaddr,
614 SCTP_ADDR_NEW, GFP_ATOMIC);
615 addr_buf += af->sockaddr_len;
617 if (asoc->src_out_of_asoc_ok) {
618 struct sctp_transport *trans;
620 list_for_each_entry(trans,
621 &asoc->peer.transport_addr_list, transports) {
622 /* Clear the source and route cache */
623 dst_release(trans->dst);
624 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
625 2*asoc->pathmtu, 4380));
626 trans->ssthresh = asoc->peer.i.a_rwnd;
627 trans->rto = asoc->rto_initial;
628 sctp_max_rto(asoc, trans);
629 trans->rtt = trans->srtt = trans->rttvar = 0;
630 sctp_transport_route(trans, NULL,
631 sctp_sk(asoc->base.sk));
634 retval = sctp_send_asconf(asoc, chunk);
637 out:
638 return retval;
641 /* Remove a list of addresses from bind addresses list. Do not remove the
642 * last address.
644 * Basically run through each address specified in the addrs/addrcnt
645 * array/length pair, determine if it is IPv6 or IPv4 and call
646 * sctp_del_bind() on it.
648 * If any of them fails, then the operation will be reversed and the
649 * ones that were removed will be added back.
651 * At least one address has to be left; if only one address is
652 * available, the operation will return -EBUSY.
654 * Only sctp_setsockopt_bindx() is supposed to call this function.
656 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
658 struct sctp_sock *sp = sctp_sk(sk);
659 struct sctp_endpoint *ep = sp->ep;
660 int cnt;
661 struct sctp_bind_addr *bp = &ep->base.bind_addr;
662 int retval = 0;
663 void *addr_buf;
664 union sctp_addr *sa_addr;
665 struct sctp_af *af;
667 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
668 __func__, sk, addrs, addrcnt);
670 addr_buf = addrs;
671 for (cnt = 0; cnt < addrcnt; cnt++) {
672 /* If the bind address list is empty or if there is only one
673 * bind address, there is nothing more to be removed (we need
674 * at least one address here).
676 if (list_empty(&bp->address_list) ||
677 (sctp_list_single_entry(&bp->address_list))) {
678 retval = -EBUSY;
679 goto err_bindx_rem;
682 sa_addr = addr_buf;
683 af = sctp_get_af_specific(sa_addr->sa.sa_family);
684 if (!af) {
685 retval = -EINVAL;
686 goto err_bindx_rem;
689 if (!af->addr_valid(sa_addr, sp, NULL)) {
690 retval = -EADDRNOTAVAIL;
691 goto err_bindx_rem;
694 if (sa_addr->v4.sin_port &&
695 sa_addr->v4.sin_port != htons(bp->port)) {
696 retval = -EINVAL;
697 goto err_bindx_rem;
700 if (!sa_addr->v4.sin_port)
701 sa_addr->v4.sin_port = htons(bp->port);
703 /* FIXME - There is probably a need to check if sk->sk_saddr and
704 * sk->sk_rcv_addr are currently set to one of the addresses to
705 * be removed. This is something which needs to be looked into
706 * when we are fixing the outstanding issues with multi-homing
707 * socket routing and failover schemes. Refer to comments in
708 * sctp_do_bind(). -daisy
710 retval = sctp_del_bind_addr(bp, sa_addr);
712 addr_buf += af->sockaddr_len;
713 err_bindx_rem:
714 if (retval < 0) {
715 /* Failed. Add the ones that has been removed back */
716 if (cnt > 0)
717 sctp_bindx_add(sk, addrs, cnt);
718 return retval;
722 return retval;
725 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
726 * the associations that are part of the endpoint indicating that a list of
727 * local addresses are removed from the endpoint.
729 * If any of the addresses is already in the bind address list of the
730 * association, we do not send the chunk for that association. But it will not
731 * affect other associations.
733 * Only sctp_setsockopt_bindx() is supposed to call this function.
735 static int sctp_send_asconf_del_ip(struct sock *sk,
736 struct sockaddr *addrs,
737 int addrcnt)
739 struct net *net = sock_net(sk);
740 struct sctp_sock *sp;
741 struct sctp_endpoint *ep;
742 struct sctp_association *asoc;
743 struct sctp_transport *transport;
744 struct sctp_bind_addr *bp;
745 struct sctp_chunk *chunk;
746 union sctp_addr *laddr;
747 void *addr_buf;
748 struct sctp_af *af;
749 struct sctp_sockaddr_entry *saddr;
750 int i;
751 int retval = 0;
752 int stored = 0;
754 chunk = NULL;
755 if (!net->sctp.addip_enable)
756 return retval;
758 sp = sctp_sk(sk);
759 ep = sp->ep;
761 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
762 __func__, sk, addrs, addrcnt);
764 list_for_each_entry(asoc, &ep->asocs, asocs) {
766 if (!asoc->peer.asconf_capable)
767 continue;
769 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
770 continue;
772 if (!sctp_state(asoc, ESTABLISHED))
773 continue;
775 /* Check if any address in the packed array of addresses is
776 * not present in the bind address list of the association.
777 * If so, do not send the asconf chunk to its peer, but
778 * continue with other associations.
780 addr_buf = addrs;
781 for (i = 0; i < addrcnt; i++) {
782 laddr = addr_buf;
783 af = sctp_get_af_specific(laddr->v4.sin_family);
784 if (!af) {
785 retval = -EINVAL;
786 goto out;
789 if (!sctp_assoc_lookup_laddr(asoc, laddr))
790 break;
792 addr_buf += af->sockaddr_len;
794 if (i < addrcnt)
795 continue;
797 /* Find one address in the association's bind address list
798 * that is not in the packed array of addresses. This is to
799 * make sure that we do not delete all the addresses in the
800 * association.
802 bp = &asoc->base.bind_addr;
803 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
804 addrcnt, sp);
805 if ((laddr == NULL) && (addrcnt == 1)) {
806 if (asoc->asconf_addr_del_pending)
807 continue;
808 asoc->asconf_addr_del_pending =
809 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
810 if (asoc->asconf_addr_del_pending == NULL) {
811 retval = -ENOMEM;
812 goto out;
814 asoc->asconf_addr_del_pending->sa.sa_family =
815 addrs->sa_family;
816 asoc->asconf_addr_del_pending->v4.sin_port =
817 htons(bp->port);
818 if (addrs->sa_family == AF_INET) {
819 struct sockaddr_in *sin;
821 sin = (struct sockaddr_in *)addrs;
822 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
823 } else if (addrs->sa_family == AF_INET6) {
824 struct sockaddr_in6 *sin6;
826 sin6 = (struct sockaddr_in6 *)addrs;
827 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
830 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
831 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
832 asoc->asconf_addr_del_pending);
834 asoc->src_out_of_asoc_ok = 1;
835 stored = 1;
836 goto skip_mkasconf;
839 if (laddr == NULL)
840 return -EINVAL;
842 /* We do not need RCU protection throughout this loop
843 * because this is done under a socket lock from the
844 * setsockopt call.
846 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
847 SCTP_PARAM_DEL_IP);
848 if (!chunk) {
849 retval = -ENOMEM;
850 goto out;
853 skip_mkasconf:
854 /* Reset use_as_src flag for the addresses in the bind address
855 * list that are to be deleted.
857 addr_buf = addrs;
858 for (i = 0; i < addrcnt; i++) {
859 laddr = addr_buf;
860 af = sctp_get_af_specific(laddr->v4.sin_family);
861 list_for_each_entry(saddr, &bp->address_list, list) {
862 if (sctp_cmp_addr_exact(&saddr->a, laddr))
863 saddr->state = SCTP_ADDR_DEL;
865 addr_buf += af->sockaddr_len;
868 /* Update the route and saddr entries for all the transports
869 * as some of the addresses in the bind address list are
870 * about to be deleted and cannot be used as source addresses.
872 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
873 transports) {
874 dst_release(transport->dst);
875 sctp_transport_route(transport, NULL,
876 sctp_sk(asoc->base.sk));
879 if (stored)
880 /* We don't need to transmit ASCONF */
881 continue;
882 retval = sctp_send_asconf(asoc, chunk);
884 out:
885 return retval;
888 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
889 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
891 struct sock *sk = sctp_opt2sk(sp);
892 union sctp_addr *addr;
893 struct sctp_af *af;
895 /* It is safe to write port space in caller. */
896 addr = &addrw->a;
897 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
898 af = sctp_get_af_specific(addr->sa.sa_family);
899 if (!af)
900 return -EINVAL;
901 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
902 return -EINVAL;
904 if (addrw->state == SCTP_ADDR_NEW)
905 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
906 else
907 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
910 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
912 * API 8.1
913 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
914 * int flags);
916 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
917 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
918 * or IPv6 addresses.
920 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
921 * Section 3.1.2 for this usage.
923 * addrs is a pointer to an array of one or more socket addresses. Each
924 * address is contained in its appropriate structure (i.e. struct
925 * sockaddr_in or struct sockaddr_in6) the family of the address type
926 * must be used to distinguish the address length (note that this
927 * representation is termed a "packed array" of addresses). The caller
928 * specifies the number of addresses in the array with addrcnt.
930 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
931 * -1, and sets errno to the appropriate error code.
933 * For SCTP, the port given in each socket address must be the same, or
934 * sctp_bindx() will fail, setting errno to EINVAL.
936 * The flags parameter is formed from the bitwise OR of zero or more of
937 * the following currently defined flags:
939 * SCTP_BINDX_ADD_ADDR
941 * SCTP_BINDX_REM_ADDR
943 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
944 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
945 * addresses from the association. The two flags are mutually exclusive;
946 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
947 * not remove all addresses from an association; sctp_bindx() will
948 * reject such an attempt with EINVAL.
950 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
951 * additional addresses with an endpoint after calling bind(). Or use
952 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
953 * socket is associated with so that no new association accepted will be
954 * associated with those addresses. If the endpoint supports dynamic
955 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
956 * endpoint to send the appropriate message to the peer to change the
957 * peers address lists.
959 * Adding and removing addresses from a connected association is
960 * optional functionality. Implementations that do not support this
961 * functionality should return EOPNOTSUPP.
963 * Basically do nothing but copying the addresses from user to kernel
964 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
965 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
966 * from userspace.
968 * We don't use copy_from_user() for optimization: we first do the
969 * sanity checks (buffer size -fast- and access check-healthy
970 * pointer); if all of those succeed, then we can alloc the memory
971 * (expensive operation) needed to copy the data to kernel. Then we do
972 * the copying without checking the user space area
973 * (__copy_from_user()).
975 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
976 * it.
978 * sk The sk of the socket
979 * addrs The pointer to the addresses in user land
980 * addrssize Size of the addrs buffer
981 * op Operation to perform (add or remove, see the flags of
982 * sctp_bindx)
984 * Returns 0 if ok, <0 errno code on error.
986 static int sctp_setsockopt_bindx(struct sock *sk,
987 struct sockaddr __user *addrs,
988 int addrs_size, int op)
990 struct sockaddr *kaddrs;
991 int err;
992 int addrcnt = 0;
993 int walk_size = 0;
994 struct sockaddr *sa_addr;
995 void *addr_buf;
996 struct sctp_af *af;
998 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
999 __func__, sk, addrs, addrs_size, op);
1001 if (unlikely(addrs_size <= 0))
1002 return -EINVAL;
1004 /* Check the user passed a healthy pointer. */
1005 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1006 return -EFAULT;
1008 /* Alloc space for the address array in kernel memory. */
1009 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
1010 if (unlikely(!kaddrs))
1011 return -ENOMEM;
1013 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1014 kfree(kaddrs);
1015 return -EFAULT;
1018 /* Walk through the addrs buffer and count the number of addresses. */
1019 addr_buf = kaddrs;
1020 while (walk_size < addrs_size) {
1021 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1022 kfree(kaddrs);
1023 return -EINVAL;
1026 sa_addr = addr_buf;
1027 af = sctp_get_af_specific(sa_addr->sa_family);
1029 /* If the address family is not supported or if this address
1030 * causes the address buffer to overflow return EINVAL.
1032 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1033 kfree(kaddrs);
1034 return -EINVAL;
1036 addrcnt++;
1037 addr_buf += af->sockaddr_len;
1038 walk_size += af->sockaddr_len;
1041 /* Do the work. */
1042 switch (op) {
1043 case SCTP_BINDX_ADD_ADDR:
1044 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1045 if (err)
1046 goto out;
1047 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1048 break;
1050 case SCTP_BINDX_REM_ADDR:
1051 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1052 if (err)
1053 goto out;
1054 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1055 break;
1057 default:
1058 err = -EINVAL;
1059 break;
1062 out:
1063 kfree(kaddrs);
1065 return err;
1068 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1070 * Common routine for handling connect() and sctp_connectx().
1071 * Connect will come in with just a single address.
1073 static int __sctp_connect(struct sock *sk,
1074 struct sockaddr *kaddrs,
1075 int addrs_size,
1076 sctp_assoc_t *assoc_id)
1078 struct net *net = sock_net(sk);
1079 struct sctp_sock *sp;
1080 struct sctp_endpoint *ep;
1081 struct sctp_association *asoc = NULL;
1082 struct sctp_association *asoc2;
1083 struct sctp_transport *transport;
1084 union sctp_addr to;
1085 sctp_scope_t scope;
1086 long timeo;
1087 int err = 0;
1088 int addrcnt = 0;
1089 int walk_size = 0;
1090 union sctp_addr *sa_addr = NULL;
1091 void *addr_buf;
1092 unsigned short port;
1093 unsigned int f_flags = 0;
1095 sp = sctp_sk(sk);
1096 ep = sp->ep;
1098 /* connect() cannot be done on a socket that is already in ESTABLISHED
1099 * state - UDP-style peeled off socket or a TCP-style socket that
1100 * is already connected.
1101 * It cannot be done even on a TCP-style listening socket.
1103 if (sctp_sstate(sk, ESTABLISHED) ||
1104 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1105 err = -EISCONN;
1106 goto out_free;
1109 /* Walk through the addrs buffer and count the number of addresses. */
1110 addr_buf = kaddrs;
1111 while (walk_size < addrs_size) {
1112 struct sctp_af *af;
1114 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1115 err = -EINVAL;
1116 goto out_free;
1119 sa_addr = addr_buf;
1120 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1122 /* If the address family is not supported or if this address
1123 * causes the address buffer to overflow return EINVAL.
1125 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1126 err = -EINVAL;
1127 goto out_free;
1130 port = ntohs(sa_addr->v4.sin_port);
1132 /* Save current address so we can work with it */
1133 memcpy(&to, sa_addr, af->sockaddr_len);
1135 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1136 if (err)
1137 goto out_free;
1139 /* Make sure the destination port is correctly set
1140 * in all addresses.
1142 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1143 err = -EINVAL;
1144 goto out_free;
1147 /* Check if there already is a matching association on the
1148 * endpoint (other than the one created here).
1150 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1151 if (asoc2 && asoc2 != asoc) {
1152 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1153 err = -EISCONN;
1154 else
1155 err = -EALREADY;
1156 goto out_free;
1159 /* If we could not find a matching association on the endpoint,
1160 * make sure that there is no peeled-off association matching
1161 * the peer address even on another socket.
1163 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1164 err = -EADDRNOTAVAIL;
1165 goto out_free;
1168 if (!asoc) {
1169 /* If a bind() or sctp_bindx() is not called prior to
1170 * an sctp_connectx() call, the system picks an
1171 * ephemeral port and will choose an address set
1172 * equivalent to binding with a wildcard address.
1174 if (!ep->base.bind_addr.port) {
1175 if (sctp_autobind(sk)) {
1176 err = -EAGAIN;
1177 goto out_free;
1179 } else {
1181 * If an unprivileged user inherits a 1-many
1182 * style socket with open associations on a
1183 * privileged port, it MAY be permitted to
1184 * accept new associations, but it SHOULD NOT
1185 * be permitted to open new associations.
1187 if (ep->base.bind_addr.port < PROT_SOCK &&
1188 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1189 err = -EACCES;
1190 goto out_free;
1194 scope = sctp_scope(&to);
1195 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1196 if (!asoc) {
1197 err = -ENOMEM;
1198 goto out_free;
1201 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1202 GFP_KERNEL);
1203 if (err < 0) {
1204 goto out_free;
1209 /* Prime the peer's transport structures. */
1210 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1211 SCTP_UNKNOWN);
1212 if (!transport) {
1213 err = -ENOMEM;
1214 goto out_free;
1217 addrcnt++;
1218 addr_buf += af->sockaddr_len;
1219 walk_size += af->sockaddr_len;
1222 /* In case the user of sctp_connectx() wants an association
1223 * id back, assign one now.
1225 if (assoc_id) {
1226 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1227 if (err < 0)
1228 goto out_free;
1231 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1232 if (err < 0) {
1233 goto out_free;
1236 /* Initialize sk's dport and daddr for getpeername() */
1237 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1238 sp->pf->to_sk_daddr(sa_addr, sk);
1239 sk->sk_err = 0;
1241 /* in-kernel sockets don't generally have a file allocated to them
1242 * if all they do is call sock_create_kern().
1244 if (sk->sk_socket->file)
1245 f_flags = sk->sk_socket->file->f_flags;
1247 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1249 if (assoc_id)
1250 *assoc_id = asoc->assoc_id;
1251 err = sctp_wait_for_connect(asoc, &timeo);
1252 /* Note: the asoc may be freed after the return of
1253 * sctp_wait_for_connect.
1256 /* Don't free association on exit. */
1257 asoc = NULL;
1259 out_free:
1260 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1261 __func__, asoc, kaddrs, err);
1263 if (asoc) {
1264 /* sctp_primitive_ASSOCIATE may have added this association
1265 * To the hash table, try to unhash it, just in case, its a noop
1266 * if it wasn't hashed so we're safe
1268 sctp_unhash_established(asoc);
1269 sctp_association_free(asoc);
1271 return err;
1274 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1276 * API 8.9
1277 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1278 * sctp_assoc_t *asoc);
1280 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1281 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1282 * or IPv6 addresses.
1284 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1285 * Section 3.1.2 for this usage.
1287 * addrs is a pointer to an array of one or more socket addresses. Each
1288 * address is contained in its appropriate structure (i.e. struct
1289 * sockaddr_in or struct sockaddr_in6) the family of the address type
1290 * must be used to distengish the address length (note that this
1291 * representation is termed a "packed array" of addresses). The caller
1292 * specifies the number of addresses in the array with addrcnt.
1294 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1295 * the association id of the new association. On failure, sctp_connectx()
1296 * returns -1, and sets errno to the appropriate error code. The assoc_id
1297 * is not touched by the kernel.
1299 * For SCTP, the port given in each socket address must be the same, or
1300 * sctp_connectx() will fail, setting errno to EINVAL.
1302 * An application can use sctp_connectx to initiate an association with
1303 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1304 * allows a caller to specify multiple addresses at which a peer can be
1305 * reached. The way the SCTP stack uses the list of addresses to set up
1306 * the association is implementation dependent. This function only
1307 * specifies that the stack will try to make use of all the addresses in
1308 * the list when needed.
1310 * Note that the list of addresses passed in is only used for setting up
1311 * the association. It does not necessarily equal the set of addresses
1312 * the peer uses for the resulting association. If the caller wants to
1313 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1314 * retrieve them after the association has been set up.
1316 * Basically do nothing but copying the addresses from user to kernel
1317 * land and invoking either sctp_connectx(). This is used for tunneling
1318 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1320 * We don't use copy_from_user() for optimization: we first do the
1321 * sanity checks (buffer size -fast- and access check-healthy
1322 * pointer); if all of those succeed, then we can alloc the memory
1323 * (expensive operation) needed to copy the data to kernel. Then we do
1324 * the copying without checking the user space area
1325 * (__copy_from_user()).
1327 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1328 * it.
1330 * sk The sk of the socket
1331 * addrs The pointer to the addresses in user land
1332 * addrssize Size of the addrs buffer
1334 * Returns >=0 if ok, <0 errno code on error.
1336 static int __sctp_setsockopt_connectx(struct sock *sk,
1337 struct sockaddr __user *addrs,
1338 int addrs_size,
1339 sctp_assoc_t *assoc_id)
1341 struct sockaddr *kaddrs;
1342 gfp_t gfp = GFP_KERNEL;
1343 int err = 0;
1345 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1346 __func__, sk, addrs, addrs_size);
1348 if (unlikely(addrs_size <= 0))
1349 return -EINVAL;
1351 /* Check the user passed a healthy pointer. */
1352 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1353 return -EFAULT;
1355 /* Alloc space for the address array in kernel memory. */
1356 if (sk->sk_socket->file)
1357 gfp = GFP_USER | __GFP_NOWARN;
1358 kaddrs = kmalloc(addrs_size, gfp);
1359 if (unlikely(!kaddrs))
1360 return -ENOMEM;
1362 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1363 err = -EFAULT;
1364 } else {
1365 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1368 kfree(kaddrs);
1370 return err;
1374 * This is an older interface. It's kept for backward compatibility
1375 * to the option that doesn't provide association id.
1377 static int sctp_setsockopt_connectx_old(struct sock *sk,
1378 struct sockaddr __user *addrs,
1379 int addrs_size)
1381 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1385 * New interface for the API. The since the API is done with a socket
1386 * option, to make it simple we feed back the association id is as a return
1387 * indication to the call. Error is always negative and association id is
1388 * always positive.
1390 static int sctp_setsockopt_connectx(struct sock *sk,
1391 struct sockaddr __user *addrs,
1392 int addrs_size)
1394 sctp_assoc_t assoc_id = 0;
1395 int err = 0;
1397 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1399 if (err)
1400 return err;
1401 else
1402 return assoc_id;
1406 * New (hopefully final) interface for the API.
1407 * We use the sctp_getaddrs_old structure so that use-space library
1408 * can avoid any unnecessary allocations. The only different part
1409 * is that we store the actual length of the address buffer into the
1410 * addrs_num structure member. That way we can re-use the existing
1411 * code.
1413 #ifdef CONFIG_COMPAT
1414 struct compat_sctp_getaddrs_old {
1415 sctp_assoc_t assoc_id;
1416 s32 addr_num;
1417 compat_uptr_t addrs; /* struct sockaddr * */
1419 #endif
1421 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1422 char __user *optval,
1423 int __user *optlen)
1425 struct sctp_getaddrs_old param;
1426 sctp_assoc_t assoc_id = 0;
1427 int err = 0;
1429 #ifdef CONFIG_COMPAT
1430 if (is_compat_task()) {
1431 struct compat_sctp_getaddrs_old param32;
1433 if (len < sizeof(param32))
1434 return -EINVAL;
1435 if (copy_from_user(&param32, optval, sizeof(param32)))
1436 return -EFAULT;
1438 param.assoc_id = param32.assoc_id;
1439 param.addr_num = param32.addr_num;
1440 param.addrs = compat_ptr(param32.addrs);
1441 } else
1442 #endif
1444 if (len < sizeof(param))
1445 return -EINVAL;
1446 if (copy_from_user(&param, optval, sizeof(param)))
1447 return -EFAULT;
1450 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1451 param.addrs, param.addr_num,
1452 &assoc_id);
1453 if (err == 0 || err == -EINPROGRESS) {
1454 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1455 return -EFAULT;
1456 if (put_user(sizeof(assoc_id), optlen))
1457 return -EFAULT;
1460 return err;
1463 /* API 3.1.4 close() - UDP Style Syntax
1464 * Applications use close() to perform graceful shutdown (as described in
1465 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1466 * by a UDP-style socket.
1468 * The syntax is
1470 * ret = close(int sd);
1472 * sd - the socket descriptor of the associations to be closed.
1474 * To gracefully shutdown a specific association represented by the
1475 * UDP-style socket, an application should use the sendmsg() call,
1476 * passing no user data, but including the appropriate flag in the
1477 * ancillary data (see Section xxxx).
1479 * If sd in the close() call is a branched-off socket representing only
1480 * one association, the shutdown is performed on that association only.
1482 * 4.1.6 close() - TCP Style Syntax
1484 * Applications use close() to gracefully close down an association.
1486 * The syntax is:
1488 * int close(int sd);
1490 * sd - the socket descriptor of the association to be closed.
1492 * After an application calls close() on a socket descriptor, no further
1493 * socket operations will succeed on that descriptor.
1495 * API 7.1.4 SO_LINGER
1497 * An application using the TCP-style socket can use this option to
1498 * perform the SCTP ABORT primitive. The linger option structure is:
1500 * struct linger {
1501 * int l_onoff; // option on/off
1502 * int l_linger; // linger time
1503 * };
1505 * To enable the option, set l_onoff to 1. If the l_linger value is set
1506 * to 0, calling close() is the same as the ABORT primitive. If the
1507 * value is set to a negative value, the setsockopt() call will return
1508 * an error. If the value is set to a positive value linger_time, the
1509 * close() can be blocked for at most linger_time ms. If the graceful
1510 * shutdown phase does not finish during this period, close() will
1511 * return but the graceful shutdown phase continues in the system.
1513 static void sctp_close(struct sock *sk, long timeout)
1515 struct net *net = sock_net(sk);
1516 struct sctp_endpoint *ep;
1517 struct sctp_association *asoc;
1518 struct list_head *pos, *temp;
1519 unsigned int data_was_unread;
1521 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1523 lock_sock(sk);
1524 sk->sk_shutdown = SHUTDOWN_MASK;
1525 sk->sk_state = SCTP_SS_CLOSING;
1527 ep = sctp_sk(sk)->ep;
1529 /* Clean up any skbs sitting on the receive queue. */
1530 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1531 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1533 /* Walk all associations on an endpoint. */
1534 list_for_each_safe(pos, temp, &ep->asocs) {
1535 asoc = list_entry(pos, struct sctp_association, asocs);
1537 if (sctp_style(sk, TCP)) {
1538 /* A closed association can still be in the list if
1539 * it belongs to a TCP-style listening socket that is
1540 * not yet accepted. If so, free it. If not, send an
1541 * ABORT or SHUTDOWN based on the linger options.
1543 if (sctp_state(asoc, CLOSED)) {
1544 sctp_unhash_established(asoc);
1545 sctp_association_free(asoc);
1546 continue;
1550 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1551 !skb_queue_empty(&asoc->ulpq.reasm) ||
1552 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1553 struct sctp_chunk *chunk;
1555 chunk = sctp_make_abort_user(asoc, NULL, 0);
1556 sctp_primitive_ABORT(net, asoc, chunk);
1557 } else
1558 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1561 /* On a TCP-style socket, block for at most linger_time if set. */
1562 if (sctp_style(sk, TCP) && timeout)
1563 sctp_wait_for_close(sk, timeout);
1565 /* This will run the backlog queue. */
1566 release_sock(sk);
1568 /* Supposedly, no process has access to the socket, but
1569 * the net layers still may.
1570 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1571 * held and that should be grabbed before socket lock.
1573 spin_lock_bh(&net->sctp.addr_wq_lock);
1574 bh_lock_sock(sk);
1576 /* Hold the sock, since sk_common_release() will put sock_put()
1577 * and we have just a little more cleanup.
1579 sock_hold(sk);
1580 sk_common_release(sk);
1582 bh_unlock_sock(sk);
1583 spin_unlock_bh(&net->sctp.addr_wq_lock);
1585 sock_put(sk);
1587 SCTP_DBG_OBJCNT_DEC(sock);
1590 /* Handle EPIPE error. */
1591 static int sctp_error(struct sock *sk, int flags, int err)
1593 if (err == -EPIPE)
1594 err = sock_error(sk) ? : -EPIPE;
1595 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1596 send_sig(SIGPIPE, current, 0);
1597 return err;
1600 /* API 3.1.3 sendmsg() - UDP Style Syntax
1602 * An application uses sendmsg() and recvmsg() calls to transmit data to
1603 * and receive data from its peer.
1605 * ssize_t sendmsg(int socket, const struct msghdr *message,
1606 * int flags);
1608 * socket - the socket descriptor of the endpoint.
1609 * message - pointer to the msghdr structure which contains a single
1610 * user message and possibly some ancillary data.
1612 * See Section 5 for complete description of the data
1613 * structures.
1615 * flags - flags sent or received with the user message, see Section
1616 * 5 for complete description of the flags.
1618 * Note: This function could use a rewrite especially when explicit
1619 * connect support comes in.
1621 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1623 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1625 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1627 struct net *net = sock_net(sk);
1628 struct sctp_sock *sp;
1629 struct sctp_endpoint *ep;
1630 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1631 struct sctp_transport *transport, *chunk_tp;
1632 struct sctp_chunk *chunk;
1633 union sctp_addr to;
1634 struct sockaddr *msg_name = NULL;
1635 struct sctp_sndrcvinfo default_sinfo;
1636 struct sctp_sndrcvinfo *sinfo;
1637 struct sctp_initmsg *sinit;
1638 sctp_assoc_t associd = 0;
1639 sctp_cmsgs_t cmsgs = { NULL };
1640 sctp_scope_t scope;
1641 bool fill_sinfo_ttl = false, wait_connect = false;
1642 struct sctp_datamsg *datamsg;
1643 int msg_flags = msg->msg_flags;
1644 __u16 sinfo_flags = 0;
1645 long timeo;
1646 int err;
1648 err = 0;
1649 sp = sctp_sk(sk);
1650 ep = sp->ep;
1652 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1653 msg, msg_len, ep);
1655 /* We cannot send a message over a TCP-style listening socket. */
1656 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1657 err = -EPIPE;
1658 goto out_nounlock;
1661 /* Parse out the SCTP CMSGs. */
1662 err = sctp_msghdr_parse(msg, &cmsgs);
1663 if (err) {
1664 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1665 goto out_nounlock;
1668 /* Fetch the destination address for this packet. This
1669 * address only selects the association--it is not necessarily
1670 * the address we will send to.
1671 * For a peeled-off socket, msg_name is ignored.
1673 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1674 int msg_namelen = msg->msg_namelen;
1676 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1677 msg_namelen);
1678 if (err)
1679 return err;
1681 if (msg_namelen > sizeof(to))
1682 msg_namelen = sizeof(to);
1683 memcpy(&to, msg->msg_name, msg_namelen);
1684 msg_name = msg->msg_name;
1687 sinit = cmsgs.init;
1688 if (cmsgs.sinfo != NULL) {
1689 memset(&default_sinfo, 0, sizeof(default_sinfo));
1690 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1691 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1692 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1693 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1694 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1696 sinfo = &default_sinfo;
1697 fill_sinfo_ttl = true;
1698 } else {
1699 sinfo = cmsgs.srinfo;
1701 /* Did the user specify SNDINFO/SNDRCVINFO? */
1702 if (sinfo) {
1703 sinfo_flags = sinfo->sinfo_flags;
1704 associd = sinfo->sinfo_assoc_id;
1707 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1708 msg_len, sinfo_flags);
1710 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1711 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1712 err = -EINVAL;
1713 goto out_nounlock;
1716 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1717 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1718 * If SCTP_ABORT is set, the message length could be non zero with
1719 * the msg_iov set to the user abort reason.
1721 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1722 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1723 err = -EINVAL;
1724 goto out_nounlock;
1727 /* If SCTP_ADDR_OVER is set, there must be an address
1728 * specified in msg_name.
1730 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1731 err = -EINVAL;
1732 goto out_nounlock;
1735 transport = NULL;
1737 pr_debug("%s: about to look up association\n", __func__);
1739 lock_sock(sk);
1741 /* If a msg_name has been specified, assume this is to be used. */
1742 if (msg_name) {
1743 /* Look for a matching association on the endpoint. */
1744 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1745 if (!asoc) {
1746 /* If we could not find a matching association on the
1747 * endpoint, make sure that it is not a TCP-style
1748 * socket that already has an association or there is
1749 * no peeled-off association on another socket.
1751 if ((sctp_style(sk, TCP) &&
1752 sctp_sstate(sk, ESTABLISHED)) ||
1753 sctp_endpoint_is_peeled_off(ep, &to)) {
1754 err = -EADDRNOTAVAIL;
1755 goto out_unlock;
1758 } else {
1759 asoc = sctp_id2assoc(sk, associd);
1760 if (!asoc) {
1761 err = -EPIPE;
1762 goto out_unlock;
1766 if (asoc) {
1767 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1769 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1770 * socket that has an association in CLOSED state. This can
1771 * happen when an accepted socket has an association that is
1772 * already CLOSED.
1774 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1775 err = -EPIPE;
1776 goto out_unlock;
1779 if (sinfo_flags & SCTP_EOF) {
1780 pr_debug("%s: shutting down association:%p\n",
1781 __func__, asoc);
1783 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1784 err = 0;
1785 goto out_unlock;
1787 if (sinfo_flags & SCTP_ABORT) {
1789 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1790 if (!chunk) {
1791 err = -ENOMEM;
1792 goto out_unlock;
1795 pr_debug("%s: aborting association:%p\n",
1796 __func__, asoc);
1798 sctp_primitive_ABORT(net, asoc, chunk);
1799 err = 0;
1800 goto out_unlock;
1804 /* Do we need to create the association? */
1805 if (!asoc) {
1806 pr_debug("%s: there is no association yet\n", __func__);
1808 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1809 err = -EINVAL;
1810 goto out_unlock;
1813 /* Check for invalid stream against the stream counts,
1814 * either the default or the user specified stream counts.
1816 if (sinfo) {
1817 if (!sinit || !sinit->sinit_num_ostreams) {
1818 /* Check against the defaults. */
1819 if (sinfo->sinfo_stream >=
1820 sp->initmsg.sinit_num_ostreams) {
1821 err = -EINVAL;
1822 goto out_unlock;
1824 } else {
1825 /* Check against the requested. */
1826 if (sinfo->sinfo_stream >=
1827 sinit->sinit_num_ostreams) {
1828 err = -EINVAL;
1829 goto out_unlock;
1835 * API 3.1.2 bind() - UDP Style Syntax
1836 * If a bind() or sctp_bindx() is not called prior to a
1837 * sendmsg() call that initiates a new association, the
1838 * system picks an ephemeral port and will choose an address
1839 * set equivalent to binding with a wildcard address.
1841 if (!ep->base.bind_addr.port) {
1842 if (sctp_autobind(sk)) {
1843 err = -EAGAIN;
1844 goto out_unlock;
1846 } else {
1848 * If an unprivileged user inherits a one-to-many
1849 * style socket with open associations on a privileged
1850 * port, it MAY be permitted to accept new associations,
1851 * but it SHOULD NOT be permitted to open new
1852 * associations.
1854 if (ep->base.bind_addr.port < PROT_SOCK &&
1855 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1856 err = -EACCES;
1857 goto out_unlock;
1861 scope = sctp_scope(&to);
1862 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1863 if (!new_asoc) {
1864 err = -ENOMEM;
1865 goto out_unlock;
1867 asoc = new_asoc;
1868 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1869 if (err < 0) {
1870 err = -ENOMEM;
1871 goto out_free;
1874 /* If the SCTP_INIT ancillary data is specified, set all
1875 * the association init values accordingly.
1877 if (sinit) {
1878 if (sinit->sinit_num_ostreams) {
1879 asoc->c.sinit_num_ostreams =
1880 sinit->sinit_num_ostreams;
1882 if (sinit->sinit_max_instreams) {
1883 asoc->c.sinit_max_instreams =
1884 sinit->sinit_max_instreams;
1886 if (sinit->sinit_max_attempts) {
1887 asoc->max_init_attempts
1888 = sinit->sinit_max_attempts;
1890 if (sinit->sinit_max_init_timeo) {
1891 asoc->max_init_timeo =
1892 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1896 /* Prime the peer's transport structures. */
1897 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1898 if (!transport) {
1899 err = -ENOMEM;
1900 goto out_free;
1904 /* ASSERT: we have a valid association at this point. */
1905 pr_debug("%s: we have a valid association\n", __func__);
1907 if (!sinfo) {
1908 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1909 * one with some defaults.
1911 memset(&default_sinfo, 0, sizeof(default_sinfo));
1912 default_sinfo.sinfo_stream = asoc->default_stream;
1913 default_sinfo.sinfo_flags = asoc->default_flags;
1914 default_sinfo.sinfo_ppid = asoc->default_ppid;
1915 default_sinfo.sinfo_context = asoc->default_context;
1916 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1917 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1919 sinfo = &default_sinfo;
1920 } else if (fill_sinfo_ttl) {
1921 /* In case SNDINFO was specified, we still need to fill
1922 * it with a default ttl from the assoc here.
1924 sinfo->sinfo_timetolive = asoc->default_timetolive;
1927 /* API 7.1.7, the sndbuf size per association bounds the
1928 * maximum size of data that can be sent in a single send call.
1930 if (msg_len > sk->sk_sndbuf) {
1931 err = -EMSGSIZE;
1932 goto out_free;
1935 if (asoc->pmtu_pending)
1936 sctp_assoc_pending_pmtu(sk, asoc);
1938 /* If fragmentation is disabled and the message length exceeds the
1939 * association fragmentation point, return EMSGSIZE. The I-D
1940 * does not specify what this error is, but this looks like
1941 * a great fit.
1943 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1944 err = -EMSGSIZE;
1945 goto out_free;
1948 /* Check for invalid stream. */
1949 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1950 err = -EINVAL;
1951 goto out_free;
1954 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1955 if (!sctp_wspace(asoc)) {
1956 /* sk can be changed by peel off when waiting for buf. */
1957 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len, &sk);
1958 if (err) {
1959 if (err == -ESRCH) {
1960 /* asoc is already dead. */
1961 new_asoc = NULL;
1962 err = -EPIPE;
1964 goto out_free;
1968 /* If an address is passed with the sendto/sendmsg call, it is used
1969 * to override the primary destination address in the TCP model, or
1970 * when SCTP_ADDR_OVER flag is set in the UDP model.
1972 if ((sctp_style(sk, TCP) && msg_name) ||
1973 (sinfo_flags & SCTP_ADDR_OVER)) {
1974 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1975 if (!chunk_tp) {
1976 err = -EINVAL;
1977 goto out_free;
1979 } else
1980 chunk_tp = NULL;
1982 /* Auto-connect, if we aren't connected already. */
1983 if (sctp_state(asoc, CLOSED)) {
1984 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1985 if (err < 0)
1986 goto out_free;
1988 wait_connect = true;
1989 pr_debug("%s: we associated primitively\n", __func__);
1992 /* Break the message into multiple chunks of maximum size. */
1993 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1994 if (IS_ERR(datamsg)) {
1995 err = PTR_ERR(datamsg);
1996 goto out_free;
1999 /* Now send the (possibly) fragmented message. */
2000 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
2001 /* Do accounting for the write space. */
2002 sctp_set_owner_w(chunk);
2004 chunk->transport = chunk_tp;
2007 /* Send it to the lower layers. Note: all chunks
2008 * must either fail or succeed. The lower layer
2009 * works that way today. Keep it that way or this
2010 * breaks.
2012 err = sctp_primitive_SEND(net, asoc, datamsg);
2013 sctp_datamsg_put(datamsg);
2014 /* Did the lower layer accept the chunk? */
2015 if (err)
2016 goto out_free;
2018 pr_debug("%s: we sent primitively\n", __func__);
2020 err = msg_len;
2022 if (unlikely(wait_connect)) {
2023 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2024 sctp_wait_for_connect(asoc, &timeo);
2027 /* If we are already past ASSOCIATE, the lower
2028 * layers are responsible for association cleanup.
2030 goto out_unlock;
2032 out_free:
2033 if (new_asoc) {
2034 sctp_unhash_established(asoc);
2035 sctp_association_free(asoc);
2037 out_unlock:
2038 release_sock(sk);
2040 out_nounlock:
2041 return sctp_error(sk, msg_flags, err);
2043 #if 0
2044 do_sock_err:
2045 if (msg_len)
2046 err = msg_len;
2047 else
2048 err = sock_error(sk);
2049 goto out;
2051 do_interrupted:
2052 if (msg_len)
2053 err = msg_len;
2054 goto out;
2055 #endif /* 0 */
2058 /* This is an extended version of skb_pull() that removes the data from the
2059 * start of a skb even when data is spread across the list of skb's in the
2060 * frag_list. len specifies the total amount of data that needs to be removed.
2061 * when 'len' bytes could be removed from the skb, it returns 0.
2062 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2063 * could not be removed.
2065 static int sctp_skb_pull(struct sk_buff *skb, int len)
2067 struct sk_buff *list;
2068 int skb_len = skb_headlen(skb);
2069 int rlen;
2071 if (len <= skb_len) {
2072 __skb_pull(skb, len);
2073 return 0;
2075 len -= skb_len;
2076 __skb_pull(skb, skb_len);
2078 skb_walk_frags(skb, list) {
2079 rlen = sctp_skb_pull(list, len);
2080 skb->len -= (len-rlen);
2081 skb->data_len -= (len-rlen);
2083 if (!rlen)
2084 return 0;
2086 len = rlen;
2089 return len;
2092 /* API 3.1.3 recvmsg() - UDP Style Syntax
2094 * ssize_t recvmsg(int socket, struct msghdr *message,
2095 * int flags);
2097 * socket - the socket descriptor of the endpoint.
2098 * message - pointer to the msghdr structure which contains a single
2099 * user message and possibly some ancillary data.
2101 * See Section 5 for complete description of the data
2102 * structures.
2104 * flags - flags sent or received with the user message, see Section
2105 * 5 for complete description of the flags.
2107 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2108 int noblock, int flags, int *addr_len)
2110 struct sctp_ulpevent *event = NULL;
2111 struct sctp_sock *sp = sctp_sk(sk);
2112 struct sk_buff *skb;
2113 int copied;
2114 int err = 0;
2115 int skb_len;
2117 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2118 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2119 addr_len);
2121 lock_sock(sk);
2123 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2124 err = -ENOTCONN;
2125 goto out;
2128 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2129 if (!skb)
2130 goto out;
2132 /* Get the total length of the skb including any skb's in the
2133 * frag_list.
2135 skb_len = skb->len;
2137 copied = skb_len;
2138 if (copied > len)
2139 copied = len;
2141 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2143 event = sctp_skb2event(skb);
2145 if (err)
2146 goto out_free;
2148 sock_recv_ts_and_drops(msg, sk, skb);
2149 if (sctp_ulpevent_is_notification(event)) {
2150 msg->msg_flags |= MSG_NOTIFICATION;
2151 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2152 } else {
2153 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2156 /* Check if we allow SCTP_NXTINFO. */
2157 if (sp->recvnxtinfo)
2158 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2159 /* Check if we allow SCTP_RCVINFO. */
2160 if (sp->recvrcvinfo)
2161 sctp_ulpevent_read_rcvinfo(event, msg);
2162 /* Check if we allow SCTP_SNDRCVINFO. */
2163 if (sp->subscribe.sctp_data_io_event)
2164 sctp_ulpevent_read_sndrcvinfo(event, msg);
2166 err = copied;
2168 /* If skb's length exceeds the user's buffer, update the skb and
2169 * push it back to the receive_queue so that the next call to
2170 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2172 if (skb_len > copied) {
2173 msg->msg_flags &= ~MSG_EOR;
2174 if (flags & MSG_PEEK)
2175 goto out_free;
2176 sctp_skb_pull(skb, copied);
2177 skb_queue_head(&sk->sk_receive_queue, skb);
2179 /* When only partial message is copied to the user, increase
2180 * rwnd by that amount. If all the data in the skb is read,
2181 * rwnd is updated when the event is freed.
2183 if (!sctp_ulpevent_is_notification(event))
2184 sctp_assoc_rwnd_increase(event->asoc, copied);
2185 goto out;
2186 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2187 (event->msg_flags & MSG_EOR))
2188 msg->msg_flags |= MSG_EOR;
2189 else
2190 msg->msg_flags &= ~MSG_EOR;
2192 out_free:
2193 if (flags & MSG_PEEK) {
2194 /* Release the skb reference acquired after peeking the skb in
2195 * sctp_skb_recv_datagram().
2197 kfree_skb(skb);
2198 } else {
2199 /* Free the event which includes releasing the reference to
2200 * the owner of the skb, freeing the skb and updating the
2201 * rwnd.
2203 sctp_ulpevent_free(event);
2205 out:
2206 release_sock(sk);
2207 return err;
2210 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2212 * This option is a on/off flag. If enabled no SCTP message
2213 * fragmentation will be performed. Instead if a message being sent
2214 * exceeds the current PMTU size, the message will NOT be sent and
2215 * instead a error will be indicated to the user.
2217 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2218 char __user *optval,
2219 unsigned int optlen)
2221 int val;
2223 if (optlen < sizeof(int))
2224 return -EINVAL;
2226 if (get_user(val, (int __user *)optval))
2227 return -EFAULT;
2229 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2231 return 0;
2234 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2235 unsigned int optlen)
2237 struct sctp_association *asoc;
2238 struct sctp_ulpevent *event;
2240 if (optlen > sizeof(struct sctp_event_subscribe))
2241 return -EINVAL;
2242 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2243 return -EFAULT;
2245 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2246 * if there is no data to be sent or retransmit, the stack will
2247 * immediately send up this notification.
2249 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2250 &sctp_sk(sk)->subscribe)) {
2251 asoc = sctp_id2assoc(sk, 0);
2253 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2254 event = sctp_ulpevent_make_sender_dry_event(asoc,
2255 GFP_ATOMIC);
2256 if (!event)
2257 return -ENOMEM;
2259 sctp_ulpq_tail_event(&asoc->ulpq, event);
2263 return 0;
2266 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2268 * This socket option is applicable to the UDP-style socket only. When
2269 * set it will cause associations that are idle for more than the
2270 * specified number of seconds to automatically close. An association
2271 * being idle is defined an association that has NOT sent or received
2272 * user data. The special value of '0' indicates that no automatic
2273 * close of any associations should be performed. The option expects an
2274 * integer defining the number of seconds of idle time before an
2275 * association is closed.
2277 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2278 unsigned int optlen)
2280 struct sctp_sock *sp = sctp_sk(sk);
2281 struct net *net = sock_net(sk);
2283 /* Applicable to UDP-style socket only */
2284 if (sctp_style(sk, TCP))
2285 return -EOPNOTSUPP;
2286 if (optlen != sizeof(int))
2287 return -EINVAL;
2288 if (copy_from_user(&sp->autoclose, optval, optlen))
2289 return -EFAULT;
2291 if (sp->autoclose > net->sctp.max_autoclose)
2292 sp->autoclose = net->sctp.max_autoclose;
2294 return 0;
2297 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2299 * Applications can enable or disable heartbeats for any peer address of
2300 * an association, modify an address's heartbeat interval, force a
2301 * heartbeat to be sent immediately, and adjust the address's maximum
2302 * number of retransmissions sent before an address is considered
2303 * unreachable. The following structure is used to access and modify an
2304 * address's parameters:
2306 * struct sctp_paddrparams {
2307 * sctp_assoc_t spp_assoc_id;
2308 * struct sockaddr_storage spp_address;
2309 * uint32_t spp_hbinterval;
2310 * uint16_t spp_pathmaxrxt;
2311 * uint32_t spp_pathmtu;
2312 * uint32_t spp_sackdelay;
2313 * uint32_t spp_flags;
2314 * };
2316 * spp_assoc_id - (one-to-many style socket) This is filled in the
2317 * application, and identifies the association for
2318 * this query.
2319 * spp_address - This specifies which address is of interest.
2320 * spp_hbinterval - This contains the value of the heartbeat interval,
2321 * in milliseconds. If a value of zero
2322 * is present in this field then no changes are to
2323 * be made to this parameter.
2324 * spp_pathmaxrxt - This contains the maximum number of
2325 * retransmissions before this address shall be
2326 * considered unreachable. If a value of zero
2327 * is present in this field then no changes are to
2328 * be made to this parameter.
2329 * spp_pathmtu - When Path MTU discovery is disabled the value
2330 * specified here will be the "fixed" path mtu.
2331 * Note that if the spp_address field is empty
2332 * then all associations on this address will
2333 * have this fixed path mtu set upon them.
2335 * spp_sackdelay - When delayed sack is enabled, this value specifies
2336 * the number of milliseconds that sacks will be delayed
2337 * for. This value will apply to all addresses of an
2338 * association if the spp_address field is empty. Note
2339 * also, that if delayed sack is enabled and this
2340 * value is set to 0, no change is made to the last
2341 * recorded delayed sack timer value.
2343 * spp_flags - These flags are used to control various features
2344 * on an association. The flag field may contain
2345 * zero or more of the following options.
2347 * SPP_HB_ENABLE - Enable heartbeats on the
2348 * specified address. Note that if the address
2349 * field is empty all addresses for the association
2350 * have heartbeats enabled upon them.
2352 * SPP_HB_DISABLE - Disable heartbeats on the
2353 * speicifed address. Note that if the address
2354 * field is empty all addresses for the association
2355 * will have their heartbeats disabled. Note also
2356 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2357 * mutually exclusive, only one of these two should
2358 * be specified. Enabling both fields will have
2359 * undetermined results.
2361 * SPP_HB_DEMAND - Request a user initiated heartbeat
2362 * to be made immediately.
2364 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2365 * heartbeat delayis to be set to the value of 0
2366 * milliseconds.
2368 * SPP_PMTUD_ENABLE - This field will enable PMTU
2369 * discovery upon the specified address. Note that
2370 * if the address feild is empty then all addresses
2371 * on the association are effected.
2373 * SPP_PMTUD_DISABLE - This field will disable PMTU
2374 * discovery upon the specified address. Note that
2375 * if the address feild is empty then all addresses
2376 * on the association are effected. Not also that
2377 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2378 * exclusive. Enabling both will have undetermined
2379 * results.
2381 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2382 * on delayed sack. The time specified in spp_sackdelay
2383 * is used to specify the sack delay for this address. Note
2384 * that if spp_address is empty then all addresses will
2385 * enable delayed sack and take on the sack delay
2386 * value specified in spp_sackdelay.
2387 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2388 * off delayed sack. If the spp_address field is blank then
2389 * delayed sack is disabled for the entire association. Note
2390 * also that this field is mutually exclusive to
2391 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2392 * results.
2394 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2395 struct sctp_transport *trans,
2396 struct sctp_association *asoc,
2397 struct sctp_sock *sp,
2398 int hb_change,
2399 int pmtud_change,
2400 int sackdelay_change)
2402 int error;
2404 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2405 struct net *net = sock_net(trans->asoc->base.sk);
2407 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2408 if (error)
2409 return error;
2412 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2413 * this field is ignored. Note also that a value of zero indicates
2414 * the current setting should be left unchanged.
2416 if (params->spp_flags & SPP_HB_ENABLE) {
2418 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2419 * set. This lets us use 0 value when this flag
2420 * is set.
2422 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2423 params->spp_hbinterval = 0;
2425 if (params->spp_hbinterval ||
2426 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2427 if (trans) {
2428 trans->hbinterval =
2429 msecs_to_jiffies(params->spp_hbinterval);
2430 } else if (asoc) {
2431 asoc->hbinterval =
2432 msecs_to_jiffies(params->spp_hbinterval);
2433 } else {
2434 sp->hbinterval = params->spp_hbinterval;
2439 if (hb_change) {
2440 if (trans) {
2441 trans->param_flags =
2442 (trans->param_flags & ~SPP_HB) | hb_change;
2443 } else if (asoc) {
2444 asoc->param_flags =
2445 (asoc->param_flags & ~SPP_HB) | hb_change;
2446 } else {
2447 sp->param_flags =
2448 (sp->param_flags & ~SPP_HB) | hb_change;
2452 /* When Path MTU discovery is disabled the value specified here will
2453 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2454 * include the flag SPP_PMTUD_DISABLE for this field to have any
2455 * effect).
2457 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2458 if (trans) {
2459 trans->pathmtu = params->spp_pathmtu;
2460 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2461 } else if (asoc) {
2462 asoc->pathmtu = params->spp_pathmtu;
2463 sctp_frag_point(asoc, params->spp_pathmtu);
2464 } else {
2465 sp->pathmtu = params->spp_pathmtu;
2469 if (pmtud_change) {
2470 if (trans) {
2471 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2472 (params->spp_flags & SPP_PMTUD_ENABLE);
2473 trans->param_flags =
2474 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2475 if (update) {
2476 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2477 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2479 } else if (asoc) {
2480 asoc->param_flags =
2481 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2482 } else {
2483 sp->param_flags =
2484 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2488 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2489 * value of this field is ignored. Note also that a value of zero
2490 * indicates the current setting should be left unchanged.
2492 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2493 if (trans) {
2494 trans->sackdelay =
2495 msecs_to_jiffies(params->spp_sackdelay);
2496 } else if (asoc) {
2497 asoc->sackdelay =
2498 msecs_to_jiffies(params->spp_sackdelay);
2499 } else {
2500 sp->sackdelay = params->spp_sackdelay;
2504 if (sackdelay_change) {
2505 if (trans) {
2506 trans->param_flags =
2507 (trans->param_flags & ~SPP_SACKDELAY) |
2508 sackdelay_change;
2509 } else if (asoc) {
2510 asoc->param_flags =
2511 (asoc->param_flags & ~SPP_SACKDELAY) |
2512 sackdelay_change;
2513 } else {
2514 sp->param_flags =
2515 (sp->param_flags & ~SPP_SACKDELAY) |
2516 sackdelay_change;
2520 /* Note that a value of zero indicates the current setting should be
2521 left unchanged.
2523 if (params->spp_pathmaxrxt) {
2524 if (trans) {
2525 trans->pathmaxrxt = params->spp_pathmaxrxt;
2526 } else if (asoc) {
2527 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2528 } else {
2529 sp->pathmaxrxt = params->spp_pathmaxrxt;
2533 return 0;
2536 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2537 char __user *optval,
2538 unsigned int optlen)
2540 struct sctp_paddrparams params;
2541 struct sctp_transport *trans = NULL;
2542 struct sctp_association *asoc = NULL;
2543 struct sctp_sock *sp = sctp_sk(sk);
2544 int error;
2545 int hb_change, pmtud_change, sackdelay_change;
2547 if (optlen != sizeof(struct sctp_paddrparams))
2548 return -EINVAL;
2550 if (copy_from_user(&params, optval, optlen))
2551 return -EFAULT;
2553 /* Validate flags and value parameters. */
2554 hb_change = params.spp_flags & SPP_HB;
2555 pmtud_change = params.spp_flags & SPP_PMTUD;
2556 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2558 if (hb_change == SPP_HB ||
2559 pmtud_change == SPP_PMTUD ||
2560 sackdelay_change == SPP_SACKDELAY ||
2561 params.spp_sackdelay > 500 ||
2562 (params.spp_pathmtu &&
2563 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2564 return -EINVAL;
2566 /* If an address other than INADDR_ANY is specified, and
2567 * no transport is found, then the request is invalid.
2569 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2570 trans = sctp_addr_id2transport(sk, &params.spp_address,
2571 params.spp_assoc_id);
2572 if (!trans)
2573 return -EINVAL;
2576 /* Get association, if assoc_id != 0 and the socket is a one
2577 * to many style socket, and an association was not found, then
2578 * the id was invalid.
2580 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2581 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2582 return -EINVAL;
2584 /* Heartbeat demand can only be sent on a transport or
2585 * association, but not a socket.
2587 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2588 return -EINVAL;
2590 /* Process parameters. */
2591 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2592 hb_change, pmtud_change,
2593 sackdelay_change);
2595 if (error)
2596 return error;
2598 /* If changes are for association, also apply parameters to each
2599 * transport.
2601 if (!trans && asoc) {
2602 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2603 transports) {
2604 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2605 hb_change, pmtud_change,
2606 sackdelay_change);
2610 return 0;
2613 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2615 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2618 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2620 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2624 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2626 * This option will effect the way delayed acks are performed. This
2627 * option allows you to get or set the delayed ack time, in
2628 * milliseconds. It also allows changing the delayed ack frequency.
2629 * Changing the frequency to 1 disables the delayed sack algorithm. If
2630 * the assoc_id is 0, then this sets or gets the endpoints default
2631 * values. If the assoc_id field is non-zero, then the set or get
2632 * effects the specified association for the one to many model (the
2633 * assoc_id field is ignored by the one to one model). Note that if
2634 * sack_delay or sack_freq are 0 when setting this option, then the
2635 * current values will remain unchanged.
2637 * struct sctp_sack_info {
2638 * sctp_assoc_t sack_assoc_id;
2639 * uint32_t sack_delay;
2640 * uint32_t sack_freq;
2641 * };
2643 * sack_assoc_id - This parameter, indicates which association the user
2644 * is performing an action upon. Note that if this field's value is
2645 * zero then the endpoints default value is changed (effecting future
2646 * associations only).
2648 * sack_delay - This parameter contains the number of milliseconds that
2649 * the user is requesting the delayed ACK timer be set to. Note that
2650 * this value is defined in the standard to be between 200 and 500
2651 * milliseconds.
2653 * sack_freq - This parameter contains the number of packets that must
2654 * be received before a sack is sent without waiting for the delay
2655 * timer to expire. The default value for this is 2, setting this
2656 * value to 1 will disable the delayed sack algorithm.
2659 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2660 char __user *optval, unsigned int optlen)
2662 struct sctp_sack_info params;
2663 struct sctp_transport *trans = NULL;
2664 struct sctp_association *asoc = NULL;
2665 struct sctp_sock *sp = sctp_sk(sk);
2667 if (optlen == sizeof(struct sctp_sack_info)) {
2668 if (copy_from_user(&params, optval, optlen))
2669 return -EFAULT;
2671 if (params.sack_delay == 0 && params.sack_freq == 0)
2672 return 0;
2673 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2674 pr_warn_ratelimited(DEPRECATED
2675 "%s (pid %d) "
2676 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2677 "Use struct sctp_sack_info instead\n",
2678 current->comm, task_pid_nr(current));
2679 if (copy_from_user(&params, optval, optlen))
2680 return -EFAULT;
2682 if (params.sack_delay == 0)
2683 params.sack_freq = 1;
2684 else
2685 params.sack_freq = 0;
2686 } else
2687 return -EINVAL;
2689 /* Validate value parameter. */
2690 if (params.sack_delay > 500)
2691 return -EINVAL;
2693 /* Get association, if sack_assoc_id != 0 and the socket is a one
2694 * to many style socket, and an association was not found, then
2695 * the id was invalid.
2697 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2698 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2699 return -EINVAL;
2701 if (params.sack_delay) {
2702 if (asoc) {
2703 asoc->sackdelay =
2704 msecs_to_jiffies(params.sack_delay);
2705 asoc->param_flags =
2706 sctp_spp_sackdelay_enable(asoc->param_flags);
2707 } else {
2708 sp->sackdelay = params.sack_delay;
2709 sp->param_flags =
2710 sctp_spp_sackdelay_enable(sp->param_flags);
2714 if (params.sack_freq == 1) {
2715 if (asoc) {
2716 asoc->param_flags =
2717 sctp_spp_sackdelay_disable(asoc->param_flags);
2718 } else {
2719 sp->param_flags =
2720 sctp_spp_sackdelay_disable(sp->param_flags);
2722 } else if (params.sack_freq > 1) {
2723 if (asoc) {
2724 asoc->sackfreq = params.sack_freq;
2725 asoc->param_flags =
2726 sctp_spp_sackdelay_enable(asoc->param_flags);
2727 } else {
2728 sp->sackfreq = params.sack_freq;
2729 sp->param_flags =
2730 sctp_spp_sackdelay_enable(sp->param_flags);
2734 /* If change is for association, also apply to each transport. */
2735 if (asoc) {
2736 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2737 transports) {
2738 if (params.sack_delay) {
2739 trans->sackdelay =
2740 msecs_to_jiffies(params.sack_delay);
2741 trans->param_flags =
2742 sctp_spp_sackdelay_enable(trans->param_flags);
2744 if (params.sack_freq == 1) {
2745 trans->param_flags =
2746 sctp_spp_sackdelay_disable(trans->param_flags);
2747 } else if (params.sack_freq > 1) {
2748 trans->sackfreq = params.sack_freq;
2749 trans->param_flags =
2750 sctp_spp_sackdelay_enable(trans->param_flags);
2755 return 0;
2758 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2760 * Applications can specify protocol parameters for the default association
2761 * initialization. The option name argument to setsockopt() and getsockopt()
2762 * is SCTP_INITMSG.
2764 * Setting initialization parameters is effective only on an unconnected
2765 * socket (for UDP-style sockets only future associations are effected
2766 * by the change). With TCP-style sockets, this option is inherited by
2767 * sockets derived from a listener socket.
2769 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2771 struct sctp_initmsg sinit;
2772 struct sctp_sock *sp = sctp_sk(sk);
2774 if (optlen != sizeof(struct sctp_initmsg))
2775 return -EINVAL;
2776 if (copy_from_user(&sinit, optval, optlen))
2777 return -EFAULT;
2779 if (sinit.sinit_num_ostreams)
2780 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2781 if (sinit.sinit_max_instreams)
2782 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2783 if (sinit.sinit_max_attempts)
2784 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2785 if (sinit.sinit_max_init_timeo)
2786 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2788 return 0;
2792 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2794 * Applications that wish to use the sendto() system call may wish to
2795 * specify a default set of parameters that would normally be supplied
2796 * through the inclusion of ancillary data. This socket option allows
2797 * such an application to set the default sctp_sndrcvinfo structure.
2798 * The application that wishes to use this socket option simply passes
2799 * in to this call the sctp_sndrcvinfo structure defined in Section
2800 * 5.2.2) The input parameters accepted by this call include
2801 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2802 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2803 * to this call if the caller is using the UDP model.
2805 static int sctp_setsockopt_default_send_param(struct sock *sk,
2806 char __user *optval,
2807 unsigned int optlen)
2809 struct sctp_sock *sp = sctp_sk(sk);
2810 struct sctp_association *asoc;
2811 struct sctp_sndrcvinfo info;
2813 if (optlen != sizeof(info))
2814 return -EINVAL;
2815 if (copy_from_user(&info, optval, optlen))
2816 return -EFAULT;
2817 if (info.sinfo_flags &
2818 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2819 SCTP_ABORT | SCTP_EOF))
2820 return -EINVAL;
2822 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2823 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2824 return -EINVAL;
2825 if (asoc) {
2826 asoc->default_stream = info.sinfo_stream;
2827 asoc->default_flags = info.sinfo_flags;
2828 asoc->default_ppid = info.sinfo_ppid;
2829 asoc->default_context = info.sinfo_context;
2830 asoc->default_timetolive = info.sinfo_timetolive;
2831 } else {
2832 sp->default_stream = info.sinfo_stream;
2833 sp->default_flags = info.sinfo_flags;
2834 sp->default_ppid = info.sinfo_ppid;
2835 sp->default_context = info.sinfo_context;
2836 sp->default_timetolive = info.sinfo_timetolive;
2839 return 0;
2842 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2843 * (SCTP_DEFAULT_SNDINFO)
2845 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2846 char __user *optval,
2847 unsigned int optlen)
2849 struct sctp_sock *sp = sctp_sk(sk);
2850 struct sctp_association *asoc;
2851 struct sctp_sndinfo info;
2853 if (optlen != sizeof(info))
2854 return -EINVAL;
2855 if (copy_from_user(&info, optval, optlen))
2856 return -EFAULT;
2857 if (info.snd_flags &
2858 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2859 SCTP_ABORT | SCTP_EOF))
2860 return -EINVAL;
2862 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2863 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2864 return -EINVAL;
2865 if (asoc) {
2866 asoc->default_stream = info.snd_sid;
2867 asoc->default_flags = info.snd_flags;
2868 asoc->default_ppid = info.snd_ppid;
2869 asoc->default_context = info.snd_context;
2870 } else {
2871 sp->default_stream = info.snd_sid;
2872 sp->default_flags = info.snd_flags;
2873 sp->default_ppid = info.snd_ppid;
2874 sp->default_context = info.snd_context;
2877 return 0;
2880 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2882 * Requests that the local SCTP stack use the enclosed peer address as
2883 * the association primary. The enclosed address must be one of the
2884 * association peer's addresses.
2886 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2887 unsigned int optlen)
2889 struct sctp_prim prim;
2890 struct sctp_transport *trans;
2892 if (optlen != sizeof(struct sctp_prim))
2893 return -EINVAL;
2895 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2896 return -EFAULT;
2898 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2899 if (!trans)
2900 return -EINVAL;
2902 sctp_assoc_set_primary(trans->asoc, trans);
2904 return 0;
2908 * 7.1.5 SCTP_NODELAY
2910 * Turn on/off any Nagle-like algorithm. This means that packets are
2911 * generally sent as soon as possible and no unnecessary delays are
2912 * introduced, at the cost of more packets in the network. Expects an
2913 * integer boolean flag.
2915 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2916 unsigned int optlen)
2918 int val;
2920 if (optlen < sizeof(int))
2921 return -EINVAL;
2922 if (get_user(val, (int __user *)optval))
2923 return -EFAULT;
2925 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2926 return 0;
2931 * 7.1.1 SCTP_RTOINFO
2933 * The protocol parameters used to initialize and bound retransmission
2934 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2935 * and modify these parameters.
2936 * All parameters are time values, in milliseconds. A value of 0, when
2937 * modifying the parameters, indicates that the current value should not
2938 * be changed.
2941 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2943 struct sctp_rtoinfo rtoinfo;
2944 struct sctp_association *asoc;
2945 unsigned long rto_min, rto_max;
2946 struct sctp_sock *sp = sctp_sk(sk);
2948 if (optlen != sizeof (struct sctp_rtoinfo))
2949 return -EINVAL;
2951 if (copy_from_user(&rtoinfo, optval, optlen))
2952 return -EFAULT;
2954 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2956 /* Set the values to the specific association */
2957 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2958 return -EINVAL;
2960 rto_max = rtoinfo.srto_max;
2961 rto_min = rtoinfo.srto_min;
2963 if (rto_max)
2964 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2965 else
2966 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2968 if (rto_min)
2969 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2970 else
2971 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2973 if (rto_min > rto_max)
2974 return -EINVAL;
2976 if (asoc) {
2977 if (rtoinfo.srto_initial != 0)
2978 asoc->rto_initial =
2979 msecs_to_jiffies(rtoinfo.srto_initial);
2980 asoc->rto_max = rto_max;
2981 asoc->rto_min = rto_min;
2982 } else {
2983 /* If there is no association or the association-id = 0
2984 * set the values to the endpoint.
2986 if (rtoinfo.srto_initial != 0)
2987 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2988 sp->rtoinfo.srto_max = rto_max;
2989 sp->rtoinfo.srto_min = rto_min;
2992 return 0;
2997 * 7.1.2 SCTP_ASSOCINFO
2999 * This option is used to tune the maximum retransmission attempts
3000 * of the association.
3001 * Returns an error if the new association retransmission value is
3002 * greater than the sum of the retransmission value of the peer.
3003 * See [SCTP] for more information.
3006 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3009 struct sctp_assocparams assocparams;
3010 struct sctp_association *asoc;
3012 if (optlen != sizeof(struct sctp_assocparams))
3013 return -EINVAL;
3014 if (copy_from_user(&assocparams, optval, optlen))
3015 return -EFAULT;
3017 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3019 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3020 return -EINVAL;
3022 /* Set the values to the specific association */
3023 if (asoc) {
3024 if (assocparams.sasoc_asocmaxrxt != 0) {
3025 __u32 path_sum = 0;
3026 int paths = 0;
3027 struct sctp_transport *peer_addr;
3029 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3030 transports) {
3031 path_sum += peer_addr->pathmaxrxt;
3032 paths++;
3035 /* Only validate asocmaxrxt if we have more than
3036 * one path/transport. We do this because path
3037 * retransmissions are only counted when we have more
3038 * then one path.
3040 if (paths > 1 &&
3041 assocparams.sasoc_asocmaxrxt > path_sum)
3042 return -EINVAL;
3044 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3047 if (assocparams.sasoc_cookie_life != 0)
3048 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3049 } else {
3050 /* Set the values to the endpoint */
3051 struct sctp_sock *sp = sctp_sk(sk);
3053 if (assocparams.sasoc_asocmaxrxt != 0)
3054 sp->assocparams.sasoc_asocmaxrxt =
3055 assocparams.sasoc_asocmaxrxt;
3056 if (assocparams.sasoc_cookie_life != 0)
3057 sp->assocparams.sasoc_cookie_life =
3058 assocparams.sasoc_cookie_life;
3060 return 0;
3064 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3066 * This socket option is a boolean flag which turns on or off mapped V4
3067 * addresses. If this option is turned on and the socket is type
3068 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3069 * If this option is turned off, then no mapping will be done of V4
3070 * addresses and a user will receive both PF_INET6 and PF_INET type
3071 * addresses on the socket.
3073 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3075 int val;
3076 struct sctp_sock *sp = sctp_sk(sk);
3078 if (optlen < sizeof(int))
3079 return -EINVAL;
3080 if (get_user(val, (int __user *)optval))
3081 return -EFAULT;
3082 if (val)
3083 sp->v4mapped = 1;
3084 else
3085 sp->v4mapped = 0;
3087 return 0;
3091 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3092 * This option will get or set the maximum size to put in any outgoing
3093 * SCTP DATA chunk. If a message is larger than this size it will be
3094 * fragmented by SCTP into the specified size. Note that the underlying
3095 * SCTP implementation may fragment into smaller sized chunks when the
3096 * PMTU of the underlying association is smaller than the value set by
3097 * the user. The default value for this option is '0' which indicates
3098 * the user is NOT limiting fragmentation and only the PMTU will effect
3099 * SCTP's choice of DATA chunk size. Note also that values set larger
3100 * than the maximum size of an IP datagram will effectively let SCTP
3101 * control fragmentation (i.e. the same as setting this option to 0).
3103 * The following structure is used to access and modify this parameter:
3105 * struct sctp_assoc_value {
3106 * sctp_assoc_t assoc_id;
3107 * uint32_t assoc_value;
3108 * };
3110 * assoc_id: This parameter is ignored for one-to-one style sockets.
3111 * For one-to-many style sockets this parameter indicates which
3112 * association the user is performing an action upon. Note that if
3113 * this field's value is zero then the endpoints default value is
3114 * changed (effecting future associations only).
3115 * assoc_value: This parameter specifies the maximum size in bytes.
3117 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3119 struct sctp_assoc_value params;
3120 struct sctp_association *asoc;
3121 struct sctp_sock *sp = sctp_sk(sk);
3122 int val;
3124 if (optlen == sizeof(int)) {
3125 pr_warn_ratelimited(DEPRECATED
3126 "%s (pid %d) "
3127 "Use of int in maxseg socket option.\n"
3128 "Use struct sctp_assoc_value instead\n",
3129 current->comm, task_pid_nr(current));
3130 if (copy_from_user(&val, optval, optlen))
3131 return -EFAULT;
3132 params.assoc_id = 0;
3133 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3134 if (copy_from_user(&params, optval, optlen))
3135 return -EFAULT;
3136 val = params.assoc_value;
3137 } else
3138 return -EINVAL;
3140 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3141 return -EINVAL;
3143 asoc = sctp_id2assoc(sk, params.assoc_id);
3144 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3145 return -EINVAL;
3147 if (asoc) {
3148 if (val == 0) {
3149 val = asoc->pathmtu;
3150 val -= sp->pf->af->net_header_len;
3151 val -= sizeof(struct sctphdr) +
3152 sizeof(struct sctp_data_chunk);
3154 asoc->user_frag = val;
3155 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3156 } else {
3157 sp->user_frag = val;
3160 return 0;
3165 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3167 * Requests that the peer mark the enclosed address as the association
3168 * primary. The enclosed address must be one of the association's
3169 * locally bound addresses. The following structure is used to make a
3170 * set primary request:
3172 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3173 unsigned int optlen)
3175 struct net *net = sock_net(sk);
3176 struct sctp_sock *sp;
3177 struct sctp_association *asoc = NULL;
3178 struct sctp_setpeerprim prim;
3179 struct sctp_chunk *chunk;
3180 struct sctp_af *af;
3181 int err;
3183 sp = sctp_sk(sk);
3185 if (!net->sctp.addip_enable)
3186 return -EPERM;
3188 if (optlen != sizeof(struct sctp_setpeerprim))
3189 return -EINVAL;
3191 if (copy_from_user(&prim, optval, optlen))
3192 return -EFAULT;
3194 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3195 if (!asoc)
3196 return -EINVAL;
3198 if (!asoc->peer.asconf_capable)
3199 return -EPERM;
3201 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3202 return -EPERM;
3204 if (!sctp_state(asoc, ESTABLISHED))
3205 return -ENOTCONN;
3207 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3208 if (!af)
3209 return -EINVAL;
3211 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3212 return -EADDRNOTAVAIL;
3214 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3215 return -EADDRNOTAVAIL;
3217 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3218 chunk = sctp_make_asconf_set_prim(asoc,
3219 (union sctp_addr *)&prim.sspp_addr);
3220 if (!chunk)
3221 return -ENOMEM;
3223 err = sctp_send_asconf(asoc, chunk);
3225 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3227 return err;
3230 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3231 unsigned int optlen)
3233 struct sctp_setadaptation adaptation;
3235 if (optlen != sizeof(struct sctp_setadaptation))
3236 return -EINVAL;
3237 if (copy_from_user(&adaptation, optval, optlen))
3238 return -EFAULT;
3240 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3242 return 0;
3246 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3248 * The context field in the sctp_sndrcvinfo structure is normally only
3249 * used when a failed message is retrieved holding the value that was
3250 * sent down on the actual send call. This option allows the setting of
3251 * a default context on an association basis that will be received on
3252 * reading messages from the peer. This is especially helpful in the
3253 * one-2-many model for an application to keep some reference to an
3254 * internal state machine that is processing messages on the
3255 * association. Note that the setting of this value only effects
3256 * received messages from the peer and does not effect the value that is
3257 * saved with outbound messages.
3259 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3260 unsigned int optlen)
3262 struct sctp_assoc_value params;
3263 struct sctp_sock *sp;
3264 struct sctp_association *asoc;
3266 if (optlen != sizeof(struct sctp_assoc_value))
3267 return -EINVAL;
3268 if (copy_from_user(&params, optval, optlen))
3269 return -EFAULT;
3271 sp = sctp_sk(sk);
3273 if (params.assoc_id != 0) {
3274 asoc = sctp_id2assoc(sk, params.assoc_id);
3275 if (!asoc)
3276 return -EINVAL;
3277 asoc->default_rcv_context = params.assoc_value;
3278 } else {
3279 sp->default_rcv_context = params.assoc_value;
3282 return 0;
3286 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3288 * This options will at a minimum specify if the implementation is doing
3289 * fragmented interleave. Fragmented interleave, for a one to many
3290 * socket, is when subsequent calls to receive a message may return
3291 * parts of messages from different associations. Some implementations
3292 * may allow you to turn this value on or off. If so, when turned off,
3293 * no fragment interleave will occur (which will cause a head of line
3294 * blocking amongst multiple associations sharing the same one to many
3295 * socket). When this option is turned on, then each receive call may
3296 * come from a different association (thus the user must receive data
3297 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3298 * association each receive belongs to.
3300 * This option takes a boolean value. A non-zero value indicates that
3301 * fragmented interleave is on. A value of zero indicates that
3302 * fragmented interleave is off.
3304 * Note that it is important that an implementation that allows this
3305 * option to be turned on, have it off by default. Otherwise an unaware
3306 * application using the one to many model may become confused and act
3307 * incorrectly.
3309 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3310 char __user *optval,
3311 unsigned int optlen)
3313 int val;
3315 if (optlen != sizeof(int))
3316 return -EINVAL;
3317 if (get_user(val, (int __user *)optval))
3318 return -EFAULT;
3320 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3322 return 0;
3326 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3327 * (SCTP_PARTIAL_DELIVERY_POINT)
3329 * This option will set or get the SCTP partial delivery point. This
3330 * point is the size of a message where the partial delivery API will be
3331 * invoked to help free up rwnd space for the peer. Setting this to a
3332 * lower value will cause partial deliveries to happen more often. The
3333 * calls argument is an integer that sets or gets the partial delivery
3334 * point. Note also that the call will fail if the user attempts to set
3335 * this value larger than the socket receive buffer size.
3337 * Note that any single message having a length smaller than or equal to
3338 * the SCTP partial delivery point will be delivered in one single read
3339 * call as long as the user provided buffer is large enough to hold the
3340 * message.
3342 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3343 char __user *optval,
3344 unsigned int optlen)
3346 u32 val;
3348 if (optlen != sizeof(u32))
3349 return -EINVAL;
3350 if (get_user(val, (int __user *)optval))
3351 return -EFAULT;
3353 /* Note: We double the receive buffer from what the user sets
3354 * it to be, also initial rwnd is based on rcvbuf/2.
3356 if (val > (sk->sk_rcvbuf >> 1))
3357 return -EINVAL;
3359 sctp_sk(sk)->pd_point = val;
3361 return 0; /* is this the right error code? */
3365 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3367 * This option will allow a user to change the maximum burst of packets
3368 * that can be emitted by this association. Note that the default value
3369 * is 4, and some implementations may restrict this setting so that it
3370 * can only be lowered.
3372 * NOTE: This text doesn't seem right. Do this on a socket basis with
3373 * future associations inheriting the socket value.
3375 static int sctp_setsockopt_maxburst(struct sock *sk,
3376 char __user *optval,
3377 unsigned int optlen)
3379 struct sctp_assoc_value params;
3380 struct sctp_sock *sp;
3381 struct sctp_association *asoc;
3382 int val;
3383 int assoc_id = 0;
3385 if (optlen == sizeof(int)) {
3386 pr_warn_ratelimited(DEPRECATED
3387 "%s (pid %d) "
3388 "Use of int in max_burst socket option deprecated.\n"
3389 "Use struct sctp_assoc_value instead\n",
3390 current->comm, task_pid_nr(current));
3391 if (copy_from_user(&val, optval, optlen))
3392 return -EFAULT;
3393 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3394 if (copy_from_user(&params, optval, optlen))
3395 return -EFAULT;
3396 val = params.assoc_value;
3397 assoc_id = params.assoc_id;
3398 } else
3399 return -EINVAL;
3401 sp = sctp_sk(sk);
3403 if (assoc_id != 0) {
3404 asoc = sctp_id2assoc(sk, assoc_id);
3405 if (!asoc)
3406 return -EINVAL;
3407 asoc->max_burst = val;
3408 } else
3409 sp->max_burst = val;
3411 return 0;
3415 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3417 * This set option adds a chunk type that the user is requesting to be
3418 * received only in an authenticated way. Changes to the list of chunks
3419 * will only effect future associations on the socket.
3421 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3422 char __user *optval,
3423 unsigned int optlen)
3425 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3426 struct sctp_authchunk val;
3428 if (!ep->auth_enable)
3429 return -EACCES;
3431 if (optlen != sizeof(struct sctp_authchunk))
3432 return -EINVAL;
3433 if (copy_from_user(&val, optval, optlen))
3434 return -EFAULT;
3436 switch (val.sauth_chunk) {
3437 case SCTP_CID_INIT:
3438 case SCTP_CID_INIT_ACK:
3439 case SCTP_CID_SHUTDOWN_COMPLETE:
3440 case SCTP_CID_AUTH:
3441 return -EINVAL;
3444 /* add this chunk id to the endpoint */
3445 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3449 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3451 * This option gets or sets the list of HMAC algorithms that the local
3452 * endpoint requires the peer to use.
3454 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3455 char __user *optval,
3456 unsigned int optlen)
3458 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3459 struct sctp_hmacalgo *hmacs;
3460 u32 idents;
3461 int err;
3463 if (!ep->auth_enable)
3464 return -EACCES;
3466 if (optlen < sizeof(struct sctp_hmacalgo))
3467 return -EINVAL;
3469 hmacs = memdup_user(optval, optlen);
3470 if (IS_ERR(hmacs))
3471 return PTR_ERR(hmacs);
3473 idents = hmacs->shmac_num_idents;
3474 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3475 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3476 err = -EINVAL;
3477 goto out;
3480 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3481 out:
3482 kfree(hmacs);
3483 return err;
3487 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3489 * This option will set a shared secret key which is used to build an
3490 * association shared key.
3492 static int sctp_setsockopt_auth_key(struct sock *sk,
3493 char __user *optval,
3494 unsigned int optlen)
3496 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3497 struct sctp_authkey *authkey;
3498 struct sctp_association *asoc;
3499 int ret;
3501 if (!ep->auth_enable)
3502 return -EACCES;
3504 if (optlen <= sizeof(struct sctp_authkey))
3505 return -EINVAL;
3507 authkey = memdup_user(optval, optlen);
3508 if (IS_ERR(authkey))
3509 return PTR_ERR(authkey);
3511 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3512 ret = -EINVAL;
3513 goto out;
3516 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3517 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3518 ret = -EINVAL;
3519 goto out;
3522 ret = sctp_auth_set_key(ep, asoc, authkey);
3523 out:
3524 kzfree(authkey);
3525 return ret;
3529 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3531 * This option will get or set the active shared key to be used to build
3532 * the association shared key.
3534 static int sctp_setsockopt_active_key(struct sock *sk,
3535 char __user *optval,
3536 unsigned int optlen)
3538 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3539 struct sctp_authkeyid val;
3540 struct sctp_association *asoc;
3542 if (!ep->auth_enable)
3543 return -EACCES;
3545 if (optlen != sizeof(struct sctp_authkeyid))
3546 return -EINVAL;
3547 if (copy_from_user(&val, optval, optlen))
3548 return -EFAULT;
3550 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3551 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3552 return -EINVAL;
3554 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3558 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3560 * This set option will delete a shared secret key from use.
3562 static int sctp_setsockopt_del_key(struct sock *sk,
3563 char __user *optval,
3564 unsigned int optlen)
3566 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3567 struct sctp_authkeyid val;
3568 struct sctp_association *asoc;
3570 if (!ep->auth_enable)
3571 return -EACCES;
3573 if (optlen != sizeof(struct sctp_authkeyid))
3574 return -EINVAL;
3575 if (copy_from_user(&val, optval, optlen))
3576 return -EFAULT;
3578 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3579 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3580 return -EINVAL;
3582 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3587 * 8.1.23 SCTP_AUTO_ASCONF
3589 * This option will enable or disable the use of the automatic generation of
3590 * ASCONF chunks to add and delete addresses to an existing association. Note
3591 * that this option has two caveats namely: a) it only affects sockets that
3592 * are bound to all addresses available to the SCTP stack, and b) the system
3593 * administrator may have an overriding control that turns the ASCONF feature
3594 * off no matter what setting the socket option may have.
3595 * This option expects an integer boolean flag, where a non-zero value turns on
3596 * the option, and a zero value turns off the option.
3597 * Note. In this implementation, socket operation overrides default parameter
3598 * being set by sysctl as well as FreeBSD implementation
3600 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3601 unsigned int optlen)
3603 int val;
3604 struct sctp_sock *sp = sctp_sk(sk);
3606 if (optlen < sizeof(int))
3607 return -EINVAL;
3608 if (get_user(val, (int __user *)optval))
3609 return -EFAULT;
3610 if (!sctp_is_ep_boundall(sk) && val)
3611 return -EINVAL;
3612 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3613 return 0;
3615 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3616 if (val == 0 && sp->do_auto_asconf) {
3617 list_del(&sp->auto_asconf_list);
3618 sp->do_auto_asconf = 0;
3619 } else if (val && !sp->do_auto_asconf) {
3620 list_add_tail(&sp->auto_asconf_list,
3621 &sock_net(sk)->sctp.auto_asconf_splist);
3622 sp->do_auto_asconf = 1;
3624 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3625 return 0;
3629 * SCTP_PEER_ADDR_THLDS
3631 * This option allows us to alter the partially failed threshold for one or all
3632 * transports in an association. See Section 6.1 of:
3633 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3635 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3636 char __user *optval,
3637 unsigned int optlen)
3639 struct sctp_paddrthlds val;
3640 struct sctp_transport *trans;
3641 struct sctp_association *asoc;
3643 if (optlen < sizeof(struct sctp_paddrthlds))
3644 return -EINVAL;
3645 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3646 sizeof(struct sctp_paddrthlds)))
3647 return -EFAULT;
3650 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3651 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3652 if (!asoc)
3653 return -ENOENT;
3654 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3655 transports) {
3656 if (val.spt_pathmaxrxt)
3657 trans->pathmaxrxt = val.spt_pathmaxrxt;
3658 trans->pf_retrans = val.spt_pathpfthld;
3661 if (val.spt_pathmaxrxt)
3662 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3663 asoc->pf_retrans = val.spt_pathpfthld;
3664 } else {
3665 trans = sctp_addr_id2transport(sk, &val.spt_address,
3666 val.spt_assoc_id);
3667 if (!trans)
3668 return -ENOENT;
3670 if (val.spt_pathmaxrxt)
3671 trans->pathmaxrxt = val.spt_pathmaxrxt;
3672 trans->pf_retrans = val.spt_pathpfthld;
3675 return 0;
3678 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3679 char __user *optval,
3680 unsigned int optlen)
3682 int val;
3684 if (optlen < sizeof(int))
3685 return -EINVAL;
3686 if (get_user(val, (int __user *) optval))
3687 return -EFAULT;
3689 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3691 return 0;
3694 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3695 char __user *optval,
3696 unsigned int optlen)
3698 int val;
3700 if (optlen < sizeof(int))
3701 return -EINVAL;
3702 if (get_user(val, (int __user *) optval))
3703 return -EFAULT;
3705 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3707 return 0;
3710 /* API 6.2 setsockopt(), getsockopt()
3712 * Applications use setsockopt() and getsockopt() to set or retrieve
3713 * socket options. Socket options are used to change the default
3714 * behavior of sockets calls. They are described in Section 7.
3716 * The syntax is:
3718 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3719 * int __user *optlen);
3720 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3721 * int optlen);
3723 * sd - the socket descript.
3724 * level - set to IPPROTO_SCTP for all SCTP options.
3725 * optname - the option name.
3726 * optval - the buffer to store the value of the option.
3727 * optlen - the size of the buffer.
3729 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3730 char __user *optval, unsigned int optlen)
3732 int retval = 0;
3734 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3736 /* I can hardly begin to describe how wrong this is. This is
3737 * so broken as to be worse than useless. The API draft
3738 * REALLY is NOT helpful here... I am not convinced that the
3739 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3740 * are at all well-founded.
3742 if (level != SOL_SCTP) {
3743 struct sctp_af *af = sctp_sk(sk)->pf->af;
3744 retval = af->setsockopt(sk, level, optname, optval, optlen);
3745 goto out_nounlock;
3748 lock_sock(sk);
3750 switch (optname) {
3751 case SCTP_SOCKOPT_BINDX_ADD:
3752 /* 'optlen' is the size of the addresses buffer. */
3753 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3754 optlen, SCTP_BINDX_ADD_ADDR);
3755 break;
3757 case SCTP_SOCKOPT_BINDX_REM:
3758 /* 'optlen' is the size of the addresses buffer. */
3759 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3760 optlen, SCTP_BINDX_REM_ADDR);
3761 break;
3763 case SCTP_SOCKOPT_CONNECTX_OLD:
3764 /* 'optlen' is the size of the addresses buffer. */
3765 retval = sctp_setsockopt_connectx_old(sk,
3766 (struct sockaddr __user *)optval,
3767 optlen);
3768 break;
3770 case SCTP_SOCKOPT_CONNECTX:
3771 /* 'optlen' is the size of the addresses buffer. */
3772 retval = sctp_setsockopt_connectx(sk,
3773 (struct sockaddr __user *)optval,
3774 optlen);
3775 break;
3777 case SCTP_DISABLE_FRAGMENTS:
3778 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3779 break;
3781 case SCTP_EVENTS:
3782 retval = sctp_setsockopt_events(sk, optval, optlen);
3783 break;
3785 case SCTP_AUTOCLOSE:
3786 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3787 break;
3789 case SCTP_PEER_ADDR_PARAMS:
3790 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3791 break;
3793 case SCTP_DELAYED_SACK:
3794 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3795 break;
3796 case SCTP_PARTIAL_DELIVERY_POINT:
3797 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3798 break;
3800 case SCTP_INITMSG:
3801 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3802 break;
3803 case SCTP_DEFAULT_SEND_PARAM:
3804 retval = sctp_setsockopt_default_send_param(sk, optval,
3805 optlen);
3806 break;
3807 case SCTP_DEFAULT_SNDINFO:
3808 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3809 break;
3810 case SCTP_PRIMARY_ADDR:
3811 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3812 break;
3813 case SCTP_SET_PEER_PRIMARY_ADDR:
3814 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3815 break;
3816 case SCTP_NODELAY:
3817 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3818 break;
3819 case SCTP_RTOINFO:
3820 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3821 break;
3822 case SCTP_ASSOCINFO:
3823 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3824 break;
3825 case SCTP_I_WANT_MAPPED_V4_ADDR:
3826 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3827 break;
3828 case SCTP_MAXSEG:
3829 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3830 break;
3831 case SCTP_ADAPTATION_LAYER:
3832 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3833 break;
3834 case SCTP_CONTEXT:
3835 retval = sctp_setsockopt_context(sk, optval, optlen);
3836 break;
3837 case SCTP_FRAGMENT_INTERLEAVE:
3838 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3839 break;
3840 case SCTP_MAX_BURST:
3841 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3842 break;
3843 case SCTP_AUTH_CHUNK:
3844 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3845 break;
3846 case SCTP_HMAC_IDENT:
3847 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3848 break;
3849 case SCTP_AUTH_KEY:
3850 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3851 break;
3852 case SCTP_AUTH_ACTIVE_KEY:
3853 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3854 break;
3855 case SCTP_AUTH_DELETE_KEY:
3856 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3857 break;
3858 case SCTP_AUTO_ASCONF:
3859 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3860 break;
3861 case SCTP_PEER_ADDR_THLDS:
3862 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3863 break;
3864 case SCTP_RECVRCVINFO:
3865 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3866 break;
3867 case SCTP_RECVNXTINFO:
3868 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3869 break;
3870 default:
3871 retval = -ENOPROTOOPT;
3872 break;
3875 release_sock(sk);
3877 out_nounlock:
3878 return retval;
3881 /* API 3.1.6 connect() - UDP Style Syntax
3883 * An application may use the connect() call in the UDP model to initiate an
3884 * association without sending data.
3886 * The syntax is:
3888 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3890 * sd: the socket descriptor to have a new association added to.
3892 * nam: the address structure (either struct sockaddr_in or struct
3893 * sockaddr_in6 defined in RFC2553 [7]).
3895 * len: the size of the address.
3897 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3898 int addr_len)
3900 int err = 0;
3901 struct sctp_af *af;
3903 lock_sock(sk);
3905 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3906 addr, addr_len);
3908 /* Validate addr_len before calling common connect/connectx routine. */
3909 af = sctp_get_af_specific(addr->sa_family);
3910 if (!af || addr_len < af->sockaddr_len) {
3911 err = -EINVAL;
3912 } else {
3913 /* Pass correct addr len to common routine (so it knows there
3914 * is only one address being passed.
3916 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3919 release_sock(sk);
3920 return err;
3923 /* FIXME: Write comments. */
3924 static int sctp_disconnect(struct sock *sk, int flags)
3926 return -EOPNOTSUPP; /* STUB */
3929 /* 4.1.4 accept() - TCP Style Syntax
3931 * Applications use accept() call to remove an established SCTP
3932 * association from the accept queue of the endpoint. A new socket
3933 * descriptor will be returned from accept() to represent the newly
3934 * formed association.
3936 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3938 struct sctp_sock *sp;
3939 struct sctp_endpoint *ep;
3940 struct sock *newsk = NULL;
3941 struct sctp_association *asoc;
3942 long timeo;
3943 int error = 0;
3945 lock_sock(sk);
3947 sp = sctp_sk(sk);
3948 ep = sp->ep;
3950 if (!sctp_style(sk, TCP)) {
3951 error = -EOPNOTSUPP;
3952 goto out;
3955 if (!sctp_sstate(sk, LISTENING)) {
3956 error = -EINVAL;
3957 goto out;
3960 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3962 error = sctp_wait_for_accept(sk, timeo);
3963 if (error)
3964 goto out;
3966 /* We treat the list of associations on the endpoint as the accept
3967 * queue and pick the first association on the list.
3969 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3971 newsk = sp->pf->create_accept_sk(sk, asoc);
3972 if (!newsk) {
3973 error = -ENOMEM;
3974 goto out;
3977 /* Populate the fields of the newsk from the oldsk and migrate the
3978 * asoc to the newsk.
3980 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3982 out:
3983 release_sock(sk);
3984 *err = error;
3985 return newsk;
3988 /* The SCTP ioctl handler. */
3989 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3991 int rc = -ENOTCONN;
3993 lock_sock(sk);
3996 * SEQPACKET-style sockets in LISTENING state are valid, for
3997 * SCTP, so only discard TCP-style sockets in LISTENING state.
3999 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4000 goto out;
4002 switch (cmd) {
4003 case SIOCINQ: {
4004 struct sk_buff *skb;
4005 unsigned int amount = 0;
4007 skb = skb_peek(&sk->sk_receive_queue);
4008 if (skb != NULL) {
4010 * We will only return the amount of this packet since
4011 * that is all that will be read.
4013 amount = skb->len;
4015 rc = put_user(amount, (int __user *)arg);
4016 break;
4018 default:
4019 rc = -ENOIOCTLCMD;
4020 break;
4022 out:
4023 release_sock(sk);
4024 return rc;
4027 /* This is the function which gets called during socket creation to
4028 * initialized the SCTP-specific portion of the sock.
4029 * The sock structure should already be zero-filled memory.
4031 static int sctp_init_sock(struct sock *sk)
4033 struct net *net = sock_net(sk);
4034 struct sctp_sock *sp;
4036 pr_debug("%s: sk:%p\n", __func__, sk);
4038 sp = sctp_sk(sk);
4040 /* Initialize the SCTP per socket area. */
4041 switch (sk->sk_type) {
4042 case SOCK_SEQPACKET:
4043 sp->type = SCTP_SOCKET_UDP;
4044 break;
4045 case SOCK_STREAM:
4046 sp->type = SCTP_SOCKET_TCP;
4047 break;
4048 default:
4049 return -ESOCKTNOSUPPORT;
4052 /* Initialize default send parameters. These parameters can be
4053 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4055 sp->default_stream = 0;
4056 sp->default_ppid = 0;
4057 sp->default_flags = 0;
4058 sp->default_context = 0;
4059 sp->default_timetolive = 0;
4061 sp->default_rcv_context = 0;
4062 sp->max_burst = net->sctp.max_burst;
4064 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4066 /* Initialize default setup parameters. These parameters
4067 * can be modified with the SCTP_INITMSG socket option or
4068 * overridden by the SCTP_INIT CMSG.
4070 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4071 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4072 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4073 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4075 /* Initialize default RTO related parameters. These parameters can
4076 * be modified for with the SCTP_RTOINFO socket option.
4078 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4079 sp->rtoinfo.srto_max = net->sctp.rto_max;
4080 sp->rtoinfo.srto_min = net->sctp.rto_min;
4082 /* Initialize default association related parameters. These parameters
4083 * can be modified with the SCTP_ASSOCINFO socket option.
4085 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4086 sp->assocparams.sasoc_number_peer_destinations = 0;
4087 sp->assocparams.sasoc_peer_rwnd = 0;
4088 sp->assocparams.sasoc_local_rwnd = 0;
4089 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4091 /* Initialize default event subscriptions. By default, all the
4092 * options are off.
4094 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4096 /* Default Peer Address Parameters. These defaults can
4097 * be modified via SCTP_PEER_ADDR_PARAMS
4099 sp->hbinterval = net->sctp.hb_interval;
4100 sp->pathmaxrxt = net->sctp.max_retrans_path;
4101 sp->pathmtu = 0; /* allow default discovery */
4102 sp->sackdelay = net->sctp.sack_timeout;
4103 sp->sackfreq = 2;
4104 sp->param_flags = SPP_HB_ENABLE |
4105 SPP_PMTUD_ENABLE |
4106 SPP_SACKDELAY_ENABLE;
4108 /* If enabled no SCTP message fragmentation will be performed.
4109 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4111 sp->disable_fragments = 0;
4113 /* Enable Nagle algorithm by default. */
4114 sp->nodelay = 0;
4116 sp->recvrcvinfo = 0;
4117 sp->recvnxtinfo = 0;
4119 /* Enable by default. */
4120 sp->v4mapped = 1;
4122 /* Auto-close idle associations after the configured
4123 * number of seconds. A value of 0 disables this
4124 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4125 * for UDP-style sockets only.
4127 sp->autoclose = 0;
4129 /* User specified fragmentation limit. */
4130 sp->user_frag = 0;
4132 sp->adaptation_ind = 0;
4134 sp->pf = sctp_get_pf_specific(sk->sk_family);
4136 /* Control variables for partial data delivery. */
4137 atomic_set(&sp->pd_mode, 0);
4138 skb_queue_head_init(&sp->pd_lobby);
4139 sp->frag_interleave = 0;
4141 /* Create a per socket endpoint structure. Even if we
4142 * change the data structure relationships, this may still
4143 * be useful for storing pre-connect address information.
4145 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4146 if (!sp->ep)
4147 return -ENOMEM;
4149 sp->hmac = NULL;
4151 sk->sk_destruct = sctp_destruct_sock;
4153 SCTP_DBG_OBJCNT_INC(sock);
4155 local_bh_disable();
4156 sk_sockets_allocated_inc(sk);
4157 sock_prot_inuse_add(net, sk->sk_prot, 1);
4159 /* Nothing can fail after this block, otherwise
4160 * sctp_destroy_sock() will be called without addr_wq_lock held
4162 if (net->sctp.default_auto_asconf) {
4163 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4164 list_add_tail(&sp->auto_asconf_list,
4165 &net->sctp.auto_asconf_splist);
4166 sp->do_auto_asconf = 1;
4167 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4168 } else {
4169 sp->do_auto_asconf = 0;
4172 local_bh_enable();
4174 return 0;
4177 /* Cleanup any SCTP per socket resources. Must be called with
4178 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4180 static void sctp_destroy_sock(struct sock *sk)
4182 struct sctp_sock *sp;
4184 pr_debug("%s: sk:%p\n", __func__, sk);
4186 /* Release our hold on the endpoint. */
4187 sp = sctp_sk(sk);
4188 /* This could happen during socket init, thus we bail out
4189 * early, since the rest of the below is not setup either.
4191 if (sp->ep == NULL)
4192 return;
4194 if (sp->do_auto_asconf) {
4195 sp->do_auto_asconf = 0;
4196 list_del(&sp->auto_asconf_list);
4198 sctp_endpoint_free(sp->ep);
4199 local_bh_disable();
4200 sk_sockets_allocated_dec(sk);
4201 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4202 local_bh_enable();
4205 /* Triggered when there are no references on the socket anymore */
4206 static void sctp_destruct_sock(struct sock *sk)
4208 struct sctp_sock *sp = sctp_sk(sk);
4210 /* Free up the HMAC transform. */
4211 crypto_free_hash(sp->hmac);
4213 inet_sock_destruct(sk);
4216 /* API 4.1.7 shutdown() - TCP Style Syntax
4217 * int shutdown(int socket, int how);
4219 * sd - the socket descriptor of the association to be closed.
4220 * how - Specifies the type of shutdown. The values are
4221 * as follows:
4222 * SHUT_RD
4223 * Disables further receive operations. No SCTP
4224 * protocol action is taken.
4225 * SHUT_WR
4226 * Disables further send operations, and initiates
4227 * the SCTP shutdown sequence.
4228 * SHUT_RDWR
4229 * Disables further send and receive operations
4230 * and initiates the SCTP shutdown sequence.
4232 static void sctp_shutdown(struct sock *sk, int how)
4234 struct net *net = sock_net(sk);
4235 struct sctp_endpoint *ep;
4236 struct sctp_association *asoc;
4238 if (!sctp_style(sk, TCP))
4239 return;
4241 if (how & SEND_SHUTDOWN) {
4242 ep = sctp_sk(sk)->ep;
4243 if (!list_empty(&ep->asocs)) {
4244 asoc = list_entry(ep->asocs.next,
4245 struct sctp_association, asocs);
4246 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4251 /* 7.2.1 Association Status (SCTP_STATUS)
4253 * Applications can retrieve current status information about an
4254 * association, including association state, peer receiver window size,
4255 * number of unacked data chunks, and number of data chunks pending
4256 * receipt. This information is read-only.
4258 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4259 char __user *optval,
4260 int __user *optlen)
4262 struct sctp_status status;
4263 struct sctp_association *asoc = NULL;
4264 struct sctp_transport *transport;
4265 sctp_assoc_t associd;
4266 int retval = 0;
4268 if (len < sizeof(status)) {
4269 retval = -EINVAL;
4270 goto out;
4273 len = sizeof(status);
4274 if (copy_from_user(&status, optval, len)) {
4275 retval = -EFAULT;
4276 goto out;
4279 associd = status.sstat_assoc_id;
4280 asoc = sctp_id2assoc(sk, associd);
4281 if (!asoc) {
4282 retval = -EINVAL;
4283 goto out;
4286 transport = asoc->peer.primary_path;
4288 status.sstat_assoc_id = sctp_assoc2id(asoc);
4289 status.sstat_state = sctp_assoc_to_state(asoc);
4290 status.sstat_rwnd = asoc->peer.rwnd;
4291 status.sstat_unackdata = asoc->unack_data;
4293 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4294 status.sstat_instrms = asoc->c.sinit_max_instreams;
4295 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4296 status.sstat_fragmentation_point = asoc->frag_point;
4297 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4298 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4299 transport->af_specific->sockaddr_len);
4300 /* Map ipv4 address into v4-mapped-on-v6 address. */
4301 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4302 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4303 status.sstat_primary.spinfo_state = transport->state;
4304 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4305 status.sstat_primary.spinfo_srtt = transport->srtt;
4306 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4307 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4309 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4310 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4312 if (put_user(len, optlen)) {
4313 retval = -EFAULT;
4314 goto out;
4317 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4318 __func__, len, status.sstat_state, status.sstat_rwnd,
4319 status.sstat_assoc_id);
4321 if (copy_to_user(optval, &status, len)) {
4322 retval = -EFAULT;
4323 goto out;
4326 out:
4327 return retval;
4331 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4333 * Applications can retrieve information about a specific peer address
4334 * of an association, including its reachability state, congestion
4335 * window, and retransmission timer values. This information is
4336 * read-only.
4338 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4339 char __user *optval,
4340 int __user *optlen)
4342 struct sctp_paddrinfo pinfo;
4343 struct sctp_transport *transport;
4344 int retval = 0;
4346 if (len < sizeof(pinfo)) {
4347 retval = -EINVAL;
4348 goto out;
4351 len = sizeof(pinfo);
4352 if (copy_from_user(&pinfo, optval, len)) {
4353 retval = -EFAULT;
4354 goto out;
4357 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4358 pinfo.spinfo_assoc_id);
4359 if (!transport)
4360 return -EINVAL;
4362 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4363 pinfo.spinfo_state = transport->state;
4364 pinfo.spinfo_cwnd = transport->cwnd;
4365 pinfo.spinfo_srtt = transport->srtt;
4366 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4367 pinfo.spinfo_mtu = transport->pathmtu;
4369 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4370 pinfo.spinfo_state = SCTP_ACTIVE;
4372 if (put_user(len, optlen)) {
4373 retval = -EFAULT;
4374 goto out;
4377 if (copy_to_user(optval, &pinfo, len)) {
4378 retval = -EFAULT;
4379 goto out;
4382 out:
4383 return retval;
4386 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4388 * This option is a on/off flag. If enabled no SCTP message
4389 * fragmentation will be performed. Instead if a message being sent
4390 * exceeds the current PMTU size, the message will NOT be sent and
4391 * instead a error will be indicated to the user.
4393 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4394 char __user *optval, int __user *optlen)
4396 int val;
4398 if (len < sizeof(int))
4399 return -EINVAL;
4401 len = sizeof(int);
4402 val = (sctp_sk(sk)->disable_fragments == 1);
4403 if (put_user(len, optlen))
4404 return -EFAULT;
4405 if (copy_to_user(optval, &val, len))
4406 return -EFAULT;
4407 return 0;
4410 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4412 * This socket option is used to specify various notifications and
4413 * ancillary data the user wishes to receive.
4415 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4416 int __user *optlen)
4418 if (len == 0)
4419 return -EINVAL;
4420 if (len > sizeof(struct sctp_event_subscribe))
4421 len = sizeof(struct sctp_event_subscribe);
4422 if (put_user(len, optlen))
4423 return -EFAULT;
4424 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4425 return -EFAULT;
4426 return 0;
4429 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4431 * This socket option is applicable to the UDP-style socket only. When
4432 * set it will cause associations that are idle for more than the
4433 * specified number of seconds to automatically close. An association
4434 * being idle is defined an association that has NOT sent or received
4435 * user data. The special value of '0' indicates that no automatic
4436 * close of any associations should be performed. The option expects an
4437 * integer defining the number of seconds of idle time before an
4438 * association is closed.
4440 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4442 /* Applicable to UDP-style socket only */
4443 if (sctp_style(sk, TCP))
4444 return -EOPNOTSUPP;
4445 if (len < sizeof(int))
4446 return -EINVAL;
4447 len = sizeof(int);
4448 if (put_user(len, optlen))
4449 return -EFAULT;
4450 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4451 return -EFAULT;
4452 return 0;
4455 /* Helper routine to branch off an association to a new socket. */
4456 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4458 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4459 struct sctp_sock *sp = sctp_sk(sk);
4460 struct socket *sock;
4461 int err = 0;
4463 /* Do not peel off from one netns to another one. */
4464 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
4465 return -EINVAL;
4467 if (!asoc)
4468 return -EINVAL;
4470 /* An association cannot be branched off from an already peeled-off
4471 * socket, nor is this supported for tcp style sockets.
4473 if (!sctp_style(sk, UDP))
4474 return -EINVAL;
4476 /* Create a new socket. */
4477 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4478 if (err < 0)
4479 return err;
4481 sctp_copy_sock(sock->sk, sk, asoc);
4483 /* Make peeled-off sockets more like 1-1 accepted sockets.
4484 * Set the daddr and initialize id to something more random
4486 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4488 /* Populate the fields of the newsk from the oldsk and migrate the
4489 * asoc to the newsk.
4491 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4493 *sockp = sock;
4495 return err;
4497 EXPORT_SYMBOL(sctp_do_peeloff);
4499 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4501 sctp_peeloff_arg_t peeloff;
4502 struct socket *newsock;
4503 struct file *newfile;
4504 int retval = 0;
4506 if (len < sizeof(sctp_peeloff_arg_t))
4507 return -EINVAL;
4508 len = sizeof(sctp_peeloff_arg_t);
4509 if (copy_from_user(&peeloff, optval, len))
4510 return -EFAULT;
4512 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4513 if (retval < 0)
4514 goto out;
4516 /* Map the socket to an unused fd that can be returned to the user. */
4517 retval = get_unused_fd_flags(0);
4518 if (retval < 0) {
4519 sock_release(newsock);
4520 goto out;
4523 newfile = sock_alloc_file(newsock, 0, NULL);
4524 if (IS_ERR(newfile)) {
4525 put_unused_fd(retval);
4526 sock_release(newsock);
4527 return PTR_ERR(newfile);
4530 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4531 retval);
4533 /* Return the fd mapped to the new socket. */
4534 if (put_user(len, optlen)) {
4535 fput(newfile);
4536 put_unused_fd(retval);
4537 return -EFAULT;
4539 peeloff.sd = retval;
4540 if (copy_to_user(optval, &peeloff, len)) {
4541 fput(newfile);
4542 put_unused_fd(retval);
4543 return -EFAULT;
4545 fd_install(retval, newfile);
4546 out:
4547 return retval;
4550 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4552 * Applications can enable or disable heartbeats for any peer address of
4553 * an association, modify an address's heartbeat interval, force a
4554 * heartbeat to be sent immediately, and adjust the address's maximum
4555 * number of retransmissions sent before an address is considered
4556 * unreachable. The following structure is used to access and modify an
4557 * address's parameters:
4559 * struct sctp_paddrparams {
4560 * sctp_assoc_t spp_assoc_id;
4561 * struct sockaddr_storage spp_address;
4562 * uint32_t spp_hbinterval;
4563 * uint16_t spp_pathmaxrxt;
4564 * uint32_t spp_pathmtu;
4565 * uint32_t spp_sackdelay;
4566 * uint32_t spp_flags;
4567 * };
4569 * spp_assoc_id - (one-to-many style socket) This is filled in the
4570 * application, and identifies the association for
4571 * this query.
4572 * spp_address - This specifies which address is of interest.
4573 * spp_hbinterval - This contains the value of the heartbeat interval,
4574 * in milliseconds. If a value of zero
4575 * is present in this field then no changes are to
4576 * be made to this parameter.
4577 * spp_pathmaxrxt - This contains the maximum number of
4578 * retransmissions before this address shall be
4579 * considered unreachable. If a value of zero
4580 * is present in this field then no changes are to
4581 * be made to this parameter.
4582 * spp_pathmtu - When Path MTU discovery is disabled the value
4583 * specified here will be the "fixed" path mtu.
4584 * Note that if the spp_address field is empty
4585 * then all associations on this address will
4586 * have this fixed path mtu set upon them.
4588 * spp_sackdelay - When delayed sack is enabled, this value specifies
4589 * the number of milliseconds that sacks will be delayed
4590 * for. This value will apply to all addresses of an
4591 * association if the spp_address field is empty. Note
4592 * also, that if delayed sack is enabled and this
4593 * value is set to 0, no change is made to the last
4594 * recorded delayed sack timer value.
4596 * spp_flags - These flags are used to control various features
4597 * on an association. The flag field may contain
4598 * zero or more of the following options.
4600 * SPP_HB_ENABLE - Enable heartbeats on the
4601 * specified address. Note that if the address
4602 * field is empty all addresses for the association
4603 * have heartbeats enabled upon them.
4605 * SPP_HB_DISABLE - Disable heartbeats on the
4606 * speicifed address. Note that if the address
4607 * field is empty all addresses for the association
4608 * will have their heartbeats disabled. Note also
4609 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4610 * mutually exclusive, only one of these two should
4611 * be specified. Enabling both fields will have
4612 * undetermined results.
4614 * SPP_HB_DEMAND - Request a user initiated heartbeat
4615 * to be made immediately.
4617 * SPP_PMTUD_ENABLE - This field will enable PMTU
4618 * discovery upon the specified address. Note that
4619 * if the address feild is empty then all addresses
4620 * on the association are effected.
4622 * SPP_PMTUD_DISABLE - This field will disable PMTU
4623 * discovery upon the specified address. Note that
4624 * if the address feild is empty then all addresses
4625 * on the association are effected. Not also that
4626 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4627 * exclusive. Enabling both will have undetermined
4628 * results.
4630 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4631 * on delayed sack. The time specified in spp_sackdelay
4632 * is used to specify the sack delay for this address. Note
4633 * that if spp_address is empty then all addresses will
4634 * enable delayed sack and take on the sack delay
4635 * value specified in spp_sackdelay.
4636 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4637 * off delayed sack. If the spp_address field is blank then
4638 * delayed sack is disabled for the entire association. Note
4639 * also that this field is mutually exclusive to
4640 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4641 * results.
4643 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4644 char __user *optval, int __user *optlen)
4646 struct sctp_paddrparams params;
4647 struct sctp_transport *trans = NULL;
4648 struct sctp_association *asoc = NULL;
4649 struct sctp_sock *sp = sctp_sk(sk);
4651 if (len < sizeof(struct sctp_paddrparams))
4652 return -EINVAL;
4653 len = sizeof(struct sctp_paddrparams);
4654 if (copy_from_user(&params, optval, len))
4655 return -EFAULT;
4657 /* If an address other than INADDR_ANY is specified, and
4658 * no transport is found, then the request is invalid.
4660 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4661 trans = sctp_addr_id2transport(sk, &params.spp_address,
4662 params.spp_assoc_id);
4663 if (!trans) {
4664 pr_debug("%s: failed no transport\n", __func__);
4665 return -EINVAL;
4669 /* Get association, if assoc_id != 0 and the socket is a one
4670 * to many style socket, and an association was not found, then
4671 * the id was invalid.
4673 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4674 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4675 pr_debug("%s: failed no association\n", __func__);
4676 return -EINVAL;
4679 if (trans) {
4680 /* Fetch transport values. */
4681 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4682 params.spp_pathmtu = trans->pathmtu;
4683 params.spp_pathmaxrxt = trans->pathmaxrxt;
4684 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4686 /*draft-11 doesn't say what to return in spp_flags*/
4687 params.spp_flags = trans->param_flags;
4688 } else if (asoc) {
4689 /* Fetch association values. */
4690 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4691 params.spp_pathmtu = asoc->pathmtu;
4692 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4693 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4695 /*draft-11 doesn't say what to return in spp_flags*/
4696 params.spp_flags = asoc->param_flags;
4697 } else {
4698 /* Fetch socket values. */
4699 params.spp_hbinterval = sp->hbinterval;
4700 params.spp_pathmtu = sp->pathmtu;
4701 params.spp_sackdelay = sp->sackdelay;
4702 params.spp_pathmaxrxt = sp->pathmaxrxt;
4704 /*draft-11 doesn't say what to return in spp_flags*/
4705 params.spp_flags = sp->param_flags;
4708 if (copy_to_user(optval, &params, len))
4709 return -EFAULT;
4711 if (put_user(len, optlen))
4712 return -EFAULT;
4714 return 0;
4718 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4720 * This option will effect the way delayed acks are performed. This
4721 * option allows you to get or set the delayed ack time, in
4722 * milliseconds. It also allows changing the delayed ack frequency.
4723 * Changing the frequency to 1 disables the delayed sack algorithm. If
4724 * the assoc_id is 0, then this sets or gets the endpoints default
4725 * values. If the assoc_id field is non-zero, then the set or get
4726 * effects the specified association for the one to many model (the
4727 * assoc_id field is ignored by the one to one model). Note that if
4728 * sack_delay or sack_freq are 0 when setting this option, then the
4729 * current values will remain unchanged.
4731 * struct sctp_sack_info {
4732 * sctp_assoc_t sack_assoc_id;
4733 * uint32_t sack_delay;
4734 * uint32_t sack_freq;
4735 * };
4737 * sack_assoc_id - This parameter, indicates which association the user
4738 * is performing an action upon. Note that if this field's value is
4739 * zero then the endpoints default value is changed (effecting future
4740 * associations only).
4742 * sack_delay - This parameter contains the number of milliseconds that
4743 * the user is requesting the delayed ACK timer be set to. Note that
4744 * this value is defined in the standard to be between 200 and 500
4745 * milliseconds.
4747 * sack_freq - This parameter contains the number of packets that must
4748 * be received before a sack is sent without waiting for the delay
4749 * timer to expire. The default value for this is 2, setting this
4750 * value to 1 will disable the delayed sack algorithm.
4752 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4753 char __user *optval,
4754 int __user *optlen)
4756 struct sctp_sack_info params;
4757 struct sctp_association *asoc = NULL;
4758 struct sctp_sock *sp = sctp_sk(sk);
4760 if (len >= sizeof(struct sctp_sack_info)) {
4761 len = sizeof(struct sctp_sack_info);
4763 if (copy_from_user(&params, optval, len))
4764 return -EFAULT;
4765 } else if (len == sizeof(struct sctp_assoc_value)) {
4766 pr_warn_ratelimited(DEPRECATED
4767 "%s (pid %d) "
4768 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4769 "Use struct sctp_sack_info instead\n",
4770 current->comm, task_pid_nr(current));
4771 if (copy_from_user(&params, optval, len))
4772 return -EFAULT;
4773 } else
4774 return -EINVAL;
4776 /* Get association, if sack_assoc_id != 0 and the socket is a one
4777 * to many style socket, and an association was not found, then
4778 * the id was invalid.
4780 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4781 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4782 return -EINVAL;
4784 if (asoc) {
4785 /* Fetch association values. */
4786 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4787 params.sack_delay = jiffies_to_msecs(
4788 asoc->sackdelay);
4789 params.sack_freq = asoc->sackfreq;
4791 } else {
4792 params.sack_delay = 0;
4793 params.sack_freq = 1;
4795 } else {
4796 /* Fetch socket values. */
4797 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4798 params.sack_delay = sp->sackdelay;
4799 params.sack_freq = sp->sackfreq;
4800 } else {
4801 params.sack_delay = 0;
4802 params.sack_freq = 1;
4806 if (copy_to_user(optval, &params, len))
4807 return -EFAULT;
4809 if (put_user(len, optlen))
4810 return -EFAULT;
4812 return 0;
4815 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4817 * Applications can specify protocol parameters for the default association
4818 * initialization. The option name argument to setsockopt() and getsockopt()
4819 * is SCTP_INITMSG.
4821 * Setting initialization parameters is effective only on an unconnected
4822 * socket (for UDP-style sockets only future associations are effected
4823 * by the change). With TCP-style sockets, this option is inherited by
4824 * sockets derived from a listener socket.
4826 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4828 if (len < sizeof(struct sctp_initmsg))
4829 return -EINVAL;
4830 len = sizeof(struct sctp_initmsg);
4831 if (put_user(len, optlen))
4832 return -EFAULT;
4833 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4834 return -EFAULT;
4835 return 0;
4839 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4840 char __user *optval, int __user *optlen)
4842 struct sctp_association *asoc;
4843 int cnt = 0;
4844 struct sctp_getaddrs getaddrs;
4845 struct sctp_transport *from;
4846 void __user *to;
4847 union sctp_addr temp;
4848 struct sctp_sock *sp = sctp_sk(sk);
4849 int addrlen;
4850 size_t space_left;
4851 int bytes_copied;
4853 if (len < sizeof(struct sctp_getaddrs))
4854 return -EINVAL;
4856 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4857 return -EFAULT;
4859 /* For UDP-style sockets, id specifies the association to query. */
4860 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4861 if (!asoc)
4862 return -EINVAL;
4864 to = optval + offsetof(struct sctp_getaddrs, addrs);
4865 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4867 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4868 transports) {
4869 memcpy(&temp, &from->ipaddr, sizeof(temp));
4870 addrlen = sctp_get_pf_specific(sk->sk_family)
4871 ->addr_to_user(sp, &temp);
4872 if (space_left < addrlen)
4873 return -ENOMEM;
4874 if (copy_to_user(to, &temp, addrlen))
4875 return -EFAULT;
4876 to += addrlen;
4877 cnt++;
4878 space_left -= addrlen;
4881 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4882 return -EFAULT;
4883 bytes_copied = ((char __user *)to) - optval;
4884 if (put_user(bytes_copied, optlen))
4885 return -EFAULT;
4887 return 0;
4890 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4891 size_t space_left, int *bytes_copied)
4893 struct sctp_sockaddr_entry *addr;
4894 union sctp_addr temp;
4895 int cnt = 0;
4896 int addrlen;
4897 struct net *net = sock_net(sk);
4899 rcu_read_lock();
4900 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4901 if (!addr->valid)
4902 continue;
4904 if ((PF_INET == sk->sk_family) &&
4905 (AF_INET6 == addr->a.sa.sa_family))
4906 continue;
4907 if ((PF_INET6 == sk->sk_family) &&
4908 inet_v6_ipv6only(sk) &&
4909 (AF_INET == addr->a.sa.sa_family))
4910 continue;
4911 memcpy(&temp, &addr->a, sizeof(temp));
4912 if (!temp.v4.sin_port)
4913 temp.v4.sin_port = htons(port);
4915 addrlen = sctp_get_pf_specific(sk->sk_family)
4916 ->addr_to_user(sctp_sk(sk), &temp);
4918 if (space_left < addrlen) {
4919 cnt = -ENOMEM;
4920 break;
4922 memcpy(to, &temp, addrlen);
4924 to += addrlen;
4925 cnt++;
4926 space_left -= addrlen;
4927 *bytes_copied += addrlen;
4929 rcu_read_unlock();
4931 return cnt;
4935 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4936 char __user *optval, int __user *optlen)
4938 struct sctp_bind_addr *bp;
4939 struct sctp_association *asoc;
4940 int cnt = 0;
4941 struct sctp_getaddrs getaddrs;
4942 struct sctp_sockaddr_entry *addr;
4943 void __user *to;
4944 union sctp_addr temp;
4945 struct sctp_sock *sp = sctp_sk(sk);
4946 int addrlen;
4947 int err = 0;
4948 size_t space_left;
4949 int bytes_copied = 0;
4950 void *addrs;
4951 void *buf;
4953 if (len < sizeof(struct sctp_getaddrs))
4954 return -EINVAL;
4956 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4957 return -EFAULT;
4960 * For UDP-style sockets, id specifies the association to query.
4961 * If the id field is set to the value '0' then the locally bound
4962 * addresses are returned without regard to any particular
4963 * association.
4965 if (0 == getaddrs.assoc_id) {
4966 bp = &sctp_sk(sk)->ep->base.bind_addr;
4967 } else {
4968 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4969 if (!asoc)
4970 return -EINVAL;
4971 bp = &asoc->base.bind_addr;
4974 to = optval + offsetof(struct sctp_getaddrs, addrs);
4975 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4977 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
4978 if (!addrs)
4979 return -ENOMEM;
4981 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4982 * addresses from the global local address list.
4984 if (sctp_list_single_entry(&bp->address_list)) {
4985 addr = list_entry(bp->address_list.next,
4986 struct sctp_sockaddr_entry, list);
4987 if (sctp_is_any(sk, &addr->a)) {
4988 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4989 space_left, &bytes_copied);
4990 if (cnt < 0) {
4991 err = cnt;
4992 goto out;
4994 goto copy_getaddrs;
4998 buf = addrs;
4999 /* Protection on the bound address list is not needed since
5000 * in the socket option context we hold a socket lock and
5001 * thus the bound address list can't change.
5003 list_for_each_entry(addr, &bp->address_list, list) {
5004 memcpy(&temp, &addr->a, sizeof(temp));
5005 addrlen = sctp_get_pf_specific(sk->sk_family)
5006 ->addr_to_user(sp, &temp);
5007 if (space_left < addrlen) {
5008 err = -ENOMEM; /*fixme: right error?*/
5009 goto out;
5011 memcpy(buf, &temp, addrlen);
5012 buf += addrlen;
5013 bytes_copied += addrlen;
5014 cnt++;
5015 space_left -= addrlen;
5018 copy_getaddrs:
5019 if (copy_to_user(to, addrs, bytes_copied)) {
5020 err = -EFAULT;
5021 goto out;
5023 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5024 err = -EFAULT;
5025 goto out;
5027 if (put_user(bytes_copied, optlen))
5028 err = -EFAULT;
5029 out:
5030 kfree(addrs);
5031 return err;
5034 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5036 * Requests that the local SCTP stack use the enclosed peer address as
5037 * the association primary. The enclosed address must be one of the
5038 * association peer's addresses.
5040 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5041 char __user *optval, int __user *optlen)
5043 struct sctp_prim prim;
5044 struct sctp_association *asoc;
5045 struct sctp_sock *sp = sctp_sk(sk);
5047 if (len < sizeof(struct sctp_prim))
5048 return -EINVAL;
5050 len = sizeof(struct sctp_prim);
5052 if (copy_from_user(&prim, optval, len))
5053 return -EFAULT;
5055 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5056 if (!asoc)
5057 return -EINVAL;
5059 if (!asoc->peer.primary_path)
5060 return -ENOTCONN;
5062 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5063 asoc->peer.primary_path->af_specific->sockaddr_len);
5065 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5066 (union sctp_addr *)&prim.ssp_addr);
5068 if (put_user(len, optlen))
5069 return -EFAULT;
5070 if (copy_to_user(optval, &prim, len))
5071 return -EFAULT;
5073 return 0;
5077 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5079 * Requests that the local endpoint set the specified Adaptation Layer
5080 * Indication parameter for all future INIT and INIT-ACK exchanges.
5082 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5083 char __user *optval, int __user *optlen)
5085 struct sctp_setadaptation adaptation;
5087 if (len < sizeof(struct sctp_setadaptation))
5088 return -EINVAL;
5090 len = sizeof(struct sctp_setadaptation);
5092 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5094 if (put_user(len, optlen))
5095 return -EFAULT;
5096 if (copy_to_user(optval, &adaptation, len))
5097 return -EFAULT;
5099 return 0;
5104 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5106 * Applications that wish to use the sendto() system call may wish to
5107 * specify a default set of parameters that would normally be supplied
5108 * through the inclusion of ancillary data. This socket option allows
5109 * such an application to set the default sctp_sndrcvinfo structure.
5112 * The application that wishes to use this socket option simply passes
5113 * in to this call the sctp_sndrcvinfo structure defined in Section
5114 * 5.2.2) The input parameters accepted by this call include
5115 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5116 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5117 * to this call if the caller is using the UDP model.
5119 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5121 static int sctp_getsockopt_default_send_param(struct sock *sk,
5122 int len, char __user *optval,
5123 int __user *optlen)
5125 struct sctp_sock *sp = sctp_sk(sk);
5126 struct sctp_association *asoc;
5127 struct sctp_sndrcvinfo info;
5129 if (len < sizeof(info))
5130 return -EINVAL;
5132 len = sizeof(info);
5134 if (copy_from_user(&info, optval, len))
5135 return -EFAULT;
5137 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5138 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5139 return -EINVAL;
5140 if (asoc) {
5141 info.sinfo_stream = asoc->default_stream;
5142 info.sinfo_flags = asoc->default_flags;
5143 info.sinfo_ppid = asoc->default_ppid;
5144 info.sinfo_context = asoc->default_context;
5145 info.sinfo_timetolive = asoc->default_timetolive;
5146 } else {
5147 info.sinfo_stream = sp->default_stream;
5148 info.sinfo_flags = sp->default_flags;
5149 info.sinfo_ppid = sp->default_ppid;
5150 info.sinfo_context = sp->default_context;
5151 info.sinfo_timetolive = sp->default_timetolive;
5154 if (put_user(len, optlen))
5155 return -EFAULT;
5156 if (copy_to_user(optval, &info, len))
5157 return -EFAULT;
5159 return 0;
5162 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5163 * (SCTP_DEFAULT_SNDINFO)
5165 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5166 char __user *optval,
5167 int __user *optlen)
5169 struct sctp_sock *sp = sctp_sk(sk);
5170 struct sctp_association *asoc;
5171 struct sctp_sndinfo info;
5173 if (len < sizeof(info))
5174 return -EINVAL;
5176 len = sizeof(info);
5178 if (copy_from_user(&info, optval, len))
5179 return -EFAULT;
5181 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5182 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5183 return -EINVAL;
5184 if (asoc) {
5185 info.snd_sid = asoc->default_stream;
5186 info.snd_flags = asoc->default_flags;
5187 info.snd_ppid = asoc->default_ppid;
5188 info.snd_context = asoc->default_context;
5189 } else {
5190 info.snd_sid = sp->default_stream;
5191 info.snd_flags = sp->default_flags;
5192 info.snd_ppid = sp->default_ppid;
5193 info.snd_context = sp->default_context;
5196 if (put_user(len, optlen))
5197 return -EFAULT;
5198 if (copy_to_user(optval, &info, len))
5199 return -EFAULT;
5201 return 0;
5206 * 7.1.5 SCTP_NODELAY
5208 * Turn on/off any Nagle-like algorithm. This means that packets are
5209 * generally sent as soon as possible and no unnecessary delays are
5210 * introduced, at the cost of more packets in the network. Expects an
5211 * integer boolean flag.
5214 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5215 char __user *optval, int __user *optlen)
5217 int val;
5219 if (len < sizeof(int))
5220 return -EINVAL;
5222 len = sizeof(int);
5223 val = (sctp_sk(sk)->nodelay == 1);
5224 if (put_user(len, optlen))
5225 return -EFAULT;
5226 if (copy_to_user(optval, &val, len))
5227 return -EFAULT;
5228 return 0;
5233 * 7.1.1 SCTP_RTOINFO
5235 * The protocol parameters used to initialize and bound retransmission
5236 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5237 * and modify these parameters.
5238 * All parameters are time values, in milliseconds. A value of 0, when
5239 * modifying the parameters, indicates that the current value should not
5240 * be changed.
5243 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5244 char __user *optval,
5245 int __user *optlen) {
5246 struct sctp_rtoinfo rtoinfo;
5247 struct sctp_association *asoc;
5249 if (len < sizeof (struct sctp_rtoinfo))
5250 return -EINVAL;
5252 len = sizeof(struct sctp_rtoinfo);
5254 if (copy_from_user(&rtoinfo, optval, len))
5255 return -EFAULT;
5257 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5259 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5260 return -EINVAL;
5262 /* Values corresponding to the specific association. */
5263 if (asoc) {
5264 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5265 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5266 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5267 } else {
5268 /* Values corresponding to the endpoint. */
5269 struct sctp_sock *sp = sctp_sk(sk);
5271 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5272 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5273 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5276 if (put_user(len, optlen))
5277 return -EFAULT;
5279 if (copy_to_user(optval, &rtoinfo, len))
5280 return -EFAULT;
5282 return 0;
5287 * 7.1.2 SCTP_ASSOCINFO
5289 * This option is used to tune the maximum retransmission attempts
5290 * of the association.
5291 * Returns an error if the new association retransmission value is
5292 * greater than the sum of the retransmission value of the peer.
5293 * See [SCTP] for more information.
5296 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5297 char __user *optval,
5298 int __user *optlen)
5301 struct sctp_assocparams assocparams;
5302 struct sctp_association *asoc;
5303 struct list_head *pos;
5304 int cnt = 0;
5306 if (len < sizeof (struct sctp_assocparams))
5307 return -EINVAL;
5309 len = sizeof(struct sctp_assocparams);
5311 if (copy_from_user(&assocparams, optval, len))
5312 return -EFAULT;
5314 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5316 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5317 return -EINVAL;
5319 /* Values correspoinding to the specific association */
5320 if (asoc) {
5321 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5322 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5323 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5324 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5326 list_for_each(pos, &asoc->peer.transport_addr_list) {
5327 cnt++;
5330 assocparams.sasoc_number_peer_destinations = cnt;
5331 } else {
5332 /* Values corresponding to the endpoint */
5333 struct sctp_sock *sp = sctp_sk(sk);
5335 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5336 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5337 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5338 assocparams.sasoc_cookie_life =
5339 sp->assocparams.sasoc_cookie_life;
5340 assocparams.sasoc_number_peer_destinations =
5341 sp->assocparams.
5342 sasoc_number_peer_destinations;
5345 if (put_user(len, optlen))
5346 return -EFAULT;
5348 if (copy_to_user(optval, &assocparams, len))
5349 return -EFAULT;
5351 return 0;
5355 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5357 * This socket option is a boolean flag which turns on or off mapped V4
5358 * addresses. If this option is turned on and the socket is type
5359 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5360 * If this option is turned off, then no mapping will be done of V4
5361 * addresses and a user will receive both PF_INET6 and PF_INET type
5362 * addresses on the socket.
5364 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5365 char __user *optval, int __user *optlen)
5367 int val;
5368 struct sctp_sock *sp = sctp_sk(sk);
5370 if (len < sizeof(int))
5371 return -EINVAL;
5373 len = sizeof(int);
5374 val = sp->v4mapped;
5375 if (put_user(len, optlen))
5376 return -EFAULT;
5377 if (copy_to_user(optval, &val, len))
5378 return -EFAULT;
5380 return 0;
5384 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5385 * (chapter and verse is quoted at sctp_setsockopt_context())
5387 static int sctp_getsockopt_context(struct sock *sk, int len,
5388 char __user *optval, int __user *optlen)
5390 struct sctp_assoc_value params;
5391 struct sctp_sock *sp;
5392 struct sctp_association *asoc;
5394 if (len < sizeof(struct sctp_assoc_value))
5395 return -EINVAL;
5397 len = sizeof(struct sctp_assoc_value);
5399 if (copy_from_user(&params, optval, len))
5400 return -EFAULT;
5402 sp = sctp_sk(sk);
5404 if (params.assoc_id != 0) {
5405 asoc = sctp_id2assoc(sk, params.assoc_id);
5406 if (!asoc)
5407 return -EINVAL;
5408 params.assoc_value = asoc->default_rcv_context;
5409 } else {
5410 params.assoc_value = sp->default_rcv_context;
5413 if (put_user(len, optlen))
5414 return -EFAULT;
5415 if (copy_to_user(optval, &params, len))
5416 return -EFAULT;
5418 return 0;
5422 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5423 * This option will get or set the maximum size to put in any outgoing
5424 * SCTP DATA chunk. If a message is larger than this size it will be
5425 * fragmented by SCTP into the specified size. Note that the underlying
5426 * SCTP implementation may fragment into smaller sized chunks when the
5427 * PMTU of the underlying association is smaller than the value set by
5428 * the user. The default value for this option is '0' which indicates
5429 * the user is NOT limiting fragmentation and only the PMTU will effect
5430 * SCTP's choice of DATA chunk size. Note also that values set larger
5431 * than the maximum size of an IP datagram will effectively let SCTP
5432 * control fragmentation (i.e. the same as setting this option to 0).
5434 * The following structure is used to access and modify this parameter:
5436 * struct sctp_assoc_value {
5437 * sctp_assoc_t assoc_id;
5438 * uint32_t assoc_value;
5439 * };
5441 * assoc_id: This parameter is ignored for one-to-one style sockets.
5442 * For one-to-many style sockets this parameter indicates which
5443 * association the user is performing an action upon. Note that if
5444 * this field's value is zero then the endpoints default value is
5445 * changed (effecting future associations only).
5446 * assoc_value: This parameter specifies the maximum size in bytes.
5448 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5449 char __user *optval, int __user *optlen)
5451 struct sctp_assoc_value params;
5452 struct sctp_association *asoc;
5454 if (len == sizeof(int)) {
5455 pr_warn_ratelimited(DEPRECATED
5456 "%s (pid %d) "
5457 "Use of int in maxseg socket option.\n"
5458 "Use struct sctp_assoc_value instead\n",
5459 current->comm, task_pid_nr(current));
5460 params.assoc_id = 0;
5461 } else if (len >= sizeof(struct sctp_assoc_value)) {
5462 len = sizeof(struct sctp_assoc_value);
5463 if (copy_from_user(&params, optval, sizeof(params)))
5464 return -EFAULT;
5465 } else
5466 return -EINVAL;
5468 asoc = sctp_id2assoc(sk, params.assoc_id);
5469 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5470 return -EINVAL;
5472 if (asoc)
5473 params.assoc_value = asoc->frag_point;
5474 else
5475 params.assoc_value = sctp_sk(sk)->user_frag;
5477 if (put_user(len, optlen))
5478 return -EFAULT;
5479 if (len == sizeof(int)) {
5480 if (copy_to_user(optval, &params.assoc_value, len))
5481 return -EFAULT;
5482 } else {
5483 if (copy_to_user(optval, &params, len))
5484 return -EFAULT;
5487 return 0;
5491 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5492 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5494 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5495 char __user *optval, int __user *optlen)
5497 int val;
5499 if (len < sizeof(int))
5500 return -EINVAL;
5502 len = sizeof(int);
5504 val = sctp_sk(sk)->frag_interleave;
5505 if (put_user(len, optlen))
5506 return -EFAULT;
5507 if (copy_to_user(optval, &val, len))
5508 return -EFAULT;
5510 return 0;
5514 * 7.1.25. Set or Get the sctp partial delivery point
5515 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5517 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5518 char __user *optval,
5519 int __user *optlen)
5521 u32 val;
5523 if (len < sizeof(u32))
5524 return -EINVAL;
5526 len = sizeof(u32);
5528 val = sctp_sk(sk)->pd_point;
5529 if (put_user(len, optlen))
5530 return -EFAULT;
5531 if (copy_to_user(optval, &val, len))
5532 return -EFAULT;
5534 return 0;
5538 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5539 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5541 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5542 char __user *optval,
5543 int __user *optlen)
5545 struct sctp_assoc_value params;
5546 struct sctp_sock *sp;
5547 struct sctp_association *asoc;
5549 if (len == sizeof(int)) {
5550 pr_warn_ratelimited(DEPRECATED
5551 "%s (pid %d) "
5552 "Use of int in max_burst socket option.\n"
5553 "Use struct sctp_assoc_value instead\n",
5554 current->comm, task_pid_nr(current));
5555 params.assoc_id = 0;
5556 } else if (len >= sizeof(struct sctp_assoc_value)) {
5557 len = sizeof(struct sctp_assoc_value);
5558 if (copy_from_user(&params, optval, len))
5559 return -EFAULT;
5560 } else
5561 return -EINVAL;
5563 sp = sctp_sk(sk);
5565 if (params.assoc_id != 0) {
5566 asoc = sctp_id2assoc(sk, params.assoc_id);
5567 if (!asoc)
5568 return -EINVAL;
5569 params.assoc_value = asoc->max_burst;
5570 } else
5571 params.assoc_value = sp->max_burst;
5573 if (len == sizeof(int)) {
5574 if (copy_to_user(optval, &params.assoc_value, len))
5575 return -EFAULT;
5576 } else {
5577 if (copy_to_user(optval, &params, len))
5578 return -EFAULT;
5581 return 0;
5585 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5586 char __user *optval, int __user *optlen)
5588 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5589 struct sctp_hmacalgo __user *p = (void __user *)optval;
5590 struct sctp_hmac_algo_param *hmacs;
5591 __u16 data_len = 0;
5592 u32 num_idents;
5593 int i;
5595 if (!ep->auth_enable)
5596 return -EACCES;
5598 hmacs = ep->auth_hmacs_list;
5599 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5601 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5602 return -EINVAL;
5604 len = sizeof(struct sctp_hmacalgo) + data_len;
5605 num_idents = data_len / sizeof(u16);
5607 if (put_user(len, optlen))
5608 return -EFAULT;
5609 if (put_user(num_idents, &p->shmac_num_idents))
5610 return -EFAULT;
5611 for (i = 0; i < num_idents; i++) {
5612 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5614 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5615 return -EFAULT;
5617 return 0;
5620 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5621 char __user *optval, int __user *optlen)
5623 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5624 struct sctp_authkeyid val;
5625 struct sctp_association *asoc;
5627 if (!ep->auth_enable)
5628 return -EACCES;
5630 if (len < sizeof(struct sctp_authkeyid))
5631 return -EINVAL;
5632 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5633 return -EFAULT;
5635 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5636 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5637 return -EINVAL;
5639 if (asoc)
5640 val.scact_keynumber = asoc->active_key_id;
5641 else
5642 val.scact_keynumber = ep->active_key_id;
5644 len = sizeof(struct sctp_authkeyid);
5645 if (put_user(len, optlen))
5646 return -EFAULT;
5647 if (copy_to_user(optval, &val, len))
5648 return -EFAULT;
5650 return 0;
5653 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5654 char __user *optval, int __user *optlen)
5656 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5657 struct sctp_authchunks __user *p = (void __user *)optval;
5658 struct sctp_authchunks val;
5659 struct sctp_association *asoc;
5660 struct sctp_chunks_param *ch;
5661 u32 num_chunks = 0;
5662 char __user *to;
5664 if (!ep->auth_enable)
5665 return -EACCES;
5667 if (len < sizeof(struct sctp_authchunks))
5668 return -EINVAL;
5670 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5671 return -EFAULT;
5673 to = p->gauth_chunks;
5674 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5675 if (!asoc)
5676 return -EINVAL;
5678 ch = asoc->peer.peer_chunks;
5679 if (!ch)
5680 goto num;
5682 /* See if the user provided enough room for all the data */
5683 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5684 if (len < num_chunks)
5685 return -EINVAL;
5687 if (copy_to_user(to, ch->chunks, num_chunks))
5688 return -EFAULT;
5689 num:
5690 len = sizeof(struct sctp_authchunks) + num_chunks;
5691 if (put_user(len, optlen))
5692 return -EFAULT;
5693 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5694 return -EFAULT;
5695 return 0;
5698 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5699 char __user *optval, int __user *optlen)
5701 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5702 struct sctp_authchunks __user *p = (void __user *)optval;
5703 struct sctp_authchunks val;
5704 struct sctp_association *asoc;
5705 struct sctp_chunks_param *ch;
5706 u32 num_chunks = 0;
5707 char __user *to;
5709 if (!ep->auth_enable)
5710 return -EACCES;
5712 if (len < sizeof(struct sctp_authchunks))
5713 return -EINVAL;
5715 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5716 return -EFAULT;
5718 to = p->gauth_chunks;
5719 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5720 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5721 return -EINVAL;
5723 if (asoc)
5724 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5725 else
5726 ch = ep->auth_chunk_list;
5728 if (!ch)
5729 goto num;
5731 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5732 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5733 return -EINVAL;
5735 if (copy_to_user(to, ch->chunks, num_chunks))
5736 return -EFAULT;
5737 num:
5738 len = sizeof(struct sctp_authchunks) + num_chunks;
5739 if (put_user(len, optlen))
5740 return -EFAULT;
5741 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5742 return -EFAULT;
5744 return 0;
5748 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5749 * This option gets the current number of associations that are attached
5750 * to a one-to-many style socket. The option value is an uint32_t.
5752 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5753 char __user *optval, int __user *optlen)
5755 struct sctp_sock *sp = sctp_sk(sk);
5756 struct sctp_association *asoc;
5757 u32 val = 0;
5759 if (sctp_style(sk, TCP))
5760 return -EOPNOTSUPP;
5762 if (len < sizeof(u32))
5763 return -EINVAL;
5765 len = sizeof(u32);
5767 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5768 val++;
5771 if (put_user(len, optlen))
5772 return -EFAULT;
5773 if (copy_to_user(optval, &val, len))
5774 return -EFAULT;
5776 return 0;
5780 * 8.1.23 SCTP_AUTO_ASCONF
5781 * See the corresponding setsockopt entry as description
5783 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5784 char __user *optval, int __user *optlen)
5786 int val = 0;
5788 if (len < sizeof(int))
5789 return -EINVAL;
5791 len = sizeof(int);
5792 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5793 val = 1;
5794 if (put_user(len, optlen))
5795 return -EFAULT;
5796 if (copy_to_user(optval, &val, len))
5797 return -EFAULT;
5798 return 0;
5802 * 8.2.6. Get the Current Identifiers of Associations
5803 * (SCTP_GET_ASSOC_ID_LIST)
5805 * This option gets the current list of SCTP association identifiers of
5806 * the SCTP associations handled by a one-to-many style socket.
5808 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5809 char __user *optval, int __user *optlen)
5811 struct sctp_sock *sp = sctp_sk(sk);
5812 struct sctp_association *asoc;
5813 struct sctp_assoc_ids *ids;
5814 u32 num = 0;
5816 if (sctp_style(sk, TCP))
5817 return -EOPNOTSUPP;
5819 if (len < sizeof(struct sctp_assoc_ids))
5820 return -EINVAL;
5822 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5823 num++;
5826 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5827 return -EINVAL;
5829 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5831 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
5832 if (unlikely(!ids))
5833 return -ENOMEM;
5835 ids->gaids_number_of_ids = num;
5836 num = 0;
5837 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5838 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5841 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5842 kfree(ids);
5843 return -EFAULT;
5846 kfree(ids);
5847 return 0;
5851 * SCTP_PEER_ADDR_THLDS
5853 * This option allows us to fetch the partially failed threshold for one or all
5854 * transports in an association. See Section 6.1 of:
5855 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5857 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5858 char __user *optval,
5859 int len,
5860 int __user *optlen)
5862 struct sctp_paddrthlds val;
5863 struct sctp_transport *trans;
5864 struct sctp_association *asoc;
5866 if (len < sizeof(struct sctp_paddrthlds))
5867 return -EINVAL;
5868 len = sizeof(struct sctp_paddrthlds);
5869 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5870 return -EFAULT;
5872 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5873 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5874 if (!asoc)
5875 return -ENOENT;
5877 val.spt_pathpfthld = asoc->pf_retrans;
5878 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5879 } else {
5880 trans = sctp_addr_id2transport(sk, &val.spt_address,
5881 val.spt_assoc_id);
5882 if (!trans)
5883 return -ENOENT;
5885 val.spt_pathmaxrxt = trans->pathmaxrxt;
5886 val.spt_pathpfthld = trans->pf_retrans;
5889 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5890 return -EFAULT;
5892 return 0;
5896 * SCTP_GET_ASSOC_STATS
5898 * This option retrieves local per endpoint statistics. It is modeled
5899 * after OpenSolaris' implementation
5901 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5902 char __user *optval,
5903 int __user *optlen)
5905 struct sctp_assoc_stats sas;
5906 struct sctp_association *asoc = NULL;
5908 /* User must provide at least the assoc id */
5909 if (len < sizeof(sctp_assoc_t))
5910 return -EINVAL;
5912 /* Allow the struct to grow and fill in as much as possible */
5913 len = min_t(size_t, len, sizeof(sas));
5915 if (copy_from_user(&sas, optval, len))
5916 return -EFAULT;
5918 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5919 if (!asoc)
5920 return -EINVAL;
5922 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5923 sas.sas_gapcnt = asoc->stats.gapcnt;
5924 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5925 sas.sas_osacks = asoc->stats.osacks;
5926 sas.sas_isacks = asoc->stats.isacks;
5927 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5928 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5929 sas.sas_oodchunks = asoc->stats.oodchunks;
5930 sas.sas_iodchunks = asoc->stats.iodchunks;
5931 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5932 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5933 sas.sas_idupchunks = asoc->stats.idupchunks;
5934 sas.sas_opackets = asoc->stats.opackets;
5935 sas.sas_ipackets = asoc->stats.ipackets;
5937 /* New high max rto observed, will return 0 if not a single
5938 * RTO update took place. obs_rto_ipaddr will be bogus
5939 * in such a case
5941 sas.sas_maxrto = asoc->stats.max_obs_rto;
5942 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5943 sizeof(struct sockaddr_storage));
5945 /* Mark beginning of a new observation period */
5946 asoc->stats.max_obs_rto = asoc->rto_min;
5948 if (put_user(len, optlen))
5949 return -EFAULT;
5951 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5953 if (copy_to_user(optval, &sas, len))
5954 return -EFAULT;
5956 return 0;
5959 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
5960 char __user *optval,
5961 int __user *optlen)
5963 int val = 0;
5965 if (len < sizeof(int))
5966 return -EINVAL;
5968 len = sizeof(int);
5969 if (sctp_sk(sk)->recvrcvinfo)
5970 val = 1;
5971 if (put_user(len, optlen))
5972 return -EFAULT;
5973 if (copy_to_user(optval, &val, len))
5974 return -EFAULT;
5976 return 0;
5979 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
5980 char __user *optval,
5981 int __user *optlen)
5983 int val = 0;
5985 if (len < sizeof(int))
5986 return -EINVAL;
5988 len = sizeof(int);
5989 if (sctp_sk(sk)->recvnxtinfo)
5990 val = 1;
5991 if (put_user(len, optlen))
5992 return -EFAULT;
5993 if (copy_to_user(optval, &val, len))
5994 return -EFAULT;
5996 return 0;
5999 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6000 char __user *optval, int __user *optlen)
6002 int retval = 0;
6003 int len;
6005 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6007 /* I can hardly begin to describe how wrong this is. This is
6008 * so broken as to be worse than useless. The API draft
6009 * REALLY is NOT helpful here... I am not convinced that the
6010 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6011 * are at all well-founded.
6013 if (level != SOL_SCTP) {
6014 struct sctp_af *af = sctp_sk(sk)->pf->af;
6016 retval = af->getsockopt(sk, level, optname, optval, optlen);
6017 return retval;
6020 if (get_user(len, optlen))
6021 return -EFAULT;
6023 if (len < 0)
6024 return -EINVAL;
6026 lock_sock(sk);
6028 switch (optname) {
6029 case SCTP_STATUS:
6030 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6031 break;
6032 case SCTP_DISABLE_FRAGMENTS:
6033 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6034 optlen);
6035 break;
6036 case SCTP_EVENTS:
6037 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6038 break;
6039 case SCTP_AUTOCLOSE:
6040 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6041 break;
6042 case SCTP_SOCKOPT_PEELOFF:
6043 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6044 break;
6045 case SCTP_PEER_ADDR_PARAMS:
6046 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6047 optlen);
6048 break;
6049 case SCTP_DELAYED_SACK:
6050 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6051 optlen);
6052 break;
6053 case SCTP_INITMSG:
6054 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6055 break;
6056 case SCTP_GET_PEER_ADDRS:
6057 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6058 optlen);
6059 break;
6060 case SCTP_GET_LOCAL_ADDRS:
6061 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6062 optlen);
6063 break;
6064 case SCTP_SOCKOPT_CONNECTX3:
6065 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6066 break;
6067 case SCTP_DEFAULT_SEND_PARAM:
6068 retval = sctp_getsockopt_default_send_param(sk, len,
6069 optval, optlen);
6070 break;
6071 case SCTP_DEFAULT_SNDINFO:
6072 retval = sctp_getsockopt_default_sndinfo(sk, len,
6073 optval, optlen);
6074 break;
6075 case SCTP_PRIMARY_ADDR:
6076 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6077 break;
6078 case SCTP_NODELAY:
6079 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6080 break;
6081 case SCTP_RTOINFO:
6082 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6083 break;
6084 case SCTP_ASSOCINFO:
6085 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6086 break;
6087 case SCTP_I_WANT_MAPPED_V4_ADDR:
6088 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6089 break;
6090 case SCTP_MAXSEG:
6091 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6092 break;
6093 case SCTP_GET_PEER_ADDR_INFO:
6094 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6095 optlen);
6096 break;
6097 case SCTP_ADAPTATION_LAYER:
6098 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6099 optlen);
6100 break;
6101 case SCTP_CONTEXT:
6102 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6103 break;
6104 case SCTP_FRAGMENT_INTERLEAVE:
6105 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6106 optlen);
6107 break;
6108 case SCTP_PARTIAL_DELIVERY_POINT:
6109 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6110 optlen);
6111 break;
6112 case SCTP_MAX_BURST:
6113 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6114 break;
6115 case SCTP_AUTH_KEY:
6116 case SCTP_AUTH_CHUNK:
6117 case SCTP_AUTH_DELETE_KEY:
6118 retval = -EOPNOTSUPP;
6119 break;
6120 case SCTP_HMAC_IDENT:
6121 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6122 break;
6123 case SCTP_AUTH_ACTIVE_KEY:
6124 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6125 break;
6126 case SCTP_PEER_AUTH_CHUNKS:
6127 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6128 optlen);
6129 break;
6130 case SCTP_LOCAL_AUTH_CHUNKS:
6131 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6132 optlen);
6133 break;
6134 case SCTP_GET_ASSOC_NUMBER:
6135 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6136 break;
6137 case SCTP_GET_ASSOC_ID_LIST:
6138 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6139 break;
6140 case SCTP_AUTO_ASCONF:
6141 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6142 break;
6143 case SCTP_PEER_ADDR_THLDS:
6144 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6145 break;
6146 case SCTP_GET_ASSOC_STATS:
6147 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6148 break;
6149 case SCTP_RECVRCVINFO:
6150 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6151 break;
6152 case SCTP_RECVNXTINFO:
6153 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6154 break;
6155 default:
6156 retval = -ENOPROTOOPT;
6157 break;
6160 release_sock(sk);
6161 return retval;
6164 static void sctp_hash(struct sock *sk)
6166 /* STUB */
6169 static void sctp_unhash(struct sock *sk)
6171 /* STUB */
6174 /* Check if port is acceptable. Possibly find first available port.
6176 * The port hash table (contained in the 'global' SCTP protocol storage
6177 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6178 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6179 * list (the list number is the port number hashed out, so as you
6180 * would expect from a hash function, all the ports in a given list have
6181 * such a number that hashes out to the same list number; you were
6182 * expecting that, right?); so each list has a set of ports, with a
6183 * link to the socket (struct sock) that uses it, the port number and
6184 * a fastreuse flag (FIXME: NPI ipg).
6186 static struct sctp_bind_bucket *sctp_bucket_create(
6187 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6189 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6191 struct sctp_bind_hashbucket *head; /* hash list */
6192 struct sctp_bind_bucket *pp;
6193 unsigned short snum;
6194 int ret;
6196 snum = ntohs(addr->v4.sin_port);
6198 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6200 local_bh_disable();
6202 if (snum == 0) {
6203 /* Search for an available port. */
6204 int low, high, remaining, index;
6205 unsigned int rover;
6206 struct net *net = sock_net(sk);
6208 inet_get_local_port_range(net, &low, &high);
6209 remaining = (high - low) + 1;
6210 rover = prandom_u32() % remaining + low;
6212 do {
6213 rover++;
6214 if ((rover < low) || (rover > high))
6215 rover = low;
6216 if (inet_is_local_reserved_port(net, rover))
6217 continue;
6218 index = sctp_phashfn(sock_net(sk), rover);
6219 head = &sctp_port_hashtable[index];
6220 spin_lock(&head->lock);
6221 sctp_for_each_hentry(pp, &head->chain)
6222 if ((pp->port == rover) &&
6223 net_eq(sock_net(sk), pp->net))
6224 goto next;
6225 break;
6226 next:
6227 spin_unlock(&head->lock);
6228 } while (--remaining > 0);
6230 /* Exhausted local port range during search? */
6231 ret = 1;
6232 if (remaining <= 0)
6233 goto fail;
6235 /* OK, here is the one we will use. HEAD (the port
6236 * hash table list entry) is non-NULL and we hold it's
6237 * mutex.
6239 snum = rover;
6240 } else {
6241 /* We are given an specific port number; we verify
6242 * that it is not being used. If it is used, we will
6243 * exahust the search in the hash list corresponding
6244 * to the port number (snum) - we detect that with the
6245 * port iterator, pp being NULL.
6247 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6248 spin_lock(&head->lock);
6249 sctp_for_each_hentry(pp, &head->chain) {
6250 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6251 goto pp_found;
6254 pp = NULL;
6255 goto pp_not_found;
6256 pp_found:
6257 if (!hlist_empty(&pp->owner)) {
6258 /* We had a port hash table hit - there is an
6259 * available port (pp != NULL) and it is being
6260 * used by other socket (pp->owner not empty); that other
6261 * socket is going to be sk2.
6263 int reuse = sk->sk_reuse;
6264 struct sock *sk2;
6266 pr_debug("%s: found a possible match\n", __func__);
6268 if (pp->fastreuse && sk->sk_reuse &&
6269 sk->sk_state != SCTP_SS_LISTENING)
6270 goto success;
6272 /* Run through the list of sockets bound to the port
6273 * (pp->port) [via the pointers bind_next and
6274 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6275 * we get the endpoint they describe and run through
6276 * the endpoint's list of IP (v4 or v6) addresses,
6277 * comparing each of the addresses with the address of
6278 * the socket sk. If we find a match, then that means
6279 * that this port/socket (sk) combination are already
6280 * in an endpoint.
6282 sk_for_each_bound(sk2, &pp->owner) {
6283 struct sctp_endpoint *ep2;
6284 ep2 = sctp_sk(sk2)->ep;
6286 if (sk == sk2 ||
6287 (reuse && sk2->sk_reuse &&
6288 sk2->sk_state != SCTP_SS_LISTENING))
6289 continue;
6291 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6292 sctp_sk(sk2), sctp_sk(sk))) {
6293 ret = (long)sk2;
6294 goto fail_unlock;
6298 pr_debug("%s: found a match\n", __func__);
6300 pp_not_found:
6301 /* If there was a hash table miss, create a new port. */
6302 ret = 1;
6303 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6304 goto fail_unlock;
6306 /* In either case (hit or miss), make sure fastreuse is 1 only
6307 * if sk->sk_reuse is too (that is, if the caller requested
6308 * SO_REUSEADDR on this socket -sk-).
6310 if (hlist_empty(&pp->owner)) {
6311 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6312 pp->fastreuse = 1;
6313 else
6314 pp->fastreuse = 0;
6315 } else if (pp->fastreuse &&
6316 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6317 pp->fastreuse = 0;
6319 /* We are set, so fill up all the data in the hash table
6320 * entry, tie the socket list information with the rest of the
6321 * sockets FIXME: Blurry, NPI (ipg).
6323 success:
6324 if (!sctp_sk(sk)->bind_hash) {
6325 inet_sk(sk)->inet_num = snum;
6326 sk_add_bind_node(sk, &pp->owner);
6327 sctp_sk(sk)->bind_hash = pp;
6329 ret = 0;
6331 fail_unlock:
6332 spin_unlock(&head->lock);
6334 fail:
6335 local_bh_enable();
6336 return ret;
6339 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6340 * port is requested.
6342 static int sctp_get_port(struct sock *sk, unsigned short snum)
6344 union sctp_addr addr;
6345 struct sctp_af *af = sctp_sk(sk)->pf->af;
6347 /* Set up a dummy address struct from the sk. */
6348 af->from_sk(&addr, sk);
6349 addr.v4.sin_port = htons(snum);
6351 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6352 return !!sctp_get_port_local(sk, &addr);
6356 * Move a socket to LISTENING state.
6358 static int sctp_listen_start(struct sock *sk, int backlog)
6360 struct sctp_sock *sp = sctp_sk(sk);
6361 struct sctp_endpoint *ep = sp->ep;
6362 struct crypto_hash *tfm = NULL;
6363 char alg[32];
6365 /* Allocate HMAC for generating cookie. */
6366 if (!sp->hmac && sp->sctp_hmac_alg) {
6367 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6368 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6369 if (IS_ERR(tfm)) {
6370 net_info_ratelimited("failed to load transform for %s: %ld\n",
6371 sp->sctp_hmac_alg, PTR_ERR(tfm));
6372 return -ENOSYS;
6374 sctp_sk(sk)->hmac = tfm;
6378 * If a bind() or sctp_bindx() is not called prior to a listen()
6379 * call that allows new associations to be accepted, the system
6380 * picks an ephemeral port and will choose an address set equivalent
6381 * to binding with a wildcard address.
6383 * This is not currently spelled out in the SCTP sockets
6384 * extensions draft, but follows the practice as seen in TCP
6385 * sockets.
6388 sk->sk_state = SCTP_SS_LISTENING;
6389 if (!ep->base.bind_addr.port) {
6390 if (sctp_autobind(sk))
6391 return -EAGAIN;
6392 } else {
6393 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6394 sk->sk_state = SCTP_SS_CLOSED;
6395 return -EADDRINUSE;
6399 sk->sk_max_ack_backlog = backlog;
6400 sctp_hash_endpoint(ep);
6401 return 0;
6405 * 4.1.3 / 5.1.3 listen()
6407 * By default, new associations are not accepted for UDP style sockets.
6408 * An application uses listen() to mark a socket as being able to
6409 * accept new associations.
6411 * On TCP style sockets, applications use listen() to ready the SCTP
6412 * endpoint for accepting inbound associations.
6414 * On both types of endpoints a backlog of '0' disables listening.
6416 * Move a socket to LISTENING state.
6418 int sctp_inet_listen(struct socket *sock, int backlog)
6420 struct sock *sk = sock->sk;
6421 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6422 int err = -EINVAL;
6424 if (unlikely(backlog < 0))
6425 return err;
6427 lock_sock(sk);
6429 /* Peeled-off sockets are not allowed to listen(). */
6430 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6431 goto out;
6433 if (sock->state != SS_UNCONNECTED)
6434 goto out;
6436 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
6437 goto out;
6439 /* If backlog is zero, disable listening. */
6440 if (!backlog) {
6441 if (sctp_sstate(sk, CLOSED))
6442 goto out;
6444 err = 0;
6445 sctp_unhash_endpoint(ep);
6446 sk->sk_state = SCTP_SS_CLOSED;
6447 if (sk->sk_reuse)
6448 sctp_sk(sk)->bind_hash->fastreuse = 1;
6449 goto out;
6452 /* If we are already listening, just update the backlog */
6453 if (sctp_sstate(sk, LISTENING))
6454 sk->sk_max_ack_backlog = backlog;
6455 else {
6456 err = sctp_listen_start(sk, backlog);
6457 if (err)
6458 goto out;
6461 err = 0;
6462 out:
6463 release_sock(sk);
6464 return err;
6468 * This function is done by modeling the current datagram_poll() and the
6469 * tcp_poll(). Note that, based on these implementations, we don't
6470 * lock the socket in this function, even though it seems that,
6471 * ideally, locking or some other mechanisms can be used to ensure
6472 * the integrity of the counters (sndbuf and wmem_alloc) used
6473 * in this place. We assume that we don't need locks either until proven
6474 * otherwise.
6476 * Another thing to note is that we include the Async I/O support
6477 * here, again, by modeling the current TCP/UDP code. We don't have
6478 * a good way to test with it yet.
6480 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6482 struct sock *sk = sock->sk;
6483 struct sctp_sock *sp = sctp_sk(sk);
6484 unsigned int mask;
6486 poll_wait(file, sk_sleep(sk), wait);
6488 /* A TCP-style listening socket becomes readable when the accept queue
6489 * is not empty.
6491 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6492 return (!list_empty(&sp->ep->asocs)) ?
6493 (POLLIN | POLLRDNORM) : 0;
6495 mask = 0;
6497 /* Is there any exceptional events? */
6498 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6499 mask |= POLLERR |
6500 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6501 if (sk->sk_shutdown & RCV_SHUTDOWN)
6502 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6503 if (sk->sk_shutdown == SHUTDOWN_MASK)
6504 mask |= POLLHUP;
6506 /* Is it readable? Reconsider this code with TCP-style support. */
6507 if (!skb_queue_empty(&sk->sk_receive_queue))
6508 mask |= POLLIN | POLLRDNORM;
6510 /* The association is either gone or not ready. */
6511 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6512 return mask;
6514 /* Is it writable? */
6515 if (sctp_writeable(sk)) {
6516 mask |= POLLOUT | POLLWRNORM;
6517 } else {
6518 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
6520 * Since the socket is not locked, the buffer
6521 * might be made available after the writeable check and
6522 * before the bit is set. This could cause a lost I/O
6523 * signal. tcp_poll() has a race breaker for this race
6524 * condition. Based on their implementation, we put
6525 * in the following code to cover it as well.
6527 if (sctp_writeable(sk))
6528 mask |= POLLOUT | POLLWRNORM;
6530 return mask;
6533 /********************************************************************
6534 * 2nd Level Abstractions
6535 ********************************************************************/
6537 static struct sctp_bind_bucket *sctp_bucket_create(
6538 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6540 struct sctp_bind_bucket *pp;
6542 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6543 if (pp) {
6544 SCTP_DBG_OBJCNT_INC(bind_bucket);
6545 pp->port = snum;
6546 pp->fastreuse = 0;
6547 INIT_HLIST_HEAD(&pp->owner);
6548 pp->net = net;
6549 hlist_add_head(&pp->node, &head->chain);
6551 return pp;
6554 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6555 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6557 if (pp && hlist_empty(&pp->owner)) {
6558 __hlist_del(&pp->node);
6559 kmem_cache_free(sctp_bucket_cachep, pp);
6560 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6564 /* Release this socket's reference to a local port. */
6565 static inline void __sctp_put_port(struct sock *sk)
6567 struct sctp_bind_hashbucket *head =
6568 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6569 inet_sk(sk)->inet_num)];
6570 struct sctp_bind_bucket *pp;
6572 spin_lock(&head->lock);
6573 pp = sctp_sk(sk)->bind_hash;
6574 __sk_del_bind_node(sk);
6575 sctp_sk(sk)->bind_hash = NULL;
6576 inet_sk(sk)->inet_num = 0;
6577 sctp_bucket_destroy(pp);
6578 spin_unlock(&head->lock);
6581 void sctp_put_port(struct sock *sk)
6583 local_bh_disable();
6584 __sctp_put_port(sk);
6585 local_bh_enable();
6589 * The system picks an ephemeral port and choose an address set equivalent
6590 * to binding with a wildcard address.
6591 * One of those addresses will be the primary address for the association.
6592 * This automatically enables the multihoming capability of SCTP.
6594 static int sctp_autobind(struct sock *sk)
6596 union sctp_addr autoaddr;
6597 struct sctp_af *af;
6598 __be16 port;
6600 /* Initialize a local sockaddr structure to INADDR_ANY. */
6601 af = sctp_sk(sk)->pf->af;
6603 port = htons(inet_sk(sk)->inet_num);
6604 af->inaddr_any(&autoaddr, port);
6606 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6609 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6611 * From RFC 2292
6612 * 4.2 The cmsghdr Structure *
6614 * When ancillary data is sent or received, any number of ancillary data
6615 * objects can be specified by the msg_control and msg_controllen members of
6616 * the msghdr structure, because each object is preceded by
6617 * a cmsghdr structure defining the object's length (the cmsg_len member).
6618 * Historically Berkeley-derived implementations have passed only one object
6619 * at a time, but this API allows multiple objects to be
6620 * passed in a single call to sendmsg() or recvmsg(). The following example
6621 * shows two ancillary data objects in a control buffer.
6623 * |<--------------------------- msg_controllen -------------------------->|
6624 * | |
6626 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6628 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6629 * | | |
6631 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6633 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6634 * | | | | |
6636 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6637 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6639 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6641 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6645 * msg_control
6646 * points here
6648 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6650 struct cmsghdr *cmsg;
6651 struct msghdr *my_msg = (struct msghdr *)msg;
6653 for_each_cmsghdr(cmsg, my_msg) {
6654 if (!CMSG_OK(my_msg, cmsg))
6655 return -EINVAL;
6657 /* Should we parse this header or ignore? */
6658 if (cmsg->cmsg_level != IPPROTO_SCTP)
6659 continue;
6661 /* Strictly check lengths following example in SCM code. */
6662 switch (cmsg->cmsg_type) {
6663 case SCTP_INIT:
6664 /* SCTP Socket API Extension
6665 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6667 * This cmsghdr structure provides information for
6668 * initializing new SCTP associations with sendmsg().
6669 * The SCTP_INITMSG socket option uses this same data
6670 * structure. This structure is not used for
6671 * recvmsg().
6673 * cmsg_level cmsg_type cmsg_data[]
6674 * ------------ ------------ ----------------------
6675 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6677 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
6678 return -EINVAL;
6680 cmsgs->init = CMSG_DATA(cmsg);
6681 break;
6683 case SCTP_SNDRCV:
6684 /* SCTP Socket API Extension
6685 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6687 * This cmsghdr structure specifies SCTP options for
6688 * sendmsg() and describes SCTP header information
6689 * about a received message through recvmsg().
6691 * cmsg_level cmsg_type cmsg_data[]
6692 * ------------ ------------ ----------------------
6693 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6695 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6696 return -EINVAL;
6698 cmsgs->srinfo = CMSG_DATA(cmsg);
6700 if (cmsgs->srinfo->sinfo_flags &
6701 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6702 SCTP_SACK_IMMEDIATELY |
6703 SCTP_ABORT | SCTP_EOF))
6704 return -EINVAL;
6705 break;
6707 case SCTP_SNDINFO:
6708 /* SCTP Socket API Extension
6709 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6711 * This cmsghdr structure specifies SCTP options for
6712 * sendmsg(). This structure and SCTP_RCVINFO replaces
6713 * SCTP_SNDRCV which has been deprecated.
6715 * cmsg_level cmsg_type cmsg_data[]
6716 * ------------ ------------ ---------------------
6717 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6719 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
6720 return -EINVAL;
6722 cmsgs->sinfo = CMSG_DATA(cmsg);
6724 if (cmsgs->sinfo->snd_flags &
6725 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6726 SCTP_SACK_IMMEDIATELY |
6727 SCTP_ABORT | SCTP_EOF))
6728 return -EINVAL;
6729 break;
6730 default:
6731 return -EINVAL;
6735 return 0;
6739 * Wait for a packet..
6740 * Note: This function is the same function as in core/datagram.c
6741 * with a few modifications to make lksctp work.
6743 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6745 int error;
6746 DEFINE_WAIT(wait);
6748 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6750 /* Socket errors? */
6751 error = sock_error(sk);
6752 if (error)
6753 goto out;
6755 if (!skb_queue_empty(&sk->sk_receive_queue))
6756 goto ready;
6758 /* Socket shut down? */
6759 if (sk->sk_shutdown & RCV_SHUTDOWN)
6760 goto out;
6762 /* Sequenced packets can come disconnected. If so we report the
6763 * problem.
6765 error = -ENOTCONN;
6767 /* Is there a good reason to think that we may receive some data? */
6768 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6769 goto out;
6771 /* Handle signals. */
6772 if (signal_pending(current))
6773 goto interrupted;
6775 /* Let another process have a go. Since we are going to sleep
6776 * anyway. Note: This may cause odd behaviors if the message
6777 * does not fit in the user's buffer, but this seems to be the
6778 * only way to honor MSG_DONTWAIT realistically.
6780 release_sock(sk);
6781 *timeo_p = schedule_timeout(*timeo_p);
6782 lock_sock(sk);
6784 ready:
6785 finish_wait(sk_sleep(sk), &wait);
6786 return 0;
6788 interrupted:
6789 error = sock_intr_errno(*timeo_p);
6791 out:
6792 finish_wait(sk_sleep(sk), &wait);
6793 *err = error;
6794 return error;
6797 /* Receive a datagram.
6798 * Note: This is pretty much the same routine as in core/datagram.c
6799 * with a few changes to make lksctp work.
6801 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6802 int noblock, int *err)
6804 int error;
6805 struct sk_buff *skb;
6806 long timeo;
6808 timeo = sock_rcvtimeo(sk, noblock);
6810 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6811 MAX_SCHEDULE_TIMEOUT);
6813 do {
6814 /* Again only user level code calls this function,
6815 * so nothing interrupt level
6816 * will suddenly eat the receive_queue.
6818 * Look at current nfs client by the way...
6819 * However, this function was correct in any case. 8)
6821 if (flags & MSG_PEEK) {
6822 spin_lock_bh(&sk->sk_receive_queue.lock);
6823 skb = skb_peek(&sk->sk_receive_queue);
6824 if (skb)
6825 atomic_inc(&skb->users);
6826 spin_unlock_bh(&sk->sk_receive_queue.lock);
6827 } else {
6828 skb = skb_dequeue(&sk->sk_receive_queue);
6831 if (skb)
6832 return skb;
6834 /* Caller is allowed not to check sk->sk_err before calling. */
6835 error = sock_error(sk);
6836 if (error)
6837 goto no_packet;
6839 if (sk->sk_shutdown & RCV_SHUTDOWN)
6840 break;
6842 if (sk_can_busy_loop(sk) &&
6843 sk_busy_loop(sk, noblock))
6844 continue;
6846 /* User doesn't want to wait. */
6847 error = -EAGAIN;
6848 if (!timeo)
6849 goto no_packet;
6850 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6852 return NULL;
6854 no_packet:
6855 *err = error;
6856 return NULL;
6859 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6860 static void __sctp_write_space(struct sctp_association *asoc)
6862 struct sock *sk = asoc->base.sk;
6864 if (sctp_wspace(asoc) <= 0)
6865 return;
6867 if (waitqueue_active(&asoc->wait))
6868 wake_up_interruptible(&asoc->wait);
6870 if (sctp_writeable(sk)) {
6871 struct socket_wq *wq;
6873 rcu_read_lock();
6874 wq = rcu_dereference(sk->sk_wq);
6875 if (wq) {
6876 if (waitqueue_active(&wq->wait))
6877 wake_up_interruptible(&wq->wait);
6879 /* Note that we try to include the Async I/O support
6880 * here by modeling from the current TCP/UDP code.
6881 * We have not tested with it yet.
6883 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6884 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
6886 rcu_read_unlock();
6890 static void sctp_wake_up_waiters(struct sock *sk,
6891 struct sctp_association *asoc)
6893 struct sctp_association *tmp = asoc;
6895 /* We do accounting for the sndbuf space per association,
6896 * so we only need to wake our own association.
6898 if (asoc->ep->sndbuf_policy)
6899 return __sctp_write_space(asoc);
6901 /* If association goes down and is just flushing its
6902 * outq, then just normally notify others.
6904 if (asoc->base.dead)
6905 return sctp_write_space(sk);
6907 /* Accounting for the sndbuf space is per socket, so we
6908 * need to wake up others, try to be fair and in case of
6909 * other associations, let them have a go first instead
6910 * of just doing a sctp_write_space() call.
6912 * Note that we reach sctp_wake_up_waiters() only when
6913 * associations free up queued chunks, thus we are under
6914 * lock and the list of associations on a socket is
6915 * guaranteed not to change.
6917 for (tmp = list_next_entry(tmp, asocs); 1;
6918 tmp = list_next_entry(tmp, asocs)) {
6919 /* Manually skip the head element. */
6920 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6921 continue;
6922 /* Wake up association. */
6923 __sctp_write_space(tmp);
6924 /* We've reached the end. */
6925 if (tmp == asoc)
6926 break;
6930 /* Do accounting for the sndbuf space.
6931 * Decrement the used sndbuf space of the corresponding association by the
6932 * data size which was just transmitted(freed).
6934 static void sctp_wfree(struct sk_buff *skb)
6936 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
6937 struct sctp_association *asoc = chunk->asoc;
6938 struct sock *sk = asoc->base.sk;
6940 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6941 sizeof(struct sk_buff) +
6942 sizeof(struct sctp_chunk);
6944 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6947 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6949 sk->sk_wmem_queued -= skb->truesize;
6950 sk_mem_uncharge(sk, skb->truesize);
6952 sock_wfree(skb);
6953 sctp_wake_up_waiters(sk, asoc);
6955 sctp_association_put(asoc);
6958 /* Do accounting for the receive space on the socket.
6959 * Accounting for the association is done in ulpevent.c
6960 * We set this as a destructor for the cloned data skbs so that
6961 * accounting is done at the correct time.
6963 void sctp_sock_rfree(struct sk_buff *skb)
6965 struct sock *sk = skb->sk;
6966 struct sctp_ulpevent *event = sctp_skb2event(skb);
6968 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6971 * Mimic the behavior of sock_rfree
6973 sk_mem_uncharge(sk, event->rmem_len);
6977 /* Helper function to wait for space in the sndbuf. */
6978 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6979 size_t msg_len, struct sock **orig_sk)
6981 struct sock *sk = asoc->base.sk;
6982 int err = 0;
6983 long current_timeo = *timeo_p;
6984 DEFINE_WAIT(wait);
6986 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6987 *timeo_p, msg_len);
6989 /* Increment the association's refcnt. */
6990 sctp_association_hold(asoc);
6992 /* Wait on the association specific sndbuf space. */
6993 for (;;) {
6994 prepare_to_wait_exclusive(&asoc->wait, &wait,
6995 TASK_INTERRUPTIBLE);
6996 if (asoc->base.dead)
6997 goto do_dead;
6998 if (!*timeo_p)
6999 goto do_nonblock;
7000 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
7001 goto do_error;
7002 if (signal_pending(current))
7003 goto do_interrupted;
7004 if (msg_len <= sctp_wspace(asoc))
7005 break;
7007 /* Let another process have a go. Since we are going
7008 * to sleep anyway.
7010 release_sock(sk);
7011 current_timeo = schedule_timeout(current_timeo);
7012 lock_sock(sk);
7013 if (sk != asoc->base.sk) {
7014 release_sock(sk);
7015 sk = asoc->base.sk;
7016 lock_sock(sk);
7019 *timeo_p = current_timeo;
7022 out:
7023 *orig_sk = sk;
7024 finish_wait(&asoc->wait, &wait);
7026 /* Release the association's refcnt. */
7027 sctp_association_put(asoc);
7029 return err;
7031 do_dead:
7032 err = -ESRCH;
7033 goto out;
7035 do_error:
7036 err = -EPIPE;
7037 goto out;
7039 do_interrupted:
7040 err = sock_intr_errno(*timeo_p);
7041 goto out;
7043 do_nonblock:
7044 err = -EAGAIN;
7045 goto out;
7048 void sctp_data_ready(struct sock *sk)
7050 struct socket_wq *wq;
7052 rcu_read_lock();
7053 wq = rcu_dereference(sk->sk_wq);
7054 if (wq_has_sleeper(wq))
7055 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7056 POLLRDNORM | POLLRDBAND);
7057 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7058 rcu_read_unlock();
7061 /* If socket sndbuf has changed, wake up all per association waiters. */
7062 void sctp_write_space(struct sock *sk)
7064 struct sctp_association *asoc;
7066 /* Wake up the tasks in each wait queue. */
7067 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7068 __sctp_write_space(asoc);
7072 /* Is there any sndbuf space available on the socket?
7074 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7075 * associations on the same socket. For a UDP-style socket with
7076 * multiple associations, it is possible for it to be "unwriteable"
7077 * prematurely. I assume that this is acceptable because
7078 * a premature "unwriteable" is better than an accidental "writeable" which
7079 * would cause an unwanted block under certain circumstances. For the 1-1
7080 * UDP-style sockets or TCP-style sockets, this code should work.
7081 * - Daisy
7083 static int sctp_writeable(struct sock *sk)
7085 int amt = 0;
7087 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7088 if (amt < 0)
7089 amt = 0;
7090 return amt;
7093 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7094 * returns immediately with EINPROGRESS.
7096 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7098 struct sock *sk = asoc->base.sk;
7099 int err = 0;
7100 long current_timeo = *timeo_p;
7101 DEFINE_WAIT(wait);
7103 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7105 /* Increment the association's refcnt. */
7106 sctp_association_hold(asoc);
7108 for (;;) {
7109 prepare_to_wait_exclusive(&asoc->wait, &wait,
7110 TASK_INTERRUPTIBLE);
7111 if (!*timeo_p)
7112 goto do_nonblock;
7113 if (sk->sk_shutdown & RCV_SHUTDOWN)
7114 break;
7115 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7116 asoc->base.dead)
7117 goto do_error;
7118 if (signal_pending(current))
7119 goto do_interrupted;
7121 if (sctp_state(asoc, ESTABLISHED))
7122 break;
7124 /* Let another process have a go. Since we are going
7125 * to sleep anyway.
7127 release_sock(sk);
7128 current_timeo = schedule_timeout(current_timeo);
7129 lock_sock(sk);
7131 *timeo_p = current_timeo;
7134 out:
7135 finish_wait(&asoc->wait, &wait);
7137 /* Release the association's refcnt. */
7138 sctp_association_put(asoc);
7140 return err;
7142 do_error:
7143 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7144 err = -ETIMEDOUT;
7145 else
7146 err = -ECONNREFUSED;
7147 goto out;
7149 do_interrupted:
7150 err = sock_intr_errno(*timeo_p);
7151 goto out;
7153 do_nonblock:
7154 err = -EINPROGRESS;
7155 goto out;
7158 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7160 struct sctp_endpoint *ep;
7161 int err = 0;
7162 DEFINE_WAIT(wait);
7164 ep = sctp_sk(sk)->ep;
7167 for (;;) {
7168 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7169 TASK_INTERRUPTIBLE);
7171 if (list_empty(&ep->asocs)) {
7172 release_sock(sk);
7173 timeo = schedule_timeout(timeo);
7174 lock_sock(sk);
7177 err = -EINVAL;
7178 if (!sctp_sstate(sk, LISTENING))
7179 break;
7181 err = 0;
7182 if (!list_empty(&ep->asocs))
7183 break;
7185 err = sock_intr_errno(timeo);
7186 if (signal_pending(current))
7187 break;
7189 err = -EAGAIN;
7190 if (!timeo)
7191 break;
7194 finish_wait(sk_sleep(sk), &wait);
7196 return err;
7199 static void sctp_wait_for_close(struct sock *sk, long timeout)
7201 DEFINE_WAIT(wait);
7203 do {
7204 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7205 if (list_empty(&sctp_sk(sk)->ep->asocs))
7206 break;
7207 release_sock(sk);
7208 timeout = schedule_timeout(timeout);
7209 lock_sock(sk);
7210 } while (!signal_pending(current) && timeout);
7212 finish_wait(sk_sleep(sk), &wait);
7215 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7217 struct sk_buff *frag;
7219 if (!skb->data_len)
7220 goto done;
7222 /* Don't forget the fragments. */
7223 skb_walk_frags(skb, frag)
7224 sctp_skb_set_owner_r_frag(frag, sk);
7226 done:
7227 sctp_skb_set_owner_r(skb, sk);
7230 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7231 struct sctp_association *asoc)
7233 struct inet_sock *inet = inet_sk(sk);
7234 struct inet_sock *newinet;
7236 newsk->sk_type = sk->sk_type;
7237 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7238 newsk->sk_flags = sk->sk_flags;
7239 newsk->sk_tsflags = sk->sk_tsflags;
7240 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7241 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7242 newsk->sk_reuse = sk->sk_reuse;
7244 newsk->sk_shutdown = sk->sk_shutdown;
7245 newsk->sk_destruct = sctp_destruct_sock;
7246 newsk->sk_family = sk->sk_family;
7247 newsk->sk_protocol = IPPROTO_SCTP;
7248 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7249 newsk->sk_sndbuf = sk->sk_sndbuf;
7250 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7251 newsk->sk_lingertime = sk->sk_lingertime;
7252 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7253 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7255 newinet = inet_sk(newsk);
7257 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7258 * getsockname() and getpeername()
7260 newinet->inet_sport = inet->inet_sport;
7261 newinet->inet_saddr = inet->inet_saddr;
7262 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7263 newinet->inet_dport = htons(asoc->peer.port);
7264 newinet->pmtudisc = inet->pmtudisc;
7265 newinet->inet_id = asoc->next_tsn ^ jiffies;
7267 newinet->uc_ttl = inet->uc_ttl;
7268 newinet->mc_loop = 1;
7269 newinet->mc_ttl = 1;
7270 newinet->mc_index = 0;
7271 newinet->mc_list = NULL;
7273 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7274 net_enable_timestamp();
7276 security_sk_clone(sk, newsk);
7279 static inline void sctp_copy_descendant(struct sock *sk_to,
7280 const struct sock *sk_from)
7282 int ancestor_size = sizeof(struct inet_sock) +
7283 sizeof(struct sctp_sock) -
7284 offsetof(struct sctp_sock, auto_asconf_list);
7286 if (sk_from->sk_family == PF_INET6)
7287 ancestor_size += sizeof(struct ipv6_pinfo);
7289 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7292 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7293 * and its messages to the newsk.
7295 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7296 struct sctp_association *assoc,
7297 sctp_socket_type_t type)
7299 struct sctp_sock *oldsp = sctp_sk(oldsk);
7300 struct sctp_sock *newsp = sctp_sk(newsk);
7301 struct sctp_bind_bucket *pp; /* hash list port iterator */
7302 struct sctp_endpoint *newep = newsp->ep;
7303 struct sk_buff *skb, *tmp;
7304 struct sctp_ulpevent *event;
7305 struct sctp_bind_hashbucket *head;
7307 /* Migrate socket buffer sizes and all the socket level options to the
7308 * new socket.
7310 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7311 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7312 /* Brute force copy old sctp opt. */
7313 sctp_copy_descendant(newsk, oldsk);
7315 /* Restore the ep value that was overwritten with the above structure
7316 * copy.
7318 newsp->ep = newep;
7319 newsp->hmac = NULL;
7321 /* Hook this new socket in to the bind_hash list. */
7322 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7323 inet_sk(oldsk)->inet_num)];
7324 local_bh_disable();
7325 spin_lock(&head->lock);
7326 pp = sctp_sk(oldsk)->bind_hash;
7327 sk_add_bind_node(newsk, &pp->owner);
7328 sctp_sk(newsk)->bind_hash = pp;
7329 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7330 spin_unlock(&head->lock);
7331 local_bh_enable();
7333 /* Copy the bind_addr list from the original endpoint to the new
7334 * endpoint so that we can handle restarts properly
7336 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7337 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7339 /* Move any messages in the old socket's receive queue that are for the
7340 * peeled off association to the new socket's receive queue.
7342 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7343 event = sctp_skb2event(skb);
7344 if (event->asoc == assoc) {
7345 __skb_unlink(skb, &oldsk->sk_receive_queue);
7346 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7347 sctp_skb_set_owner_r_frag(skb, newsk);
7351 /* Clean up any messages pending delivery due to partial
7352 * delivery. Three cases:
7353 * 1) No partial deliver; no work.
7354 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7355 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7357 skb_queue_head_init(&newsp->pd_lobby);
7358 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7360 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7361 struct sk_buff_head *queue;
7363 /* Decide which queue to move pd_lobby skbs to. */
7364 if (assoc->ulpq.pd_mode) {
7365 queue = &newsp->pd_lobby;
7366 } else
7367 queue = &newsk->sk_receive_queue;
7369 /* Walk through the pd_lobby, looking for skbs that
7370 * need moved to the new socket.
7372 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7373 event = sctp_skb2event(skb);
7374 if (event->asoc == assoc) {
7375 __skb_unlink(skb, &oldsp->pd_lobby);
7376 __skb_queue_tail(queue, skb);
7377 sctp_skb_set_owner_r_frag(skb, newsk);
7381 /* Clear up any skbs waiting for the partial
7382 * delivery to finish.
7384 if (assoc->ulpq.pd_mode)
7385 sctp_clear_pd(oldsk, NULL);
7389 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7390 sctp_skb_set_owner_r_frag(skb, newsk);
7392 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7393 sctp_skb_set_owner_r_frag(skb, newsk);
7395 /* Set the type of socket to indicate that it is peeled off from the
7396 * original UDP-style socket or created with the accept() call on a
7397 * TCP-style socket..
7399 newsp->type = type;
7401 /* Mark the new socket "in-use" by the user so that any packets
7402 * that may arrive on the association after we've moved it are
7403 * queued to the backlog. This prevents a potential race between
7404 * backlog processing on the old socket and new-packet processing
7405 * on the new socket.
7407 * The caller has just allocated newsk so we can guarantee that other
7408 * paths won't try to lock it and then oldsk.
7410 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7411 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
7412 sctp_assoc_migrate(assoc, newsk);
7413 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
7415 /* If the association on the newsk is already closed before accept()
7416 * is called, set RCV_SHUTDOWN flag.
7418 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7419 newsk->sk_shutdown |= RCV_SHUTDOWN;
7421 newsk->sk_state = SCTP_SS_ESTABLISHED;
7422 release_sock(newsk);
7426 /* This proto struct describes the ULP interface for SCTP. */
7427 struct proto sctp_prot = {
7428 .name = "SCTP",
7429 .owner = THIS_MODULE,
7430 .close = sctp_close,
7431 .connect = sctp_connect,
7432 .disconnect = sctp_disconnect,
7433 .accept = sctp_accept,
7434 .ioctl = sctp_ioctl,
7435 .init = sctp_init_sock,
7436 .destroy = sctp_destroy_sock,
7437 .shutdown = sctp_shutdown,
7438 .setsockopt = sctp_setsockopt,
7439 .getsockopt = sctp_getsockopt,
7440 .sendmsg = sctp_sendmsg,
7441 .recvmsg = sctp_recvmsg,
7442 .bind = sctp_bind,
7443 .backlog_rcv = sctp_backlog_rcv,
7444 .hash = sctp_hash,
7445 .unhash = sctp_unhash,
7446 .get_port = sctp_get_port,
7447 .obj_size = sizeof(struct sctp_sock),
7448 .sysctl_mem = sysctl_sctp_mem,
7449 .sysctl_rmem = sysctl_sctp_rmem,
7450 .sysctl_wmem = sysctl_sctp_wmem,
7451 .memory_pressure = &sctp_memory_pressure,
7452 .enter_memory_pressure = sctp_enter_memory_pressure,
7453 .memory_allocated = &sctp_memory_allocated,
7454 .sockets_allocated = &sctp_sockets_allocated,
7457 #if IS_ENABLED(CONFIG_IPV6)
7459 #include <net/transp_v6.h>
7460 static void sctp_v6_destroy_sock(struct sock *sk)
7462 sctp_destroy_sock(sk);
7463 inet6_destroy_sock(sk);
7466 struct proto sctpv6_prot = {
7467 .name = "SCTPv6",
7468 .owner = THIS_MODULE,
7469 .close = sctp_close,
7470 .connect = sctp_connect,
7471 .disconnect = sctp_disconnect,
7472 .accept = sctp_accept,
7473 .ioctl = sctp_ioctl,
7474 .init = sctp_init_sock,
7475 .destroy = sctp_v6_destroy_sock,
7476 .shutdown = sctp_shutdown,
7477 .setsockopt = sctp_setsockopt,
7478 .getsockopt = sctp_getsockopt,
7479 .sendmsg = sctp_sendmsg,
7480 .recvmsg = sctp_recvmsg,
7481 .bind = sctp_bind,
7482 .backlog_rcv = sctp_backlog_rcv,
7483 .hash = sctp_hash,
7484 .unhash = sctp_unhash,
7485 .get_port = sctp_get_port,
7486 .obj_size = sizeof(struct sctp6_sock),
7487 .sysctl_mem = sysctl_sctp_mem,
7488 .sysctl_rmem = sysctl_sctp_rmem,
7489 .sysctl_wmem = sysctl_sctp_wmem,
7490 .memory_pressure = &sctp_memory_pressure,
7491 .enter_memory_pressure = sctp_enter_memory_pressure,
7492 .memory_allocated = &sctp_memory_allocated,
7493 .sockets_allocated = &sctp_sockets_allocated,
7495 #endif /* IS_ENABLED(CONFIG_IPV6) */