dt-bindings: add Marvell core PLL and clock divider PMU documentation
[linux/fpc-iii.git] / net / sctp / socket.c
blob897c01c029cab3d5805cc56b0964c70e06f4143a
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
35 * email address(es):
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <linux/types.h>
56 #include <linux/kernel.h>
57 #include <linux/wait.h>
58 #include <linux/time.h>
59 #include <linux/ip.h>
60 #include <linux/capability.h>
61 #include <linux/fcntl.h>
62 #include <linux/poll.h>
63 #include <linux/init.h>
64 #include <linux/crypto.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
69 #include <net/ip.h>
70 #include <net/icmp.h>
71 #include <net/route.h>
72 #include <net/ipv6.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
78 #include <net/sock.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
86 size_t msg_len);
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
105 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 /* Verify that this is a valid address. */
172 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
173 int len)
175 struct sctp_af *af;
177 /* Verify basic sockaddr. */
178 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
179 if (!af)
180 return -EINVAL;
182 /* Is this a valid SCTP address? */
183 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
184 return -EINVAL;
186 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
187 return -EINVAL;
189 return 0;
192 /* Look up the association by its id. If this is not a UDP-style
193 * socket, the ID field is always ignored.
195 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
197 struct sctp_association *asoc = NULL;
199 /* If this is not a UDP-style socket, assoc id should be ignored. */
200 if (!sctp_style(sk, UDP)) {
201 /* Return NULL if the socket state is not ESTABLISHED. It
202 * could be a TCP-style listening socket or a socket which
203 * hasn't yet called connect() to establish an association.
205 if (!sctp_sstate(sk, ESTABLISHED))
206 return NULL;
208 /* Get the first and the only association from the list. */
209 if (!list_empty(&sctp_sk(sk)->ep->asocs))
210 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
211 struct sctp_association, asocs);
212 return asoc;
215 /* Otherwise this is a UDP-style socket. */
216 if (!id || (id == (sctp_assoc_t)-1))
217 return NULL;
219 spin_lock_bh(&sctp_assocs_id_lock);
220 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
221 spin_unlock_bh(&sctp_assocs_id_lock);
223 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
224 return NULL;
226 return asoc;
229 /* Look up the transport from an address and an assoc id. If both address and
230 * id are specified, the associations matching the address and the id should be
231 * the same.
233 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
234 struct sockaddr_storage *addr,
235 sctp_assoc_t id)
237 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
238 struct sctp_transport *transport;
239 union sctp_addr *laddr = (union sctp_addr *)addr;
241 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
242 laddr,
243 &transport);
245 if (!addr_asoc)
246 return NULL;
248 id_asoc = sctp_id2assoc(sk, id);
249 if (id_asoc && (id_asoc != addr_asoc))
250 return NULL;
252 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
253 (union sctp_addr *)addr);
255 return transport;
258 /* API 3.1.2 bind() - UDP Style Syntax
259 * The syntax of bind() is,
261 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
263 * sd - the socket descriptor returned by socket().
264 * addr - the address structure (struct sockaddr_in or struct
265 * sockaddr_in6 [RFC 2553]),
266 * addr_len - the size of the address structure.
268 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
270 int retval = 0;
272 lock_sock(sk);
274 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
275 addr, addr_len);
277 /* Disallow binding twice. */
278 if (!sctp_sk(sk)->ep->base.bind_addr.port)
279 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
280 addr_len);
281 else
282 retval = -EINVAL;
284 release_sock(sk);
286 return retval;
289 static long sctp_get_port_local(struct sock *, union sctp_addr *);
291 /* Verify this is a valid sockaddr. */
292 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
293 union sctp_addr *addr, int len)
295 struct sctp_af *af;
297 /* Check minimum size. */
298 if (len < sizeof (struct sockaddr))
299 return NULL;
301 /* V4 mapped address are really of AF_INET family */
302 if (addr->sa.sa_family == AF_INET6 &&
303 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
304 if (!opt->pf->af_supported(AF_INET, opt))
305 return NULL;
306 } else {
307 /* Does this PF support this AF? */
308 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
309 return NULL;
312 /* If we get this far, af is valid. */
313 af = sctp_get_af_specific(addr->sa.sa_family);
315 if (len < af->sockaddr_len)
316 return NULL;
318 return af;
321 /* Bind a local address either to an endpoint or to an association. */
322 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
324 struct net *net = sock_net(sk);
325 struct sctp_sock *sp = sctp_sk(sk);
326 struct sctp_endpoint *ep = sp->ep;
327 struct sctp_bind_addr *bp = &ep->base.bind_addr;
328 struct sctp_af *af;
329 unsigned short snum;
330 int ret = 0;
332 /* Common sockaddr verification. */
333 af = sctp_sockaddr_af(sp, addr, len);
334 if (!af) {
335 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
336 __func__, sk, addr, len);
337 return -EINVAL;
340 snum = ntohs(addr->v4.sin_port);
342 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
343 __func__, sk, &addr->sa, bp->port, snum, len);
345 /* PF specific bind() address verification. */
346 if (!sp->pf->bind_verify(sp, addr))
347 return -EADDRNOTAVAIL;
349 /* We must either be unbound, or bind to the same port.
350 * It's OK to allow 0 ports if we are already bound.
351 * We'll just inhert an already bound port in this case
353 if (bp->port) {
354 if (!snum)
355 snum = bp->port;
356 else if (snum != bp->port) {
357 pr_debug("%s: new port %d doesn't match existing port "
358 "%d\n", __func__, snum, bp->port);
359 return -EINVAL;
363 if (snum && snum < PROT_SOCK &&
364 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
365 return -EACCES;
367 /* See if the address matches any of the addresses we may have
368 * already bound before checking against other endpoints.
370 if (sctp_bind_addr_match(bp, addr, sp))
371 return -EINVAL;
373 /* Make sure we are allowed to bind here.
374 * The function sctp_get_port_local() does duplicate address
375 * detection.
377 addr->v4.sin_port = htons(snum);
378 if ((ret = sctp_get_port_local(sk, addr))) {
379 return -EADDRINUSE;
382 /* Refresh ephemeral port. */
383 if (!bp->port)
384 bp->port = inet_sk(sk)->inet_num;
386 /* Add the address to the bind address list.
387 * Use GFP_ATOMIC since BHs will be disabled.
389 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
391 /* Copy back into socket for getsockname() use. */
392 if (!ret) {
393 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
394 sp->pf->to_sk_saddr(addr, sk);
397 return ret;
400 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
402 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
403 * at any one time. If a sender, after sending an ASCONF chunk, decides
404 * it needs to transfer another ASCONF Chunk, it MUST wait until the
405 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
406 * subsequent ASCONF. Note this restriction binds each side, so at any
407 * time two ASCONF may be in-transit on any given association (one sent
408 * from each endpoint).
410 static int sctp_send_asconf(struct sctp_association *asoc,
411 struct sctp_chunk *chunk)
413 struct net *net = sock_net(asoc->base.sk);
414 int retval = 0;
416 /* If there is an outstanding ASCONF chunk, queue it for later
417 * transmission.
419 if (asoc->addip_last_asconf) {
420 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
421 goto out;
424 /* Hold the chunk until an ASCONF_ACK is received. */
425 sctp_chunk_hold(chunk);
426 retval = sctp_primitive_ASCONF(net, asoc, chunk);
427 if (retval)
428 sctp_chunk_free(chunk);
429 else
430 asoc->addip_last_asconf = chunk;
432 out:
433 return retval;
436 /* Add a list of addresses as bind addresses to local endpoint or
437 * association.
439 * Basically run through each address specified in the addrs/addrcnt
440 * array/length pair, determine if it is IPv6 or IPv4 and call
441 * sctp_do_bind() on it.
443 * If any of them fails, then the operation will be reversed and the
444 * ones that were added will be removed.
446 * Only sctp_setsockopt_bindx() is supposed to call this function.
448 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
450 int cnt;
451 int retval = 0;
452 void *addr_buf;
453 struct sockaddr *sa_addr;
454 struct sctp_af *af;
456 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
457 addrs, addrcnt);
459 addr_buf = addrs;
460 for (cnt = 0; cnt < addrcnt; cnt++) {
461 /* The list may contain either IPv4 or IPv6 address;
462 * determine the address length for walking thru the list.
464 sa_addr = addr_buf;
465 af = sctp_get_af_specific(sa_addr->sa_family);
466 if (!af) {
467 retval = -EINVAL;
468 goto err_bindx_add;
471 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
472 af->sockaddr_len);
474 addr_buf += af->sockaddr_len;
476 err_bindx_add:
477 if (retval < 0) {
478 /* Failed. Cleanup the ones that have been added */
479 if (cnt > 0)
480 sctp_bindx_rem(sk, addrs, cnt);
481 return retval;
485 return retval;
488 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
489 * associations that are part of the endpoint indicating that a list of local
490 * addresses are added to the endpoint.
492 * If any of the addresses is already in the bind address list of the
493 * association, we do not send the chunk for that association. But it will not
494 * affect other associations.
496 * Only sctp_setsockopt_bindx() is supposed to call this function.
498 static int sctp_send_asconf_add_ip(struct sock *sk,
499 struct sockaddr *addrs,
500 int addrcnt)
502 struct net *net = sock_net(sk);
503 struct sctp_sock *sp;
504 struct sctp_endpoint *ep;
505 struct sctp_association *asoc;
506 struct sctp_bind_addr *bp;
507 struct sctp_chunk *chunk;
508 struct sctp_sockaddr_entry *laddr;
509 union sctp_addr *addr;
510 union sctp_addr saveaddr;
511 void *addr_buf;
512 struct sctp_af *af;
513 struct list_head *p;
514 int i;
515 int retval = 0;
517 if (!net->sctp.addip_enable)
518 return retval;
520 sp = sctp_sk(sk);
521 ep = sp->ep;
523 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
524 __func__, sk, addrs, addrcnt);
526 list_for_each_entry(asoc, &ep->asocs, asocs) {
527 if (!asoc->peer.asconf_capable)
528 continue;
530 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
531 continue;
533 if (!sctp_state(asoc, ESTABLISHED))
534 continue;
536 /* Check if any address in the packed array of addresses is
537 * in the bind address list of the association. If so,
538 * do not send the asconf chunk to its peer, but continue with
539 * other associations.
541 addr_buf = addrs;
542 for (i = 0; i < addrcnt; i++) {
543 addr = addr_buf;
544 af = sctp_get_af_specific(addr->v4.sin_family);
545 if (!af) {
546 retval = -EINVAL;
547 goto out;
550 if (sctp_assoc_lookup_laddr(asoc, addr))
551 break;
553 addr_buf += af->sockaddr_len;
555 if (i < addrcnt)
556 continue;
558 /* Use the first valid address in bind addr list of
559 * association as Address Parameter of ASCONF CHUNK.
561 bp = &asoc->base.bind_addr;
562 p = bp->address_list.next;
563 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
564 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
565 addrcnt, SCTP_PARAM_ADD_IP);
566 if (!chunk) {
567 retval = -ENOMEM;
568 goto out;
571 /* Add the new addresses to the bind address list with
572 * use_as_src set to 0.
574 addr_buf = addrs;
575 for (i = 0; i < addrcnt; i++) {
576 addr = addr_buf;
577 af = sctp_get_af_specific(addr->v4.sin_family);
578 memcpy(&saveaddr, addr, af->sockaddr_len);
579 retval = sctp_add_bind_addr(bp, &saveaddr,
580 SCTP_ADDR_NEW, GFP_ATOMIC);
581 addr_buf += af->sockaddr_len;
583 if (asoc->src_out_of_asoc_ok) {
584 struct sctp_transport *trans;
586 list_for_each_entry(trans,
587 &asoc->peer.transport_addr_list, transports) {
588 /* Clear the source and route cache */
589 dst_release(trans->dst);
590 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
591 2*asoc->pathmtu, 4380));
592 trans->ssthresh = asoc->peer.i.a_rwnd;
593 trans->rto = asoc->rto_initial;
594 sctp_max_rto(asoc, trans);
595 trans->rtt = trans->srtt = trans->rttvar = 0;
596 sctp_transport_route(trans, NULL,
597 sctp_sk(asoc->base.sk));
600 retval = sctp_send_asconf(asoc, chunk);
603 out:
604 return retval;
607 /* Remove a list of addresses from bind addresses list. Do not remove the
608 * last address.
610 * Basically run through each address specified in the addrs/addrcnt
611 * array/length pair, determine if it is IPv6 or IPv4 and call
612 * sctp_del_bind() on it.
614 * If any of them fails, then the operation will be reversed and the
615 * ones that were removed will be added back.
617 * At least one address has to be left; if only one address is
618 * available, the operation will return -EBUSY.
620 * Only sctp_setsockopt_bindx() is supposed to call this function.
622 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
624 struct sctp_sock *sp = sctp_sk(sk);
625 struct sctp_endpoint *ep = sp->ep;
626 int cnt;
627 struct sctp_bind_addr *bp = &ep->base.bind_addr;
628 int retval = 0;
629 void *addr_buf;
630 union sctp_addr *sa_addr;
631 struct sctp_af *af;
633 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
634 __func__, sk, addrs, addrcnt);
636 addr_buf = addrs;
637 for (cnt = 0; cnt < addrcnt; cnt++) {
638 /* If the bind address list is empty or if there is only one
639 * bind address, there is nothing more to be removed (we need
640 * at least one address here).
642 if (list_empty(&bp->address_list) ||
643 (sctp_list_single_entry(&bp->address_list))) {
644 retval = -EBUSY;
645 goto err_bindx_rem;
648 sa_addr = addr_buf;
649 af = sctp_get_af_specific(sa_addr->sa.sa_family);
650 if (!af) {
651 retval = -EINVAL;
652 goto err_bindx_rem;
655 if (!af->addr_valid(sa_addr, sp, NULL)) {
656 retval = -EADDRNOTAVAIL;
657 goto err_bindx_rem;
660 if (sa_addr->v4.sin_port &&
661 sa_addr->v4.sin_port != htons(bp->port)) {
662 retval = -EINVAL;
663 goto err_bindx_rem;
666 if (!sa_addr->v4.sin_port)
667 sa_addr->v4.sin_port = htons(bp->port);
669 /* FIXME - There is probably a need to check if sk->sk_saddr and
670 * sk->sk_rcv_addr are currently set to one of the addresses to
671 * be removed. This is something which needs to be looked into
672 * when we are fixing the outstanding issues with multi-homing
673 * socket routing and failover schemes. Refer to comments in
674 * sctp_do_bind(). -daisy
676 retval = sctp_del_bind_addr(bp, sa_addr);
678 addr_buf += af->sockaddr_len;
679 err_bindx_rem:
680 if (retval < 0) {
681 /* Failed. Add the ones that has been removed back */
682 if (cnt > 0)
683 sctp_bindx_add(sk, addrs, cnt);
684 return retval;
688 return retval;
691 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
692 * the associations that are part of the endpoint indicating that a list of
693 * local addresses are removed from the endpoint.
695 * If any of the addresses is already in the bind address list of the
696 * association, we do not send the chunk for that association. But it will not
697 * affect other associations.
699 * Only sctp_setsockopt_bindx() is supposed to call this function.
701 static int sctp_send_asconf_del_ip(struct sock *sk,
702 struct sockaddr *addrs,
703 int addrcnt)
705 struct net *net = sock_net(sk);
706 struct sctp_sock *sp;
707 struct sctp_endpoint *ep;
708 struct sctp_association *asoc;
709 struct sctp_transport *transport;
710 struct sctp_bind_addr *bp;
711 struct sctp_chunk *chunk;
712 union sctp_addr *laddr;
713 void *addr_buf;
714 struct sctp_af *af;
715 struct sctp_sockaddr_entry *saddr;
716 int i;
717 int retval = 0;
718 int stored = 0;
720 chunk = NULL;
721 if (!net->sctp.addip_enable)
722 return retval;
724 sp = sctp_sk(sk);
725 ep = sp->ep;
727 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
728 __func__, sk, addrs, addrcnt);
730 list_for_each_entry(asoc, &ep->asocs, asocs) {
732 if (!asoc->peer.asconf_capable)
733 continue;
735 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
736 continue;
738 if (!sctp_state(asoc, ESTABLISHED))
739 continue;
741 /* Check if any address in the packed array of addresses is
742 * not present in the bind address list of the association.
743 * If so, do not send the asconf chunk to its peer, but
744 * continue with other associations.
746 addr_buf = addrs;
747 for (i = 0; i < addrcnt; i++) {
748 laddr = addr_buf;
749 af = sctp_get_af_specific(laddr->v4.sin_family);
750 if (!af) {
751 retval = -EINVAL;
752 goto out;
755 if (!sctp_assoc_lookup_laddr(asoc, laddr))
756 break;
758 addr_buf += af->sockaddr_len;
760 if (i < addrcnt)
761 continue;
763 /* Find one address in the association's bind address list
764 * that is not in the packed array of addresses. This is to
765 * make sure that we do not delete all the addresses in the
766 * association.
768 bp = &asoc->base.bind_addr;
769 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
770 addrcnt, sp);
771 if ((laddr == NULL) && (addrcnt == 1)) {
772 if (asoc->asconf_addr_del_pending)
773 continue;
774 asoc->asconf_addr_del_pending =
775 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
776 if (asoc->asconf_addr_del_pending == NULL) {
777 retval = -ENOMEM;
778 goto out;
780 asoc->asconf_addr_del_pending->sa.sa_family =
781 addrs->sa_family;
782 asoc->asconf_addr_del_pending->v4.sin_port =
783 htons(bp->port);
784 if (addrs->sa_family == AF_INET) {
785 struct sockaddr_in *sin;
787 sin = (struct sockaddr_in *)addrs;
788 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
789 } else if (addrs->sa_family == AF_INET6) {
790 struct sockaddr_in6 *sin6;
792 sin6 = (struct sockaddr_in6 *)addrs;
793 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
796 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
797 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
798 asoc->asconf_addr_del_pending);
800 asoc->src_out_of_asoc_ok = 1;
801 stored = 1;
802 goto skip_mkasconf;
805 if (laddr == NULL)
806 return -EINVAL;
808 /* We do not need RCU protection throughout this loop
809 * because this is done under a socket lock from the
810 * setsockopt call.
812 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
813 SCTP_PARAM_DEL_IP);
814 if (!chunk) {
815 retval = -ENOMEM;
816 goto out;
819 skip_mkasconf:
820 /* Reset use_as_src flag for the addresses in the bind address
821 * list that are to be deleted.
823 addr_buf = addrs;
824 for (i = 0; i < addrcnt; i++) {
825 laddr = addr_buf;
826 af = sctp_get_af_specific(laddr->v4.sin_family);
827 list_for_each_entry(saddr, &bp->address_list, list) {
828 if (sctp_cmp_addr_exact(&saddr->a, laddr))
829 saddr->state = SCTP_ADDR_DEL;
831 addr_buf += af->sockaddr_len;
834 /* Update the route and saddr entries for all the transports
835 * as some of the addresses in the bind address list are
836 * about to be deleted and cannot be used as source addresses.
838 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
839 transports) {
840 dst_release(transport->dst);
841 sctp_transport_route(transport, NULL,
842 sctp_sk(asoc->base.sk));
845 if (stored)
846 /* We don't need to transmit ASCONF */
847 continue;
848 retval = sctp_send_asconf(asoc, chunk);
850 out:
851 return retval;
854 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
855 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
857 struct sock *sk = sctp_opt2sk(sp);
858 union sctp_addr *addr;
859 struct sctp_af *af;
861 /* It is safe to write port space in caller. */
862 addr = &addrw->a;
863 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
864 af = sctp_get_af_specific(addr->sa.sa_family);
865 if (!af)
866 return -EINVAL;
867 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
868 return -EINVAL;
870 if (addrw->state == SCTP_ADDR_NEW)
871 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
872 else
873 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
876 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
878 * API 8.1
879 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
880 * int flags);
882 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
883 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
884 * or IPv6 addresses.
886 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
887 * Section 3.1.2 for this usage.
889 * addrs is a pointer to an array of one or more socket addresses. Each
890 * address is contained in its appropriate structure (i.e. struct
891 * sockaddr_in or struct sockaddr_in6) the family of the address type
892 * must be used to distinguish the address length (note that this
893 * representation is termed a "packed array" of addresses). The caller
894 * specifies the number of addresses in the array with addrcnt.
896 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
897 * -1, and sets errno to the appropriate error code.
899 * For SCTP, the port given in each socket address must be the same, or
900 * sctp_bindx() will fail, setting errno to EINVAL.
902 * The flags parameter is formed from the bitwise OR of zero or more of
903 * the following currently defined flags:
905 * SCTP_BINDX_ADD_ADDR
907 * SCTP_BINDX_REM_ADDR
909 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
910 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
911 * addresses from the association. The two flags are mutually exclusive;
912 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
913 * not remove all addresses from an association; sctp_bindx() will
914 * reject such an attempt with EINVAL.
916 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
917 * additional addresses with an endpoint after calling bind(). Or use
918 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
919 * socket is associated with so that no new association accepted will be
920 * associated with those addresses. If the endpoint supports dynamic
921 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
922 * endpoint to send the appropriate message to the peer to change the
923 * peers address lists.
925 * Adding and removing addresses from a connected association is
926 * optional functionality. Implementations that do not support this
927 * functionality should return EOPNOTSUPP.
929 * Basically do nothing but copying the addresses from user to kernel
930 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
931 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
932 * from userspace.
934 * We don't use copy_from_user() for optimization: we first do the
935 * sanity checks (buffer size -fast- and access check-healthy
936 * pointer); if all of those succeed, then we can alloc the memory
937 * (expensive operation) needed to copy the data to kernel. Then we do
938 * the copying without checking the user space area
939 * (__copy_from_user()).
941 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
942 * it.
944 * sk The sk of the socket
945 * addrs The pointer to the addresses in user land
946 * addrssize Size of the addrs buffer
947 * op Operation to perform (add or remove, see the flags of
948 * sctp_bindx)
950 * Returns 0 if ok, <0 errno code on error.
952 static int sctp_setsockopt_bindx(struct sock *sk,
953 struct sockaddr __user *addrs,
954 int addrs_size, int op)
956 struct sockaddr *kaddrs;
957 int err;
958 int addrcnt = 0;
959 int walk_size = 0;
960 struct sockaddr *sa_addr;
961 void *addr_buf;
962 struct sctp_af *af;
964 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
965 __func__, sk, addrs, addrs_size, op);
967 if (unlikely(addrs_size <= 0))
968 return -EINVAL;
970 /* Check the user passed a healthy pointer. */
971 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
972 return -EFAULT;
974 /* Alloc space for the address array in kernel memory. */
975 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
976 if (unlikely(!kaddrs))
977 return -ENOMEM;
979 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
980 kfree(kaddrs);
981 return -EFAULT;
984 /* Walk through the addrs buffer and count the number of addresses. */
985 addr_buf = kaddrs;
986 while (walk_size < addrs_size) {
987 if (walk_size + sizeof(sa_family_t) > addrs_size) {
988 kfree(kaddrs);
989 return -EINVAL;
992 sa_addr = addr_buf;
993 af = sctp_get_af_specific(sa_addr->sa_family);
995 /* If the address family is not supported or if this address
996 * causes the address buffer to overflow return EINVAL.
998 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
999 kfree(kaddrs);
1000 return -EINVAL;
1002 addrcnt++;
1003 addr_buf += af->sockaddr_len;
1004 walk_size += af->sockaddr_len;
1007 /* Do the work. */
1008 switch (op) {
1009 case SCTP_BINDX_ADD_ADDR:
1010 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1011 if (err)
1012 goto out;
1013 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1014 break;
1016 case SCTP_BINDX_REM_ADDR:
1017 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1018 if (err)
1019 goto out;
1020 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1021 break;
1023 default:
1024 err = -EINVAL;
1025 break;
1028 out:
1029 kfree(kaddrs);
1031 return err;
1034 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1036 * Common routine for handling connect() and sctp_connectx().
1037 * Connect will come in with just a single address.
1039 static int __sctp_connect(struct sock *sk,
1040 struct sockaddr *kaddrs,
1041 int addrs_size,
1042 sctp_assoc_t *assoc_id)
1044 struct net *net = sock_net(sk);
1045 struct sctp_sock *sp;
1046 struct sctp_endpoint *ep;
1047 struct sctp_association *asoc = NULL;
1048 struct sctp_association *asoc2;
1049 struct sctp_transport *transport;
1050 union sctp_addr to;
1051 sctp_scope_t scope;
1052 long timeo;
1053 int err = 0;
1054 int addrcnt = 0;
1055 int walk_size = 0;
1056 union sctp_addr *sa_addr = NULL;
1057 void *addr_buf;
1058 unsigned short port;
1059 unsigned int f_flags = 0;
1061 sp = sctp_sk(sk);
1062 ep = sp->ep;
1064 /* connect() cannot be done on a socket that is already in ESTABLISHED
1065 * state - UDP-style peeled off socket or a TCP-style socket that
1066 * is already connected.
1067 * It cannot be done even on a TCP-style listening socket.
1069 if (sctp_sstate(sk, ESTABLISHED) ||
1070 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1071 err = -EISCONN;
1072 goto out_free;
1075 /* Walk through the addrs buffer and count the number of addresses. */
1076 addr_buf = kaddrs;
1077 while (walk_size < addrs_size) {
1078 struct sctp_af *af;
1080 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1081 err = -EINVAL;
1082 goto out_free;
1085 sa_addr = addr_buf;
1086 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1088 /* If the address family is not supported or if this address
1089 * causes the address buffer to overflow return EINVAL.
1091 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1092 err = -EINVAL;
1093 goto out_free;
1096 port = ntohs(sa_addr->v4.sin_port);
1098 /* Save current address so we can work with it */
1099 memcpy(&to, sa_addr, af->sockaddr_len);
1101 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1102 if (err)
1103 goto out_free;
1105 /* Make sure the destination port is correctly set
1106 * in all addresses.
1108 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1109 err = -EINVAL;
1110 goto out_free;
1113 /* Check if there already is a matching association on the
1114 * endpoint (other than the one created here).
1116 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1117 if (asoc2 && asoc2 != asoc) {
1118 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1119 err = -EISCONN;
1120 else
1121 err = -EALREADY;
1122 goto out_free;
1125 /* If we could not find a matching association on the endpoint,
1126 * make sure that there is no peeled-off association matching
1127 * the peer address even on another socket.
1129 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1130 err = -EADDRNOTAVAIL;
1131 goto out_free;
1134 if (!asoc) {
1135 /* If a bind() or sctp_bindx() is not called prior to
1136 * an sctp_connectx() call, the system picks an
1137 * ephemeral port and will choose an address set
1138 * equivalent to binding with a wildcard address.
1140 if (!ep->base.bind_addr.port) {
1141 if (sctp_autobind(sk)) {
1142 err = -EAGAIN;
1143 goto out_free;
1145 } else {
1147 * If an unprivileged user inherits a 1-many
1148 * style socket with open associations on a
1149 * privileged port, it MAY be permitted to
1150 * accept new associations, but it SHOULD NOT
1151 * be permitted to open new associations.
1153 if (ep->base.bind_addr.port < PROT_SOCK &&
1154 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1155 err = -EACCES;
1156 goto out_free;
1160 scope = sctp_scope(&to);
1161 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1162 if (!asoc) {
1163 err = -ENOMEM;
1164 goto out_free;
1167 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1168 GFP_KERNEL);
1169 if (err < 0) {
1170 goto out_free;
1175 /* Prime the peer's transport structures. */
1176 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1177 SCTP_UNKNOWN);
1178 if (!transport) {
1179 err = -ENOMEM;
1180 goto out_free;
1183 addrcnt++;
1184 addr_buf += af->sockaddr_len;
1185 walk_size += af->sockaddr_len;
1188 /* In case the user of sctp_connectx() wants an association
1189 * id back, assign one now.
1191 if (assoc_id) {
1192 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1193 if (err < 0)
1194 goto out_free;
1197 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1198 if (err < 0) {
1199 goto out_free;
1202 /* Initialize sk's dport and daddr for getpeername() */
1203 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1204 sp->pf->to_sk_daddr(sa_addr, sk);
1205 sk->sk_err = 0;
1207 /* in-kernel sockets don't generally have a file allocated to them
1208 * if all they do is call sock_create_kern().
1210 if (sk->sk_socket->file)
1211 f_flags = sk->sk_socket->file->f_flags;
1213 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1215 err = sctp_wait_for_connect(asoc, &timeo);
1216 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1217 *assoc_id = asoc->assoc_id;
1219 /* Don't free association on exit. */
1220 asoc = NULL;
1222 out_free:
1223 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1224 __func__, asoc, kaddrs, err);
1226 if (asoc) {
1227 /* sctp_primitive_ASSOCIATE may have added this association
1228 * To the hash table, try to unhash it, just in case, its a noop
1229 * if it wasn't hashed so we're safe
1231 sctp_unhash_established(asoc);
1232 sctp_association_free(asoc);
1234 return err;
1237 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1239 * API 8.9
1240 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1241 * sctp_assoc_t *asoc);
1243 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1244 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1245 * or IPv6 addresses.
1247 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1248 * Section 3.1.2 for this usage.
1250 * addrs is a pointer to an array of one or more socket addresses. Each
1251 * address is contained in its appropriate structure (i.e. struct
1252 * sockaddr_in or struct sockaddr_in6) the family of the address type
1253 * must be used to distengish the address length (note that this
1254 * representation is termed a "packed array" of addresses). The caller
1255 * specifies the number of addresses in the array with addrcnt.
1257 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1258 * the association id of the new association. On failure, sctp_connectx()
1259 * returns -1, and sets errno to the appropriate error code. The assoc_id
1260 * is not touched by the kernel.
1262 * For SCTP, the port given in each socket address must be the same, or
1263 * sctp_connectx() will fail, setting errno to EINVAL.
1265 * An application can use sctp_connectx to initiate an association with
1266 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1267 * allows a caller to specify multiple addresses at which a peer can be
1268 * reached. The way the SCTP stack uses the list of addresses to set up
1269 * the association is implementation dependent. This function only
1270 * specifies that the stack will try to make use of all the addresses in
1271 * the list when needed.
1273 * Note that the list of addresses passed in is only used for setting up
1274 * the association. It does not necessarily equal the set of addresses
1275 * the peer uses for the resulting association. If the caller wants to
1276 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1277 * retrieve them after the association has been set up.
1279 * Basically do nothing but copying the addresses from user to kernel
1280 * land and invoking either sctp_connectx(). This is used for tunneling
1281 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1283 * We don't use copy_from_user() for optimization: we first do the
1284 * sanity checks (buffer size -fast- and access check-healthy
1285 * pointer); if all of those succeed, then we can alloc the memory
1286 * (expensive operation) needed to copy the data to kernel. Then we do
1287 * the copying without checking the user space area
1288 * (__copy_from_user()).
1290 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1291 * it.
1293 * sk The sk of the socket
1294 * addrs The pointer to the addresses in user land
1295 * addrssize Size of the addrs buffer
1297 * Returns >=0 if ok, <0 errno code on error.
1299 static int __sctp_setsockopt_connectx(struct sock *sk,
1300 struct sockaddr __user *addrs,
1301 int addrs_size,
1302 sctp_assoc_t *assoc_id)
1304 int err = 0;
1305 struct sockaddr *kaddrs;
1307 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1308 __func__, sk, addrs, addrs_size);
1310 if (unlikely(addrs_size <= 0))
1311 return -EINVAL;
1313 /* Check the user passed a healthy pointer. */
1314 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1315 return -EFAULT;
1317 /* Alloc space for the address array in kernel memory. */
1318 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1319 if (unlikely(!kaddrs))
1320 return -ENOMEM;
1322 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1323 err = -EFAULT;
1324 } else {
1325 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1328 kfree(kaddrs);
1330 return err;
1334 * This is an older interface. It's kept for backward compatibility
1335 * to the option that doesn't provide association id.
1337 static int sctp_setsockopt_connectx_old(struct sock *sk,
1338 struct sockaddr __user *addrs,
1339 int addrs_size)
1341 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1345 * New interface for the API. The since the API is done with a socket
1346 * option, to make it simple we feed back the association id is as a return
1347 * indication to the call. Error is always negative and association id is
1348 * always positive.
1350 static int sctp_setsockopt_connectx(struct sock *sk,
1351 struct sockaddr __user *addrs,
1352 int addrs_size)
1354 sctp_assoc_t assoc_id = 0;
1355 int err = 0;
1357 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1359 if (err)
1360 return err;
1361 else
1362 return assoc_id;
1366 * New (hopefully final) interface for the API.
1367 * We use the sctp_getaddrs_old structure so that use-space library
1368 * can avoid any unnecessary allocations. The only different part
1369 * is that we store the actual length of the address buffer into the
1370 * addrs_num structure member. That way we can re-use the existing
1371 * code.
1373 #ifdef CONFIG_COMPAT
1374 struct compat_sctp_getaddrs_old {
1375 sctp_assoc_t assoc_id;
1376 s32 addr_num;
1377 compat_uptr_t addrs; /* struct sockaddr * */
1379 #endif
1381 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1382 char __user *optval,
1383 int __user *optlen)
1385 struct sctp_getaddrs_old param;
1386 sctp_assoc_t assoc_id = 0;
1387 int err = 0;
1389 #ifdef CONFIG_COMPAT
1390 if (is_compat_task()) {
1391 struct compat_sctp_getaddrs_old param32;
1393 if (len < sizeof(param32))
1394 return -EINVAL;
1395 if (copy_from_user(&param32, optval, sizeof(param32)))
1396 return -EFAULT;
1398 param.assoc_id = param32.assoc_id;
1399 param.addr_num = param32.addr_num;
1400 param.addrs = compat_ptr(param32.addrs);
1401 } else
1402 #endif
1404 if (len < sizeof(param))
1405 return -EINVAL;
1406 if (copy_from_user(&param, optval, sizeof(param)))
1407 return -EFAULT;
1410 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1411 param.addrs, param.addr_num,
1412 &assoc_id);
1413 if (err == 0 || err == -EINPROGRESS) {
1414 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1415 return -EFAULT;
1416 if (put_user(sizeof(assoc_id), optlen))
1417 return -EFAULT;
1420 return err;
1423 /* API 3.1.4 close() - UDP Style Syntax
1424 * Applications use close() to perform graceful shutdown (as described in
1425 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1426 * by a UDP-style socket.
1428 * The syntax is
1430 * ret = close(int sd);
1432 * sd - the socket descriptor of the associations to be closed.
1434 * To gracefully shutdown a specific association represented by the
1435 * UDP-style socket, an application should use the sendmsg() call,
1436 * passing no user data, but including the appropriate flag in the
1437 * ancillary data (see Section xxxx).
1439 * If sd in the close() call is a branched-off socket representing only
1440 * one association, the shutdown is performed on that association only.
1442 * 4.1.6 close() - TCP Style Syntax
1444 * Applications use close() to gracefully close down an association.
1446 * The syntax is:
1448 * int close(int sd);
1450 * sd - the socket descriptor of the association to be closed.
1452 * After an application calls close() on a socket descriptor, no further
1453 * socket operations will succeed on that descriptor.
1455 * API 7.1.4 SO_LINGER
1457 * An application using the TCP-style socket can use this option to
1458 * perform the SCTP ABORT primitive. The linger option structure is:
1460 * struct linger {
1461 * int l_onoff; // option on/off
1462 * int l_linger; // linger time
1463 * };
1465 * To enable the option, set l_onoff to 1. If the l_linger value is set
1466 * to 0, calling close() is the same as the ABORT primitive. If the
1467 * value is set to a negative value, the setsockopt() call will return
1468 * an error. If the value is set to a positive value linger_time, the
1469 * close() can be blocked for at most linger_time ms. If the graceful
1470 * shutdown phase does not finish during this period, close() will
1471 * return but the graceful shutdown phase continues in the system.
1473 static void sctp_close(struct sock *sk, long timeout)
1475 struct net *net = sock_net(sk);
1476 struct sctp_endpoint *ep;
1477 struct sctp_association *asoc;
1478 struct list_head *pos, *temp;
1479 unsigned int data_was_unread;
1481 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1483 lock_sock(sk);
1484 sk->sk_shutdown = SHUTDOWN_MASK;
1485 sk->sk_state = SCTP_SS_CLOSING;
1487 ep = sctp_sk(sk)->ep;
1489 /* Clean up any skbs sitting on the receive queue. */
1490 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1491 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1493 /* Walk all associations on an endpoint. */
1494 list_for_each_safe(pos, temp, &ep->asocs) {
1495 asoc = list_entry(pos, struct sctp_association, asocs);
1497 if (sctp_style(sk, TCP)) {
1498 /* A closed association can still be in the list if
1499 * it belongs to a TCP-style listening socket that is
1500 * not yet accepted. If so, free it. If not, send an
1501 * ABORT or SHUTDOWN based on the linger options.
1503 if (sctp_state(asoc, CLOSED)) {
1504 sctp_unhash_established(asoc);
1505 sctp_association_free(asoc);
1506 continue;
1510 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1511 !skb_queue_empty(&asoc->ulpq.reasm) ||
1512 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1513 struct sctp_chunk *chunk;
1515 chunk = sctp_make_abort_user(asoc, NULL, 0);
1516 if (chunk)
1517 sctp_primitive_ABORT(net, asoc, chunk);
1518 } else
1519 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1522 /* On a TCP-style socket, block for at most linger_time if set. */
1523 if (sctp_style(sk, TCP) && timeout)
1524 sctp_wait_for_close(sk, timeout);
1526 /* This will run the backlog queue. */
1527 release_sock(sk);
1529 /* Supposedly, no process has access to the socket, but
1530 * the net layers still may.
1531 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1532 * held and that should be grabbed before socket lock.
1534 spin_lock_bh(&net->sctp.addr_wq_lock);
1535 bh_lock_sock(sk);
1537 /* Hold the sock, since sk_common_release() will put sock_put()
1538 * and we have just a little more cleanup.
1540 sock_hold(sk);
1541 sk_common_release(sk);
1543 bh_unlock_sock(sk);
1544 spin_unlock_bh(&net->sctp.addr_wq_lock);
1546 sock_put(sk);
1548 SCTP_DBG_OBJCNT_DEC(sock);
1551 /* Handle EPIPE error. */
1552 static int sctp_error(struct sock *sk, int flags, int err)
1554 if (err == -EPIPE)
1555 err = sock_error(sk) ? : -EPIPE;
1556 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1557 send_sig(SIGPIPE, current, 0);
1558 return err;
1561 /* API 3.1.3 sendmsg() - UDP Style Syntax
1563 * An application uses sendmsg() and recvmsg() calls to transmit data to
1564 * and receive data from its peer.
1566 * ssize_t sendmsg(int socket, const struct msghdr *message,
1567 * int flags);
1569 * socket - the socket descriptor of the endpoint.
1570 * message - pointer to the msghdr structure which contains a single
1571 * user message and possibly some ancillary data.
1573 * See Section 5 for complete description of the data
1574 * structures.
1576 * flags - flags sent or received with the user message, see Section
1577 * 5 for complete description of the flags.
1579 * Note: This function could use a rewrite especially when explicit
1580 * connect support comes in.
1582 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1584 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1586 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1588 struct net *net = sock_net(sk);
1589 struct sctp_sock *sp;
1590 struct sctp_endpoint *ep;
1591 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1592 struct sctp_transport *transport, *chunk_tp;
1593 struct sctp_chunk *chunk;
1594 union sctp_addr to;
1595 struct sockaddr *msg_name = NULL;
1596 struct sctp_sndrcvinfo default_sinfo;
1597 struct sctp_sndrcvinfo *sinfo;
1598 struct sctp_initmsg *sinit;
1599 sctp_assoc_t associd = 0;
1600 sctp_cmsgs_t cmsgs = { NULL };
1601 sctp_scope_t scope;
1602 bool fill_sinfo_ttl = false, wait_connect = false;
1603 struct sctp_datamsg *datamsg;
1604 int msg_flags = msg->msg_flags;
1605 __u16 sinfo_flags = 0;
1606 long timeo;
1607 int err;
1609 err = 0;
1610 sp = sctp_sk(sk);
1611 ep = sp->ep;
1613 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1614 msg, msg_len, ep);
1616 /* We cannot send a message over a TCP-style listening socket. */
1617 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1618 err = -EPIPE;
1619 goto out_nounlock;
1622 /* Parse out the SCTP CMSGs. */
1623 err = sctp_msghdr_parse(msg, &cmsgs);
1624 if (err) {
1625 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1626 goto out_nounlock;
1629 /* Fetch the destination address for this packet. This
1630 * address only selects the association--it is not necessarily
1631 * the address we will send to.
1632 * For a peeled-off socket, msg_name is ignored.
1634 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1635 int msg_namelen = msg->msg_namelen;
1637 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1638 msg_namelen);
1639 if (err)
1640 return err;
1642 if (msg_namelen > sizeof(to))
1643 msg_namelen = sizeof(to);
1644 memcpy(&to, msg->msg_name, msg_namelen);
1645 msg_name = msg->msg_name;
1648 sinit = cmsgs.init;
1649 if (cmsgs.sinfo != NULL) {
1650 memset(&default_sinfo, 0, sizeof(default_sinfo));
1651 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1652 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1653 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1654 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1655 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1657 sinfo = &default_sinfo;
1658 fill_sinfo_ttl = true;
1659 } else {
1660 sinfo = cmsgs.srinfo;
1662 /* Did the user specify SNDINFO/SNDRCVINFO? */
1663 if (sinfo) {
1664 sinfo_flags = sinfo->sinfo_flags;
1665 associd = sinfo->sinfo_assoc_id;
1668 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1669 msg_len, sinfo_flags);
1671 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1672 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1673 err = -EINVAL;
1674 goto out_nounlock;
1677 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1678 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1679 * If SCTP_ABORT is set, the message length could be non zero with
1680 * the msg_iov set to the user abort reason.
1682 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1683 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1684 err = -EINVAL;
1685 goto out_nounlock;
1688 /* If SCTP_ADDR_OVER is set, there must be an address
1689 * specified in msg_name.
1691 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1692 err = -EINVAL;
1693 goto out_nounlock;
1696 transport = NULL;
1698 pr_debug("%s: about to look up association\n", __func__);
1700 lock_sock(sk);
1702 /* If a msg_name has been specified, assume this is to be used. */
1703 if (msg_name) {
1704 /* Look for a matching association on the endpoint. */
1705 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1706 if (!asoc) {
1707 /* If we could not find a matching association on the
1708 * endpoint, make sure that it is not a TCP-style
1709 * socket that already has an association or there is
1710 * no peeled-off association on another socket.
1712 if ((sctp_style(sk, TCP) &&
1713 sctp_sstate(sk, ESTABLISHED)) ||
1714 sctp_endpoint_is_peeled_off(ep, &to)) {
1715 err = -EADDRNOTAVAIL;
1716 goto out_unlock;
1719 } else {
1720 asoc = sctp_id2assoc(sk, associd);
1721 if (!asoc) {
1722 err = -EPIPE;
1723 goto out_unlock;
1727 if (asoc) {
1728 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1730 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1731 * socket that has an association in CLOSED state. This can
1732 * happen when an accepted socket has an association that is
1733 * already CLOSED.
1735 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1736 err = -EPIPE;
1737 goto out_unlock;
1740 if (sinfo_flags & SCTP_EOF) {
1741 pr_debug("%s: shutting down association:%p\n",
1742 __func__, asoc);
1744 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1745 err = 0;
1746 goto out_unlock;
1748 if (sinfo_flags & SCTP_ABORT) {
1750 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1751 if (!chunk) {
1752 err = -ENOMEM;
1753 goto out_unlock;
1756 pr_debug("%s: aborting association:%p\n",
1757 __func__, asoc);
1759 sctp_primitive_ABORT(net, asoc, chunk);
1760 err = 0;
1761 goto out_unlock;
1765 /* Do we need to create the association? */
1766 if (!asoc) {
1767 pr_debug("%s: there is no association yet\n", __func__);
1769 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1770 err = -EINVAL;
1771 goto out_unlock;
1774 /* Check for invalid stream against the stream counts,
1775 * either the default or the user specified stream counts.
1777 if (sinfo) {
1778 if (!sinit || !sinit->sinit_num_ostreams) {
1779 /* Check against the defaults. */
1780 if (sinfo->sinfo_stream >=
1781 sp->initmsg.sinit_num_ostreams) {
1782 err = -EINVAL;
1783 goto out_unlock;
1785 } else {
1786 /* Check against the requested. */
1787 if (sinfo->sinfo_stream >=
1788 sinit->sinit_num_ostreams) {
1789 err = -EINVAL;
1790 goto out_unlock;
1796 * API 3.1.2 bind() - UDP Style Syntax
1797 * If a bind() or sctp_bindx() is not called prior to a
1798 * sendmsg() call that initiates a new association, the
1799 * system picks an ephemeral port and will choose an address
1800 * set equivalent to binding with a wildcard address.
1802 if (!ep->base.bind_addr.port) {
1803 if (sctp_autobind(sk)) {
1804 err = -EAGAIN;
1805 goto out_unlock;
1807 } else {
1809 * If an unprivileged user inherits a one-to-many
1810 * style socket with open associations on a privileged
1811 * port, it MAY be permitted to accept new associations,
1812 * but it SHOULD NOT be permitted to open new
1813 * associations.
1815 if (ep->base.bind_addr.port < PROT_SOCK &&
1816 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1817 err = -EACCES;
1818 goto out_unlock;
1822 scope = sctp_scope(&to);
1823 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1824 if (!new_asoc) {
1825 err = -ENOMEM;
1826 goto out_unlock;
1828 asoc = new_asoc;
1829 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1830 if (err < 0) {
1831 err = -ENOMEM;
1832 goto out_free;
1835 /* If the SCTP_INIT ancillary data is specified, set all
1836 * the association init values accordingly.
1838 if (sinit) {
1839 if (sinit->sinit_num_ostreams) {
1840 asoc->c.sinit_num_ostreams =
1841 sinit->sinit_num_ostreams;
1843 if (sinit->sinit_max_instreams) {
1844 asoc->c.sinit_max_instreams =
1845 sinit->sinit_max_instreams;
1847 if (sinit->sinit_max_attempts) {
1848 asoc->max_init_attempts
1849 = sinit->sinit_max_attempts;
1851 if (sinit->sinit_max_init_timeo) {
1852 asoc->max_init_timeo =
1853 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1857 /* Prime the peer's transport structures. */
1858 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1859 if (!transport) {
1860 err = -ENOMEM;
1861 goto out_free;
1865 /* ASSERT: we have a valid association at this point. */
1866 pr_debug("%s: we have a valid association\n", __func__);
1868 if (!sinfo) {
1869 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1870 * one with some defaults.
1872 memset(&default_sinfo, 0, sizeof(default_sinfo));
1873 default_sinfo.sinfo_stream = asoc->default_stream;
1874 default_sinfo.sinfo_flags = asoc->default_flags;
1875 default_sinfo.sinfo_ppid = asoc->default_ppid;
1876 default_sinfo.sinfo_context = asoc->default_context;
1877 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1878 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1880 sinfo = &default_sinfo;
1881 } else if (fill_sinfo_ttl) {
1882 /* In case SNDINFO was specified, we still need to fill
1883 * it with a default ttl from the assoc here.
1885 sinfo->sinfo_timetolive = asoc->default_timetolive;
1888 /* API 7.1.7, the sndbuf size per association bounds the
1889 * maximum size of data that can be sent in a single send call.
1891 if (msg_len > sk->sk_sndbuf) {
1892 err = -EMSGSIZE;
1893 goto out_free;
1896 if (asoc->pmtu_pending)
1897 sctp_assoc_pending_pmtu(sk, asoc);
1899 /* If fragmentation is disabled and the message length exceeds the
1900 * association fragmentation point, return EMSGSIZE. The I-D
1901 * does not specify what this error is, but this looks like
1902 * a great fit.
1904 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1905 err = -EMSGSIZE;
1906 goto out_free;
1909 /* Check for invalid stream. */
1910 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1911 err = -EINVAL;
1912 goto out_free;
1915 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1916 if (!sctp_wspace(asoc)) {
1917 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1918 if (err)
1919 goto out_free;
1922 /* If an address is passed with the sendto/sendmsg call, it is used
1923 * to override the primary destination address in the TCP model, or
1924 * when SCTP_ADDR_OVER flag is set in the UDP model.
1926 if ((sctp_style(sk, TCP) && msg_name) ||
1927 (sinfo_flags & SCTP_ADDR_OVER)) {
1928 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1929 if (!chunk_tp) {
1930 err = -EINVAL;
1931 goto out_free;
1933 } else
1934 chunk_tp = NULL;
1936 /* Auto-connect, if we aren't connected already. */
1937 if (sctp_state(asoc, CLOSED)) {
1938 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1939 if (err < 0)
1940 goto out_free;
1942 wait_connect = true;
1943 pr_debug("%s: we associated primitively\n", __func__);
1946 /* Break the message into multiple chunks of maximum size. */
1947 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1948 if (IS_ERR(datamsg)) {
1949 err = PTR_ERR(datamsg);
1950 goto out_free;
1953 /* Now send the (possibly) fragmented message. */
1954 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1955 sctp_chunk_hold(chunk);
1957 /* Do accounting for the write space. */
1958 sctp_set_owner_w(chunk);
1960 chunk->transport = chunk_tp;
1963 /* Send it to the lower layers. Note: all chunks
1964 * must either fail or succeed. The lower layer
1965 * works that way today. Keep it that way or this
1966 * breaks.
1968 err = sctp_primitive_SEND(net, asoc, datamsg);
1969 /* Did the lower layer accept the chunk? */
1970 if (err) {
1971 sctp_datamsg_free(datamsg);
1972 goto out_free;
1975 pr_debug("%s: we sent primitively\n", __func__);
1977 sctp_datamsg_put(datamsg);
1978 err = msg_len;
1980 if (unlikely(wait_connect)) {
1981 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
1982 sctp_wait_for_connect(asoc, &timeo);
1985 /* If we are already past ASSOCIATE, the lower
1986 * layers are responsible for association cleanup.
1988 goto out_unlock;
1990 out_free:
1991 if (new_asoc) {
1992 sctp_unhash_established(asoc);
1993 sctp_association_free(asoc);
1995 out_unlock:
1996 release_sock(sk);
1998 out_nounlock:
1999 return sctp_error(sk, msg_flags, err);
2001 #if 0
2002 do_sock_err:
2003 if (msg_len)
2004 err = msg_len;
2005 else
2006 err = sock_error(sk);
2007 goto out;
2009 do_interrupted:
2010 if (msg_len)
2011 err = msg_len;
2012 goto out;
2013 #endif /* 0 */
2016 /* This is an extended version of skb_pull() that removes the data from the
2017 * start of a skb even when data is spread across the list of skb's in the
2018 * frag_list. len specifies the total amount of data that needs to be removed.
2019 * when 'len' bytes could be removed from the skb, it returns 0.
2020 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2021 * could not be removed.
2023 static int sctp_skb_pull(struct sk_buff *skb, int len)
2025 struct sk_buff *list;
2026 int skb_len = skb_headlen(skb);
2027 int rlen;
2029 if (len <= skb_len) {
2030 __skb_pull(skb, len);
2031 return 0;
2033 len -= skb_len;
2034 __skb_pull(skb, skb_len);
2036 skb_walk_frags(skb, list) {
2037 rlen = sctp_skb_pull(list, len);
2038 skb->len -= (len-rlen);
2039 skb->data_len -= (len-rlen);
2041 if (!rlen)
2042 return 0;
2044 len = rlen;
2047 return len;
2050 /* API 3.1.3 recvmsg() - UDP Style Syntax
2052 * ssize_t recvmsg(int socket, struct msghdr *message,
2053 * int flags);
2055 * socket - the socket descriptor of the endpoint.
2056 * message - pointer to the msghdr structure which contains a single
2057 * user message and possibly some ancillary data.
2059 * See Section 5 for complete description of the data
2060 * structures.
2062 * flags - flags sent or received with the user message, see Section
2063 * 5 for complete description of the flags.
2065 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2066 int noblock, int flags, int *addr_len)
2068 struct sctp_ulpevent *event = NULL;
2069 struct sctp_sock *sp = sctp_sk(sk);
2070 struct sk_buff *skb;
2071 int copied;
2072 int err = 0;
2073 int skb_len;
2075 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2076 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2077 addr_len);
2079 lock_sock(sk);
2081 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2082 err = -ENOTCONN;
2083 goto out;
2086 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2087 if (!skb)
2088 goto out;
2090 /* Get the total length of the skb including any skb's in the
2091 * frag_list.
2093 skb_len = skb->len;
2095 copied = skb_len;
2096 if (copied > len)
2097 copied = len;
2099 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2101 event = sctp_skb2event(skb);
2103 if (err)
2104 goto out_free;
2106 sock_recv_ts_and_drops(msg, sk, skb);
2107 if (sctp_ulpevent_is_notification(event)) {
2108 msg->msg_flags |= MSG_NOTIFICATION;
2109 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2110 } else {
2111 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2114 /* Check if we allow SCTP_NXTINFO. */
2115 if (sp->recvnxtinfo)
2116 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2117 /* Check if we allow SCTP_RCVINFO. */
2118 if (sp->recvrcvinfo)
2119 sctp_ulpevent_read_rcvinfo(event, msg);
2120 /* Check if we allow SCTP_SNDRCVINFO. */
2121 if (sp->subscribe.sctp_data_io_event)
2122 sctp_ulpevent_read_sndrcvinfo(event, msg);
2124 err = copied;
2126 /* If skb's length exceeds the user's buffer, update the skb and
2127 * push it back to the receive_queue so that the next call to
2128 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2130 if (skb_len > copied) {
2131 msg->msg_flags &= ~MSG_EOR;
2132 if (flags & MSG_PEEK)
2133 goto out_free;
2134 sctp_skb_pull(skb, copied);
2135 skb_queue_head(&sk->sk_receive_queue, skb);
2137 /* When only partial message is copied to the user, increase
2138 * rwnd by that amount. If all the data in the skb is read,
2139 * rwnd is updated when the event is freed.
2141 if (!sctp_ulpevent_is_notification(event))
2142 sctp_assoc_rwnd_increase(event->asoc, copied);
2143 goto out;
2144 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2145 (event->msg_flags & MSG_EOR))
2146 msg->msg_flags |= MSG_EOR;
2147 else
2148 msg->msg_flags &= ~MSG_EOR;
2150 out_free:
2151 if (flags & MSG_PEEK) {
2152 /* Release the skb reference acquired after peeking the skb in
2153 * sctp_skb_recv_datagram().
2155 kfree_skb(skb);
2156 } else {
2157 /* Free the event which includes releasing the reference to
2158 * the owner of the skb, freeing the skb and updating the
2159 * rwnd.
2161 sctp_ulpevent_free(event);
2163 out:
2164 release_sock(sk);
2165 return err;
2168 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2170 * This option is a on/off flag. If enabled no SCTP message
2171 * fragmentation will be performed. Instead if a message being sent
2172 * exceeds the current PMTU size, the message will NOT be sent and
2173 * instead a error will be indicated to the user.
2175 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2176 char __user *optval,
2177 unsigned int optlen)
2179 int val;
2181 if (optlen < sizeof(int))
2182 return -EINVAL;
2184 if (get_user(val, (int __user *)optval))
2185 return -EFAULT;
2187 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2189 return 0;
2192 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2193 unsigned int optlen)
2195 struct sctp_association *asoc;
2196 struct sctp_ulpevent *event;
2198 if (optlen > sizeof(struct sctp_event_subscribe))
2199 return -EINVAL;
2200 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2201 return -EFAULT;
2203 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2204 * if there is no data to be sent or retransmit, the stack will
2205 * immediately send up this notification.
2207 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2208 &sctp_sk(sk)->subscribe)) {
2209 asoc = sctp_id2assoc(sk, 0);
2211 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2212 event = sctp_ulpevent_make_sender_dry_event(asoc,
2213 GFP_ATOMIC);
2214 if (!event)
2215 return -ENOMEM;
2217 sctp_ulpq_tail_event(&asoc->ulpq, event);
2221 return 0;
2224 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2226 * This socket option is applicable to the UDP-style socket only. When
2227 * set it will cause associations that are idle for more than the
2228 * specified number of seconds to automatically close. An association
2229 * being idle is defined an association that has NOT sent or received
2230 * user data. The special value of '0' indicates that no automatic
2231 * close of any associations should be performed. The option expects an
2232 * integer defining the number of seconds of idle time before an
2233 * association is closed.
2235 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2236 unsigned int optlen)
2238 struct sctp_sock *sp = sctp_sk(sk);
2239 struct net *net = sock_net(sk);
2241 /* Applicable to UDP-style socket only */
2242 if (sctp_style(sk, TCP))
2243 return -EOPNOTSUPP;
2244 if (optlen != sizeof(int))
2245 return -EINVAL;
2246 if (copy_from_user(&sp->autoclose, optval, optlen))
2247 return -EFAULT;
2249 if (sp->autoclose > net->sctp.max_autoclose)
2250 sp->autoclose = net->sctp.max_autoclose;
2252 return 0;
2255 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2257 * Applications can enable or disable heartbeats for any peer address of
2258 * an association, modify an address's heartbeat interval, force a
2259 * heartbeat to be sent immediately, and adjust the address's maximum
2260 * number of retransmissions sent before an address is considered
2261 * unreachable. The following structure is used to access and modify an
2262 * address's parameters:
2264 * struct sctp_paddrparams {
2265 * sctp_assoc_t spp_assoc_id;
2266 * struct sockaddr_storage spp_address;
2267 * uint32_t spp_hbinterval;
2268 * uint16_t spp_pathmaxrxt;
2269 * uint32_t spp_pathmtu;
2270 * uint32_t spp_sackdelay;
2271 * uint32_t spp_flags;
2272 * };
2274 * spp_assoc_id - (one-to-many style socket) This is filled in the
2275 * application, and identifies the association for
2276 * this query.
2277 * spp_address - This specifies which address is of interest.
2278 * spp_hbinterval - This contains the value of the heartbeat interval,
2279 * in milliseconds. If a value of zero
2280 * is present in this field then no changes are to
2281 * be made to this parameter.
2282 * spp_pathmaxrxt - This contains the maximum number of
2283 * retransmissions before this address shall be
2284 * considered unreachable. If a value of zero
2285 * is present in this field then no changes are to
2286 * be made to this parameter.
2287 * spp_pathmtu - When Path MTU discovery is disabled the value
2288 * specified here will be the "fixed" path mtu.
2289 * Note that if the spp_address field is empty
2290 * then all associations on this address will
2291 * have this fixed path mtu set upon them.
2293 * spp_sackdelay - When delayed sack is enabled, this value specifies
2294 * the number of milliseconds that sacks will be delayed
2295 * for. This value will apply to all addresses of an
2296 * association if the spp_address field is empty. Note
2297 * also, that if delayed sack is enabled and this
2298 * value is set to 0, no change is made to the last
2299 * recorded delayed sack timer value.
2301 * spp_flags - These flags are used to control various features
2302 * on an association. The flag field may contain
2303 * zero or more of the following options.
2305 * SPP_HB_ENABLE - Enable heartbeats on the
2306 * specified address. Note that if the address
2307 * field is empty all addresses for the association
2308 * have heartbeats enabled upon them.
2310 * SPP_HB_DISABLE - Disable heartbeats on the
2311 * speicifed address. Note that if the address
2312 * field is empty all addresses for the association
2313 * will have their heartbeats disabled. Note also
2314 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2315 * mutually exclusive, only one of these two should
2316 * be specified. Enabling both fields will have
2317 * undetermined results.
2319 * SPP_HB_DEMAND - Request a user initiated heartbeat
2320 * to be made immediately.
2322 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2323 * heartbeat delayis to be set to the value of 0
2324 * milliseconds.
2326 * SPP_PMTUD_ENABLE - This field will enable PMTU
2327 * discovery upon the specified address. Note that
2328 * if the address feild is empty then all addresses
2329 * on the association are effected.
2331 * SPP_PMTUD_DISABLE - This field will disable PMTU
2332 * discovery upon the specified address. Note that
2333 * if the address feild is empty then all addresses
2334 * on the association are effected. Not also that
2335 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2336 * exclusive. Enabling both will have undetermined
2337 * results.
2339 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2340 * on delayed sack. The time specified in spp_sackdelay
2341 * is used to specify the sack delay for this address. Note
2342 * that if spp_address is empty then all addresses will
2343 * enable delayed sack and take on the sack delay
2344 * value specified in spp_sackdelay.
2345 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2346 * off delayed sack. If the spp_address field is blank then
2347 * delayed sack is disabled for the entire association. Note
2348 * also that this field is mutually exclusive to
2349 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2350 * results.
2352 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2353 struct sctp_transport *trans,
2354 struct sctp_association *asoc,
2355 struct sctp_sock *sp,
2356 int hb_change,
2357 int pmtud_change,
2358 int sackdelay_change)
2360 int error;
2362 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2363 struct net *net = sock_net(trans->asoc->base.sk);
2365 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2366 if (error)
2367 return error;
2370 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2371 * this field is ignored. Note also that a value of zero indicates
2372 * the current setting should be left unchanged.
2374 if (params->spp_flags & SPP_HB_ENABLE) {
2376 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2377 * set. This lets us use 0 value when this flag
2378 * is set.
2380 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2381 params->spp_hbinterval = 0;
2383 if (params->spp_hbinterval ||
2384 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2385 if (trans) {
2386 trans->hbinterval =
2387 msecs_to_jiffies(params->spp_hbinterval);
2388 } else if (asoc) {
2389 asoc->hbinterval =
2390 msecs_to_jiffies(params->spp_hbinterval);
2391 } else {
2392 sp->hbinterval = params->spp_hbinterval;
2397 if (hb_change) {
2398 if (trans) {
2399 trans->param_flags =
2400 (trans->param_flags & ~SPP_HB) | hb_change;
2401 } else if (asoc) {
2402 asoc->param_flags =
2403 (asoc->param_flags & ~SPP_HB) | hb_change;
2404 } else {
2405 sp->param_flags =
2406 (sp->param_flags & ~SPP_HB) | hb_change;
2410 /* When Path MTU discovery is disabled the value specified here will
2411 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2412 * include the flag SPP_PMTUD_DISABLE for this field to have any
2413 * effect).
2415 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2416 if (trans) {
2417 trans->pathmtu = params->spp_pathmtu;
2418 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2419 } else if (asoc) {
2420 asoc->pathmtu = params->spp_pathmtu;
2421 sctp_frag_point(asoc, params->spp_pathmtu);
2422 } else {
2423 sp->pathmtu = params->spp_pathmtu;
2427 if (pmtud_change) {
2428 if (trans) {
2429 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2430 (params->spp_flags & SPP_PMTUD_ENABLE);
2431 trans->param_flags =
2432 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2433 if (update) {
2434 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2435 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2437 } else if (asoc) {
2438 asoc->param_flags =
2439 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2440 } else {
2441 sp->param_flags =
2442 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2446 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2447 * value of this field is ignored. Note also that a value of zero
2448 * indicates the current setting should be left unchanged.
2450 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2451 if (trans) {
2452 trans->sackdelay =
2453 msecs_to_jiffies(params->spp_sackdelay);
2454 } else if (asoc) {
2455 asoc->sackdelay =
2456 msecs_to_jiffies(params->spp_sackdelay);
2457 } else {
2458 sp->sackdelay = params->spp_sackdelay;
2462 if (sackdelay_change) {
2463 if (trans) {
2464 trans->param_flags =
2465 (trans->param_flags & ~SPP_SACKDELAY) |
2466 sackdelay_change;
2467 } else if (asoc) {
2468 asoc->param_flags =
2469 (asoc->param_flags & ~SPP_SACKDELAY) |
2470 sackdelay_change;
2471 } else {
2472 sp->param_flags =
2473 (sp->param_flags & ~SPP_SACKDELAY) |
2474 sackdelay_change;
2478 /* Note that a value of zero indicates the current setting should be
2479 left unchanged.
2481 if (params->spp_pathmaxrxt) {
2482 if (trans) {
2483 trans->pathmaxrxt = params->spp_pathmaxrxt;
2484 } else if (asoc) {
2485 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2486 } else {
2487 sp->pathmaxrxt = params->spp_pathmaxrxt;
2491 return 0;
2494 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2495 char __user *optval,
2496 unsigned int optlen)
2498 struct sctp_paddrparams params;
2499 struct sctp_transport *trans = NULL;
2500 struct sctp_association *asoc = NULL;
2501 struct sctp_sock *sp = sctp_sk(sk);
2502 int error;
2503 int hb_change, pmtud_change, sackdelay_change;
2505 if (optlen != sizeof(struct sctp_paddrparams))
2506 return -EINVAL;
2508 if (copy_from_user(&params, optval, optlen))
2509 return -EFAULT;
2511 /* Validate flags and value parameters. */
2512 hb_change = params.spp_flags & SPP_HB;
2513 pmtud_change = params.spp_flags & SPP_PMTUD;
2514 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2516 if (hb_change == SPP_HB ||
2517 pmtud_change == SPP_PMTUD ||
2518 sackdelay_change == SPP_SACKDELAY ||
2519 params.spp_sackdelay > 500 ||
2520 (params.spp_pathmtu &&
2521 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2522 return -EINVAL;
2524 /* If an address other than INADDR_ANY is specified, and
2525 * no transport is found, then the request is invalid.
2527 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2528 trans = sctp_addr_id2transport(sk, &params.spp_address,
2529 params.spp_assoc_id);
2530 if (!trans)
2531 return -EINVAL;
2534 /* Get association, if assoc_id != 0 and the socket is a one
2535 * to many style socket, and an association was not found, then
2536 * the id was invalid.
2538 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2539 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2540 return -EINVAL;
2542 /* Heartbeat demand can only be sent on a transport or
2543 * association, but not a socket.
2545 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2546 return -EINVAL;
2548 /* Process parameters. */
2549 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2550 hb_change, pmtud_change,
2551 sackdelay_change);
2553 if (error)
2554 return error;
2556 /* If changes are for association, also apply parameters to each
2557 * transport.
2559 if (!trans && asoc) {
2560 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2561 transports) {
2562 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2563 hb_change, pmtud_change,
2564 sackdelay_change);
2568 return 0;
2571 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2573 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2576 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2578 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2582 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2584 * This option will effect the way delayed acks are performed. This
2585 * option allows you to get or set the delayed ack time, in
2586 * milliseconds. It also allows changing the delayed ack frequency.
2587 * Changing the frequency to 1 disables the delayed sack algorithm. If
2588 * the assoc_id is 0, then this sets or gets the endpoints default
2589 * values. If the assoc_id field is non-zero, then the set or get
2590 * effects the specified association for the one to many model (the
2591 * assoc_id field is ignored by the one to one model). Note that if
2592 * sack_delay or sack_freq are 0 when setting this option, then the
2593 * current values will remain unchanged.
2595 * struct sctp_sack_info {
2596 * sctp_assoc_t sack_assoc_id;
2597 * uint32_t sack_delay;
2598 * uint32_t sack_freq;
2599 * };
2601 * sack_assoc_id - This parameter, indicates which association the user
2602 * is performing an action upon. Note that if this field's value is
2603 * zero then the endpoints default value is changed (effecting future
2604 * associations only).
2606 * sack_delay - This parameter contains the number of milliseconds that
2607 * the user is requesting the delayed ACK timer be set to. Note that
2608 * this value is defined in the standard to be between 200 and 500
2609 * milliseconds.
2611 * sack_freq - This parameter contains the number of packets that must
2612 * be received before a sack is sent without waiting for the delay
2613 * timer to expire. The default value for this is 2, setting this
2614 * value to 1 will disable the delayed sack algorithm.
2617 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2618 char __user *optval, unsigned int optlen)
2620 struct sctp_sack_info params;
2621 struct sctp_transport *trans = NULL;
2622 struct sctp_association *asoc = NULL;
2623 struct sctp_sock *sp = sctp_sk(sk);
2625 if (optlen == sizeof(struct sctp_sack_info)) {
2626 if (copy_from_user(&params, optval, optlen))
2627 return -EFAULT;
2629 if (params.sack_delay == 0 && params.sack_freq == 0)
2630 return 0;
2631 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2632 pr_warn_ratelimited(DEPRECATED
2633 "%s (pid %d) "
2634 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2635 "Use struct sctp_sack_info instead\n",
2636 current->comm, task_pid_nr(current));
2637 if (copy_from_user(&params, optval, optlen))
2638 return -EFAULT;
2640 if (params.sack_delay == 0)
2641 params.sack_freq = 1;
2642 else
2643 params.sack_freq = 0;
2644 } else
2645 return -EINVAL;
2647 /* Validate value parameter. */
2648 if (params.sack_delay > 500)
2649 return -EINVAL;
2651 /* Get association, if sack_assoc_id != 0 and the socket is a one
2652 * to many style socket, and an association was not found, then
2653 * the id was invalid.
2655 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2656 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2657 return -EINVAL;
2659 if (params.sack_delay) {
2660 if (asoc) {
2661 asoc->sackdelay =
2662 msecs_to_jiffies(params.sack_delay);
2663 asoc->param_flags =
2664 sctp_spp_sackdelay_enable(asoc->param_flags);
2665 } else {
2666 sp->sackdelay = params.sack_delay;
2667 sp->param_flags =
2668 sctp_spp_sackdelay_enable(sp->param_flags);
2672 if (params.sack_freq == 1) {
2673 if (asoc) {
2674 asoc->param_flags =
2675 sctp_spp_sackdelay_disable(asoc->param_flags);
2676 } else {
2677 sp->param_flags =
2678 sctp_spp_sackdelay_disable(sp->param_flags);
2680 } else if (params.sack_freq > 1) {
2681 if (asoc) {
2682 asoc->sackfreq = params.sack_freq;
2683 asoc->param_flags =
2684 sctp_spp_sackdelay_enable(asoc->param_flags);
2685 } else {
2686 sp->sackfreq = params.sack_freq;
2687 sp->param_flags =
2688 sctp_spp_sackdelay_enable(sp->param_flags);
2692 /* If change is for association, also apply to each transport. */
2693 if (asoc) {
2694 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2695 transports) {
2696 if (params.sack_delay) {
2697 trans->sackdelay =
2698 msecs_to_jiffies(params.sack_delay);
2699 trans->param_flags =
2700 sctp_spp_sackdelay_enable(trans->param_flags);
2702 if (params.sack_freq == 1) {
2703 trans->param_flags =
2704 sctp_spp_sackdelay_disable(trans->param_flags);
2705 } else if (params.sack_freq > 1) {
2706 trans->sackfreq = params.sack_freq;
2707 trans->param_flags =
2708 sctp_spp_sackdelay_enable(trans->param_flags);
2713 return 0;
2716 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2718 * Applications can specify protocol parameters for the default association
2719 * initialization. The option name argument to setsockopt() and getsockopt()
2720 * is SCTP_INITMSG.
2722 * Setting initialization parameters is effective only on an unconnected
2723 * socket (for UDP-style sockets only future associations are effected
2724 * by the change). With TCP-style sockets, this option is inherited by
2725 * sockets derived from a listener socket.
2727 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2729 struct sctp_initmsg sinit;
2730 struct sctp_sock *sp = sctp_sk(sk);
2732 if (optlen != sizeof(struct sctp_initmsg))
2733 return -EINVAL;
2734 if (copy_from_user(&sinit, optval, optlen))
2735 return -EFAULT;
2737 if (sinit.sinit_num_ostreams)
2738 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2739 if (sinit.sinit_max_instreams)
2740 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2741 if (sinit.sinit_max_attempts)
2742 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2743 if (sinit.sinit_max_init_timeo)
2744 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2746 return 0;
2750 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2752 * Applications that wish to use the sendto() system call may wish to
2753 * specify a default set of parameters that would normally be supplied
2754 * through the inclusion of ancillary data. This socket option allows
2755 * such an application to set the default sctp_sndrcvinfo structure.
2756 * The application that wishes to use this socket option simply passes
2757 * in to this call the sctp_sndrcvinfo structure defined in Section
2758 * 5.2.2) The input parameters accepted by this call include
2759 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2760 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2761 * to this call if the caller is using the UDP model.
2763 static int sctp_setsockopt_default_send_param(struct sock *sk,
2764 char __user *optval,
2765 unsigned int optlen)
2767 struct sctp_sock *sp = sctp_sk(sk);
2768 struct sctp_association *asoc;
2769 struct sctp_sndrcvinfo info;
2771 if (optlen != sizeof(info))
2772 return -EINVAL;
2773 if (copy_from_user(&info, optval, optlen))
2774 return -EFAULT;
2775 if (info.sinfo_flags &
2776 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2777 SCTP_ABORT | SCTP_EOF))
2778 return -EINVAL;
2780 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2781 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2782 return -EINVAL;
2783 if (asoc) {
2784 asoc->default_stream = info.sinfo_stream;
2785 asoc->default_flags = info.sinfo_flags;
2786 asoc->default_ppid = info.sinfo_ppid;
2787 asoc->default_context = info.sinfo_context;
2788 asoc->default_timetolive = info.sinfo_timetolive;
2789 } else {
2790 sp->default_stream = info.sinfo_stream;
2791 sp->default_flags = info.sinfo_flags;
2792 sp->default_ppid = info.sinfo_ppid;
2793 sp->default_context = info.sinfo_context;
2794 sp->default_timetolive = info.sinfo_timetolive;
2797 return 0;
2800 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2801 * (SCTP_DEFAULT_SNDINFO)
2803 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2804 char __user *optval,
2805 unsigned int optlen)
2807 struct sctp_sock *sp = sctp_sk(sk);
2808 struct sctp_association *asoc;
2809 struct sctp_sndinfo info;
2811 if (optlen != sizeof(info))
2812 return -EINVAL;
2813 if (copy_from_user(&info, optval, optlen))
2814 return -EFAULT;
2815 if (info.snd_flags &
2816 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2817 SCTP_ABORT | SCTP_EOF))
2818 return -EINVAL;
2820 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2821 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2822 return -EINVAL;
2823 if (asoc) {
2824 asoc->default_stream = info.snd_sid;
2825 asoc->default_flags = info.snd_flags;
2826 asoc->default_ppid = info.snd_ppid;
2827 asoc->default_context = info.snd_context;
2828 } else {
2829 sp->default_stream = info.snd_sid;
2830 sp->default_flags = info.snd_flags;
2831 sp->default_ppid = info.snd_ppid;
2832 sp->default_context = info.snd_context;
2835 return 0;
2838 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2840 * Requests that the local SCTP stack use the enclosed peer address as
2841 * the association primary. The enclosed address must be one of the
2842 * association peer's addresses.
2844 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2845 unsigned int optlen)
2847 struct sctp_prim prim;
2848 struct sctp_transport *trans;
2850 if (optlen != sizeof(struct sctp_prim))
2851 return -EINVAL;
2853 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2854 return -EFAULT;
2856 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2857 if (!trans)
2858 return -EINVAL;
2860 sctp_assoc_set_primary(trans->asoc, trans);
2862 return 0;
2866 * 7.1.5 SCTP_NODELAY
2868 * Turn on/off any Nagle-like algorithm. This means that packets are
2869 * generally sent as soon as possible and no unnecessary delays are
2870 * introduced, at the cost of more packets in the network. Expects an
2871 * integer boolean flag.
2873 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2874 unsigned int optlen)
2876 int val;
2878 if (optlen < sizeof(int))
2879 return -EINVAL;
2880 if (get_user(val, (int __user *)optval))
2881 return -EFAULT;
2883 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2884 return 0;
2889 * 7.1.1 SCTP_RTOINFO
2891 * The protocol parameters used to initialize and bound retransmission
2892 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2893 * and modify these parameters.
2894 * All parameters are time values, in milliseconds. A value of 0, when
2895 * modifying the parameters, indicates that the current value should not
2896 * be changed.
2899 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2901 struct sctp_rtoinfo rtoinfo;
2902 struct sctp_association *asoc;
2903 unsigned long rto_min, rto_max;
2904 struct sctp_sock *sp = sctp_sk(sk);
2906 if (optlen != sizeof (struct sctp_rtoinfo))
2907 return -EINVAL;
2909 if (copy_from_user(&rtoinfo, optval, optlen))
2910 return -EFAULT;
2912 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2914 /* Set the values to the specific association */
2915 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2916 return -EINVAL;
2918 rto_max = rtoinfo.srto_max;
2919 rto_min = rtoinfo.srto_min;
2921 if (rto_max)
2922 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2923 else
2924 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2926 if (rto_min)
2927 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2928 else
2929 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2931 if (rto_min > rto_max)
2932 return -EINVAL;
2934 if (asoc) {
2935 if (rtoinfo.srto_initial != 0)
2936 asoc->rto_initial =
2937 msecs_to_jiffies(rtoinfo.srto_initial);
2938 asoc->rto_max = rto_max;
2939 asoc->rto_min = rto_min;
2940 } else {
2941 /* If there is no association or the association-id = 0
2942 * set the values to the endpoint.
2944 if (rtoinfo.srto_initial != 0)
2945 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2946 sp->rtoinfo.srto_max = rto_max;
2947 sp->rtoinfo.srto_min = rto_min;
2950 return 0;
2955 * 7.1.2 SCTP_ASSOCINFO
2957 * This option is used to tune the maximum retransmission attempts
2958 * of the association.
2959 * Returns an error if the new association retransmission value is
2960 * greater than the sum of the retransmission value of the peer.
2961 * See [SCTP] for more information.
2964 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2967 struct sctp_assocparams assocparams;
2968 struct sctp_association *asoc;
2970 if (optlen != sizeof(struct sctp_assocparams))
2971 return -EINVAL;
2972 if (copy_from_user(&assocparams, optval, optlen))
2973 return -EFAULT;
2975 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2977 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2978 return -EINVAL;
2980 /* Set the values to the specific association */
2981 if (asoc) {
2982 if (assocparams.sasoc_asocmaxrxt != 0) {
2983 __u32 path_sum = 0;
2984 int paths = 0;
2985 struct sctp_transport *peer_addr;
2987 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2988 transports) {
2989 path_sum += peer_addr->pathmaxrxt;
2990 paths++;
2993 /* Only validate asocmaxrxt if we have more than
2994 * one path/transport. We do this because path
2995 * retransmissions are only counted when we have more
2996 * then one path.
2998 if (paths > 1 &&
2999 assocparams.sasoc_asocmaxrxt > path_sum)
3000 return -EINVAL;
3002 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3005 if (assocparams.sasoc_cookie_life != 0)
3006 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3007 } else {
3008 /* Set the values to the endpoint */
3009 struct sctp_sock *sp = sctp_sk(sk);
3011 if (assocparams.sasoc_asocmaxrxt != 0)
3012 sp->assocparams.sasoc_asocmaxrxt =
3013 assocparams.sasoc_asocmaxrxt;
3014 if (assocparams.sasoc_cookie_life != 0)
3015 sp->assocparams.sasoc_cookie_life =
3016 assocparams.sasoc_cookie_life;
3018 return 0;
3022 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3024 * This socket option is a boolean flag which turns on or off mapped V4
3025 * addresses. If this option is turned on and the socket is type
3026 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3027 * If this option is turned off, then no mapping will be done of V4
3028 * addresses and a user will receive both PF_INET6 and PF_INET type
3029 * addresses on the socket.
3031 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3033 int val;
3034 struct sctp_sock *sp = sctp_sk(sk);
3036 if (optlen < sizeof(int))
3037 return -EINVAL;
3038 if (get_user(val, (int __user *)optval))
3039 return -EFAULT;
3040 if (val)
3041 sp->v4mapped = 1;
3042 else
3043 sp->v4mapped = 0;
3045 return 0;
3049 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3050 * This option will get or set the maximum size to put in any outgoing
3051 * SCTP DATA chunk. If a message is larger than this size it will be
3052 * fragmented by SCTP into the specified size. Note that the underlying
3053 * SCTP implementation may fragment into smaller sized chunks when the
3054 * PMTU of the underlying association is smaller than the value set by
3055 * the user. The default value for this option is '0' which indicates
3056 * the user is NOT limiting fragmentation and only the PMTU will effect
3057 * SCTP's choice of DATA chunk size. Note also that values set larger
3058 * than the maximum size of an IP datagram will effectively let SCTP
3059 * control fragmentation (i.e. the same as setting this option to 0).
3061 * The following structure is used to access and modify this parameter:
3063 * struct sctp_assoc_value {
3064 * sctp_assoc_t assoc_id;
3065 * uint32_t assoc_value;
3066 * };
3068 * assoc_id: This parameter is ignored for one-to-one style sockets.
3069 * For one-to-many style sockets this parameter indicates which
3070 * association the user is performing an action upon. Note that if
3071 * this field's value is zero then the endpoints default value is
3072 * changed (effecting future associations only).
3073 * assoc_value: This parameter specifies the maximum size in bytes.
3075 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3077 struct sctp_assoc_value params;
3078 struct sctp_association *asoc;
3079 struct sctp_sock *sp = sctp_sk(sk);
3080 int val;
3082 if (optlen == sizeof(int)) {
3083 pr_warn_ratelimited(DEPRECATED
3084 "%s (pid %d) "
3085 "Use of int in maxseg socket option.\n"
3086 "Use struct sctp_assoc_value instead\n",
3087 current->comm, task_pid_nr(current));
3088 if (copy_from_user(&val, optval, optlen))
3089 return -EFAULT;
3090 params.assoc_id = 0;
3091 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3092 if (copy_from_user(&params, optval, optlen))
3093 return -EFAULT;
3094 val = params.assoc_value;
3095 } else
3096 return -EINVAL;
3098 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3099 return -EINVAL;
3101 asoc = sctp_id2assoc(sk, params.assoc_id);
3102 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3103 return -EINVAL;
3105 if (asoc) {
3106 if (val == 0) {
3107 val = asoc->pathmtu;
3108 val -= sp->pf->af->net_header_len;
3109 val -= sizeof(struct sctphdr) +
3110 sizeof(struct sctp_data_chunk);
3112 asoc->user_frag = val;
3113 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3114 } else {
3115 sp->user_frag = val;
3118 return 0;
3123 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3125 * Requests that the peer mark the enclosed address as the association
3126 * primary. The enclosed address must be one of the association's
3127 * locally bound addresses. The following structure is used to make a
3128 * set primary request:
3130 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3131 unsigned int optlen)
3133 struct net *net = sock_net(sk);
3134 struct sctp_sock *sp;
3135 struct sctp_association *asoc = NULL;
3136 struct sctp_setpeerprim prim;
3137 struct sctp_chunk *chunk;
3138 struct sctp_af *af;
3139 int err;
3141 sp = sctp_sk(sk);
3143 if (!net->sctp.addip_enable)
3144 return -EPERM;
3146 if (optlen != sizeof(struct sctp_setpeerprim))
3147 return -EINVAL;
3149 if (copy_from_user(&prim, optval, optlen))
3150 return -EFAULT;
3152 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3153 if (!asoc)
3154 return -EINVAL;
3156 if (!asoc->peer.asconf_capable)
3157 return -EPERM;
3159 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3160 return -EPERM;
3162 if (!sctp_state(asoc, ESTABLISHED))
3163 return -ENOTCONN;
3165 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3166 if (!af)
3167 return -EINVAL;
3169 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3170 return -EADDRNOTAVAIL;
3172 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3173 return -EADDRNOTAVAIL;
3175 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3176 chunk = sctp_make_asconf_set_prim(asoc,
3177 (union sctp_addr *)&prim.sspp_addr);
3178 if (!chunk)
3179 return -ENOMEM;
3181 err = sctp_send_asconf(asoc, chunk);
3183 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3185 return err;
3188 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3189 unsigned int optlen)
3191 struct sctp_setadaptation adaptation;
3193 if (optlen != sizeof(struct sctp_setadaptation))
3194 return -EINVAL;
3195 if (copy_from_user(&adaptation, optval, optlen))
3196 return -EFAULT;
3198 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3200 return 0;
3204 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3206 * The context field in the sctp_sndrcvinfo structure is normally only
3207 * used when a failed message is retrieved holding the value that was
3208 * sent down on the actual send call. This option allows the setting of
3209 * a default context on an association basis that will be received on
3210 * reading messages from the peer. This is especially helpful in the
3211 * one-2-many model for an application to keep some reference to an
3212 * internal state machine that is processing messages on the
3213 * association. Note that the setting of this value only effects
3214 * received messages from the peer and does not effect the value that is
3215 * saved with outbound messages.
3217 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3218 unsigned int optlen)
3220 struct sctp_assoc_value params;
3221 struct sctp_sock *sp;
3222 struct sctp_association *asoc;
3224 if (optlen != sizeof(struct sctp_assoc_value))
3225 return -EINVAL;
3226 if (copy_from_user(&params, optval, optlen))
3227 return -EFAULT;
3229 sp = sctp_sk(sk);
3231 if (params.assoc_id != 0) {
3232 asoc = sctp_id2assoc(sk, params.assoc_id);
3233 if (!asoc)
3234 return -EINVAL;
3235 asoc->default_rcv_context = params.assoc_value;
3236 } else {
3237 sp->default_rcv_context = params.assoc_value;
3240 return 0;
3244 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3246 * This options will at a minimum specify if the implementation is doing
3247 * fragmented interleave. Fragmented interleave, for a one to many
3248 * socket, is when subsequent calls to receive a message may return
3249 * parts of messages from different associations. Some implementations
3250 * may allow you to turn this value on or off. If so, when turned off,
3251 * no fragment interleave will occur (which will cause a head of line
3252 * blocking amongst multiple associations sharing the same one to many
3253 * socket). When this option is turned on, then each receive call may
3254 * come from a different association (thus the user must receive data
3255 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3256 * association each receive belongs to.
3258 * This option takes a boolean value. A non-zero value indicates that
3259 * fragmented interleave is on. A value of zero indicates that
3260 * fragmented interleave is off.
3262 * Note that it is important that an implementation that allows this
3263 * option to be turned on, have it off by default. Otherwise an unaware
3264 * application using the one to many model may become confused and act
3265 * incorrectly.
3267 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3268 char __user *optval,
3269 unsigned int optlen)
3271 int val;
3273 if (optlen != sizeof(int))
3274 return -EINVAL;
3275 if (get_user(val, (int __user *)optval))
3276 return -EFAULT;
3278 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3280 return 0;
3284 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3285 * (SCTP_PARTIAL_DELIVERY_POINT)
3287 * This option will set or get the SCTP partial delivery point. This
3288 * point is the size of a message where the partial delivery API will be
3289 * invoked to help free up rwnd space for the peer. Setting this to a
3290 * lower value will cause partial deliveries to happen more often. The
3291 * calls argument is an integer that sets or gets the partial delivery
3292 * point. Note also that the call will fail if the user attempts to set
3293 * this value larger than the socket receive buffer size.
3295 * Note that any single message having a length smaller than or equal to
3296 * the SCTP partial delivery point will be delivered in one single read
3297 * call as long as the user provided buffer is large enough to hold the
3298 * message.
3300 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3301 char __user *optval,
3302 unsigned int optlen)
3304 u32 val;
3306 if (optlen != sizeof(u32))
3307 return -EINVAL;
3308 if (get_user(val, (int __user *)optval))
3309 return -EFAULT;
3311 /* Note: We double the receive buffer from what the user sets
3312 * it to be, also initial rwnd is based on rcvbuf/2.
3314 if (val > (sk->sk_rcvbuf >> 1))
3315 return -EINVAL;
3317 sctp_sk(sk)->pd_point = val;
3319 return 0; /* is this the right error code? */
3323 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3325 * This option will allow a user to change the maximum burst of packets
3326 * that can be emitted by this association. Note that the default value
3327 * is 4, and some implementations may restrict this setting so that it
3328 * can only be lowered.
3330 * NOTE: This text doesn't seem right. Do this on a socket basis with
3331 * future associations inheriting the socket value.
3333 static int sctp_setsockopt_maxburst(struct sock *sk,
3334 char __user *optval,
3335 unsigned int optlen)
3337 struct sctp_assoc_value params;
3338 struct sctp_sock *sp;
3339 struct sctp_association *asoc;
3340 int val;
3341 int assoc_id = 0;
3343 if (optlen == sizeof(int)) {
3344 pr_warn_ratelimited(DEPRECATED
3345 "%s (pid %d) "
3346 "Use of int in max_burst socket option deprecated.\n"
3347 "Use struct sctp_assoc_value instead\n",
3348 current->comm, task_pid_nr(current));
3349 if (copy_from_user(&val, optval, optlen))
3350 return -EFAULT;
3351 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3352 if (copy_from_user(&params, optval, optlen))
3353 return -EFAULT;
3354 val = params.assoc_value;
3355 assoc_id = params.assoc_id;
3356 } else
3357 return -EINVAL;
3359 sp = sctp_sk(sk);
3361 if (assoc_id != 0) {
3362 asoc = sctp_id2assoc(sk, assoc_id);
3363 if (!asoc)
3364 return -EINVAL;
3365 asoc->max_burst = val;
3366 } else
3367 sp->max_burst = val;
3369 return 0;
3373 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3375 * This set option adds a chunk type that the user is requesting to be
3376 * received only in an authenticated way. Changes to the list of chunks
3377 * will only effect future associations on the socket.
3379 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3380 char __user *optval,
3381 unsigned int optlen)
3383 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3384 struct sctp_authchunk val;
3386 if (!ep->auth_enable)
3387 return -EACCES;
3389 if (optlen != sizeof(struct sctp_authchunk))
3390 return -EINVAL;
3391 if (copy_from_user(&val, optval, optlen))
3392 return -EFAULT;
3394 switch (val.sauth_chunk) {
3395 case SCTP_CID_INIT:
3396 case SCTP_CID_INIT_ACK:
3397 case SCTP_CID_SHUTDOWN_COMPLETE:
3398 case SCTP_CID_AUTH:
3399 return -EINVAL;
3402 /* add this chunk id to the endpoint */
3403 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3407 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3409 * This option gets or sets the list of HMAC algorithms that the local
3410 * endpoint requires the peer to use.
3412 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3413 char __user *optval,
3414 unsigned int optlen)
3416 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3417 struct sctp_hmacalgo *hmacs;
3418 u32 idents;
3419 int err;
3421 if (!ep->auth_enable)
3422 return -EACCES;
3424 if (optlen < sizeof(struct sctp_hmacalgo))
3425 return -EINVAL;
3427 hmacs = memdup_user(optval, optlen);
3428 if (IS_ERR(hmacs))
3429 return PTR_ERR(hmacs);
3431 idents = hmacs->shmac_num_idents;
3432 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3433 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3434 err = -EINVAL;
3435 goto out;
3438 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3439 out:
3440 kfree(hmacs);
3441 return err;
3445 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3447 * This option will set a shared secret key which is used to build an
3448 * association shared key.
3450 static int sctp_setsockopt_auth_key(struct sock *sk,
3451 char __user *optval,
3452 unsigned int optlen)
3454 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3455 struct sctp_authkey *authkey;
3456 struct sctp_association *asoc;
3457 int ret;
3459 if (!ep->auth_enable)
3460 return -EACCES;
3462 if (optlen <= sizeof(struct sctp_authkey))
3463 return -EINVAL;
3465 authkey = memdup_user(optval, optlen);
3466 if (IS_ERR(authkey))
3467 return PTR_ERR(authkey);
3469 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3470 ret = -EINVAL;
3471 goto out;
3474 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3475 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3476 ret = -EINVAL;
3477 goto out;
3480 ret = sctp_auth_set_key(ep, asoc, authkey);
3481 out:
3482 kzfree(authkey);
3483 return ret;
3487 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3489 * This option will get or set the active shared key to be used to build
3490 * the association shared key.
3492 static int sctp_setsockopt_active_key(struct sock *sk,
3493 char __user *optval,
3494 unsigned int optlen)
3496 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3497 struct sctp_authkeyid val;
3498 struct sctp_association *asoc;
3500 if (!ep->auth_enable)
3501 return -EACCES;
3503 if (optlen != sizeof(struct sctp_authkeyid))
3504 return -EINVAL;
3505 if (copy_from_user(&val, optval, optlen))
3506 return -EFAULT;
3508 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3509 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3510 return -EINVAL;
3512 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3516 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3518 * This set option will delete a shared secret key from use.
3520 static int sctp_setsockopt_del_key(struct sock *sk,
3521 char __user *optval,
3522 unsigned int optlen)
3524 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3525 struct sctp_authkeyid val;
3526 struct sctp_association *asoc;
3528 if (!ep->auth_enable)
3529 return -EACCES;
3531 if (optlen != sizeof(struct sctp_authkeyid))
3532 return -EINVAL;
3533 if (copy_from_user(&val, optval, optlen))
3534 return -EFAULT;
3536 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3537 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3538 return -EINVAL;
3540 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3545 * 8.1.23 SCTP_AUTO_ASCONF
3547 * This option will enable or disable the use of the automatic generation of
3548 * ASCONF chunks to add and delete addresses to an existing association. Note
3549 * that this option has two caveats namely: a) it only affects sockets that
3550 * are bound to all addresses available to the SCTP stack, and b) the system
3551 * administrator may have an overriding control that turns the ASCONF feature
3552 * off no matter what setting the socket option may have.
3553 * This option expects an integer boolean flag, where a non-zero value turns on
3554 * the option, and a zero value turns off the option.
3555 * Note. In this implementation, socket operation overrides default parameter
3556 * being set by sysctl as well as FreeBSD implementation
3558 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3559 unsigned int optlen)
3561 int val;
3562 struct sctp_sock *sp = sctp_sk(sk);
3564 if (optlen < sizeof(int))
3565 return -EINVAL;
3566 if (get_user(val, (int __user *)optval))
3567 return -EFAULT;
3568 if (!sctp_is_ep_boundall(sk) && val)
3569 return -EINVAL;
3570 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3571 return 0;
3573 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3574 if (val == 0 && sp->do_auto_asconf) {
3575 list_del(&sp->auto_asconf_list);
3576 sp->do_auto_asconf = 0;
3577 } else if (val && !sp->do_auto_asconf) {
3578 list_add_tail(&sp->auto_asconf_list,
3579 &sock_net(sk)->sctp.auto_asconf_splist);
3580 sp->do_auto_asconf = 1;
3582 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3583 return 0;
3587 * SCTP_PEER_ADDR_THLDS
3589 * This option allows us to alter the partially failed threshold for one or all
3590 * transports in an association. See Section 6.1 of:
3591 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3593 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3594 char __user *optval,
3595 unsigned int optlen)
3597 struct sctp_paddrthlds val;
3598 struct sctp_transport *trans;
3599 struct sctp_association *asoc;
3601 if (optlen < sizeof(struct sctp_paddrthlds))
3602 return -EINVAL;
3603 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3604 sizeof(struct sctp_paddrthlds)))
3605 return -EFAULT;
3608 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3609 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3610 if (!asoc)
3611 return -ENOENT;
3612 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3613 transports) {
3614 if (val.spt_pathmaxrxt)
3615 trans->pathmaxrxt = val.spt_pathmaxrxt;
3616 trans->pf_retrans = val.spt_pathpfthld;
3619 if (val.spt_pathmaxrxt)
3620 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3621 asoc->pf_retrans = val.spt_pathpfthld;
3622 } else {
3623 trans = sctp_addr_id2transport(sk, &val.spt_address,
3624 val.spt_assoc_id);
3625 if (!trans)
3626 return -ENOENT;
3628 if (val.spt_pathmaxrxt)
3629 trans->pathmaxrxt = val.spt_pathmaxrxt;
3630 trans->pf_retrans = val.spt_pathpfthld;
3633 return 0;
3636 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3637 char __user *optval,
3638 unsigned int optlen)
3640 int val;
3642 if (optlen < sizeof(int))
3643 return -EINVAL;
3644 if (get_user(val, (int __user *) optval))
3645 return -EFAULT;
3647 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3649 return 0;
3652 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3653 char __user *optval,
3654 unsigned int optlen)
3656 int val;
3658 if (optlen < sizeof(int))
3659 return -EINVAL;
3660 if (get_user(val, (int __user *) optval))
3661 return -EFAULT;
3663 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3665 return 0;
3668 /* API 6.2 setsockopt(), getsockopt()
3670 * Applications use setsockopt() and getsockopt() to set or retrieve
3671 * socket options. Socket options are used to change the default
3672 * behavior of sockets calls. They are described in Section 7.
3674 * The syntax is:
3676 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3677 * int __user *optlen);
3678 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3679 * int optlen);
3681 * sd - the socket descript.
3682 * level - set to IPPROTO_SCTP for all SCTP options.
3683 * optname - the option name.
3684 * optval - the buffer to store the value of the option.
3685 * optlen - the size of the buffer.
3687 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3688 char __user *optval, unsigned int optlen)
3690 int retval = 0;
3692 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3694 /* I can hardly begin to describe how wrong this is. This is
3695 * so broken as to be worse than useless. The API draft
3696 * REALLY is NOT helpful here... I am not convinced that the
3697 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3698 * are at all well-founded.
3700 if (level != SOL_SCTP) {
3701 struct sctp_af *af = sctp_sk(sk)->pf->af;
3702 retval = af->setsockopt(sk, level, optname, optval, optlen);
3703 goto out_nounlock;
3706 lock_sock(sk);
3708 switch (optname) {
3709 case SCTP_SOCKOPT_BINDX_ADD:
3710 /* 'optlen' is the size of the addresses buffer. */
3711 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3712 optlen, SCTP_BINDX_ADD_ADDR);
3713 break;
3715 case SCTP_SOCKOPT_BINDX_REM:
3716 /* 'optlen' is the size of the addresses buffer. */
3717 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3718 optlen, SCTP_BINDX_REM_ADDR);
3719 break;
3721 case SCTP_SOCKOPT_CONNECTX_OLD:
3722 /* 'optlen' is the size of the addresses buffer. */
3723 retval = sctp_setsockopt_connectx_old(sk,
3724 (struct sockaddr __user *)optval,
3725 optlen);
3726 break;
3728 case SCTP_SOCKOPT_CONNECTX:
3729 /* 'optlen' is the size of the addresses buffer. */
3730 retval = sctp_setsockopt_connectx(sk,
3731 (struct sockaddr __user *)optval,
3732 optlen);
3733 break;
3735 case SCTP_DISABLE_FRAGMENTS:
3736 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3737 break;
3739 case SCTP_EVENTS:
3740 retval = sctp_setsockopt_events(sk, optval, optlen);
3741 break;
3743 case SCTP_AUTOCLOSE:
3744 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3745 break;
3747 case SCTP_PEER_ADDR_PARAMS:
3748 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3749 break;
3751 case SCTP_DELAYED_SACK:
3752 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3753 break;
3754 case SCTP_PARTIAL_DELIVERY_POINT:
3755 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3756 break;
3758 case SCTP_INITMSG:
3759 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3760 break;
3761 case SCTP_DEFAULT_SEND_PARAM:
3762 retval = sctp_setsockopt_default_send_param(sk, optval,
3763 optlen);
3764 break;
3765 case SCTP_DEFAULT_SNDINFO:
3766 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3767 break;
3768 case SCTP_PRIMARY_ADDR:
3769 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3770 break;
3771 case SCTP_SET_PEER_PRIMARY_ADDR:
3772 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3773 break;
3774 case SCTP_NODELAY:
3775 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3776 break;
3777 case SCTP_RTOINFO:
3778 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3779 break;
3780 case SCTP_ASSOCINFO:
3781 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3782 break;
3783 case SCTP_I_WANT_MAPPED_V4_ADDR:
3784 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3785 break;
3786 case SCTP_MAXSEG:
3787 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3788 break;
3789 case SCTP_ADAPTATION_LAYER:
3790 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3791 break;
3792 case SCTP_CONTEXT:
3793 retval = sctp_setsockopt_context(sk, optval, optlen);
3794 break;
3795 case SCTP_FRAGMENT_INTERLEAVE:
3796 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3797 break;
3798 case SCTP_MAX_BURST:
3799 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3800 break;
3801 case SCTP_AUTH_CHUNK:
3802 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3803 break;
3804 case SCTP_HMAC_IDENT:
3805 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3806 break;
3807 case SCTP_AUTH_KEY:
3808 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3809 break;
3810 case SCTP_AUTH_ACTIVE_KEY:
3811 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3812 break;
3813 case SCTP_AUTH_DELETE_KEY:
3814 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3815 break;
3816 case SCTP_AUTO_ASCONF:
3817 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3818 break;
3819 case SCTP_PEER_ADDR_THLDS:
3820 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3821 break;
3822 case SCTP_RECVRCVINFO:
3823 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3824 break;
3825 case SCTP_RECVNXTINFO:
3826 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3827 break;
3828 default:
3829 retval = -ENOPROTOOPT;
3830 break;
3833 release_sock(sk);
3835 out_nounlock:
3836 return retval;
3839 /* API 3.1.6 connect() - UDP Style Syntax
3841 * An application may use the connect() call in the UDP model to initiate an
3842 * association without sending data.
3844 * The syntax is:
3846 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3848 * sd: the socket descriptor to have a new association added to.
3850 * nam: the address structure (either struct sockaddr_in or struct
3851 * sockaddr_in6 defined in RFC2553 [7]).
3853 * len: the size of the address.
3855 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3856 int addr_len)
3858 int err = 0;
3859 struct sctp_af *af;
3861 lock_sock(sk);
3863 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3864 addr, addr_len);
3866 /* Validate addr_len before calling common connect/connectx routine. */
3867 af = sctp_get_af_specific(addr->sa_family);
3868 if (!af || addr_len < af->sockaddr_len) {
3869 err = -EINVAL;
3870 } else {
3871 /* Pass correct addr len to common routine (so it knows there
3872 * is only one address being passed.
3874 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3877 release_sock(sk);
3878 return err;
3881 /* FIXME: Write comments. */
3882 static int sctp_disconnect(struct sock *sk, int flags)
3884 return -EOPNOTSUPP; /* STUB */
3887 /* 4.1.4 accept() - TCP Style Syntax
3889 * Applications use accept() call to remove an established SCTP
3890 * association from the accept queue of the endpoint. A new socket
3891 * descriptor will be returned from accept() to represent the newly
3892 * formed association.
3894 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3896 struct sctp_sock *sp;
3897 struct sctp_endpoint *ep;
3898 struct sock *newsk = NULL;
3899 struct sctp_association *asoc;
3900 long timeo;
3901 int error = 0;
3903 lock_sock(sk);
3905 sp = sctp_sk(sk);
3906 ep = sp->ep;
3908 if (!sctp_style(sk, TCP)) {
3909 error = -EOPNOTSUPP;
3910 goto out;
3913 if (!sctp_sstate(sk, LISTENING)) {
3914 error = -EINVAL;
3915 goto out;
3918 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3920 error = sctp_wait_for_accept(sk, timeo);
3921 if (error)
3922 goto out;
3924 /* We treat the list of associations on the endpoint as the accept
3925 * queue and pick the first association on the list.
3927 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3929 newsk = sp->pf->create_accept_sk(sk, asoc);
3930 if (!newsk) {
3931 error = -ENOMEM;
3932 goto out;
3935 /* Populate the fields of the newsk from the oldsk and migrate the
3936 * asoc to the newsk.
3938 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3940 out:
3941 release_sock(sk);
3942 *err = error;
3943 return newsk;
3946 /* The SCTP ioctl handler. */
3947 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3949 int rc = -ENOTCONN;
3951 lock_sock(sk);
3954 * SEQPACKET-style sockets in LISTENING state are valid, for
3955 * SCTP, so only discard TCP-style sockets in LISTENING state.
3957 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3958 goto out;
3960 switch (cmd) {
3961 case SIOCINQ: {
3962 struct sk_buff *skb;
3963 unsigned int amount = 0;
3965 skb = skb_peek(&sk->sk_receive_queue);
3966 if (skb != NULL) {
3968 * We will only return the amount of this packet since
3969 * that is all that will be read.
3971 amount = skb->len;
3973 rc = put_user(amount, (int __user *)arg);
3974 break;
3976 default:
3977 rc = -ENOIOCTLCMD;
3978 break;
3980 out:
3981 release_sock(sk);
3982 return rc;
3985 /* This is the function which gets called during socket creation to
3986 * initialized the SCTP-specific portion of the sock.
3987 * The sock structure should already be zero-filled memory.
3989 static int sctp_init_sock(struct sock *sk)
3991 struct net *net = sock_net(sk);
3992 struct sctp_sock *sp;
3994 pr_debug("%s: sk:%p\n", __func__, sk);
3996 sp = sctp_sk(sk);
3998 /* Initialize the SCTP per socket area. */
3999 switch (sk->sk_type) {
4000 case SOCK_SEQPACKET:
4001 sp->type = SCTP_SOCKET_UDP;
4002 break;
4003 case SOCK_STREAM:
4004 sp->type = SCTP_SOCKET_TCP;
4005 break;
4006 default:
4007 return -ESOCKTNOSUPPORT;
4010 /* Initialize default send parameters. These parameters can be
4011 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4013 sp->default_stream = 0;
4014 sp->default_ppid = 0;
4015 sp->default_flags = 0;
4016 sp->default_context = 0;
4017 sp->default_timetolive = 0;
4019 sp->default_rcv_context = 0;
4020 sp->max_burst = net->sctp.max_burst;
4022 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4024 /* Initialize default setup parameters. These parameters
4025 * can be modified with the SCTP_INITMSG socket option or
4026 * overridden by the SCTP_INIT CMSG.
4028 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4029 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4030 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4031 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4033 /* Initialize default RTO related parameters. These parameters can
4034 * be modified for with the SCTP_RTOINFO socket option.
4036 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4037 sp->rtoinfo.srto_max = net->sctp.rto_max;
4038 sp->rtoinfo.srto_min = net->sctp.rto_min;
4040 /* Initialize default association related parameters. These parameters
4041 * can be modified with the SCTP_ASSOCINFO socket option.
4043 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4044 sp->assocparams.sasoc_number_peer_destinations = 0;
4045 sp->assocparams.sasoc_peer_rwnd = 0;
4046 sp->assocparams.sasoc_local_rwnd = 0;
4047 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4049 /* Initialize default event subscriptions. By default, all the
4050 * options are off.
4052 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4054 /* Default Peer Address Parameters. These defaults can
4055 * be modified via SCTP_PEER_ADDR_PARAMS
4057 sp->hbinterval = net->sctp.hb_interval;
4058 sp->pathmaxrxt = net->sctp.max_retrans_path;
4059 sp->pathmtu = 0; /* allow default discovery */
4060 sp->sackdelay = net->sctp.sack_timeout;
4061 sp->sackfreq = 2;
4062 sp->param_flags = SPP_HB_ENABLE |
4063 SPP_PMTUD_ENABLE |
4064 SPP_SACKDELAY_ENABLE;
4066 /* If enabled no SCTP message fragmentation will be performed.
4067 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4069 sp->disable_fragments = 0;
4071 /* Enable Nagle algorithm by default. */
4072 sp->nodelay = 0;
4074 sp->recvrcvinfo = 0;
4075 sp->recvnxtinfo = 0;
4077 /* Enable by default. */
4078 sp->v4mapped = 1;
4080 /* Auto-close idle associations after the configured
4081 * number of seconds. A value of 0 disables this
4082 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4083 * for UDP-style sockets only.
4085 sp->autoclose = 0;
4087 /* User specified fragmentation limit. */
4088 sp->user_frag = 0;
4090 sp->adaptation_ind = 0;
4092 sp->pf = sctp_get_pf_specific(sk->sk_family);
4094 /* Control variables for partial data delivery. */
4095 atomic_set(&sp->pd_mode, 0);
4096 skb_queue_head_init(&sp->pd_lobby);
4097 sp->frag_interleave = 0;
4099 /* Create a per socket endpoint structure. Even if we
4100 * change the data structure relationships, this may still
4101 * be useful for storing pre-connect address information.
4103 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4104 if (!sp->ep)
4105 return -ENOMEM;
4107 sp->hmac = NULL;
4109 sk->sk_destruct = sctp_destruct_sock;
4111 SCTP_DBG_OBJCNT_INC(sock);
4113 local_bh_disable();
4114 percpu_counter_inc(&sctp_sockets_allocated);
4115 sock_prot_inuse_add(net, sk->sk_prot, 1);
4117 /* Nothing can fail after this block, otherwise
4118 * sctp_destroy_sock() will be called without addr_wq_lock held
4120 if (net->sctp.default_auto_asconf) {
4121 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4122 list_add_tail(&sp->auto_asconf_list,
4123 &net->sctp.auto_asconf_splist);
4124 sp->do_auto_asconf = 1;
4125 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4126 } else {
4127 sp->do_auto_asconf = 0;
4130 local_bh_enable();
4132 return 0;
4135 /* Cleanup any SCTP per socket resources. Must be called with
4136 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4138 static void sctp_destroy_sock(struct sock *sk)
4140 struct sctp_sock *sp;
4142 pr_debug("%s: sk:%p\n", __func__, sk);
4144 /* Release our hold on the endpoint. */
4145 sp = sctp_sk(sk);
4146 /* This could happen during socket init, thus we bail out
4147 * early, since the rest of the below is not setup either.
4149 if (sp->ep == NULL)
4150 return;
4152 if (sp->do_auto_asconf) {
4153 sp->do_auto_asconf = 0;
4154 list_del(&sp->auto_asconf_list);
4156 sctp_endpoint_free(sp->ep);
4157 local_bh_disable();
4158 percpu_counter_dec(&sctp_sockets_allocated);
4159 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4160 local_bh_enable();
4163 /* Triggered when there are no references on the socket anymore */
4164 static void sctp_destruct_sock(struct sock *sk)
4166 struct sctp_sock *sp = sctp_sk(sk);
4168 /* Free up the HMAC transform. */
4169 crypto_free_hash(sp->hmac);
4171 inet_sock_destruct(sk);
4174 /* API 4.1.7 shutdown() - TCP Style Syntax
4175 * int shutdown(int socket, int how);
4177 * sd - the socket descriptor of the association to be closed.
4178 * how - Specifies the type of shutdown. The values are
4179 * as follows:
4180 * SHUT_RD
4181 * Disables further receive operations. No SCTP
4182 * protocol action is taken.
4183 * SHUT_WR
4184 * Disables further send operations, and initiates
4185 * the SCTP shutdown sequence.
4186 * SHUT_RDWR
4187 * Disables further send and receive operations
4188 * and initiates the SCTP shutdown sequence.
4190 static void sctp_shutdown(struct sock *sk, int how)
4192 struct net *net = sock_net(sk);
4193 struct sctp_endpoint *ep;
4194 struct sctp_association *asoc;
4196 if (!sctp_style(sk, TCP))
4197 return;
4199 if (how & SEND_SHUTDOWN) {
4200 ep = sctp_sk(sk)->ep;
4201 if (!list_empty(&ep->asocs)) {
4202 asoc = list_entry(ep->asocs.next,
4203 struct sctp_association, asocs);
4204 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4209 /* 7.2.1 Association Status (SCTP_STATUS)
4211 * Applications can retrieve current status information about an
4212 * association, including association state, peer receiver window size,
4213 * number of unacked data chunks, and number of data chunks pending
4214 * receipt. This information is read-only.
4216 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4217 char __user *optval,
4218 int __user *optlen)
4220 struct sctp_status status;
4221 struct sctp_association *asoc = NULL;
4222 struct sctp_transport *transport;
4223 sctp_assoc_t associd;
4224 int retval = 0;
4226 if (len < sizeof(status)) {
4227 retval = -EINVAL;
4228 goto out;
4231 len = sizeof(status);
4232 if (copy_from_user(&status, optval, len)) {
4233 retval = -EFAULT;
4234 goto out;
4237 associd = status.sstat_assoc_id;
4238 asoc = sctp_id2assoc(sk, associd);
4239 if (!asoc) {
4240 retval = -EINVAL;
4241 goto out;
4244 transport = asoc->peer.primary_path;
4246 status.sstat_assoc_id = sctp_assoc2id(asoc);
4247 status.sstat_state = sctp_assoc_to_state(asoc);
4248 status.sstat_rwnd = asoc->peer.rwnd;
4249 status.sstat_unackdata = asoc->unack_data;
4251 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4252 status.sstat_instrms = asoc->c.sinit_max_instreams;
4253 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4254 status.sstat_fragmentation_point = asoc->frag_point;
4255 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4256 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4257 transport->af_specific->sockaddr_len);
4258 /* Map ipv4 address into v4-mapped-on-v6 address. */
4259 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4260 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4261 status.sstat_primary.spinfo_state = transport->state;
4262 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4263 status.sstat_primary.spinfo_srtt = transport->srtt;
4264 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4265 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4267 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4268 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4270 if (put_user(len, optlen)) {
4271 retval = -EFAULT;
4272 goto out;
4275 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4276 __func__, len, status.sstat_state, status.sstat_rwnd,
4277 status.sstat_assoc_id);
4279 if (copy_to_user(optval, &status, len)) {
4280 retval = -EFAULT;
4281 goto out;
4284 out:
4285 return retval;
4289 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4291 * Applications can retrieve information about a specific peer address
4292 * of an association, including its reachability state, congestion
4293 * window, and retransmission timer values. This information is
4294 * read-only.
4296 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4297 char __user *optval,
4298 int __user *optlen)
4300 struct sctp_paddrinfo pinfo;
4301 struct sctp_transport *transport;
4302 int retval = 0;
4304 if (len < sizeof(pinfo)) {
4305 retval = -EINVAL;
4306 goto out;
4309 len = sizeof(pinfo);
4310 if (copy_from_user(&pinfo, optval, len)) {
4311 retval = -EFAULT;
4312 goto out;
4315 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4316 pinfo.spinfo_assoc_id);
4317 if (!transport)
4318 return -EINVAL;
4320 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4321 pinfo.spinfo_state = transport->state;
4322 pinfo.spinfo_cwnd = transport->cwnd;
4323 pinfo.spinfo_srtt = transport->srtt;
4324 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4325 pinfo.spinfo_mtu = transport->pathmtu;
4327 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4328 pinfo.spinfo_state = SCTP_ACTIVE;
4330 if (put_user(len, optlen)) {
4331 retval = -EFAULT;
4332 goto out;
4335 if (copy_to_user(optval, &pinfo, len)) {
4336 retval = -EFAULT;
4337 goto out;
4340 out:
4341 return retval;
4344 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4346 * This option is a on/off flag. If enabled no SCTP message
4347 * fragmentation will be performed. Instead if a message being sent
4348 * exceeds the current PMTU size, the message will NOT be sent and
4349 * instead a error will be indicated to the user.
4351 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4352 char __user *optval, int __user *optlen)
4354 int val;
4356 if (len < sizeof(int))
4357 return -EINVAL;
4359 len = sizeof(int);
4360 val = (sctp_sk(sk)->disable_fragments == 1);
4361 if (put_user(len, optlen))
4362 return -EFAULT;
4363 if (copy_to_user(optval, &val, len))
4364 return -EFAULT;
4365 return 0;
4368 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4370 * This socket option is used to specify various notifications and
4371 * ancillary data the user wishes to receive.
4373 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4374 int __user *optlen)
4376 if (len <= 0)
4377 return -EINVAL;
4378 if (len > sizeof(struct sctp_event_subscribe))
4379 len = sizeof(struct sctp_event_subscribe);
4380 if (put_user(len, optlen))
4381 return -EFAULT;
4382 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4383 return -EFAULT;
4384 return 0;
4387 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4389 * This socket option is applicable to the UDP-style socket only. When
4390 * set it will cause associations that are idle for more than the
4391 * specified number of seconds to automatically close. An association
4392 * being idle is defined an association that has NOT sent or received
4393 * user data. The special value of '0' indicates that no automatic
4394 * close of any associations should be performed. The option expects an
4395 * integer defining the number of seconds of idle time before an
4396 * association is closed.
4398 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4400 /* Applicable to UDP-style socket only */
4401 if (sctp_style(sk, TCP))
4402 return -EOPNOTSUPP;
4403 if (len < sizeof(int))
4404 return -EINVAL;
4405 len = sizeof(int);
4406 if (put_user(len, optlen))
4407 return -EFAULT;
4408 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4409 return -EFAULT;
4410 return 0;
4413 /* Helper routine to branch off an association to a new socket. */
4414 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4416 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4417 struct sctp_sock *sp = sctp_sk(sk);
4418 struct socket *sock;
4419 int err = 0;
4421 if (!asoc)
4422 return -EINVAL;
4424 /* An association cannot be branched off from an already peeled-off
4425 * socket, nor is this supported for tcp style sockets.
4427 if (!sctp_style(sk, UDP))
4428 return -EINVAL;
4430 /* Create a new socket. */
4431 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4432 if (err < 0)
4433 return err;
4435 sctp_copy_sock(sock->sk, sk, asoc);
4437 /* Make peeled-off sockets more like 1-1 accepted sockets.
4438 * Set the daddr and initialize id to something more random
4440 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4442 /* Populate the fields of the newsk from the oldsk and migrate the
4443 * asoc to the newsk.
4445 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4447 *sockp = sock;
4449 return err;
4451 EXPORT_SYMBOL(sctp_do_peeloff);
4453 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4455 sctp_peeloff_arg_t peeloff;
4456 struct socket *newsock;
4457 struct file *newfile;
4458 int retval = 0;
4460 if (len < sizeof(sctp_peeloff_arg_t))
4461 return -EINVAL;
4462 len = sizeof(sctp_peeloff_arg_t);
4463 if (copy_from_user(&peeloff, optval, len))
4464 return -EFAULT;
4466 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4467 if (retval < 0)
4468 goto out;
4470 /* Map the socket to an unused fd that can be returned to the user. */
4471 retval = get_unused_fd_flags(0);
4472 if (retval < 0) {
4473 sock_release(newsock);
4474 goto out;
4477 newfile = sock_alloc_file(newsock, 0, NULL);
4478 if (IS_ERR(newfile)) {
4479 put_unused_fd(retval);
4480 sock_release(newsock);
4481 return PTR_ERR(newfile);
4484 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4485 retval);
4487 /* Return the fd mapped to the new socket. */
4488 if (put_user(len, optlen)) {
4489 fput(newfile);
4490 put_unused_fd(retval);
4491 return -EFAULT;
4493 peeloff.sd = retval;
4494 if (copy_to_user(optval, &peeloff, len)) {
4495 fput(newfile);
4496 put_unused_fd(retval);
4497 return -EFAULT;
4499 fd_install(retval, newfile);
4500 out:
4501 return retval;
4504 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4506 * Applications can enable or disable heartbeats for any peer address of
4507 * an association, modify an address's heartbeat interval, force a
4508 * heartbeat to be sent immediately, and adjust the address's maximum
4509 * number of retransmissions sent before an address is considered
4510 * unreachable. The following structure is used to access and modify an
4511 * address's parameters:
4513 * struct sctp_paddrparams {
4514 * sctp_assoc_t spp_assoc_id;
4515 * struct sockaddr_storage spp_address;
4516 * uint32_t spp_hbinterval;
4517 * uint16_t spp_pathmaxrxt;
4518 * uint32_t spp_pathmtu;
4519 * uint32_t spp_sackdelay;
4520 * uint32_t spp_flags;
4521 * };
4523 * spp_assoc_id - (one-to-many style socket) This is filled in the
4524 * application, and identifies the association for
4525 * this query.
4526 * spp_address - This specifies which address is of interest.
4527 * spp_hbinterval - This contains the value of the heartbeat interval,
4528 * in milliseconds. If a value of zero
4529 * is present in this field then no changes are to
4530 * be made to this parameter.
4531 * spp_pathmaxrxt - This contains the maximum number of
4532 * retransmissions before this address shall be
4533 * considered unreachable. If a value of zero
4534 * is present in this field then no changes are to
4535 * be made to this parameter.
4536 * spp_pathmtu - When Path MTU discovery is disabled the value
4537 * specified here will be the "fixed" path mtu.
4538 * Note that if the spp_address field is empty
4539 * then all associations on this address will
4540 * have this fixed path mtu set upon them.
4542 * spp_sackdelay - When delayed sack is enabled, this value specifies
4543 * the number of milliseconds that sacks will be delayed
4544 * for. This value will apply to all addresses of an
4545 * association if the spp_address field is empty. Note
4546 * also, that if delayed sack is enabled and this
4547 * value is set to 0, no change is made to the last
4548 * recorded delayed sack timer value.
4550 * spp_flags - These flags are used to control various features
4551 * on an association. The flag field may contain
4552 * zero or more of the following options.
4554 * SPP_HB_ENABLE - Enable heartbeats on the
4555 * specified address. Note that if the address
4556 * field is empty all addresses for the association
4557 * have heartbeats enabled upon them.
4559 * SPP_HB_DISABLE - Disable heartbeats on the
4560 * speicifed address. Note that if the address
4561 * field is empty all addresses for the association
4562 * will have their heartbeats disabled. Note also
4563 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4564 * mutually exclusive, only one of these two should
4565 * be specified. Enabling both fields will have
4566 * undetermined results.
4568 * SPP_HB_DEMAND - Request a user initiated heartbeat
4569 * to be made immediately.
4571 * SPP_PMTUD_ENABLE - This field will enable PMTU
4572 * discovery upon the specified address. Note that
4573 * if the address feild is empty then all addresses
4574 * on the association are effected.
4576 * SPP_PMTUD_DISABLE - This field will disable PMTU
4577 * discovery upon the specified address. Note that
4578 * if the address feild is empty then all addresses
4579 * on the association are effected. Not also that
4580 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4581 * exclusive. Enabling both will have undetermined
4582 * results.
4584 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4585 * on delayed sack. The time specified in spp_sackdelay
4586 * is used to specify the sack delay for this address. Note
4587 * that if spp_address is empty then all addresses will
4588 * enable delayed sack and take on the sack delay
4589 * value specified in spp_sackdelay.
4590 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4591 * off delayed sack. If the spp_address field is blank then
4592 * delayed sack is disabled for the entire association. Note
4593 * also that this field is mutually exclusive to
4594 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4595 * results.
4597 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4598 char __user *optval, int __user *optlen)
4600 struct sctp_paddrparams params;
4601 struct sctp_transport *trans = NULL;
4602 struct sctp_association *asoc = NULL;
4603 struct sctp_sock *sp = sctp_sk(sk);
4605 if (len < sizeof(struct sctp_paddrparams))
4606 return -EINVAL;
4607 len = sizeof(struct sctp_paddrparams);
4608 if (copy_from_user(&params, optval, len))
4609 return -EFAULT;
4611 /* If an address other than INADDR_ANY is specified, and
4612 * no transport is found, then the request is invalid.
4614 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4615 trans = sctp_addr_id2transport(sk, &params.spp_address,
4616 params.spp_assoc_id);
4617 if (!trans) {
4618 pr_debug("%s: failed no transport\n", __func__);
4619 return -EINVAL;
4623 /* Get association, if assoc_id != 0 and the socket is a one
4624 * to many style socket, and an association was not found, then
4625 * the id was invalid.
4627 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4628 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4629 pr_debug("%s: failed no association\n", __func__);
4630 return -EINVAL;
4633 if (trans) {
4634 /* Fetch transport values. */
4635 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4636 params.spp_pathmtu = trans->pathmtu;
4637 params.spp_pathmaxrxt = trans->pathmaxrxt;
4638 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4640 /*draft-11 doesn't say what to return in spp_flags*/
4641 params.spp_flags = trans->param_flags;
4642 } else if (asoc) {
4643 /* Fetch association values. */
4644 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4645 params.spp_pathmtu = asoc->pathmtu;
4646 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4647 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4649 /*draft-11 doesn't say what to return in spp_flags*/
4650 params.spp_flags = asoc->param_flags;
4651 } else {
4652 /* Fetch socket values. */
4653 params.spp_hbinterval = sp->hbinterval;
4654 params.spp_pathmtu = sp->pathmtu;
4655 params.spp_sackdelay = sp->sackdelay;
4656 params.spp_pathmaxrxt = sp->pathmaxrxt;
4658 /*draft-11 doesn't say what to return in spp_flags*/
4659 params.spp_flags = sp->param_flags;
4662 if (copy_to_user(optval, &params, len))
4663 return -EFAULT;
4665 if (put_user(len, optlen))
4666 return -EFAULT;
4668 return 0;
4672 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4674 * This option will effect the way delayed acks are performed. This
4675 * option allows you to get or set the delayed ack time, in
4676 * milliseconds. It also allows changing the delayed ack frequency.
4677 * Changing the frequency to 1 disables the delayed sack algorithm. If
4678 * the assoc_id is 0, then this sets or gets the endpoints default
4679 * values. If the assoc_id field is non-zero, then the set or get
4680 * effects the specified association for the one to many model (the
4681 * assoc_id field is ignored by the one to one model). Note that if
4682 * sack_delay or sack_freq are 0 when setting this option, then the
4683 * current values will remain unchanged.
4685 * struct sctp_sack_info {
4686 * sctp_assoc_t sack_assoc_id;
4687 * uint32_t sack_delay;
4688 * uint32_t sack_freq;
4689 * };
4691 * sack_assoc_id - This parameter, indicates which association the user
4692 * is performing an action upon. Note that if this field's value is
4693 * zero then the endpoints default value is changed (effecting future
4694 * associations only).
4696 * sack_delay - This parameter contains the number of milliseconds that
4697 * the user is requesting the delayed ACK timer be set to. Note that
4698 * this value is defined in the standard to be between 200 and 500
4699 * milliseconds.
4701 * sack_freq - This parameter contains the number of packets that must
4702 * be received before a sack is sent without waiting for the delay
4703 * timer to expire. The default value for this is 2, setting this
4704 * value to 1 will disable the delayed sack algorithm.
4706 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4707 char __user *optval,
4708 int __user *optlen)
4710 struct sctp_sack_info params;
4711 struct sctp_association *asoc = NULL;
4712 struct sctp_sock *sp = sctp_sk(sk);
4714 if (len >= sizeof(struct sctp_sack_info)) {
4715 len = sizeof(struct sctp_sack_info);
4717 if (copy_from_user(&params, optval, len))
4718 return -EFAULT;
4719 } else if (len == sizeof(struct sctp_assoc_value)) {
4720 pr_warn_ratelimited(DEPRECATED
4721 "%s (pid %d) "
4722 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4723 "Use struct sctp_sack_info instead\n",
4724 current->comm, task_pid_nr(current));
4725 if (copy_from_user(&params, optval, len))
4726 return -EFAULT;
4727 } else
4728 return -EINVAL;
4730 /* Get association, if sack_assoc_id != 0 and the socket is a one
4731 * to many style socket, and an association was not found, then
4732 * the id was invalid.
4734 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4735 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4736 return -EINVAL;
4738 if (asoc) {
4739 /* Fetch association values. */
4740 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4741 params.sack_delay = jiffies_to_msecs(
4742 asoc->sackdelay);
4743 params.sack_freq = asoc->sackfreq;
4745 } else {
4746 params.sack_delay = 0;
4747 params.sack_freq = 1;
4749 } else {
4750 /* Fetch socket values. */
4751 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4752 params.sack_delay = sp->sackdelay;
4753 params.sack_freq = sp->sackfreq;
4754 } else {
4755 params.sack_delay = 0;
4756 params.sack_freq = 1;
4760 if (copy_to_user(optval, &params, len))
4761 return -EFAULT;
4763 if (put_user(len, optlen))
4764 return -EFAULT;
4766 return 0;
4769 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4771 * Applications can specify protocol parameters for the default association
4772 * initialization. The option name argument to setsockopt() and getsockopt()
4773 * is SCTP_INITMSG.
4775 * Setting initialization parameters is effective only on an unconnected
4776 * socket (for UDP-style sockets only future associations are effected
4777 * by the change). With TCP-style sockets, this option is inherited by
4778 * sockets derived from a listener socket.
4780 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4782 if (len < sizeof(struct sctp_initmsg))
4783 return -EINVAL;
4784 len = sizeof(struct sctp_initmsg);
4785 if (put_user(len, optlen))
4786 return -EFAULT;
4787 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4788 return -EFAULT;
4789 return 0;
4793 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4794 char __user *optval, int __user *optlen)
4796 struct sctp_association *asoc;
4797 int cnt = 0;
4798 struct sctp_getaddrs getaddrs;
4799 struct sctp_transport *from;
4800 void __user *to;
4801 union sctp_addr temp;
4802 struct sctp_sock *sp = sctp_sk(sk);
4803 int addrlen;
4804 size_t space_left;
4805 int bytes_copied;
4807 if (len < sizeof(struct sctp_getaddrs))
4808 return -EINVAL;
4810 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4811 return -EFAULT;
4813 /* For UDP-style sockets, id specifies the association to query. */
4814 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4815 if (!asoc)
4816 return -EINVAL;
4818 to = optval + offsetof(struct sctp_getaddrs, addrs);
4819 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4821 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4822 transports) {
4823 memcpy(&temp, &from->ipaddr, sizeof(temp));
4824 addrlen = sctp_get_pf_specific(sk->sk_family)
4825 ->addr_to_user(sp, &temp);
4826 if (space_left < addrlen)
4827 return -ENOMEM;
4828 if (copy_to_user(to, &temp, addrlen))
4829 return -EFAULT;
4830 to += addrlen;
4831 cnt++;
4832 space_left -= addrlen;
4835 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4836 return -EFAULT;
4837 bytes_copied = ((char __user *)to) - optval;
4838 if (put_user(bytes_copied, optlen))
4839 return -EFAULT;
4841 return 0;
4844 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4845 size_t space_left, int *bytes_copied)
4847 struct sctp_sockaddr_entry *addr;
4848 union sctp_addr temp;
4849 int cnt = 0;
4850 int addrlen;
4851 struct net *net = sock_net(sk);
4853 rcu_read_lock();
4854 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4855 if (!addr->valid)
4856 continue;
4858 if ((PF_INET == sk->sk_family) &&
4859 (AF_INET6 == addr->a.sa.sa_family))
4860 continue;
4861 if ((PF_INET6 == sk->sk_family) &&
4862 inet_v6_ipv6only(sk) &&
4863 (AF_INET == addr->a.sa.sa_family))
4864 continue;
4865 memcpy(&temp, &addr->a, sizeof(temp));
4866 if (!temp.v4.sin_port)
4867 temp.v4.sin_port = htons(port);
4869 addrlen = sctp_get_pf_specific(sk->sk_family)
4870 ->addr_to_user(sctp_sk(sk), &temp);
4872 if (space_left < addrlen) {
4873 cnt = -ENOMEM;
4874 break;
4876 memcpy(to, &temp, addrlen);
4878 to += addrlen;
4879 cnt++;
4880 space_left -= addrlen;
4881 *bytes_copied += addrlen;
4883 rcu_read_unlock();
4885 return cnt;
4889 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4890 char __user *optval, int __user *optlen)
4892 struct sctp_bind_addr *bp;
4893 struct sctp_association *asoc;
4894 int cnt = 0;
4895 struct sctp_getaddrs getaddrs;
4896 struct sctp_sockaddr_entry *addr;
4897 void __user *to;
4898 union sctp_addr temp;
4899 struct sctp_sock *sp = sctp_sk(sk);
4900 int addrlen;
4901 int err = 0;
4902 size_t space_left;
4903 int bytes_copied = 0;
4904 void *addrs;
4905 void *buf;
4907 if (len < sizeof(struct sctp_getaddrs))
4908 return -EINVAL;
4910 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4911 return -EFAULT;
4914 * For UDP-style sockets, id specifies the association to query.
4915 * If the id field is set to the value '0' then the locally bound
4916 * addresses are returned without regard to any particular
4917 * association.
4919 if (0 == getaddrs.assoc_id) {
4920 bp = &sctp_sk(sk)->ep->base.bind_addr;
4921 } else {
4922 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4923 if (!asoc)
4924 return -EINVAL;
4925 bp = &asoc->base.bind_addr;
4928 to = optval + offsetof(struct sctp_getaddrs, addrs);
4929 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4931 addrs = kmalloc(space_left, GFP_KERNEL);
4932 if (!addrs)
4933 return -ENOMEM;
4935 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4936 * addresses from the global local address list.
4938 if (sctp_list_single_entry(&bp->address_list)) {
4939 addr = list_entry(bp->address_list.next,
4940 struct sctp_sockaddr_entry, list);
4941 if (sctp_is_any(sk, &addr->a)) {
4942 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4943 space_left, &bytes_copied);
4944 if (cnt < 0) {
4945 err = cnt;
4946 goto out;
4948 goto copy_getaddrs;
4952 buf = addrs;
4953 /* Protection on the bound address list is not needed since
4954 * in the socket option context we hold a socket lock and
4955 * thus the bound address list can't change.
4957 list_for_each_entry(addr, &bp->address_list, list) {
4958 memcpy(&temp, &addr->a, sizeof(temp));
4959 addrlen = sctp_get_pf_specific(sk->sk_family)
4960 ->addr_to_user(sp, &temp);
4961 if (space_left < addrlen) {
4962 err = -ENOMEM; /*fixme: right error?*/
4963 goto out;
4965 memcpy(buf, &temp, addrlen);
4966 buf += addrlen;
4967 bytes_copied += addrlen;
4968 cnt++;
4969 space_left -= addrlen;
4972 copy_getaddrs:
4973 if (copy_to_user(to, addrs, bytes_copied)) {
4974 err = -EFAULT;
4975 goto out;
4977 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4978 err = -EFAULT;
4979 goto out;
4981 if (put_user(bytes_copied, optlen))
4982 err = -EFAULT;
4983 out:
4984 kfree(addrs);
4985 return err;
4988 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4990 * Requests that the local SCTP stack use the enclosed peer address as
4991 * the association primary. The enclosed address must be one of the
4992 * association peer's addresses.
4994 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4995 char __user *optval, int __user *optlen)
4997 struct sctp_prim prim;
4998 struct sctp_association *asoc;
4999 struct sctp_sock *sp = sctp_sk(sk);
5001 if (len < sizeof(struct sctp_prim))
5002 return -EINVAL;
5004 len = sizeof(struct sctp_prim);
5006 if (copy_from_user(&prim, optval, len))
5007 return -EFAULT;
5009 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5010 if (!asoc)
5011 return -EINVAL;
5013 if (!asoc->peer.primary_path)
5014 return -ENOTCONN;
5016 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5017 asoc->peer.primary_path->af_specific->sockaddr_len);
5019 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5020 (union sctp_addr *)&prim.ssp_addr);
5022 if (put_user(len, optlen))
5023 return -EFAULT;
5024 if (copy_to_user(optval, &prim, len))
5025 return -EFAULT;
5027 return 0;
5031 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5033 * Requests that the local endpoint set the specified Adaptation Layer
5034 * Indication parameter for all future INIT and INIT-ACK exchanges.
5036 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5037 char __user *optval, int __user *optlen)
5039 struct sctp_setadaptation adaptation;
5041 if (len < sizeof(struct sctp_setadaptation))
5042 return -EINVAL;
5044 len = sizeof(struct sctp_setadaptation);
5046 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5048 if (put_user(len, optlen))
5049 return -EFAULT;
5050 if (copy_to_user(optval, &adaptation, len))
5051 return -EFAULT;
5053 return 0;
5058 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5060 * Applications that wish to use the sendto() system call may wish to
5061 * specify a default set of parameters that would normally be supplied
5062 * through the inclusion of ancillary data. This socket option allows
5063 * such an application to set the default sctp_sndrcvinfo structure.
5066 * The application that wishes to use this socket option simply passes
5067 * in to this call the sctp_sndrcvinfo structure defined in Section
5068 * 5.2.2) The input parameters accepted by this call include
5069 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5070 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5071 * to this call if the caller is using the UDP model.
5073 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5075 static int sctp_getsockopt_default_send_param(struct sock *sk,
5076 int len, char __user *optval,
5077 int __user *optlen)
5079 struct sctp_sock *sp = sctp_sk(sk);
5080 struct sctp_association *asoc;
5081 struct sctp_sndrcvinfo info;
5083 if (len < sizeof(info))
5084 return -EINVAL;
5086 len = sizeof(info);
5088 if (copy_from_user(&info, optval, len))
5089 return -EFAULT;
5091 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5092 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5093 return -EINVAL;
5094 if (asoc) {
5095 info.sinfo_stream = asoc->default_stream;
5096 info.sinfo_flags = asoc->default_flags;
5097 info.sinfo_ppid = asoc->default_ppid;
5098 info.sinfo_context = asoc->default_context;
5099 info.sinfo_timetolive = asoc->default_timetolive;
5100 } else {
5101 info.sinfo_stream = sp->default_stream;
5102 info.sinfo_flags = sp->default_flags;
5103 info.sinfo_ppid = sp->default_ppid;
5104 info.sinfo_context = sp->default_context;
5105 info.sinfo_timetolive = sp->default_timetolive;
5108 if (put_user(len, optlen))
5109 return -EFAULT;
5110 if (copy_to_user(optval, &info, len))
5111 return -EFAULT;
5113 return 0;
5116 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5117 * (SCTP_DEFAULT_SNDINFO)
5119 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5120 char __user *optval,
5121 int __user *optlen)
5123 struct sctp_sock *sp = sctp_sk(sk);
5124 struct sctp_association *asoc;
5125 struct sctp_sndinfo info;
5127 if (len < sizeof(info))
5128 return -EINVAL;
5130 len = sizeof(info);
5132 if (copy_from_user(&info, optval, len))
5133 return -EFAULT;
5135 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5136 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5137 return -EINVAL;
5138 if (asoc) {
5139 info.snd_sid = asoc->default_stream;
5140 info.snd_flags = asoc->default_flags;
5141 info.snd_ppid = asoc->default_ppid;
5142 info.snd_context = asoc->default_context;
5143 } else {
5144 info.snd_sid = sp->default_stream;
5145 info.snd_flags = sp->default_flags;
5146 info.snd_ppid = sp->default_ppid;
5147 info.snd_context = sp->default_context;
5150 if (put_user(len, optlen))
5151 return -EFAULT;
5152 if (copy_to_user(optval, &info, len))
5153 return -EFAULT;
5155 return 0;
5160 * 7.1.5 SCTP_NODELAY
5162 * Turn on/off any Nagle-like algorithm. This means that packets are
5163 * generally sent as soon as possible and no unnecessary delays are
5164 * introduced, at the cost of more packets in the network. Expects an
5165 * integer boolean flag.
5168 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5169 char __user *optval, int __user *optlen)
5171 int val;
5173 if (len < sizeof(int))
5174 return -EINVAL;
5176 len = sizeof(int);
5177 val = (sctp_sk(sk)->nodelay == 1);
5178 if (put_user(len, optlen))
5179 return -EFAULT;
5180 if (copy_to_user(optval, &val, len))
5181 return -EFAULT;
5182 return 0;
5187 * 7.1.1 SCTP_RTOINFO
5189 * The protocol parameters used to initialize and bound retransmission
5190 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5191 * and modify these parameters.
5192 * All parameters are time values, in milliseconds. A value of 0, when
5193 * modifying the parameters, indicates that the current value should not
5194 * be changed.
5197 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5198 char __user *optval,
5199 int __user *optlen) {
5200 struct sctp_rtoinfo rtoinfo;
5201 struct sctp_association *asoc;
5203 if (len < sizeof (struct sctp_rtoinfo))
5204 return -EINVAL;
5206 len = sizeof(struct sctp_rtoinfo);
5208 if (copy_from_user(&rtoinfo, optval, len))
5209 return -EFAULT;
5211 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5213 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5214 return -EINVAL;
5216 /* Values corresponding to the specific association. */
5217 if (asoc) {
5218 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5219 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5220 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5221 } else {
5222 /* Values corresponding to the endpoint. */
5223 struct sctp_sock *sp = sctp_sk(sk);
5225 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5226 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5227 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5230 if (put_user(len, optlen))
5231 return -EFAULT;
5233 if (copy_to_user(optval, &rtoinfo, len))
5234 return -EFAULT;
5236 return 0;
5241 * 7.1.2 SCTP_ASSOCINFO
5243 * This option is used to tune the maximum retransmission attempts
5244 * of the association.
5245 * Returns an error if the new association retransmission value is
5246 * greater than the sum of the retransmission value of the peer.
5247 * See [SCTP] for more information.
5250 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5251 char __user *optval,
5252 int __user *optlen)
5255 struct sctp_assocparams assocparams;
5256 struct sctp_association *asoc;
5257 struct list_head *pos;
5258 int cnt = 0;
5260 if (len < sizeof (struct sctp_assocparams))
5261 return -EINVAL;
5263 len = sizeof(struct sctp_assocparams);
5265 if (copy_from_user(&assocparams, optval, len))
5266 return -EFAULT;
5268 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5270 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5271 return -EINVAL;
5273 /* Values correspoinding to the specific association */
5274 if (asoc) {
5275 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5276 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5277 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5278 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5280 list_for_each(pos, &asoc->peer.transport_addr_list) {
5281 cnt++;
5284 assocparams.sasoc_number_peer_destinations = cnt;
5285 } else {
5286 /* Values corresponding to the endpoint */
5287 struct sctp_sock *sp = sctp_sk(sk);
5289 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5290 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5291 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5292 assocparams.sasoc_cookie_life =
5293 sp->assocparams.sasoc_cookie_life;
5294 assocparams.sasoc_number_peer_destinations =
5295 sp->assocparams.
5296 sasoc_number_peer_destinations;
5299 if (put_user(len, optlen))
5300 return -EFAULT;
5302 if (copy_to_user(optval, &assocparams, len))
5303 return -EFAULT;
5305 return 0;
5309 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5311 * This socket option is a boolean flag which turns on or off mapped V4
5312 * addresses. If this option is turned on and the socket is type
5313 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5314 * If this option is turned off, then no mapping will be done of V4
5315 * addresses and a user will receive both PF_INET6 and PF_INET type
5316 * addresses on the socket.
5318 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5319 char __user *optval, int __user *optlen)
5321 int val;
5322 struct sctp_sock *sp = sctp_sk(sk);
5324 if (len < sizeof(int))
5325 return -EINVAL;
5327 len = sizeof(int);
5328 val = sp->v4mapped;
5329 if (put_user(len, optlen))
5330 return -EFAULT;
5331 if (copy_to_user(optval, &val, len))
5332 return -EFAULT;
5334 return 0;
5338 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5339 * (chapter and verse is quoted at sctp_setsockopt_context())
5341 static int sctp_getsockopt_context(struct sock *sk, int len,
5342 char __user *optval, int __user *optlen)
5344 struct sctp_assoc_value params;
5345 struct sctp_sock *sp;
5346 struct sctp_association *asoc;
5348 if (len < sizeof(struct sctp_assoc_value))
5349 return -EINVAL;
5351 len = sizeof(struct sctp_assoc_value);
5353 if (copy_from_user(&params, optval, len))
5354 return -EFAULT;
5356 sp = sctp_sk(sk);
5358 if (params.assoc_id != 0) {
5359 asoc = sctp_id2assoc(sk, params.assoc_id);
5360 if (!asoc)
5361 return -EINVAL;
5362 params.assoc_value = asoc->default_rcv_context;
5363 } else {
5364 params.assoc_value = sp->default_rcv_context;
5367 if (put_user(len, optlen))
5368 return -EFAULT;
5369 if (copy_to_user(optval, &params, len))
5370 return -EFAULT;
5372 return 0;
5376 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5377 * This option will get or set the maximum size to put in any outgoing
5378 * SCTP DATA chunk. If a message is larger than this size it will be
5379 * fragmented by SCTP into the specified size. Note that the underlying
5380 * SCTP implementation may fragment into smaller sized chunks when the
5381 * PMTU of the underlying association is smaller than the value set by
5382 * the user. The default value for this option is '0' which indicates
5383 * the user is NOT limiting fragmentation and only the PMTU will effect
5384 * SCTP's choice of DATA chunk size. Note also that values set larger
5385 * than the maximum size of an IP datagram will effectively let SCTP
5386 * control fragmentation (i.e. the same as setting this option to 0).
5388 * The following structure is used to access and modify this parameter:
5390 * struct sctp_assoc_value {
5391 * sctp_assoc_t assoc_id;
5392 * uint32_t assoc_value;
5393 * };
5395 * assoc_id: This parameter is ignored for one-to-one style sockets.
5396 * For one-to-many style sockets this parameter indicates which
5397 * association the user is performing an action upon. Note that if
5398 * this field's value is zero then the endpoints default value is
5399 * changed (effecting future associations only).
5400 * assoc_value: This parameter specifies the maximum size in bytes.
5402 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5403 char __user *optval, int __user *optlen)
5405 struct sctp_assoc_value params;
5406 struct sctp_association *asoc;
5408 if (len == sizeof(int)) {
5409 pr_warn_ratelimited(DEPRECATED
5410 "%s (pid %d) "
5411 "Use of int in maxseg socket option.\n"
5412 "Use struct sctp_assoc_value instead\n",
5413 current->comm, task_pid_nr(current));
5414 params.assoc_id = 0;
5415 } else if (len >= sizeof(struct sctp_assoc_value)) {
5416 len = sizeof(struct sctp_assoc_value);
5417 if (copy_from_user(&params, optval, sizeof(params)))
5418 return -EFAULT;
5419 } else
5420 return -EINVAL;
5422 asoc = sctp_id2assoc(sk, params.assoc_id);
5423 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5424 return -EINVAL;
5426 if (asoc)
5427 params.assoc_value = asoc->frag_point;
5428 else
5429 params.assoc_value = sctp_sk(sk)->user_frag;
5431 if (put_user(len, optlen))
5432 return -EFAULT;
5433 if (len == sizeof(int)) {
5434 if (copy_to_user(optval, &params.assoc_value, len))
5435 return -EFAULT;
5436 } else {
5437 if (copy_to_user(optval, &params, len))
5438 return -EFAULT;
5441 return 0;
5445 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5446 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5448 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5449 char __user *optval, int __user *optlen)
5451 int val;
5453 if (len < sizeof(int))
5454 return -EINVAL;
5456 len = sizeof(int);
5458 val = sctp_sk(sk)->frag_interleave;
5459 if (put_user(len, optlen))
5460 return -EFAULT;
5461 if (copy_to_user(optval, &val, len))
5462 return -EFAULT;
5464 return 0;
5468 * 7.1.25. Set or Get the sctp partial delivery point
5469 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5471 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5472 char __user *optval,
5473 int __user *optlen)
5475 u32 val;
5477 if (len < sizeof(u32))
5478 return -EINVAL;
5480 len = sizeof(u32);
5482 val = sctp_sk(sk)->pd_point;
5483 if (put_user(len, optlen))
5484 return -EFAULT;
5485 if (copy_to_user(optval, &val, len))
5486 return -EFAULT;
5488 return 0;
5492 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5493 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5495 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5496 char __user *optval,
5497 int __user *optlen)
5499 struct sctp_assoc_value params;
5500 struct sctp_sock *sp;
5501 struct sctp_association *asoc;
5503 if (len == sizeof(int)) {
5504 pr_warn_ratelimited(DEPRECATED
5505 "%s (pid %d) "
5506 "Use of int in max_burst socket option.\n"
5507 "Use struct sctp_assoc_value instead\n",
5508 current->comm, task_pid_nr(current));
5509 params.assoc_id = 0;
5510 } else if (len >= sizeof(struct sctp_assoc_value)) {
5511 len = sizeof(struct sctp_assoc_value);
5512 if (copy_from_user(&params, optval, len))
5513 return -EFAULT;
5514 } else
5515 return -EINVAL;
5517 sp = sctp_sk(sk);
5519 if (params.assoc_id != 0) {
5520 asoc = sctp_id2assoc(sk, params.assoc_id);
5521 if (!asoc)
5522 return -EINVAL;
5523 params.assoc_value = asoc->max_burst;
5524 } else
5525 params.assoc_value = sp->max_burst;
5527 if (len == sizeof(int)) {
5528 if (copy_to_user(optval, &params.assoc_value, len))
5529 return -EFAULT;
5530 } else {
5531 if (copy_to_user(optval, &params, len))
5532 return -EFAULT;
5535 return 0;
5539 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5540 char __user *optval, int __user *optlen)
5542 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5543 struct sctp_hmacalgo __user *p = (void __user *)optval;
5544 struct sctp_hmac_algo_param *hmacs;
5545 __u16 data_len = 0;
5546 u32 num_idents;
5548 if (!ep->auth_enable)
5549 return -EACCES;
5551 hmacs = ep->auth_hmacs_list;
5552 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5554 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5555 return -EINVAL;
5557 len = sizeof(struct sctp_hmacalgo) + data_len;
5558 num_idents = data_len / sizeof(u16);
5560 if (put_user(len, optlen))
5561 return -EFAULT;
5562 if (put_user(num_idents, &p->shmac_num_idents))
5563 return -EFAULT;
5564 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5565 return -EFAULT;
5566 return 0;
5569 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5570 char __user *optval, int __user *optlen)
5572 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5573 struct sctp_authkeyid val;
5574 struct sctp_association *asoc;
5576 if (!ep->auth_enable)
5577 return -EACCES;
5579 if (len < sizeof(struct sctp_authkeyid))
5580 return -EINVAL;
5581 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5582 return -EFAULT;
5584 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5585 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5586 return -EINVAL;
5588 if (asoc)
5589 val.scact_keynumber = asoc->active_key_id;
5590 else
5591 val.scact_keynumber = ep->active_key_id;
5593 len = sizeof(struct sctp_authkeyid);
5594 if (put_user(len, optlen))
5595 return -EFAULT;
5596 if (copy_to_user(optval, &val, len))
5597 return -EFAULT;
5599 return 0;
5602 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5603 char __user *optval, int __user *optlen)
5605 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5606 struct sctp_authchunks __user *p = (void __user *)optval;
5607 struct sctp_authchunks val;
5608 struct sctp_association *asoc;
5609 struct sctp_chunks_param *ch;
5610 u32 num_chunks = 0;
5611 char __user *to;
5613 if (!ep->auth_enable)
5614 return -EACCES;
5616 if (len < sizeof(struct sctp_authchunks))
5617 return -EINVAL;
5619 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5620 return -EFAULT;
5622 to = p->gauth_chunks;
5623 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5624 if (!asoc)
5625 return -EINVAL;
5627 ch = asoc->peer.peer_chunks;
5628 if (!ch)
5629 goto num;
5631 /* See if the user provided enough room for all the data */
5632 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5633 if (len < num_chunks)
5634 return -EINVAL;
5636 if (copy_to_user(to, ch->chunks, num_chunks))
5637 return -EFAULT;
5638 num:
5639 len = sizeof(struct sctp_authchunks) + num_chunks;
5640 if (put_user(len, optlen))
5641 return -EFAULT;
5642 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5643 return -EFAULT;
5644 return 0;
5647 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5648 char __user *optval, int __user *optlen)
5650 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5651 struct sctp_authchunks __user *p = (void __user *)optval;
5652 struct sctp_authchunks val;
5653 struct sctp_association *asoc;
5654 struct sctp_chunks_param *ch;
5655 u32 num_chunks = 0;
5656 char __user *to;
5658 if (!ep->auth_enable)
5659 return -EACCES;
5661 if (len < sizeof(struct sctp_authchunks))
5662 return -EINVAL;
5664 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5665 return -EFAULT;
5667 to = p->gauth_chunks;
5668 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5669 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5670 return -EINVAL;
5672 if (asoc)
5673 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5674 else
5675 ch = ep->auth_chunk_list;
5677 if (!ch)
5678 goto num;
5680 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5681 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5682 return -EINVAL;
5684 if (copy_to_user(to, ch->chunks, num_chunks))
5685 return -EFAULT;
5686 num:
5687 len = sizeof(struct sctp_authchunks) + num_chunks;
5688 if (put_user(len, optlen))
5689 return -EFAULT;
5690 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5691 return -EFAULT;
5693 return 0;
5697 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5698 * This option gets the current number of associations that are attached
5699 * to a one-to-many style socket. The option value is an uint32_t.
5701 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5702 char __user *optval, int __user *optlen)
5704 struct sctp_sock *sp = sctp_sk(sk);
5705 struct sctp_association *asoc;
5706 u32 val = 0;
5708 if (sctp_style(sk, TCP))
5709 return -EOPNOTSUPP;
5711 if (len < sizeof(u32))
5712 return -EINVAL;
5714 len = sizeof(u32);
5716 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5717 val++;
5720 if (put_user(len, optlen))
5721 return -EFAULT;
5722 if (copy_to_user(optval, &val, len))
5723 return -EFAULT;
5725 return 0;
5729 * 8.1.23 SCTP_AUTO_ASCONF
5730 * See the corresponding setsockopt entry as description
5732 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5733 char __user *optval, int __user *optlen)
5735 int val = 0;
5737 if (len < sizeof(int))
5738 return -EINVAL;
5740 len = sizeof(int);
5741 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5742 val = 1;
5743 if (put_user(len, optlen))
5744 return -EFAULT;
5745 if (copy_to_user(optval, &val, len))
5746 return -EFAULT;
5747 return 0;
5751 * 8.2.6. Get the Current Identifiers of Associations
5752 * (SCTP_GET_ASSOC_ID_LIST)
5754 * This option gets the current list of SCTP association identifiers of
5755 * the SCTP associations handled by a one-to-many style socket.
5757 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5758 char __user *optval, int __user *optlen)
5760 struct sctp_sock *sp = sctp_sk(sk);
5761 struct sctp_association *asoc;
5762 struct sctp_assoc_ids *ids;
5763 u32 num = 0;
5765 if (sctp_style(sk, TCP))
5766 return -EOPNOTSUPP;
5768 if (len < sizeof(struct sctp_assoc_ids))
5769 return -EINVAL;
5771 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5772 num++;
5775 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5776 return -EINVAL;
5778 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5780 ids = kmalloc(len, GFP_KERNEL);
5781 if (unlikely(!ids))
5782 return -ENOMEM;
5784 ids->gaids_number_of_ids = num;
5785 num = 0;
5786 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5787 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5790 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5791 kfree(ids);
5792 return -EFAULT;
5795 kfree(ids);
5796 return 0;
5800 * SCTP_PEER_ADDR_THLDS
5802 * This option allows us to fetch the partially failed threshold for one or all
5803 * transports in an association. See Section 6.1 of:
5804 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5806 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5807 char __user *optval,
5808 int len,
5809 int __user *optlen)
5811 struct sctp_paddrthlds val;
5812 struct sctp_transport *trans;
5813 struct sctp_association *asoc;
5815 if (len < sizeof(struct sctp_paddrthlds))
5816 return -EINVAL;
5817 len = sizeof(struct sctp_paddrthlds);
5818 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5819 return -EFAULT;
5821 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5822 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5823 if (!asoc)
5824 return -ENOENT;
5826 val.spt_pathpfthld = asoc->pf_retrans;
5827 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5828 } else {
5829 trans = sctp_addr_id2transport(sk, &val.spt_address,
5830 val.spt_assoc_id);
5831 if (!trans)
5832 return -ENOENT;
5834 val.spt_pathmaxrxt = trans->pathmaxrxt;
5835 val.spt_pathpfthld = trans->pf_retrans;
5838 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5839 return -EFAULT;
5841 return 0;
5845 * SCTP_GET_ASSOC_STATS
5847 * This option retrieves local per endpoint statistics. It is modeled
5848 * after OpenSolaris' implementation
5850 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5851 char __user *optval,
5852 int __user *optlen)
5854 struct sctp_assoc_stats sas;
5855 struct sctp_association *asoc = NULL;
5857 /* User must provide at least the assoc id */
5858 if (len < sizeof(sctp_assoc_t))
5859 return -EINVAL;
5861 /* Allow the struct to grow and fill in as much as possible */
5862 len = min_t(size_t, len, sizeof(sas));
5864 if (copy_from_user(&sas, optval, len))
5865 return -EFAULT;
5867 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5868 if (!asoc)
5869 return -EINVAL;
5871 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5872 sas.sas_gapcnt = asoc->stats.gapcnt;
5873 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5874 sas.sas_osacks = asoc->stats.osacks;
5875 sas.sas_isacks = asoc->stats.isacks;
5876 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5877 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5878 sas.sas_oodchunks = asoc->stats.oodchunks;
5879 sas.sas_iodchunks = asoc->stats.iodchunks;
5880 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5881 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5882 sas.sas_idupchunks = asoc->stats.idupchunks;
5883 sas.sas_opackets = asoc->stats.opackets;
5884 sas.sas_ipackets = asoc->stats.ipackets;
5886 /* New high max rto observed, will return 0 if not a single
5887 * RTO update took place. obs_rto_ipaddr will be bogus
5888 * in such a case
5890 sas.sas_maxrto = asoc->stats.max_obs_rto;
5891 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5892 sizeof(struct sockaddr_storage));
5894 /* Mark beginning of a new observation period */
5895 asoc->stats.max_obs_rto = asoc->rto_min;
5897 if (put_user(len, optlen))
5898 return -EFAULT;
5900 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5902 if (copy_to_user(optval, &sas, len))
5903 return -EFAULT;
5905 return 0;
5908 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
5909 char __user *optval,
5910 int __user *optlen)
5912 int val = 0;
5914 if (len < sizeof(int))
5915 return -EINVAL;
5917 len = sizeof(int);
5918 if (sctp_sk(sk)->recvrcvinfo)
5919 val = 1;
5920 if (put_user(len, optlen))
5921 return -EFAULT;
5922 if (copy_to_user(optval, &val, len))
5923 return -EFAULT;
5925 return 0;
5928 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
5929 char __user *optval,
5930 int __user *optlen)
5932 int val = 0;
5934 if (len < sizeof(int))
5935 return -EINVAL;
5937 len = sizeof(int);
5938 if (sctp_sk(sk)->recvnxtinfo)
5939 val = 1;
5940 if (put_user(len, optlen))
5941 return -EFAULT;
5942 if (copy_to_user(optval, &val, len))
5943 return -EFAULT;
5945 return 0;
5948 static int sctp_getsockopt(struct sock *sk, int level, int optname,
5949 char __user *optval, int __user *optlen)
5951 int retval = 0;
5952 int len;
5954 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5956 /* I can hardly begin to describe how wrong this is. This is
5957 * so broken as to be worse than useless. The API draft
5958 * REALLY is NOT helpful here... I am not convinced that the
5959 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5960 * are at all well-founded.
5962 if (level != SOL_SCTP) {
5963 struct sctp_af *af = sctp_sk(sk)->pf->af;
5965 retval = af->getsockopt(sk, level, optname, optval, optlen);
5966 return retval;
5969 if (get_user(len, optlen))
5970 return -EFAULT;
5972 lock_sock(sk);
5974 switch (optname) {
5975 case SCTP_STATUS:
5976 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5977 break;
5978 case SCTP_DISABLE_FRAGMENTS:
5979 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5980 optlen);
5981 break;
5982 case SCTP_EVENTS:
5983 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5984 break;
5985 case SCTP_AUTOCLOSE:
5986 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5987 break;
5988 case SCTP_SOCKOPT_PEELOFF:
5989 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5990 break;
5991 case SCTP_PEER_ADDR_PARAMS:
5992 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5993 optlen);
5994 break;
5995 case SCTP_DELAYED_SACK:
5996 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5997 optlen);
5998 break;
5999 case SCTP_INITMSG:
6000 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6001 break;
6002 case SCTP_GET_PEER_ADDRS:
6003 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6004 optlen);
6005 break;
6006 case SCTP_GET_LOCAL_ADDRS:
6007 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6008 optlen);
6009 break;
6010 case SCTP_SOCKOPT_CONNECTX3:
6011 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6012 break;
6013 case SCTP_DEFAULT_SEND_PARAM:
6014 retval = sctp_getsockopt_default_send_param(sk, len,
6015 optval, optlen);
6016 break;
6017 case SCTP_DEFAULT_SNDINFO:
6018 retval = sctp_getsockopt_default_sndinfo(sk, len,
6019 optval, optlen);
6020 break;
6021 case SCTP_PRIMARY_ADDR:
6022 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6023 break;
6024 case SCTP_NODELAY:
6025 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6026 break;
6027 case SCTP_RTOINFO:
6028 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6029 break;
6030 case SCTP_ASSOCINFO:
6031 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6032 break;
6033 case SCTP_I_WANT_MAPPED_V4_ADDR:
6034 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6035 break;
6036 case SCTP_MAXSEG:
6037 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6038 break;
6039 case SCTP_GET_PEER_ADDR_INFO:
6040 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6041 optlen);
6042 break;
6043 case SCTP_ADAPTATION_LAYER:
6044 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6045 optlen);
6046 break;
6047 case SCTP_CONTEXT:
6048 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6049 break;
6050 case SCTP_FRAGMENT_INTERLEAVE:
6051 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6052 optlen);
6053 break;
6054 case SCTP_PARTIAL_DELIVERY_POINT:
6055 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6056 optlen);
6057 break;
6058 case SCTP_MAX_BURST:
6059 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6060 break;
6061 case SCTP_AUTH_KEY:
6062 case SCTP_AUTH_CHUNK:
6063 case SCTP_AUTH_DELETE_KEY:
6064 retval = -EOPNOTSUPP;
6065 break;
6066 case SCTP_HMAC_IDENT:
6067 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6068 break;
6069 case SCTP_AUTH_ACTIVE_KEY:
6070 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6071 break;
6072 case SCTP_PEER_AUTH_CHUNKS:
6073 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6074 optlen);
6075 break;
6076 case SCTP_LOCAL_AUTH_CHUNKS:
6077 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6078 optlen);
6079 break;
6080 case SCTP_GET_ASSOC_NUMBER:
6081 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6082 break;
6083 case SCTP_GET_ASSOC_ID_LIST:
6084 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6085 break;
6086 case SCTP_AUTO_ASCONF:
6087 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6088 break;
6089 case SCTP_PEER_ADDR_THLDS:
6090 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6091 break;
6092 case SCTP_GET_ASSOC_STATS:
6093 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6094 break;
6095 case SCTP_RECVRCVINFO:
6096 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6097 break;
6098 case SCTP_RECVNXTINFO:
6099 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6100 break;
6101 default:
6102 retval = -ENOPROTOOPT;
6103 break;
6106 release_sock(sk);
6107 return retval;
6110 static void sctp_hash(struct sock *sk)
6112 /* STUB */
6115 static void sctp_unhash(struct sock *sk)
6117 /* STUB */
6120 /* Check if port is acceptable. Possibly find first available port.
6122 * The port hash table (contained in the 'global' SCTP protocol storage
6123 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6124 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6125 * list (the list number is the port number hashed out, so as you
6126 * would expect from a hash function, all the ports in a given list have
6127 * such a number that hashes out to the same list number; you were
6128 * expecting that, right?); so each list has a set of ports, with a
6129 * link to the socket (struct sock) that uses it, the port number and
6130 * a fastreuse flag (FIXME: NPI ipg).
6132 static struct sctp_bind_bucket *sctp_bucket_create(
6133 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6135 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6137 struct sctp_bind_hashbucket *head; /* hash list */
6138 struct sctp_bind_bucket *pp;
6139 unsigned short snum;
6140 int ret;
6142 snum = ntohs(addr->v4.sin_port);
6144 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6146 local_bh_disable();
6148 if (snum == 0) {
6149 /* Search for an available port. */
6150 int low, high, remaining, index;
6151 unsigned int rover;
6152 struct net *net = sock_net(sk);
6154 inet_get_local_port_range(net, &low, &high);
6155 remaining = (high - low) + 1;
6156 rover = prandom_u32() % remaining + low;
6158 do {
6159 rover++;
6160 if ((rover < low) || (rover > high))
6161 rover = low;
6162 if (inet_is_local_reserved_port(net, rover))
6163 continue;
6164 index = sctp_phashfn(sock_net(sk), rover);
6165 head = &sctp_port_hashtable[index];
6166 spin_lock(&head->lock);
6167 sctp_for_each_hentry(pp, &head->chain)
6168 if ((pp->port == rover) &&
6169 net_eq(sock_net(sk), pp->net))
6170 goto next;
6171 break;
6172 next:
6173 spin_unlock(&head->lock);
6174 } while (--remaining > 0);
6176 /* Exhausted local port range during search? */
6177 ret = 1;
6178 if (remaining <= 0)
6179 goto fail;
6181 /* OK, here is the one we will use. HEAD (the port
6182 * hash table list entry) is non-NULL and we hold it's
6183 * mutex.
6185 snum = rover;
6186 } else {
6187 /* We are given an specific port number; we verify
6188 * that it is not being used. If it is used, we will
6189 * exahust the search in the hash list corresponding
6190 * to the port number (snum) - we detect that with the
6191 * port iterator, pp being NULL.
6193 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6194 spin_lock(&head->lock);
6195 sctp_for_each_hentry(pp, &head->chain) {
6196 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6197 goto pp_found;
6200 pp = NULL;
6201 goto pp_not_found;
6202 pp_found:
6203 if (!hlist_empty(&pp->owner)) {
6204 /* We had a port hash table hit - there is an
6205 * available port (pp != NULL) and it is being
6206 * used by other socket (pp->owner not empty); that other
6207 * socket is going to be sk2.
6209 int reuse = sk->sk_reuse;
6210 struct sock *sk2;
6212 pr_debug("%s: found a possible match\n", __func__);
6214 if (pp->fastreuse && sk->sk_reuse &&
6215 sk->sk_state != SCTP_SS_LISTENING)
6216 goto success;
6218 /* Run through the list of sockets bound to the port
6219 * (pp->port) [via the pointers bind_next and
6220 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6221 * we get the endpoint they describe and run through
6222 * the endpoint's list of IP (v4 or v6) addresses,
6223 * comparing each of the addresses with the address of
6224 * the socket sk. If we find a match, then that means
6225 * that this port/socket (sk) combination are already
6226 * in an endpoint.
6228 sk_for_each_bound(sk2, &pp->owner) {
6229 struct sctp_endpoint *ep2;
6230 ep2 = sctp_sk(sk2)->ep;
6232 if (sk == sk2 ||
6233 (reuse && sk2->sk_reuse &&
6234 sk2->sk_state != SCTP_SS_LISTENING))
6235 continue;
6237 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6238 sctp_sk(sk2), sctp_sk(sk))) {
6239 ret = (long)sk2;
6240 goto fail_unlock;
6244 pr_debug("%s: found a match\n", __func__);
6246 pp_not_found:
6247 /* If there was a hash table miss, create a new port. */
6248 ret = 1;
6249 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6250 goto fail_unlock;
6252 /* In either case (hit or miss), make sure fastreuse is 1 only
6253 * if sk->sk_reuse is too (that is, if the caller requested
6254 * SO_REUSEADDR on this socket -sk-).
6256 if (hlist_empty(&pp->owner)) {
6257 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6258 pp->fastreuse = 1;
6259 else
6260 pp->fastreuse = 0;
6261 } else if (pp->fastreuse &&
6262 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6263 pp->fastreuse = 0;
6265 /* We are set, so fill up all the data in the hash table
6266 * entry, tie the socket list information with the rest of the
6267 * sockets FIXME: Blurry, NPI (ipg).
6269 success:
6270 if (!sctp_sk(sk)->bind_hash) {
6271 inet_sk(sk)->inet_num = snum;
6272 sk_add_bind_node(sk, &pp->owner);
6273 sctp_sk(sk)->bind_hash = pp;
6275 ret = 0;
6277 fail_unlock:
6278 spin_unlock(&head->lock);
6280 fail:
6281 local_bh_enable();
6282 return ret;
6285 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6286 * port is requested.
6288 static int sctp_get_port(struct sock *sk, unsigned short snum)
6290 union sctp_addr addr;
6291 struct sctp_af *af = sctp_sk(sk)->pf->af;
6293 /* Set up a dummy address struct from the sk. */
6294 af->from_sk(&addr, sk);
6295 addr.v4.sin_port = htons(snum);
6297 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6298 return !!sctp_get_port_local(sk, &addr);
6302 * Move a socket to LISTENING state.
6304 static int sctp_listen_start(struct sock *sk, int backlog)
6306 struct sctp_sock *sp = sctp_sk(sk);
6307 struct sctp_endpoint *ep = sp->ep;
6308 struct crypto_hash *tfm = NULL;
6309 char alg[32];
6311 /* Allocate HMAC for generating cookie. */
6312 if (!sp->hmac && sp->sctp_hmac_alg) {
6313 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6314 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6315 if (IS_ERR(tfm)) {
6316 net_info_ratelimited("failed to load transform for %s: %ld\n",
6317 sp->sctp_hmac_alg, PTR_ERR(tfm));
6318 return -ENOSYS;
6320 sctp_sk(sk)->hmac = tfm;
6324 * If a bind() or sctp_bindx() is not called prior to a listen()
6325 * call that allows new associations to be accepted, the system
6326 * picks an ephemeral port and will choose an address set equivalent
6327 * to binding with a wildcard address.
6329 * This is not currently spelled out in the SCTP sockets
6330 * extensions draft, but follows the practice as seen in TCP
6331 * sockets.
6334 sk->sk_state = SCTP_SS_LISTENING;
6335 if (!ep->base.bind_addr.port) {
6336 if (sctp_autobind(sk))
6337 return -EAGAIN;
6338 } else {
6339 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6340 sk->sk_state = SCTP_SS_CLOSED;
6341 return -EADDRINUSE;
6345 sk->sk_max_ack_backlog = backlog;
6346 sctp_hash_endpoint(ep);
6347 return 0;
6351 * 4.1.3 / 5.1.3 listen()
6353 * By default, new associations are not accepted for UDP style sockets.
6354 * An application uses listen() to mark a socket as being able to
6355 * accept new associations.
6357 * On TCP style sockets, applications use listen() to ready the SCTP
6358 * endpoint for accepting inbound associations.
6360 * On both types of endpoints a backlog of '0' disables listening.
6362 * Move a socket to LISTENING state.
6364 int sctp_inet_listen(struct socket *sock, int backlog)
6366 struct sock *sk = sock->sk;
6367 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6368 int err = -EINVAL;
6370 if (unlikely(backlog < 0))
6371 return err;
6373 lock_sock(sk);
6375 /* Peeled-off sockets are not allowed to listen(). */
6376 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6377 goto out;
6379 if (sock->state != SS_UNCONNECTED)
6380 goto out;
6382 /* If backlog is zero, disable listening. */
6383 if (!backlog) {
6384 if (sctp_sstate(sk, CLOSED))
6385 goto out;
6387 err = 0;
6388 sctp_unhash_endpoint(ep);
6389 sk->sk_state = SCTP_SS_CLOSED;
6390 if (sk->sk_reuse)
6391 sctp_sk(sk)->bind_hash->fastreuse = 1;
6392 goto out;
6395 /* If we are already listening, just update the backlog */
6396 if (sctp_sstate(sk, LISTENING))
6397 sk->sk_max_ack_backlog = backlog;
6398 else {
6399 err = sctp_listen_start(sk, backlog);
6400 if (err)
6401 goto out;
6404 err = 0;
6405 out:
6406 release_sock(sk);
6407 return err;
6411 * This function is done by modeling the current datagram_poll() and the
6412 * tcp_poll(). Note that, based on these implementations, we don't
6413 * lock the socket in this function, even though it seems that,
6414 * ideally, locking or some other mechanisms can be used to ensure
6415 * the integrity of the counters (sndbuf and wmem_alloc) used
6416 * in this place. We assume that we don't need locks either until proven
6417 * otherwise.
6419 * Another thing to note is that we include the Async I/O support
6420 * here, again, by modeling the current TCP/UDP code. We don't have
6421 * a good way to test with it yet.
6423 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6425 struct sock *sk = sock->sk;
6426 struct sctp_sock *sp = sctp_sk(sk);
6427 unsigned int mask;
6429 poll_wait(file, sk_sleep(sk), wait);
6431 /* A TCP-style listening socket becomes readable when the accept queue
6432 * is not empty.
6434 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6435 return (!list_empty(&sp->ep->asocs)) ?
6436 (POLLIN | POLLRDNORM) : 0;
6438 mask = 0;
6440 /* Is there any exceptional events? */
6441 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6442 mask |= POLLERR |
6443 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6444 if (sk->sk_shutdown & RCV_SHUTDOWN)
6445 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6446 if (sk->sk_shutdown == SHUTDOWN_MASK)
6447 mask |= POLLHUP;
6449 /* Is it readable? Reconsider this code with TCP-style support. */
6450 if (!skb_queue_empty(&sk->sk_receive_queue))
6451 mask |= POLLIN | POLLRDNORM;
6453 /* The association is either gone or not ready. */
6454 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6455 return mask;
6457 /* Is it writable? */
6458 if (sctp_writeable(sk)) {
6459 mask |= POLLOUT | POLLWRNORM;
6460 } else {
6461 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6463 * Since the socket is not locked, the buffer
6464 * might be made available after the writeable check and
6465 * before the bit is set. This could cause a lost I/O
6466 * signal. tcp_poll() has a race breaker for this race
6467 * condition. Based on their implementation, we put
6468 * in the following code to cover it as well.
6470 if (sctp_writeable(sk))
6471 mask |= POLLOUT | POLLWRNORM;
6473 return mask;
6476 /********************************************************************
6477 * 2nd Level Abstractions
6478 ********************************************************************/
6480 static struct sctp_bind_bucket *sctp_bucket_create(
6481 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6483 struct sctp_bind_bucket *pp;
6485 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6486 if (pp) {
6487 SCTP_DBG_OBJCNT_INC(bind_bucket);
6488 pp->port = snum;
6489 pp->fastreuse = 0;
6490 INIT_HLIST_HEAD(&pp->owner);
6491 pp->net = net;
6492 hlist_add_head(&pp->node, &head->chain);
6494 return pp;
6497 /* Caller must hold hashbucket lock for this tb with local BH disabled */
6498 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6500 if (pp && hlist_empty(&pp->owner)) {
6501 __hlist_del(&pp->node);
6502 kmem_cache_free(sctp_bucket_cachep, pp);
6503 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6507 /* Release this socket's reference to a local port. */
6508 static inline void __sctp_put_port(struct sock *sk)
6510 struct sctp_bind_hashbucket *head =
6511 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6512 inet_sk(sk)->inet_num)];
6513 struct sctp_bind_bucket *pp;
6515 spin_lock(&head->lock);
6516 pp = sctp_sk(sk)->bind_hash;
6517 __sk_del_bind_node(sk);
6518 sctp_sk(sk)->bind_hash = NULL;
6519 inet_sk(sk)->inet_num = 0;
6520 sctp_bucket_destroy(pp);
6521 spin_unlock(&head->lock);
6524 void sctp_put_port(struct sock *sk)
6526 local_bh_disable();
6527 __sctp_put_port(sk);
6528 local_bh_enable();
6532 * The system picks an ephemeral port and choose an address set equivalent
6533 * to binding with a wildcard address.
6534 * One of those addresses will be the primary address for the association.
6535 * This automatically enables the multihoming capability of SCTP.
6537 static int sctp_autobind(struct sock *sk)
6539 union sctp_addr autoaddr;
6540 struct sctp_af *af;
6541 __be16 port;
6543 /* Initialize a local sockaddr structure to INADDR_ANY. */
6544 af = sctp_sk(sk)->pf->af;
6546 port = htons(inet_sk(sk)->inet_num);
6547 af->inaddr_any(&autoaddr, port);
6549 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6552 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6554 * From RFC 2292
6555 * 4.2 The cmsghdr Structure *
6557 * When ancillary data is sent or received, any number of ancillary data
6558 * objects can be specified by the msg_control and msg_controllen members of
6559 * the msghdr structure, because each object is preceded by
6560 * a cmsghdr structure defining the object's length (the cmsg_len member).
6561 * Historically Berkeley-derived implementations have passed only one object
6562 * at a time, but this API allows multiple objects to be
6563 * passed in a single call to sendmsg() or recvmsg(). The following example
6564 * shows two ancillary data objects in a control buffer.
6566 * |<--------------------------- msg_controllen -------------------------->|
6567 * | |
6569 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6571 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6572 * | | |
6574 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6576 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6577 * | | | | |
6579 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6580 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6582 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6584 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6588 * msg_control
6589 * points here
6591 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6593 struct cmsghdr *cmsg;
6594 struct msghdr *my_msg = (struct msghdr *)msg;
6596 for_each_cmsghdr(cmsg, my_msg) {
6597 if (!CMSG_OK(my_msg, cmsg))
6598 return -EINVAL;
6600 /* Should we parse this header or ignore? */
6601 if (cmsg->cmsg_level != IPPROTO_SCTP)
6602 continue;
6604 /* Strictly check lengths following example in SCM code. */
6605 switch (cmsg->cmsg_type) {
6606 case SCTP_INIT:
6607 /* SCTP Socket API Extension
6608 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
6610 * This cmsghdr structure provides information for
6611 * initializing new SCTP associations with sendmsg().
6612 * The SCTP_INITMSG socket option uses this same data
6613 * structure. This structure is not used for
6614 * recvmsg().
6616 * cmsg_level cmsg_type cmsg_data[]
6617 * ------------ ------------ ----------------------
6618 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6620 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
6621 return -EINVAL;
6623 cmsgs->init = CMSG_DATA(cmsg);
6624 break;
6626 case SCTP_SNDRCV:
6627 /* SCTP Socket API Extension
6628 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
6630 * This cmsghdr structure specifies SCTP options for
6631 * sendmsg() and describes SCTP header information
6632 * about a received message through recvmsg().
6634 * cmsg_level cmsg_type cmsg_data[]
6635 * ------------ ------------ ----------------------
6636 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6638 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6639 return -EINVAL;
6641 cmsgs->srinfo = CMSG_DATA(cmsg);
6643 if (cmsgs->srinfo->sinfo_flags &
6644 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6645 SCTP_ABORT | SCTP_EOF))
6646 return -EINVAL;
6647 break;
6649 case SCTP_SNDINFO:
6650 /* SCTP Socket API Extension
6651 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
6653 * This cmsghdr structure specifies SCTP options for
6654 * sendmsg(). This structure and SCTP_RCVINFO replaces
6655 * SCTP_SNDRCV which has been deprecated.
6657 * cmsg_level cmsg_type cmsg_data[]
6658 * ------------ ------------ ---------------------
6659 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
6661 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
6662 return -EINVAL;
6664 cmsgs->sinfo = CMSG_DATA(cmsg);
6666 if (cmsgs->sinfo->snd_flags &
6667 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6668 SCTP_ABORT | SCTP_EOF))
6669 return -EINVAL;
6670 break;
6671 default:
6672 return -EINVAL;
6676 return 0;
6680 * Wait for a packet..
6681 * Note: This function is the same function as in core/datagram.c
6682 * with a few modifications to make lksctp work.
6684 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6686 int error;
6687 DEFINE_WAIT(wait);
6689 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6691 /* Socket errors? */
6692 error = sock_error(sk);
6693 if (error)
6694 goto out;
6696 if (!skb_queue_empty(&sk->sk_receive_queue))
6697 goto ready;
6699 /* Socket shut down? */
6700 if (sk->sk_shutdown & RCV_SHUTDOWN)
6701 goto out;
6703 /* Sequenced packets can come disconnected. If so we report the
6704 * problem.
6706 error = -ENOTCONN;
6708 /* Is there a good reason to think that we may receive some data? */
6709 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6710 goto out;
6712 /* Handle signals. */
6713 if (signal_pending(current))
6714 goto interrupted;
6716 /* Let another process have a go. Since we are going to sleep
6717 * anyway. Note: This may cause odd behaviors if the message
6718 * does not fit in the user's buffer, but this seems to be the
6719 * only way to honor MSG_DONTWAIT realistically.
6721 release_sock(sk);
6722 *timeo_p = schedule_timeout(*timeo_p);
6723 lock_sock(sk);
6725 ready:
6726 finish_wait(sk_sleep(sk), &wait);
6727 return 0;
6729 interrupted:
6730 error = sock_intr_errno(*timeo_p);
6732 out:
6733 finish_wait(sk_sleep(sk), &wait);
6734 *err = error;
6735 return error;
6738 /* Receive a datagram.
6739 * Note: This is pretty much the same routine as in core/datagram.c
6740 * with a few changes to make lksctp work.
6742 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6743 int noblock, int *err)
6745 int error;
6746 struct sk_buff *skb;
6747 long timeo;
6749 timeo = sock_rcvtimeo(sk, noblock);
6751 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6752 MAX_SCHEDULE_TIMEOUT);
6754 do {
6755 /* Again only user level code calls this function,
6756 * so nothing interrupt level
6757 * will suddenly eat the receive_queue.
6759 * Look at current nfs client by the way...
6760 * However, this function was correct in any case. 8)
6762 if (flags & MSG_PEEK) {
6763 spin_lock_bh(&sk->sk_receive_queue.lock);
6764 skb = skb_peek(&sk->sk_receive_queue);
6765 if (skb)
6766 atomic_inc(&skb->users);
6767 spin_unlock_bh(&sk->sk_receive_queue.lock);
6768 } else {
6769 skb = skb_dequeue(&sk->sk_receive_queue);
6772 if (skb)
6773 return skb;
6775 /* Caller is allowed not to check sk->sk_err before calling. */
6776 error = sock_error(sk);
6777 if (error)
6778 goto no_packet;
6780 if (sk->sk_shutdown & RCV_SHUTDOWN)
6781 break;
6783 if (sk_can_busy_loop(sk) &&
6784 sk_busy_loop(sk, noblock))
6785 continue;
6787 /* User doesn't want to wait. */
6788 error = -EAGAIN;
6789 if (!timeo)
6790 goto no_packet;
6791 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6793 return NULL;
6795 no_packet:
6796 *err = error;
6797 return NULL;
6800 /* If sndbuf has changed, wake up per association sndbuf waiters. */
6801 static void __sctp_write_space(struct sctp_association *asoc)
6803 struct sock *sk = asoc->base.sk;
6804 struct socket *sock = sk->sk_socket;
6806 if ((sctp_wspace(asoc) > 0) && sock) {
6807 if (waitqueue_active(&asoc->wait))
6808 wake_up_interruptible(&asoc->wait);
6810 if (sctp_writeable(sk)) {
6811 wait_queue_head_t *wq = sk_sleep(sk);
6813 if (wq && waitqueue_active(wq))
6814 wake_up_interruptible(wq);
6816 /* Note that we try to include the Async I/O support
6817 * here by modeling from the current TCP/UDP code.
6818 * We have not tested with it yet.
6820 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6821 sock_wake_async(sock,
6822 SOCK_WAKE_SPACE, POLL_OUT);
6827 static void sctp_wake_up_waiters(struct sock *sk,
6828 struct sctp_association *asoc)
6830 struct sctp_association *tmp = asoc;
6832 /* We do accounting for the sndbuf space per association,
6833 * so we only need to wake our own association.
6835 if (asoc->ep->sndbuf_policy)
6836 return __sctp_write_space(asoc);
6838 /* If association goes down and is just flushing its
6839 * outq, then just normally notify others.
6841 if (asoc->base.dead)
6842 return sctp_write_space(sk);
6844 /* Accounting for the sndbuf space is per socket, so we
6845 * need to wake up others, try to be fair and in case of
6846 * other associations, let them have a go first instead
6847 * of just doing a sctp_write_space() call.
6849 * Note that we reach sctp_wake_up_waiters() only when
6850 * associations free up queued chunks, thus we are under
6851 * lock and the list of associations on a socket is
6852 * guaranteed not to change.
6854 for (tmp = list_next_entry(tmp, asocs); 1;
6855 tmp = list_next_entry(tmp, asocs)) {
6856 /* Manually skip the head element. */
6857 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6858 continue;
6859 /* Wake up association. */
6860 __sctp_write_space(tmp);
6861 /* We've reached the end. */
6862 if (tmp == asoc)
6863 break;
6867 /* Do accounting for the sndbuf space.
6868 * Decrement the used sndbuf space of the corresponding association by the
6869 * data size which was just transmitted(freed).
6871 static void sctp_wfree(struct sk_buff *skb)
6873 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
6874 struct sctp_association *asoc = chunk->asoc;
6875 struct sock *sk = asoc->base.sk;
6877 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6878 sizeof(struct sk_buff) +
6879 sizeof(struct sctp_chunk);
6881 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6884 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6886 sk->sk_wmem_queued -= skb->truesize;
6887 sk_mem_uncharge(sk, skb->truesize);
6889 sock_wfree(skb);
6890 sctp_wake_up_waiters(sk, asoc);
6892 sctp_association_put(asoc);
6895 /* Do accounting for the receive space on the socket.
6896 * Accounting for the association is done in ulpevent.c
6897 * We set this as a destructor for the cloned data skbs so that
6898 * accounting is done at the correct time.
6900 void sctp_sock_rfree(struct sk_buff *skb)
6902 struct sock *sk = skb->sk;
6903 struct sctp_ulpevent *event = sctp_skb2event(skb);
6905 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6908 * Mimic the behavior of sock_rfree
6910 sk_mem_uncharge(sk, event->rmem_len);
6914 /* Helper function to wait for space in the sndbuf. */
6915 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6916 size_t msg_len)
6918 struct sock *sk = asoc->base.sk;
6919 int err = 0;
6920 long current_timeo = *timeo_p;
6921 DEFINE_WAIT(wait);
6923 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6924 *timeo_p, msg_len);
6926 /* Increment the association's refcnt. */
6927 sctp_association_hold(asoc);
6929 /* Wait on the association specific sndbuf space. */
6930 for (;;) {
6931 prepare_to_wait_exclusive(&asoc->wait, &wait,
6932 TASK_INTERRUPTIBLE);
6933 if (!*timeo_p)
6934 goto do_nonblock;
6935 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6936 asoc->base.dead)
6937 goto do_error;
6938 if (signal_pending(current))
6939 goto do_interrupted;
6940 if (msg_len <= sctp_wspace(asoc))
6941 break;
6943 /* Let another process have a go. Since we are going
6944 * to sleep anyway.
6946 release_sock(sk);
6947 current_timeo = schedule_timeout(current_timeo);
6948 BUG_ON(sk != asoc->base.sk);
6949 lock_sock(sk);
6951 *timeo_p = current_timeo;
6954 out:
6955 finish_wait(&asoc->wait, &wait);
6957 /* Release the association's refcnt. */
6958 sctp_association_put(asoc);
6960 return err;
6962 do_error:
6963 err = -EPIPE;
6964 goto out;
6966 do_interrupted:
6967 err = sock_intr_errno(*timeo_p);
6968 goto out;
6970 do_nonblock:
6971 err = -EAGAIN;
6972 goto out;
6975 void sctp_data_ready(struct sock *sk)
6977 struct socket_wq *wq;
6979 rcu_read_lock();
6980 wq = rcu_dereference(sk->sk_wq);
6981 if (wq_has_sleeper(wq))
6982 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6983 POLLRDNORM | POLLRDBAND);
6984 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6985 rcu_read_unlock();
6988 /* If socket sndbuf has changed, wake up all per association waiters. */
6989 void sctp_write_space(struct sock *sk)
6991 struct sctp_association *asoc;
6993 /* Wake up the tasks in each wait queue. */
6994 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6995 __sctp_write_space(asoc);
6999 /* Is there any sndbuf space available on the socket?
7001 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7002 * associations on the same socket. For a UDP-style socket with
7003 * multiple associations, it is possible for it to be "unwriteable"
7004 * prematurely. I assume that this is acceptable because
7005 * a premature "unwriteable" is better than an accidental "writeable" which
7006 * would cause an unwanted block under certain circumstances. For the 1-1
7007 * UDP-style sockets or TCP-style sockets, this code should work.
7008 * - Daisy
7010 static int sctp_writeable(struct sock *sk)
7012 int amt = 0;
7014 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7015 if (amt < 0)
7016 amt = 0;
7017 return amt;
7020 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7021 * returns immediately with EINPROGRESS.
7023 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7025 struct sock *sk = asoc->base.sk;
7026 int err = 0;
7027 long current_timeo = *timeo_p;
7028 DEFINE_WAIT(wait);
7030 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7032 /* Increment the association's refcnt. */
7033 sctp_association_hold(asoc);
7035 for (;;) {
7036 prepare_to_wait_exclusive(&asoc->wait, &wait,
7037 TASK_INTERRUPTIBLE);
7038 if (!*timeo_p)
7039 goto do_nonblock;
7040 if (sk->sk_shutdown & RCV_SHUTDOWN)
7041 break;
7042 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7043 asoc->base.dead)
7044 goto do_error;
7045 if (signal_pending(current))
7046 goto do_interrupted;
7048 if (sctp_state(asoc, ESTABLISHED))
7049 break;
7051 /* Let another process have a go. Since we are going
7052 * to sleep anyway.
7054 release_sock(sk);
7055 current_timeo = schedule_timeout(current_timeo);
7056 lock_sock(sk);
7058 *timeo_p = current_timeo;
7061 out:
7062 finish_wait(&asoc->wait, &wait);
7064 /* Release the association's refcnt. */
7065 sctp_association_put(asoc);
7067 return err;
7069 do_error:
7070 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7071 err = -ETIMEDOUT;
7072 else
7073 err = -ECONNREFUSED;
7074 goto out;
7076 do_interrupted:
7077 err = sock_intr_errno(*timeo_p);
7078 goto out;
7080 do_nonblock:
7081 err = -EINPROGRESS;
7082 goto out;
7085 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7087 struct sctp_endpoint *ep;
7088 int err = 0;
7089 DEFINE_WAIT(wait);
7091 ep = sctp_sk(sk)->ep;
7094 for (;;) {
7095 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7096 TASK_INTERRUPTIBLE);
7098 if (list_empty(&ep->asocs)) {
7099 release_sock(sk);
7100 timeo = schedule_timeout(timeo);
7101 lock_sock(sk);
7104 err = -EINVAL;
7105 if (!sctp_sstate(sk, LISTENING))
7106 break;
7108 err = 0;
7109 if (!list_empty(&ep->asocs))
7110 break;
7112 err = sock_intr_errno(timeo);
7113 if (signal_pending(current))
7114 break;
7116 err = -EAGAIN;
7117 if (!timeo)
7118 break;
7121 finish_wait(sk_sleep(sk), &wait);
7123 return err;
7126 static void sctp_wait_for_close(struct sock *sk, long timeout)
7128 DEFINE_WAIT(wait);
7130 do {
7131 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7132 if (list_empty(&sctp_sk(sk)->ep->asocs))
7133 break;
7134 release_sock(sk);
7135 timeout = schedule_timeout(timeout);
7136 lock_sock(sk);
7137 } while (!signal_pending(current) && timeout);
7139 finish_wait(sk_sleep(sk), &wait);
7142 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7144 struct sk_buff *frag;
7146 if (!skb->data_len)
7147 goto done;
7149 /* Don't forget the fragments. */
7150 skb_walk_frags(skb, frag)
7151 sctp_skb_set_owner_r_frag(frag, sk);
7153 done:
7154 sctp_skb_set_owner_r(skb, sk);
7157 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7158 struct sctp_association *asoc)
7160 struct inet_sock *inet = inet_sk(sk);
7161 struct inet_sock *newinet;
7163 newsk->sk_type = sk->sk_type;
7164 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7165 newsk->sk_flags = sk->sk_flags;
7166 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7167 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7168 newsk->sk_reuse = sk->sk_reuse;
7170 newsk->sk_shutdown = sk->sk_shutdown;
7171 newsk->sk_destruct = sctp_destruct_sock;
7172 newsk->sk_family = sk->sk_family;
7173 newsk->sk_protocol = IPPROTO_SCTP;
7174 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7175 newsk->sk_sndbuf = sk->sk_sndbuf;
7176 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7177 newsk->sk_lingertime = sk->sk_lingertime;
7178 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7179 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7181 newinet = inet_sk(newsk);
7183 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7184 * getsockname() and getpeername()
7186 newinet->inet_sport = inet->inet_sport;
7187 newinet->inet_saddr = inet->inet_saddr;
7188 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7189 newinet->inet_dport = htons(asoc->peer.port);
7190 newinet->pmtudisc = inet->pmtudisc;
7191 newinet->inet_id = asoc->next_tsn ^ jiffies;
7193 newinet->uc_ttl = inet->uc_ttl;
7194 newinet->mc_loop = 1;
7195 newinet->mc_ttl = 1;
7196 newinet->mc_index = 0;
7197 newinet->mc_list = NULL;
7200 static inline void sctp_copy_descendant(struct sock *sk_to,
7201 const struct sock *sk_from)
7203 int ancestor_size = sizeof(struct inet_sock) +
7204 sizeof(struct sctp_sock) -
7205 offsetof(struct sctp_sock, auto_asconf_list);
7207 if (sk_from->sk_family == PF_INET6)
7208 ancestor_size += sizeof(struct ipv6_pinfo);
7210 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7213 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7214 * and its messages to the newsk.
7216 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7217 struct sctp_association *assoc,
7218 sctp_socket_type_t type)
7220 struct sctp_sock *oldsp = sctp_sk(oldsk);
7221 struct sctp_sock *newsp = sctp_sk(newsk);
7222 struct sctp_bind_bucket *pp; /* hash list port iterator */
7223 struct sctp_endpoint *newep = newsp->ep;
7224 struct sk_buff *skb, *tmp;
7225 struct sctp_ulpevent *event;
7226 struct sctp_bind_hashbucket *head;
7228 /* Migrate socket buffer sizes and all the socket level options to the
7229 * new socket.
7231 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7232 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7233 /* Brute force copy old sctp opt. */
7234 sctp_copy_descendant(newsk, oldsk);
7236 /* Restore the ep value that was overwritten with the above structure
7237 * copy.
7239 newsp->ep = newep;
7240 newsp->hmac = NULL;
7242 /* Hook this new socket in to the bind_hash list. */
7243 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7244 inet_sk(oldsk)->inet_num)];
7245 local_bh_disable();
7246 spin_lock(&head->lock);
7247 pp = sctp_sk(oldsk)->bind_hash;
7248 sk_add_bind_node(newsk, &pp->owner);
7249 sctp_sk(newsk)->bind_hash = pp;
7250 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7251 spin_unlock(&head->lock);
7252 local_bh_enable();
7254 /* Copy the bind_addr list from the original endpoint to the new
7255 * endpoint so that we can handle restarts properly
7257 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7258 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7260 /* Move any messages in the old socket's receive queue that are for the
7261 * peeled off association to the new socket's receive queue.
7263 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7264 event = sctp_skb2event(skb);
7265 if (event->asoc == assoc) {
7266 __skb_unlink(skb, &oldsk->sk_receive_queue);
7267 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7268 sctp_skb_set_owner_r_frag(skb, newsk);
7272 /* Clean up any messages pending delivery due to partial
7273 * delivery. Three cases:
7274 * 1) No partial deliver; no work.
7275 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7276 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7278 skb_queue_head_init(&newsp->pd_lobby);
7279 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7281 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7282 struct sk_buff_head *queue;
7284 /* Decide which queue to move pd_lobby skbs to. */
7285 if (assoc->ulpq.pd_mode) {
7286 queue = &newsp->pd_lobby;
7287 } else
7288 queue = &newsk->sk_receive_queue;
7290 /* Walk through the pd_lobby, looking for skbs that
7291 * need moved to the new socket.
7293 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7294 event = sctp_skb2event(skb);
7295 if (event->asoc == assoc) {
7296 __skb_unlink(skb, &oldsp->pd_lobby);
7297 __skb_queue_tail(queue, skb);
7298 sctp_skb_set_owner_r_frag(skb, newsk);
7302 /* Clear up any skbs waiting for the partial
7303 * delivery to finish.
7305 if (assoc->ulpq.pd_mode)
7306 sctp_clear_pd(oldsk, NULL);
7310 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7311 sctp_skb_set_owner_r_frag(skb, newsk);
7313 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7314 sctp_skb_set_owner_r_frag(skb, newsk);
7316 /* Set the type of socket to indicate that it is peeled off from the
7317 * original UDP-style socket or created with the accept() call on a
7318 * TCP-style socket..
7320 newsp->type = type;
7322 /* Mark the new socket "in-use" by the user so that any packets
7323 * that may arrive on the association after we've moved it are
7324 * queued to the backlog. This prevents a potential race between
7325 * backlog processing on the old socket and new-packet processing
7326 * on the new socket.
7328 * The caller has just allocated newsk so we can guarantee that other
7329 * paths won't try to lock it and then oldsk.
7331 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7332 sctp_assoc_migrate(assoc, newsk);
7334 /* If the association on the newsk is already closed before accept()
7335 * is called, set RCV_SHUTDOWN flag.
7337 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7338 newsk->sk_shutdown |= RCV_SHUTDOWN;
7340 newsk->sk_state = SCTP_SS_ESTABLISHED;
7341 release_sock(newsk);
7345 /* This proto struct describes the ULP interface for SCTP. */
7346 struct proto sctp_prot = {
7347 .name = "SCTP",
7348 .owner = THIS_MODULE,
7349 .close = sctp_close,
7350 .connect = sctp_connect,
7351 .disconnect = sctp_disconnect,
7352 .accept = sctp_accept,
7353 .ioctl = sctp_ioctl,
7354 .init = sctp_init_sock,
7355 .destroy = sctp_destroy_sock,
7356 .shutdown = sctp_shutdown,
7357 .setsockopt = sctp_setsockopt,
7358 .getsockopt = sctp_getsockopt,
7359 .sendmsg = sctp_sendmsg,
7360 .recvmsg = sctp_recvmsg,
7361 .bind = sctp_bind,
7362 .backlog_rcv = sctp_backlog_rcv,
7363 .hash = sctp_hash,
7364 .unhash = sctp_unhash,
7365 .get_port = sctp_get_port,
7366 .obj_size = sizeof(struct sctp_sock),
7367 .sysctl_mem = sysctl_sctp_mem,
7368 .sysctl_rmem = sysctl_sctp_rmem,
7369 .sysctl_wmem = sysctl_sctp_wmem,
7370 .memory_pressure = &sctp_memory_pressure,
7371 .enter_memory_pressure = sctp_enter_memory_pressure,
7372 .memory_allocated = &sctp_memory_allocated,
7373 .sockets_allocated = &sctp_sockets_allocated,
7376 #if IS_ENABLED(CONFIG_IPV6)
7378 struct proto sctpv6_prot = {
7379 .name = "SCTPv6",
7380 .owner = THIS_MODULE,
7381 .close = sctp_close,
7382 .connect = sctp_connect,
7383 .disconnect = sctp_disconnect,
7384 .accept = sctp_accept,
7385 .ioctl = sctp_ioctl,
7386 .init = sctp_init_sock,
7387 .destroy = sctp_destroy_sock,
7388 .shutdown = sctp_shutdown,
7389 .setsockopt = sctp_setsockopt,
7390 .getsockopt = sctp_getsockopt,
7391 .sendmsg = sctp_sendmsg,
7392 .recvmsg = sctp_recvmsg,
7393 .bind = sctp_bind,
7394 .backlog_rcv = sctp_backlog_rcv,
7395 .hash = sctp_hash,
7396 .unhash = sctp_unhash,
7397 .get_port = sctp_get_port,
7398 .obj_size = sizeof(struct sctp6_sock),
7399 .sysctl_mem = sysctl_sctp_mem,
7400 .sysctl_rmem = sysctl_sctp_rmem,
7401 .sysctl_wmem = sysctl_sctp_wmem,
7402 .memory_pressure = &sctp_memory_pressure,
7403 .enter_memory_pressure = sctp_enter_memory_pressure,
7404 .memory_allocated = &sctp_memory_allocated,
7405 .sockets_allocated = &sctp_sockets_allocated,
7407 #endif /* IS_ENABLED(CONFIG_IPV6) */