Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / net / sctp / socket.c
blobbf271f8c2dc9b28c992f9be3bc9b6a7bfa5fdb82
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 <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
60 #include <linux/sched/signal.h>
61 #include <linux/ip.h>
62 #include <linux/capability.h>
63 #include <linux/fcntl.h>
64 #include <linux/poll.h>
65 #include <linux/init.h>
66 #include <linux/slab.h>
67 #include <linux/file.h>
68 #include <linux/compat.h>
70 #include <net/ip.h>
71 #include <net/icmp.h>
72 #include <net/route.h>
73 #include <net/ipv6.h>
74 #include <net/inet_common.h>
75 #include <net/busy_poll.h>
77 #include <linux/socket.h> /* for sa_family_t */
78 #include <linux/export.h>
79 #include <net/sock.h>
80 #include <net/sctp/sctp.h>
81 #include <net/sctp/sm.h>
82 #include <net/sctp/stream_sched.h>
84 /* Forward declarations for internal helper functions. */
85 static int sctp_writeable(struct sock *sk);
86 static void sctp_wfree(struct sk_buff *skb);
87 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
88 size_t msg_len);
89 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
90 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
91 static int sctp_wait_for_accept(struct sock *sk, long timeo);
92 static void sctp_wait_for_close(struct sock *sk, long timeo);
93 static void sctp_destruct_sock(struct sock *sk);
94 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
95 union sctp_addr *addr, int len);
96 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
97 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf(struct sctp_association *asoc,
101 struct sctp_chunk *chunk);
102 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
103 static int sctp_autobind(struct sock *sk);
104 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
105 struct sctp_association *assoc,
106 enum sctp_socket_type type);
108 static unsigned long sctp_memory_pressure;
109 static atomic_long_t sctp_memory_allocated;
110 struct percpu_counter sctp_sockets_allocated;
112 static void sctp_enter_memory_pressure(struct sock *sk)
114 sctp_memory_pressure = 1;
118 /* Get the sndbuf space available at the time on the association. */
119 static inline int sctp_wspace(struct sctp_association *asoc)
121 int amt;
123 if (asoc->ep->sndbuf_policy)
124 amt = asoc->sndbuf_used;
125 else
126 amt = sk_wmem_alloc_get(asoc->base.sk);
128 if (amt >= asoc->base.sk->sk_sndbuf) {
129 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
130 amt = 0;
131 else {
132 amt = sk_stream_wspace(asoc->base.sk);
133 if (amt < 0)
134 amt = 0;
136 } else {
137 amt = asoc->base.sk->sk_sndbuf - amt;
139 return amt;
142 /* Increment the used sndbuf space count of the corresponding association by
143 * the size of the outgoing data chunk.
144 * Also, set the skb destructor for sndbuf accounting later.
146 * Since it is always 1-1 between chunk and skb, and also a new skb is always
147 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
148 * destructor in the data chunk skb for the purpose of the sndbuf space
149 * tracking.
151 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
153 struct sctp_association *asoc = chunk->asoc;
154 struct sock *sk = asoc->base.sk;
156 /* The sndbuf space is tracked per association. */
157 sctp_association_hold(asoc);
159 skb_set_owner_w(chunk->skb, sk);
161 chunk->skb->destructor = sctp_wfree;
162 /* Save the chunk pointer in skb for sctp_wfree to use later. */
163 skb_shinfo(chunk->skb)->destructor_arg = chunk;
165 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
166 sizeof(struct sk_buff) +
167 sizeof(struct sctp_chunk);
169 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
170 sk->sk_wmem_queued += chunk->skb->truesize;
171 sk_mem_charge(sk, chunk->skb->truesize);
174 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
176 skb_orphan(chunk->skb);
179 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
180 void (*cb)(struct sctp_chunk *))
183 struct sctp_outq *q = &asoc->outqueue;
184 struct sctp_transport *t;
185 struct sctp_chunk *chunk;
187 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
188 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
189 cb(chunk);
191 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
192 cb(chunk);
194 list_for_each_entry(chunk, &q->sacked, transmitted_list)
195 cb(chunk);
197 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
198 cb(chunk);
200 list_for_each_entry(chunk, &q->out_chunk_list, list)
201 cb(chunk);
204 static void sctp_for_each_rx_skb(struct sctp_association *asoc, struct sock *sk,
205 void (*cb)(struct sk_buff *, struct sock *))
208 struct sk_buff *skb, *tmp;
210 sctp_skb_for_each(skb, &asoc->ulpq.lobby, tmp)
211 cb(skb, sk);
213 sctp_skb_for_each(skb, &asoc->ulpq.reasm, tmp)
214 cb(skb, sk);
216 sctp_skb_for_each(skb, &asoc->ulpq.reasm_uo, tmp)
217 cb(skb, sk);
220 /* Verify that this is a valid address. */
221 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
222 int len)
224 struct sctp_af *af;
226 /* Verify basic sockaddr. */
227 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
228 if (!af)
229 return -EINVAL;
231 /* Is this a valid SCTP address? */
232 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
233 return -EINVAL;
235 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
236 return -EINVAL;
238 return 0;
241 /* Look up the association by its id. If this is not a UDP-style
242 * socket, the ID field is always ignored.
244 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
246 struct sctp_association *asoc = NULL;
248 /* If this is not a UDP-style socket, assoc id should be ignored. */
249 if (!sctp_style(sk, UDP)) {
250 /* Return NULL if the socket state is not ESTABLISHED. It
251 * could be a TCP-style listening socket or a socket which
252 * hasn't yet called connect() to establish an association.
254 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
255 return NULL;
257 /* Get the first and the only association from the list. */
258 if (!list_empty(&sctp_sk(sk)->ep->asocs))
259 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
260 struct sctp_association, asocs);
261 return asoc;
264 /* Otherwise this is a UDP-style socket. */
265 if (!id || (id == (sctp_assoc_t)-1))
266 return NULL;
268 spin_lock_bh(&sctp_assocs_id_lock);
269 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
270 spin_unlock_bh(&sctp_assocs_id_lock);
272 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
273 return NULL;
275 return asoc;
278 /* Look up the transport from an address and an assoc id. If both address and
279 * id are specified, the associations matching the address and the id should be
280 * the same.
282 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
283 struct sockaddr_storage *addr,
284 sctp_assoc_t id)
286 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
287 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
288 union sctp_addr *laddr = (union sctp_addr *)addr;
289 struct sctp_transport *transport;
291 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
292 return NULL;
294 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
295 laddr,
296 &transport);
298 if (!addr_asoc)
299 return NULL;
301 id_asoc = sctp_id2assoc(sk, id);
302 if (id_asoc && (id_asoc != addr_asoc))
303 return NULL;
305 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
306 (union sctp_addr *)addr);
308 return transport;
311 /* API 3.1.2 bind() - UDP Style Syntax
312 * The syntax of bind() is,
314 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
316 * sd - the socket descriptor returned by socket().
317 * addr - the address structure (struct sockaddr_in or struct
318 * sockaddr_in6 [RFC 2553]),
319 * addr_len - the size of the address structure.
321 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
323 int retval = 0;
325 lock_sock(sk);
327 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
328 addr, addr_len);
330 /* Disallow binding twice. */
331 if (!sctp_sk(sk)->ep->base.bind_addr.port)
332 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
333 addr_len);
334 else
335 retval = -EINVAL;
337 release_sock(sk);
339 return retval;
342 static long sctp_get_port_local(struct sock *, union sctp_addr *);
344 /* Verify this is a valid sockaddr. */
345 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
346 union sctp_addr *addr, int len)
348 struct sctp_af *af;
350 /* Check minimum size. */
351 if (len < sizeof (struct sockaddr))
352 return NULL;
354 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
355 return NULL;
357 /* V4 mapped address are really of AF_INET family */
358 if (addr->sa.sa_family == AF_INET6 &&
359 ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
360 !opt->pf->af_supported(AF_INET, opt))
361 return NULL;
363 /* If we get this far, af is valid. */
364 af = sctp_get_af_specific(addr->sa.sa_family);
366 if (len < af->sockaddr_len)
367 return NULL;
369 return af;
372 /* Bind a local address either to an endpoint or to an association. */
373 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
375 struct net *net = sock_net(sk);
376 struct sctp_sock *sp = sctp_sk(sk);
377 struct sctp_endpoint *ep = sp->ep;
378 struct sctp_bind_addr *bp = &ep->base.bind_addr;
379 struct sctp_af *af;
380 unsigned short snum;
381 int ret = 0;
383 /* Common sockaddr verification. */
384 af = sctp_sockaddr_af(sp, addr, len);
385 if (!af) {
386 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
387 __func__, sk, addr, len);
388 return -EINVAL;
391 snum = ntohs(addr->v4.sin_port);
393 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
394 __func__, sk, &addr->sa, bp->port, snum, len);
396 /* PF specific bind() address verification. */
397 if (!sp->pf->bind_verify(sp, addr))
398 return -EADDRNOTAVAIL;
400 /* We must either be unbound, or bind to the same port.
401 * It's OK to allow 0 ports if we are already bound.
402 * We'll just inhert an already bound port in this case
404 if (bp->port) {
405 if (!snum)
406 snum = bp->port;
407 else if (snum != bp->port) {
408 pr_debug("%s: new port %d doesn't match existing port "
409 "%d\n", __func__, snum, bp->port);
410 return -EINVAL;
414 if (snum && snum < inet_prot_sock(net) &&
415 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
416 return -EACCES;
418 /* See if the address matches any of the addresses we may have
419 * already bound before checking against other endpoints.
421 if (sctp_bind_addr_match(bp, addr, sp))
422 return -EINVAL;
424 /* Make sure we are allowed to bind here.
425 * The function sctp_get_port_local() does duplicate address
426 * detection.
428 addr->v4.sin_port = htons(snum);
429 if ((ret = sctp_get_port_local(sk, addr))) {
430 return -EADDRINUSE;
433 /* Refresh ephemeral port. */
434 if (!bp->port)
435 bp->port = inet_sk(sk)->inet_num;
437 /* Add the address to the bind address list.
438 * Use GFP_ATOMIC since BHs will be disabled.
440 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
441 SCTP_ADDR_SRC, GFP_ATOMIC);
443 /* Copy back into socket for getsockname() use. */
444 if (!ret) {
445 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
446 sp->pf->to_sk_saddr(addr, sk);
449 return ret;
452 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
454 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
455 * at any one time. If a sender, after sending an ASCONF chunk, decides
456 * it needs to transfer another ASCONF Chunk, it MUST wait until the
457 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
458 * subsequent ASCONF. Note this restriction binds each side, so at any
459 * time two ASCONF may be in-transit on any given association (one sent
460 * from each endpoint).
462 static int sctp_send_asconf(struct sctp_association *asoc,
463 struct sctp_chunk *chunk)
465 struct net *net = sock_net(asoc->base.sk);
466 int retval = 0;
468 /* If there is an outstanding ASCONF chunk, queue it for later
469 * transmission.
471 if (asoc->addip_last_asconf) {
472 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
473 goto out;
476 /* Hold the chunk until an ASCONF_ACK is received. */
477 sctp_chunk_hold(chunk);
478 retval = sctp_primitive_ASCONF(net, asoc, chunk);
479 if (retval)
480 sctp_chunk_free(chunk);
481 else
482 asoc->addip_last_asconf = chunk;
484 out:
485 return retval;
488 /* Add a list of addresses as bind addresses to local endpoint or
489 * association.
491 * Basically run through each address specified in the addrs/addrcnt
492 * array/length pair, determine if it is IPv6 or IPv4 and call
493 * sctp_do_bind() on it.
495 * If any of them fails, then the operation will be reversed and the
496 * ones that were added will be removed.
498 * Only sctp_setsockopt_bindx() is supposed to call this function.
500 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
502 int cnt;
503 int retval = 0;
504 void *addr_buf;
505 struct sockaddr *sa_addr;
506 struct sctp_af *af;
508 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
509 addrs, addrcnt);
511 addr_buf = addrs;
512 for (cnt = 0; cnt < addrcnt; cnt++) {
513 /* The list may contain either IPv4 or IPv6 address;
514 * determine the address length for walking thru the list.
516 sa_addr = addr_buf;
517 af = sctp_get_af_specific(sa_addr->sa_family);
518 if (!af) {
519 retval = -EINVAL;
520 goto err_bindx_add;
523 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
524 af->sockaddr_len);
526 addr_buf += af->sockaddr_len;
528 err_bindx_add:
529 if (retval < 0) {
530 /* Failed. Cleanup the ones that have been added */
531 if (cnt > 0)
532 sctp_bindx_rem(sk, addrs, cnt);
533 return retval;
537 return retval;
540 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
541 * associations that are part of the endpoint indicating that a list of local
542 * addresses are added to the endpoint.
544 * If any of the addresses is already in the bind address list of the
545 * association, we do not send the chunk for that association. But it will not
546 * affect other associations.
548 * Only sctp_setsockopt_bindx() is supposed to call this function.
550 static int sctp_send_asconf_add_ip(struct sock *sk,
551 struct sockaddr *addrs,
552 int addrcnt)
554 struct net *net = sock_net(sk);
555 struct sctp_sock *sp;
556 struct sctp_endpoint *ep;
557 struct sctp_association *asoc;
558 struct sctp_bind_addr *bp;
559 struct sctp_chunk *chunk;
560 struct sctp_sockaddr_entry *laddr;
561 union sctp_addr *addr;
562 union sctp_addr saveaddr;
563 void *addr_buf;
564 struct sctp_af *af;
565 struct list_head *p;
566 int i;
567 int retval = 0;
569 if (!net->sctp.addip_enable)
570 return retval;
572 sp = sctp_sk(sk);
573 ep = sp->ep;
575 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
576 __func__, sk, addrs, addrcnt);
578 list_for_each_entry(asoc, &ep->asocs, asocs) {
579 if (!asoc->peer.asconf_capable)
580 continue;
582 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
583 continue;
585 if (!sctp_state(asoc, ESTABLISHED))
586 continue;
588 /* Check if any address in the packed array of addresses is
589 * in the bind address list of the association. If so,
590 * do not send the asconf chunk to its peer, but continue with
591 * other associations.
593 addr_buf = addrs;
594 for (i = 0; i < addrcnt; i++) {
595 addr = addr_buf;
596 af = sctp_get_af_specific(addr->v4.sin_family);
597 if (!af) {
598 retval = -EINVAL;
599 goto out;
602 if (sctp_assoc_lookup_laddr(asoc, addr))
603 break;
605 addr_buf += af->sockaddr_len;
607 if (i < addrcnt)
608 continue;
610 /* Use the first valid address in bind addr list of
611 * association as Address Parameter of ASCONF CHUNK.
613 bp = &asoc->base.bind_addr;
614 p = bp->address_list.next;
615 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
616 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
617 addrcnt, SCTP_PARAM_ADD_IP);
618 if (!chunk) {
619 retval = -ENOMEM;
620 goto out;
623 /* Add the new addresses to the bind address list with
624 * use_as_src set to 0.
626 addr_buf = addrs;
627 for (i = 0; i < addrcnt; i++) {
628 addr = addr_buf;
629 af = sctp_get_af_specific(addr->v4.sin_family);
630 memcpy(&saveaddr, addr, af->sockaddr_len);
631 retval = sctp_add_bind_addr(bp, &saveaddr,
632 sizeof(saveaddr),
633 SCTP_ADDR_NEW, GFP_ATOMIC);
634 addr_buf += af->sockaddr_len;
636 if (asoc->src_out_of_asoc_ok) {
637 struct sctp_transport *trans;
639 list_for_each_entry(trans,
640 &asoc->peer.transport_addr_list, transports) {
641 /* Clear the source and route cache */
642 sctp_transport_dst_release(trans);
643 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
644 2*asoc->pathmtu, 4380));
645 trans->ssthresh = asoc->peer.i.a_rwnd;
646 trans->rto = asoc->rto_initial;
647 sctp_max_rto(asoc, trans);
648 trans->rtt = trans->srtt = trans->rttvar = 0;
649 sctp_transport_route(trans, NULL,
650 sctp_sk(asoc->base.sk));
653 retval = sctp_send_asconf(asoc, chunk);
656 out:
657 return retval;
660 /* Remove a list of addresses from bind addresses list. Do not remove the
661 * last address.
663 * Basically run through each address specified in the addrs/addrcnt
664 * array/length pair, determine if it is IPv6 or IPv4 and call
665 * sctp_del_bind() on it.
667 * If any of them fails, then the operation will be reversed and the
668 * ones that were removed will be added back.
670 * At least one address has to be left; if only one address is
671 * available, the operation will return -EBUSY.
673 * Only sctp_setsockopt_bindx() is supposed to call this function.
675 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
677 struct sctp_sock *sp = sctp_sk(sk);
678 struct sctp_endpoint *ep = sp->ep;
679 int cnt;
680 struct sctp_bind_addr *bp = &ep->base.bind_addr;
681 int retval = 0;
682 void *addr_buf;
683 union sctp_addr *sa_addr;
684 struct sctp_af *af;
686 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
687 __func__, sk, addrs, addrcnt);
689 addr_buf = addrs;
690 for (cnt = 0; cnt < addrcnt; cnt++) {
691 /* If the bind address list is empty or if there is only one
692 * bind address, there is nothing more to be removed (we need
693 * at least one address here).
695 if (list_empty(&bp->address_list) ||
696 (sctp_list_single_entry(&bp->address_list))) {
697 retval = -EBUSY;
698 goto err_bindx_rem;
701 sa_addr = addr_buf;
702 af = sctp_get_af_specific(sa_addr->sa.sa_family);
703 if (!af) {
704 retval = -EINVAL;
705 goto err_bindx_rem;
708 if (!af->addr_valid(sa_addr, sp, NULL)) {
709 retval = -EADDRNOTAVAIL;
710 goto err_bindx_rem;
713 if (sa_addr->v4.sin_port &&
714 sa_addr->v4.sin_port != htons(bp->port)) {
715 retval = -EINVAL;
716 goto err_bindx_rem;
719 if (!sa_addr->v4.sin_port)
720 sa_addr->v4.sin_port = htons(bp->port);
722 /* FIXME - There is probably a need to check if sk->sk_saddr and
723 * sk->sk_rcv_addr are currently set to one of the addresses to
724 * be removed. This is something which needs to be looked into
725 * when we are fixing the outstanding issues with multi-homing
726 * socket routing and failover schemes. Refer to comments in
727 * sctp_do_bind(). -daisy
729 retval = sctp_del_bind_addr(bp, sa_addr);
731 addr_buf += af->sockaddr_len;
732 err_bindx_rem:
733 if (retval < 0) {
734 /* Failed. Add the ones that has been removed back */
735 if (cnt > 0)
736 sctp_bindx_add(sk, addrs, cnt);
737 return retval;
741 return retval;
744 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
745 * the associations that are part of the endpoint indicating that a list of
746 * local addresses are removed from the endpoint.
748 * If any of the addresses is already in the bind address list of the
749 * association, we do not send the chunk for that association. But it will not
750 * affect other associations.
752 * Only sctp_setsockopt_bindx() is supposed to call this function.
754 static int sctp_send_asconf_del_ip(struct sock *sk,
755 struct sockaddr *addrs,
756 int addrcnt)
758 struct net *net = sock_net(sk);
759 struct sctp_sock *sp;
760 struct sctp_endpoint *ep;
761 struct sctp_association *asoc;
762 struct sctp_transport *transport;
763 struct sctp_bind_addr *bp;
764 struct sctp_chunk *chunk;
765 union sctp_addr *laddr;
766 void *addr_buf;
767 struct sctp_af *af;
768 struct sctp_sockaddr_entry *saddr;
769 int i;
770 int retval = 0;
771 int stored = 0;
773 chunk = NULL;
774 if (!net->sctp.addip_enable)
775 return retval;
777 sp = sctp_sk(sk);
778 ep = sp->ep;
780 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
781 __func__, sk, addrs, addrcnt);
783 list_for_each_entry(asoc, &ep->asocs, asocs) {
785 if (!asoc->peer.asconf_capable)
786 continue;
788 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
789 continue;
791 if (!sctp_state(asoc, ESTABLISHED))
792 continue;
794 /* Check if any address in the packed array of addresses is
795 * not present in the bind address list of the association.
796 * If so, do not send the asconf chunk to its peer, but
797 * continue with other associations.
799 addr_buf = addrs;
800 for (i = 0; i < addrcnt; i++) {
801 laddr = addr_buf;
802 af = sctp_get_af_specific(laddr->v4.sin_family);
803 if (!af) {
804 retval = -EINVAL;
805 goto out;
808 if (!sctp_assoc_lookup_laddr(asoc, laddr))
809 break;
811 addr_buf += af->sockaddr_len;
813 if (i < addrcnt)
814 continue;
816 /* Find one address in the association's bind address list
817 * that is not in the packed array of addresses. This is to
818 * make sure that we do not delete all the addresses in the
819 * association.
821 bp = &asoc->base.bind_addr;
822 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
823 addrcnt, sp);
824 if ((laddr == NULL) && (addrcnt == 1)) {
825 if (asoc->asconf_addr_del_pending)
826 continue;
827 asoc->asconf_addr_del_pending =
828 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
829 if (asoc->asconf_addr_del_pending == NULL) {
830 retval = -ENOMEM;
831 goto out;
833 asoc->asconf_addr_del_pending->sa.sa_family =
834 addrs->sa_family;
835 asoc->asconf_addr_del_pending->v4.sin_port =
836 htons(bp->port);
837 if (addrs->sa_family == AF_INET) {
838 struct sockaddr_in *sin;
840 sin = (struct sockaddr_in *)addrs;
841 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
842 } else if (addrs->sa_family == AF_INET6) {
843 struct sockaddr_in6 *sin6;
845 sin6 = (struct sockaddr_in6 *)addrs;
846 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
849 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
850 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
851 asoc->asconf_addr_del_pending);
853 asoc->src_out_of_asoc_ok = 1;
854 stored = 1;
855 goto skip_mkasconf;
858 if (laddr == NULL)
859 return -EINVAL;
861 /* We do not need RCU protection throughout this loop
862 * because this is done under a socket lock from the
863 * setsockopt call.
865 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
866 SCTP_PARAM_DEL_IP);
867 if (!chunk) {
868 retval = -ENOMEM;
869 goto out;
872 skip_mkasconf:
873 /* Reset use_as_src flag for the addresses in the bind address
874 * list that are to be deleted.
876 addr_buf = addrs;
877 for (i = 0; i < addrcnt; i++) {
878 laddr = addr_buf;
879 af = sctp_get_af_specific(laddr->v4.sin_family);
880 list_for_each_entry(saddr, &bp->address_list, list) {
881 if (sctp_cmp_addr_exact(&saddr->a, laddr))
882 saddr->state = SCTP_ADDR_DEL;
884 addr_buf += af->sockaddr_len;
887 /* Update the route and saddr entries for all the transports
888 * as some of the addresses in the bind address list are
889 * about to be deleted and cannot be used as source addresses.
891 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
892 transports) {
893 sctp_transport_dst_release(transport);
894 sctp_transport_route(transport, NULL,
895 sctp_sk(asoc->base.sk));
898 if (stored)
899 /* We don't need to transmit ASCONF */
900 continue;
901 retval = sctp_send_asconf(asoc, chunk);
903 out:
904 return retval;
907 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
908 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
910 struct sock *sk = sctp_opt2sk(sp);
911 union sctp_addr *addr;
912 struct sctp_af *af;
914 /* It is safe to write port space in caller. */
915 addr = &addrw->a;
916 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
917 af = sctp_get_af_specific(addr->sa.sa_family);
918 if (!af)
919 return -EINVAL;
920 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
921 return -EINVAL;
923 if (addrw->state == SCTP_ADDR_NEW)
924 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
925 else
926 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
929 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
931 * API 8.1
932 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
933 * int flags);
935 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
936 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
937 * or IPv6 addresses.
939 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
940 * Section 3.1.2 for this usage.
942 * addrs is a pointer to an array of one or more socket addresses. Each
943 * address is contained in its appropriate structure (i.e. struct
944 * sockaddr_in or struct sockaddr_in6) the family of the address type
945 * must be used to distinguish the address length (note that this
946 * representation is termed a "packed array" of addresses). The caller
947 * specifies the number of addresses in the array with addrcnt.
949 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
950 * -1, and sets errno to the appropriate error code.
952 * For SCTP, the port given in each socket address must be the same, or
953 * sctp_bindx() will fail, setting errno to EINVAL.
955 * The flags parameter is formed from the bitwise OR of zero or more of
956 * the following currently defined flags:
958 * SCTP_BINDX_ADD_ADDR
960 * SCTP_BINDX_REM_ADDR
962 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
963 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
964 * addresses from the association. The two flags are mutually exclusive;
965 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
966 * not remove all addresses from an association; sctp_bindx() will
967 * reject such an attempt with EINVAL.
969 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
970 * additional addresses with an endpoint after calling bind(). Or use
971 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
972 * socket is associated with so that no new association accepted will be
973 * associated with those addresses. If the endpoint supports dynamic
974 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
975 * endpoint to send the appropriate message to the peer to change the
976 * peers address lists.
978 * Adding and removing addresses from a connected association is
979 * optional functionality. Implementations that do not support this
980 * functionality should return EOPNOTSUPP.
982 * Basically do nothing but copying the addresses from user to kernel
983 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
984 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
985 * from userspace.
987 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
988 * it.
990 * sk The sk of the socket
991 * addrs The pointer to the addresses in user land
992 * addrssize Size of the addrs buffer
993 * op Operation to perform (add or remove, see the flags of
994 * sctp_bindx)
996 * Returns 0 if ok, <0 errno code on error.
998 static int sctp_setsockopt_bindx(struct sock *sk,
999 struct sockaddr __user *addrs,
1000 int addrs_size, int op)
1002 struct sockaddr *kaddrs;
1003 int err;
1004 int addrcnt = 0;
1005 int walk_size = 0;
1006 struct sockaddr *sa_addr;
1007 void *addr_buf;
1008 struct sctp_af *af;
1010 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1011 __func__, sk, addrs, addrs_size, op);
1013 if (unlikely(addrs_size <= 0))
1014 return -EINVAL;
1016 kaddrs = vmemdup_user(addrs, addrs_size);
1017 if (unlikely(IS_ERR(kaddrs)))
1018 return PTR_ERR(kaddrs);
1020 /* Walk through the addrs buffer and count the number of addresses. */
1021 addr_buf = kaddrs;
1022 while (walk_size < addrs_size) {
1023 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1024 kvfree(kaddrs);
1025 return -EINVAL;
1028 sa_addr = addr_buf;
1029 af = sctp_get_af_specific(sa_addr->sa_family);
1031 /* If the address family is not supported or if this address
1032 * causes the address buffer to overflow return EINVAL.
1034 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1035 kvfree(kaddrs);
1036 return -EINVAL;
1038 addrcnt++;
1039 addr_buf += af->sockaddr_len;
1040 walk_size += af->sockaddr_len;
1043 /* Do the work. */
1044 switch (op) {
1045 case SCTP_BINDX_ADD_ADDR:
1046 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1047 if (err)
1048 goto out;
1049 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1050 break;
1052 case SCTP_BINDX_REM_ADDR:
1053 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1054 if (err)
1055 goto out;
1056 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1057 break;
1059 default:
1060 err = -EINVAL;
1061 break;
1064 out:
1065 kvfree(kaddrs);
1067 return err;
1070 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1072 * Common routine for handling connect() and sctp_connectx().
1073 * Connect will come in with just a single address.
1075 static int __sctp_connect(struct sock *sk,
1076 struct sockaddr *kaddrs,
1077 int addrs_size,
1078 sctp_assoc_t *assoc_id)
1080 struct net *net = sock_net(sk);
1081 struct sctp_sock *sp;
1082 struct sctp_endpoint *ep;
1083 struct sctp_association *asoc = NULL;
1084 struct sctp_association *asoc2;
1085 struct sctp_transport *transport;
1086 union sctp_addr to;
1087 enum sctp_scope scope;
1088 long timeo;
1089 int err = 0;
1090 int addrcnt = 0;
1091 int walk_size = 0;
1092 union sctp_addr *sa_addr = NULL;
1093 void *addr_buf;
1094 unsigned short port;
1095 unsigned int f_flags = 0;
1097 sp = sctp_sk(sk);
1098 ep = sp->ep;
1100 /* connect() cannot be done on a socket that is already in ESTABLISHED
1101 * state - UDP-style peeled off socket or a TCP-style socket that
1102 * is already connected.
1103 * It cannot be done even on a TCP-style listening socket.
1105 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1106 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1107 err = -EISCONN;
1108 goto out_free;
1111 /* Walk through the addrs buffer and count the number of addresses. */
1112 addr_buf = kaddrs;
1113 while (walk_size < addrs_size) {
1114 struct sctp_af *af;
1116 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1117 err = -EINVAL;
1118 goto out_free;
1121 sa_addr = addr_buf;
1122 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1124 /* If the address family is not supported or if this address
1125 * causes the address buffer to overflow return EINVAL.
1127 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1128 err = -EINVAL;
1129 goto out_free;
1132 port = ntohs(sa_addr->v4.sin_port);
1134 /* Save current address so we can work with it */
1135 memcpy(&to, sa_addr, af->sockaddr_len);
1137 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1138 if (err)
1139 goto out_free;
1141 /* Make sure the destination port is correctly set
1142 * in all addresses.
1144 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1145 err = -EINVAL;
1146 goto out_free;
1149 /* Check if there already is a matching association on the
1150 * endpoint (other than the one created here).
1152 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1153 if (asoc2 && asoc2 != asoc) {
1154 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1155 err = -EISCONN;
1156 else
1157 err = -EALREADY;
1158 goto out_free;
1161 /* If we could not find a matching association on the endpoint,
1162 * make sure that there is no peeled-off association matching
1163 * the peer address even on another socket.
1165 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1166 err = -EADDRNOTAVAIL;
1167 goto out_free;
1170 if (!asoc) {
1171 /* If a bind() or sctp_bindx() is not called prior to
1172 * an sctp_connectx() call, the system picks an
1173 * ephemeral port and will choose an address set
1174 * equivalent to binding with a wildcard address.
1176 if (!ep->base.bind_addr.port) {
1177 if (sctp_autobind(sk)) {
1178 err = -EAGAIN;
1179 goto out_free;
1181 } else {
1183 * If an unprivileged user inherits a 1-many
1184 * style socket with open associations on a
1185 * privileged port, it MAY be permitted to
1186 * accept new associations, but it SHOULD NOT
1187 * be permitted to open new associations.
1189 if (ep->base.bind_addr.port <
1190 inet_prot_sock(net) &&
1191 !ns_capable(net->user_ns,
1192 CAP_NET_BIND_SERVICE)) {
1193 err = -EACCES;
1194 goto out_free;
1198 scope = sctp_scope(&to);
1199 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1200 if (!asoc) {
1201 err = -ENOMEM;
1202 goto out_free;
1205 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1206 GFP_KERNEL);
1207 if (err < 0) {
1208 goto out_free;
1213 /* Prime the peer's transport structures. */
1214 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1215 SCTP_UNKNOWN);
1216 if (!transport) {
1217 err = -ENOMEM;
1218 goto out_free;
1221 addrcnt++;
1222 addr_buf += af->sockaddr_len;
1223 walk_size += af->sockaddr_len;
1226 /* In case the user of sctp_connectx() wants an association
1227 * id back, assign one now.
1229 if (assoc_id) {
1230 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1231 if (err < 0)
1232 goto out_free;
1235 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1236 if (err < 0) {
1237 goto out_free;
1240 /* Initialize sk's dport and daddr for getpeername() */
1241 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1242 sp->pf->to_sk_daddr(sa_addr, sk);
1243 sk->sk_err = 0;
1245 /* in-kernel sockets don't generally have a file allocated to them
1246 * if all they do is call sock_create_kern().
1248 if (sk->sk_socket->file)
1249 f_flags = sk->sk_socket->file->f_flags;
1251 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1253 if (assoc_id)
1254 *assoc_id = asoc->assoc_id;
1255 err = sctp_wait_for_connect(asoc, &timeo);
1256 /* Note: the asoc may be freed after the return of
1257 * sctp_wait_for_connect.
1260 /* Don't free association on exit. */
1261 asoc = NULL;
1263 out_free:
1264 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1265 __func__, asoc, kaddrs, err);
1267 if (asoc) {
1268 /* sctp_primitive_ASSOCIATE may have added this association
1269 * To the hash table, try to unhash it, just in case, its a noop
1270 * if it wasn't hashed so we're safe
1272 sctp_association_free(asoc);
1274 return err;
1277 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1279 * API 8.9
1280 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1281 * sctp_assoc_t *asoc);
1283 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1284 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1285 * or IPv6 addresses.
1287 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1288 * Section 3.1.2 for this usage.
1290 * addrs is a pointer to an array of one or more socket addresses. Each
1291 * address is contained in its appropriate structure (i.e. struct
1292 * sockaddr_in or struct sockaddr_in6) the family of the address type
1293 * must be used to distengish the address length (note that this
1294 * representation is termed a "packed array" of addresses). The caller
1295 * specifies the number of addresses in the array with addrcnt.
1297 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1298 * the association id of the new association. On failure, sctp_connectx()
1299 * returns -1, and sets errno to the appropriate error code. The assoc_id
1300 * is not touched by the kernel.
1302 * For SCTP, the port given in each socket address must be the same, or
1303 * sctp_connectx() will fail, setting errno to EINVAL.
1305 * An application can use sctp_connectx to initiate an association with
1306 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1307 * allows a caller to specify multiple addresses at which a peer can be
1308 * reached. The way the SCTP stack uses the list of addresses to set up
1309 * the association is implementation dependent. This function only
1310 * specifies that the stack will try to make use of all the addresses in
1311 * the list when needed.
1313 * Note that the list of addresses passed in is only used for setting up
1314 * the association. It does not necessarily equal the set of addresses
1315 * the peer uses for the resulting association. If the caller wants to
1316 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1317 * retrieve them after the association has been set up.
1319 * Basically do nothing but copying the addresses from user to kernel
1320 * land and invoking either sctp_connectx(). This is used for tunneling
1321 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1323 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1324 * it.
1326 * sk The sk of the socket
1327 * addrs The pointer to the addresses in user land
1328 * addrssize Size of the addrs buffer
1330 * Returns >=0 if ok, <0 errno code on error.
1332 static int __sctp_setsockopt_connectx(struct sock *sk,
1333 struct sockaddr __user *addrs,
1334 int addrs_size,
1335 sctp_assoc_t *assoc_id)
1337 struct sockaddr *kaddrs;
1338 int err = 0;
1340 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1341 __func__, sk, addrs, addrs_size);
1343 if (unlikely(addrs_size <= 0))
1344 return -EINVAL;
1346 kaddrs = vmemdup_user(addrs, addrs_size);
1347 if (unlikely(IS_ERR(kaddrs)))
1348 return PTR_ERR(kaddrs);
1350 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1351 kvfree(kaddrs);
1353 return err;
1357 * This is an older interface. It's kept for backward compatibility
1358 * to the option that doesn't provide association id.
1360 static int sctp_setsockopt_connectx_old(struct sock *sk,
1361 struct sockaddr __user *addrs,
1362 int addrs_size)
1364 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1368 * New interface for the API. The since the API is done with a socket
1369 * option, to make it simple we feed back the association id is as a return
1370 * indication to the call. Error is always negative and association id is
1371 * always positive.
1373 static int sctp_setsockopt_connectx(struct sock *sk,
1374 struct sockaddr __user *addrs,
1375 int addrs_size)
1377 sctp_assoc_t assoc_id = 0;
1378 int err = 0;
1380 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1382 if (err)
1383 return err;
1384 else
1385 return assoc_id;
1389 * New (hopefully final) interface for the API.
1390 * We use the sctp_getaddrs_old structure so that use-space library
1391 * can avoid any unnecessary allocations. The only different part
1392 * is that we store the actual length of the address buffer into the
1393 * addrs_num structure member. That way we can re-use the existing
1394 * code.
1396 #ifdef CONFIG_COMPAT
1397 struct compat_sctp_getaddrs_old {
1398 sctp_assoc_t assoc_id;
1399 s32 addr_num;
1400 compat_uptr_t addrs; /* struct sockaddr * */
1402 #endif
1404 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1405 char __user *optval,
1406 int __user *optlen)
1408 struct sctp_getaddrs_old param;
1409 sctp_assoc_t assoc_id = 0;
1410 int err = 0;
1412 #ifdef CONFIG_COMPAT
1413 if (in_compat_syscall()) {
1414 struct compat_sctp_getaddrs_old param32;
1416 if (len < sizeof(param32))
1417 return -EINVAL;
1418 if (copy_from_user(&param32, optval, sizeof(param32)))
1419 return -EFAULT;
1421 param.assoc_id = param32.assoc_id;
1422 param.addr_num = param32.addr_num;
1423 param.addrs = compat_ptr(param32.addrs);
1424 } else
1425 #endif
1427 if (len < sizeof(param))
1428 return -EINVAL;
1429 if (copy_from_user(&param, optval, sizeof(param)))
1430 return -EFAULT;
1433 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1434 param.addrs, param.addr_num,
1435 &assoc_id);
1436 if (err == 0 || err == -EINPROGRESS) {
1437 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1438 return -EFAULT;
1439 if (put_user(sizeof(assoc_id), optlen))
1440 return -EFAULT;
1443 return err;
1446 /* API 3.1.4 close() - UDP Style Syntax
1447 * Applications use close() to perform graceful shutdown (as described in
1448 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1449 * by a UDP-style socket.
1451 * The syntax is
1453 * ret = close(int sd);
1455 * sd - the socket descriptor of the associations to be closed.
1457 * To gracefully shutdown a specific association represented by the
1458 * UDP-style socket, an application should use the sendmsg() call,
1459 * passing no user data, but including the appropriate flag in the
1460 * ancillary data (see Section xxxx).
1462 * If sd in the close() call is a branched-off socket representing only
1463 * one association, the shutdown is performed on that association only.
1465 * 4.1.6 close() - TCP Style Syntax
1467 * Applications use close() to gracefully close down an association.
1469 * The syntax is:
1471 * int close(int sd);
1473 * sd - the socket descriptor of the association to be closed.
1475 * After an application calls close() on a socket descriptor, no further
1476 * socket operations will succeed on that descriptor.
1478 * API 7.1.4 SO_LINGER
1480 * An application using the TCP-style socket can use this option to
1481 * perform the SCTP ABORT primitive. The linger option structure is:
1483 * struct linger {
1484 * int l_onoff; // option on/off
1485 * int l_linger; // linger time
1486 * };
1488 * To enable the option, set l_onoff to 1. If the l_linger value is set
1489 * to 0, calling close() is the same as the ABORT primitive. If the
1490 * value is set to a negative value, the setsockopt() call will return
1491 * an error. If the value is set to a positive value linger_time, the
1492 * close() can be blocked for at most linger_time ms. If the graceful
1493 * shutdown phase does not finish during this period, close() will
1494 * return but the graceful shutdown phase continues in the system.
1496 static void sctp_close(struct sock *sk, long timeout)
1498 struct net *net = sock_net(sk);
1499 struct sctp_endpoint *ep;
1500 struct sctp_association *asoc;
1501 struct list_head *pos, *temp;
1502 unsigned int data_was_unread;
1504 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1506 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1507 sk->sk_shutdown = SHUTDOWN_MASK;
1508 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1510 ep = sctp_sk(sk)->ep;
1512 /* Clean up any skbs sitting on the receive queue. */
1513 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1514 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1516 /* Walk all associations on an endpoint. */
1517 list_for_each_safe(pos, temp, &ep->asocs) {
1518 asoc = list_entry(pos, struct sctp_association, asocs);
1520 if (sctp_style(sk, TCP)) {
1521 /* A closed association can still be in the list if
1522 * it belongs to a TCP-style listening socket that is
1523 * not yet accepted. If so, free it. If not, send an
1524 * ABORT or SHUTDOWN based on the linger options.
1526 if (sctp_state(asoc, CLOSED)) {
1527 sctp_association_free(asoc);
1528 continue;
1532 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1533 !skb_queue_empty(&asoc->ulpq.reasm) ||
1534 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1535 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1536 struct sctp_chunk *chunk;
1538 chunk = sctp_make_abort_user(asoc, NULL, 0);
1539 sctp_primitive_ABORT(net, asoc, chunk);
1540 } else
1541 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1544 /* On a TCP-style socket, block for at most linger_time if set. */
1545 if (sctp_style(sk, TCP) && timeout)
1546 sctp_wait_for_close(sk, timeout);
1548 /* This will run the backlog queue. */
1549 release_sock(sk);
1551 /* Supposedly, no process has access to the socket, but
1552 * the net layers still may.
1553 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1554 * held and that should be grabbed before socket lock.
1556 spin_lock_bh(&net->sctp.addr_wq_lock);
1557 bh_lock_sock_nested(sk);
1559 /* Hold the sock, since sk_common_release() will put sock_put()
1560 * and we have just a little more cleanup.
1562 sock_hold(sk);
1563 sk_common_release(sk);
1565 bh_unlock_sock(sk);
1566 spin_unlock_bh(&net->sctp.addr_wq_lock);
1568 sock_put(sk);
1570 SCTP_DBG_OBJCNT_DEC(sock);
1573 /* Handle EPIPE error. */
1574 static int sctp_error(struct sock *sk, int flags, int err)
1576 if (err == -EPIPE)
1577 err = sock_error(sk) ? : -EPIPE;
1578 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1579 send_sig(SIGPIPE, current, 0);
1580 return err;
1583 /* API 3.1.3 sendmsg() - UDP Style Syntax
1585 * An application uses sendmsg() and recvmsg() calls to transmit data to
1586 * and receive data from its peer.
1588 * ssize_t sendmsg(int socket, const struct msghdr *message,
1589 * int flags);
1591 * socket - the socket descriptor of the endpoint.
1592 * message - pointer to the msghdr structure which contains a single
1593 * user message and possibly some ancillary data.
1595 * See Section 5 for complete description of the data
1596 * structures.
1598 * flags - flags sent or received with the user message, see Section
1599 * 5 for complete description of the flags.
1601 * Note: This function could use a rewrite especially when explicit
1602 * connect support comes in.
1604 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1606 static int sctp_msghdr_parse(const struct msghdr *msg,
1607 struct sctp_cmsgs *cmsgs);
1609 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1611 struct net *net = sock_net(sk);
1612 struct sctp_sock *sp;
1613 struct sctp_endpoint *ep;
1614 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1615 struct sctp_transport *transport, *chunk_tp;
1616 struct sctp_chunk *chunk;
1617 union sctp_addr to;
1618 struct sockaddr *msg_name = NULL;
1619 struct sctp_sndrcvinfo default_sinfo;
1620 struct sctp_sndrcvinfo *sinfo;
1621 struct sctp_initmsg *sinit;
1622 sctp_assoc_t associd = 0;
1623 struct sctp_cmsgs cmsgs = { NULL };
1624 enum sctp_scope scope;
1625 bool fill_sinfo_ttl = false, wait_connect = false;
1626 struct sctp_datamsg *datamsg;
1627 int msg_flags = msg->msg_flags;
1628 __u16 sinfo_flags = 0;
1629 long timeo;
1630 int err;
1632 err = 0;
1633 sp = sctp_sk(sk);
1634 ep = sp->ep;
1636 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1637 msg, msg_len, ep);
1639 /* We cannot send a message over a TCP-style listening socket. */
1640 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1641 err = -EPIPE;
1642 goto out_nounlock;
1645 /* Parse out the SCTP CMSGs. */
1646 err = sctp_msghdr_parse(msg, &cmsgs);
1647 if (err) {
1648 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1649 goto out_nounlock;
1652 /* Fetch the destination address for this packet. This
1653 * address only selects the association--it is not necessarily
1654 * the address we will send to.
1655 * For a peeled-off socket, msg_name is ignored.
1657 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1658 int msg_namelen = msg->msg_namelen;
1660 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1661 msg_namelen);
1662 if (err)
1663 return err;
1665 if (msg_namelen > sizeof(to))
1666 msg_namelen = sizeof(to);
1667 memcpy(&to, msg->msg_name, msg_namelen);
1668 msg_name = msg->msg_name;
1671 sinit = cmsgs.init;
1672 if (cmsgs.sinfo != NULL) {
1673 memset(&default_sinfo, 0, sizeof(default_sinfo));
1674 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1675 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1676 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1677 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1678 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1680 sinfo = &default_sinfo;
1681 fill_sinfo_ttl = true;
1682 } else {
1683 sinfo = cmsgs.srinfo;
1685 /* Did the user specify SNDINFO/SNDRCVINFO? */
1686 if (sinfo) {
1687 sinfo_flags = sinfo->sinfo_flags;
1688 associd = sinfo->sinfo_assoc_id;
1691 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1692 msg_len, sinfo_flags);
1694 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1695 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1696 err = -EINVAL;
1697 goto out_nounlock;
1700 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1701 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1702 * If SCTP_ABORT is set, the message length could be non zero with
1703 * the msg_iov set to the user abort reason.
1705 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1706 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1707 err = -EINVAL;
1708 goto out_nounlock;
1711 /* If SCTP_ADDR_OVER is set, there must be an address
1712 * specified in msg_name.
1714 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1715 err = -EINVAL;
1716 goto out_nounlock;
1719 transport = NULL;
1721 pr_debug("%s: about to look up association\n", __func__);
1723 lock_sock(sk);
1725 /* If a msg_name has been specified, assume this is to be used. */
1726 if (msg_name) {
1727 /* Look for a matching association on the endpoint. */
1728 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1730 /* If we could not find a matching association on the
1731 * endpoint, make sure that it is not a TCP-style
1732 * socket that already has an association or there is
1733 * no peeled-off association on another socket.
1735 if (!asoc &&
1736 ((sctp_style(sk, TCP) &&
1737 (sctp_sstate(sk, ESTABLISHED) ||
1738 sctp_sstate(sk, CLOSING))) ||
1739 sctp_endpoint_is_peeled_off(ep, &to))) {
1740 err = -EADDRNOTAVAIL;
1741 goto out_unlock;
1743 } else {
1744 asoc = sctp_id2assoc(sk, associd);
1745 if (!asoc) {
1746 err = -EPIPE;
1747 goto out_unlock;
1751 if (asoc) {
1752 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1754 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1755 * socket that has an association in CLOSED state. This can
1756 * happen when an accepted socket has an association that is
1757 * already CLOSED.
1759 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1760 err = -EPIPE;
1761 goto out_unlock;
1764 if (sinfo_flags & SCTP_EOF) {
1765 pr_debug("%s: shutting down association:%p\n",
1766 __func__, asoc);
1768 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1769 err = 0;
1770 goto out_unlock;
1772 if (sinfo_flags & SCTP_ABORT) {
1774 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1775 if (!chunk) {
1776 err = -ENOMEM;
1777 goto out_unlock;
1780 pr_debug("%s: aborting association:%p\n",
1781 __func__, asoc);
1783 sctp_primitive_ABORT(net, asoc, chunk);
1784 err = 0;
1785 goto out_unlock;
1789 /* Do we need to create the association? */
1790 if (!asoc) {
1791 pr_debug("%s: there is no association yet\n", __func__);
1793 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1794 err = -EINVAL;
1795 goto out_unlock;
1798 /* Check for invalid stream against the stream counts,
1799 * either the default or the user specified stream counts.
1801 if (sinfo) {
1802 if (!sinit || !sinit->sinit_num_ostreams) {
1803 /* Check against the defaults. */
1804 if (sinfo->sinfo_stream >=
1805 sp->initmsg.sinit_num_ostreams) {
1806 err = -EINVAL;
1807 goto out_unlock;
1809 } else {
1810 /* Check against the requested. */
1811 if (sinfo->sinfo_stream >=
1812 sinit->sinit_num_ostreams) {
1813 err = -EINVAL;
1814 goto out_unlock;
1820 * API 3.1.2 bind() - UDP Style Syntax
1821 * If a bind() or sctp_bindx() is not called prior to a
1822 * sendmsg() call that initiates a new association, the
1823 * system picks an ephemeral port and will choose an address
1824 * set equivalent to binding with a wildcard address.
1826 if (!ep->base.bind_addr.port) {
1827 if (sctp_autobind(sk)) {
1828 err = -EAGAIN;
1829 goto out_unlock;
1831 } else {
1833 * If an unprivileged user inherits a one-to-many
1834 * style socket with open associations on a privileged
1835 * port, it MAY be permitted to accept new associations,
1836 * but it SHOULD NOT be permitted to open new
1837 * associations.
1839 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1840 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1841 err = -EACCES;
1842 goto out_unlock;
1846 scope = sctp_scope(&to);
1847 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1848 if (!new_asoc) {
1849 err = -ENOMEM;
1850 goto out_unlock;
1852 asoc = new_asoc;
1853 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1854 if (err < 0) {
1855 err = -ENOMEM;
1856 goto out_free;
1859 /* If the SCTP_INIT ancillary data is specified, set all
1860 * the association init values accordingly.
1862 if (sinit) {
1863 if (sinit->sinit_num_ostreams) {
1864 __u16 outcnt = sinit->sinit_num_ostreams;
1866 asoc->c.sinit_num_ostreams = outcnt;
1867 /* outcnt has been changed, so re-init stream */
1868 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1869 GFP_KERNEL);
1870 if (err)
1871 goto out_free;
1873 if (sinit->sinit_max_instreams) {
1874 asoc->c.sinit_max_instreams =
1875 sinit->sinit_max_instreams;
1877 if (sinit->sinit_max_attempts) {
1878 asoc->max_init_attempts
1879 = sinit->sinit_max_attempts;
1881 if (sinit->sinit_max_init_timeo) {
1882 asoc->max_init_timeo =
1883 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1887 /* Prime the peer's transport structures. */
1888 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1889 if (!transport) {
1890 err = -ENOMEM;
1891 goto out_free;
1895 /* ASSERT: we have a valid association at this point. */
1896 pr_debug("%s: we have a valid association\n", __func__);
1898 if (!sinfo) {
1899 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1900 * one with some defaults.
1902 memset(&default_sinfo, 0, sizeof(default_sinfo));
1903 default_sinfo.sinfo_stream = asoc->default_stream;
1904 default_sinfo.sinfo_flags = asoc->default_flags;
1905 default_sinfo.sinfo_ppid = asoc->default_ppid;
1906 default_sinfo.sinfo_context = asoc->default_context;
1907 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1908 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1910 sinfo = &default_sinfo;
1911 } else if (fill_sinfo_ttl) {
1912 /* In case SNDINFO was specified, we still need to fill
1913 * it with a default ttl from the assoc here.
1915 sinfo->sinfo_timetolive = asoc->default_timetolive;
1918 /* API 7.1.7, the sndbuf size per association bounds the
1919 * maximum size of data that can be sent in a single send call.
1921 if (msg_len > sk->sk_sndbuf) {
1922 err = -EMSGSIZE;
1923 goto out_free;
1926 if (asoc->pmtu_pending)
1927 sctp_assoc_pending_pmtu(asoc);
1929 /* If fragmentation is disabled and the message length exceeds the
1930 * association fragmentation point, return EMSGSIZE. The I-D
1931 * does not specify what this error is, but this looks like
1932 * a great fit.
1934 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1935 err = -EMSGSIZE;
1936 goto out_free;
1939 /* Check for invalid stream. */
1940 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1941 err = -EINVAL;
1942 goto out_free;
1945 /* Allocate sctp_stream_out_ext if not already done */
1946 if (unlikely(!asoc->stream.out[sinfo->sinfo_stream].ext)) {
1947 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1948 if (err)
1949 goto out_free;
1952 if (sctp_wspace(asoc) < msg_len)
1953 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1955 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1956 if (!sctp_wspace(asoc)) {
1957 /* sk can be changed by peel off when waiting for buf. */
1958 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1959 if (err) {
1960 if (err == -ESRCH) {
1961 /* asoc is already dead. */
1962 new_asoc = NULL;
1963 err = -EPIPE;
1965 goto out_free;
1969 /* If an address is passed with the sendto/sendmsg call, it is used
1970 * to override the primary destination address in the TCP model, or
1971 * when SCTP_ADDR_OVER flag is set in the UDP model.
1973 if ((sctp_style(sk, TCP) && msg_name) ||
1974 (sinfo_flags & SCTP_ADDR_OVER)) {
1975 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1976 if (!chunk_tp) {
1977 err = -EINVAL;
1978 goto out_free;
1980 } else
1981 chunk_tp = NULL;
1983 /* Auto-connect, if we aren't connected already. */
1984 if (sctp_state(asoc, CLOSED)) {
1985 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1986 if (err < 0)
1987 goto out_free;
1989 /* If stream interleave is enabled, wait_connect has to be
1990 * done earlier than data enqueue, as it needs to make data
1991 * or idata according to asoc->intl_enable which is set
1992 * after connection is done.
1994 if (sctp_sk(asoc->base.sk)->strm_interleave) {
1995 timeo = sock_sndtimeo(sk, 0);
1996 err = sctp_wait_for_connect(asoc, &timeo);
1997 if (err)
1998 goto out_unlock;
1999 } else {
2000 wait_connect = true;
2003 pr_debug("%s: we associated primitively\n", __func__);
2006 /* Break the message into multiple chunks of maximum size. */
2007 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
2008 if (IS_ERR(datamsg)) {
2009 err = PTR_ERR(datamsg);
2010 goto out_free;
2012 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
2014 /* Now send the (possibly) fragmented message. */
2015 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
2016 sctp_chunk_hold(chunk);
2018 /* Do accounting for the write space. */
2019 sctp_set_owner_w(chunk);
2021 chunk->transport = chunk_tp;
2024 /* Send it to the lower layers. Note: all chunks
2025 * must either fail or succeed. The lower layer
2026 * works that way today. Keep it that way or this
2027 * breaks.
2029 err = sctp_primitive_SEND(net, asoc, datamsg);
2030 /* Did the lower layer accept the chunk? */
2031 if (err) {
2032 sctp_datamsg_free(datamsg);
2033 goto out_free;
2036 pr_debug("%s: we sent primitively\n", __func__);
2038 sctp_datamsg_put(datamsg);
2039 err = msg_len;
2041 if (unlikely(wait_connect)) {
2042 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2043 sctp_wait_for_connect(asoc, &timeo);
2046 /* If we are already past ASSOCIATE, the lower
2047 * layers are responsible for association cleanup.
2049 goto out_unlock;
2051 out_free:
2052 if (new_asoc)
2053 sctp_association_free(asoc);
2054 out_unlock:
2055 release_sock(sk);
2057 out_nounlock:
2058 return sctp_error(sk, msg_flags, err);
2060 #if 0
2061 do_sock_err:
2062 if (msg_len)
2063 err = msg_len;
2064 else
2065 err = sock_error(sk);
2066 goto out;
2068 do_interrupted:
2069 if (msg_len)
2070 err = msg_len;
2071 goto out;
2072 #endif /* 0 */
2075 /* This is an extended version of skb_pull() that removes the data from the
2076 * start of a skb even when data is spread across the list of skb's in the
2077 * frag_list. len specifies the total amount of data that needs to be removed.
2078 * when 'len' bytes could be removed from the skb, it returns 0.
2079 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2080 * could not be removed.
2082 static int sctp_skb_pull(struct sk_buff *skb, int len)
2084 struct sk_buff *list;
2085 int skb_len = skb_headlen(skb);
2086 int rlen;
2088 if (len <= skb_len) {
2089 __skb_pull(skb, len);
2090 return 0;
2092 len -= skb_len;
2093 __skb_pull(skb, skb_len);
2095 skb_walk_frags(skb, list) {
2096 rlen = sctp_skb_pull(list, len);
2097 skb->len -= (len-rlen);
2098 skb->data_len -= (len-rlen);
2100 if (!rlen)
2101 return 0;
2103 len = rlen;
2106 return len;
2109 /* API 3.1.3 recvmsg() - UDP Style Syntax
2111 * ssize_t recvmsg(int socket, struct msghdr *message,
2112 * int flags);
2114 * socket - the socket descriptor of the endpoint.
2115 * message - pointer to the msghdr structure which contains a single
2116 * user message and possibly some ancillary data.
2118 * See Section 5 for complete description of the data
2119 * structures.
2121 * flags - flags sent or received with the user message, see Section
2122 * 5 for complete description of the flags.
2124 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2125 int noblock, int flags, int *addr_len)
2127 struct sctp_ulpevent *event = NULL;
2128 struct sctp_sock *sp = sctp_sk(sk);
2129 struct sk_buff *skb, *head_skb;
2130 int copied;
2131 int err = 0;
2132 int skb_len;
2134 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2135 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2136 addr_len);
2138 lock_sock(sk);
2140 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2141 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2142 err = -ENOTCONN;
2143 goto out;
2146 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2147 if (!skb)
2148 goto out;
2150 /* Get the total length of the skb including any skb's in the
2151 * frag_list.
2153 skb_len = skb->len;
2155 copied = skb_len;
2156 if (copied > len)
2157 copied = len;
2159 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2161 event = sctp_skb2event(skb);
2163 if (err)
2164 goto out_free;
2166 if (event->chunk && event->chunk->head_skb)
2167 head_skb = event->chunk->head_skb;
2168 else
2169 head_skb = skb;
2170 sock_recv_ts_and_drops(msg, sk, head_skb);
2171 if (sctp_ulpevent_is_notification(event)) {
2172 msg->msg_flags |= MSG_NOTIFICATION;
2173 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2174 } else {
2175 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2178 /* Check if we allow SCTP_NXTINFO. */
2179 if (sp->recvnxtinfo)
2180 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2181 /* Check if we allow SCTP_RCVINFO. */
2182 if (sp->recvrcvinfo)
2183 sctp_ulpevent_read_rcvinfo(event, msg);
2184 /* Check if we allow SCTP_SNDRCVINFO. */
2185 if (sp->subscribe.sctp_data_io_event)
2186 sctp_ulpevent_read_sndrcvinfo(event, msg);
2188 err = copied;
2190 /* If skb's length exceeds the user's buffer, update the skb and
2191 * push it back to the receive_queue so that the next call to
2192 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2194 if (skb_len > copied) {
2195 msg->msg_flags &= ~MSG_EOR;
2196 if (flags & MSG_PEEK)
2197 goto out_free;
2198 sctp_skb_pull(skb, copied);
2199 skb_queue_head(&sk->sk_receive_queue, skb);
2201 /* When only partial message is copied to the user, increase
2202 * rwnd by that amount. If all the data in the skb is read,
2203 * rwnd is updated when the event is freed.
2205 if (!sctp_ulpevent_is_notification(event))
2206 sctp_assoc_rwnd_increase(event->asoc, copied);
2207 goto out;
2208 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2209 (event->msg_flags & MSG_EOR))
2210 msg->msg_flags |= MSG_EOR;
2211 else
2212 msg->msg_flags &= ~MSG_EOR;
2214 out_free:
2215 if (flags & MSG_PEEK) {
2216 /* Release the skb reference acquired after peeking the skb in
2217 * sctp_skb_recv_datagram().
2219 kfree_skb(skb);
2220 } else {
2221 /* Free the event which includes releasing the reference to
2222 * the owner of the skb, freeing the skb and updating the
2223 * rwnd.
2225 sctp_ulpevent_free(event);
2227 out:
2228 release_sock(sk);
2229 return err;
2232 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2234 * This option is a on/off flag. If enabled no SCTP message
2235 * fragmentation will be performed. Instead if a message being sent
2236 * exceeds the current PMTU size, the message will NOT be sent and
2237 * instead a error will be indicated to the user.
2239 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2240 char __user *optval,
2241 unsigned int optlen)
2243 int val;
2245 if (optlen < sizeof(int))
2246 return -EINVAL;
2248 if (get_user(val, (int __user *)optval))
2249 return -EFAULT;
2251 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2253 return 0;
2256 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2257 unsigned int optlen)
2259 struct sctp_association *asoc;
2260 struct sctp_ulpevent *event;
2262 if (optlen > sizeof(struct sctp_event_subscribe))
2263 return -EINVAL;
2264 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2265 return -EFAULT;
2267 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2268 * if there is no data to be sent or retransmit, the stack will
2269 * immediately send up this notification.
2271 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2272 &sctp_sk(sk)->subscribe)) {
2273 asoc = sctp_id2assoc(sk, 0);
2275 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2276 event = sctp_ulpevent_make_sender_dry_event(asoc,
2277 GFP_USER | __GFP_NOWARN);
2278 if (!event)
2279 return -ENOMEM;
2281 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2285 return 0;
2288 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2290 * This socket option is applicable to the UDP-style socket only. When
2291 * set it will cause associations that are idle for more than the
2292 * specified number of seconds to automatically close. An association
2293 * being idle is defined an association that has NOT sent or received
2294 * user data. The special value of '0' indicates that no automatic
2295 * close of any associations should be performed. The option expects an
2296 * integer defining the number of seconds of idle time before an
2297 * association is closed.
2299 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2300 unsigned int optlen)
2302 struct sctp_sock *sp = sctp_sk(sk);
2303 struct net *net = sock_net(sk);
2305 /* Applicable to UDP-style socket only */
2306 if (sctp_style(sk, TCP))
2307 return -EOPNOTSUPP;
2308 if (optlen != sizeof(int))
2309 return -EINVAL;
2310 if (copy_from_user(&sp->autoclose, optval, optlen))
2311 return -EFAULT;
2313 if (sp->autoclose > net->sctp.max_autoclose)
2314 sp->autoclose = net->sctp.max_autoclose;
2316 return 0;
2319 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2321 * Applications can enable or disable heartbeats for any peer address of
2322 * an association, modify an address's heartbeat interval, force a
2323 * heartbeat to be sent immediately, and adjust the address's maximum
2324 * number of retransmissions sent before an address is considered
2325 * unreachable. The following structure is used to access and modify an
2326 * address's parameters:
2328 * struct sctp_paddrparams {
2329 * sctp_assoc_t spp_assoc_id;
2330 * struct sockaddr_storage spp_address;
2331 * uint32_t spp_hbinterval;
2332 * uint16_t spp_pathmaxrxt;
2333 * uint32_t spp_pathmtu;
2334 * uint32_t spp_sackdelay;
2335 * uint32_t spp_flags;
2336 * };
2338 * spp_assoc_id - (one-to-many style socket) This is filled in the
2339 * application, and identifies the association for
2340 * this query.
2341 * spp_address - This specifies which address is of interest.
2342 * spp_hbinterval - This contains the value of the heartbeat interval,
2343 * in milliseconds. If a value of zero
2344 * is present in this field then no changes are to
2345 * be made to this parameter.
2346 * spp_pathmaxrxt - This contains the maximum number of
2347 * retransmissions before this address shall be
2348 * considered unreachable. If a value of zero
2349 * is present in this field then no changes are to
2350 * be made to this parameter.
2351 * spp_pathmtu - When Path MTU discovery is disabled the value
2352 * specified here will be the "fixed" path mtu.
2353 * Note that if the spp_address field is empty
2354 * then all associations on this address will
2355 * have this fixed path mtu set upon them.
2357 * spp_sackdelay - When delayed sack is enabled, this value specifies
2358 * the number of milliseconds that sacks will be delayed
2359 * for. This value will apply to all addresses of an
2360 * association if the spp_address field is empty. Note
2361 * also, that if delayed sack is enabled and this
2362 * value is set to 0, no change is made to the last
2363 * recorded delayed sack timer value.
2365 * spp_flags - These flags are used to control various features
2366 * on an association. The flag field may contain
2367 * zero or more of the following options.
2369 * SPP_HB_ENABLE - Enable heartbeats on the
2370 * specified address. Note that if the address
2371 * field is empty all addresses for the association
2372 * have heartbeats enabled upon them.
2374 * SPP_HB_DISABLE - Disable heartbeats on the
2375 * speicifed address. Note that if the address
2376 * field is empty all addresses for the association
2377 * will have their heartbeats disabled. Note also
2378 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2379 * mutually exclusive, only one of these two should
2380 * be specified. Enabling both fields will have
2381 * undetermined results.
2383 * SPP_HB_DEMAND - Request a user initiated heartbeat
2384 * to be made immediately.
2386 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2387 * heartbeat delayis to be set to the value of 0
2388 * milliseconds.
2390 * SPP_PMTUD_ENABLE - This field will enable PMTU
2391 * discovery upon the specified address. Note that
2392 * if the address feild is empty then all addresses
2393 * on the association are effected.
2395 * SPP_PMTUD_DISABLE - This field will disable PMTU
2396 * discovery upon the specified address. Note that
2397 * if the address feild is empty then all addresses
2398 * on the association are effected. Not also that
2399 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2400 * exclusive. Enabling both will have undetermined
2401 * results.
2403 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2404 * on delayed sack. The time specified in spp_sackdelay
2405 * is used to specify the sack delay for this address. Note
2406 * that if spp_address is empty then all addresses will
2407 * enable delayed sack and take on the sack delay
2408 * value specified in spp_sackdelay.
2409 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2410 * off delayed sack. If the spp_address field is blank then
2411 * delayed sack is disabled for the entire association. Note
2412 * also that this field is mutually exclusive to
2413 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2414 * results.
2416 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2417 struct sctp_transport *trans,
2418 struct sctp_association *asoc,
2419 struct sctp_sock *sp,
2420 int hb_change,
2421 int pmtud_change,
2422 int sackdelay_change)
2424 int error;
2426 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2427 struct net *net = sock_net(trans->asoc->base.sk);
2429 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2430 if (error)
2431 return error;
2434 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2435 * this field is ignored. Note also that a value of zero indicates
2436 * the current setting should be left unchanged.
2438 if (params->spp_flags & SPP_HB_ENABLE) {
2440 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2441 * set. This lets us use 0 value when this flag
2442 * is set.
2444 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2445 params->spp_hbinterval = 0;
2447 if (params->spp_hbinterval ||
2448 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2449 if (trans) {
2450 trans->hbinterval =
2451 msecs_to_jiffies(params->spp_hbinterval);
2452 } else if (asoc) {
2453 asoc->hbinterval =
2454 msecs_to_jiffies(params->spp_hbinterval);
2455 } else {
2456 sp->hbinterval = params->spp_hbinterval;
2461 if (hb_change) {
2462 if (trans) {
2463 trans->param_flags =
2464 (trans->param_flags & ~SPP_HB) | hb_change;
2465 } else if (asoc) {
2466 asoc->param_flags =
2467 (asoc->param_flags & ~SPP_HB) | hb_change;
2468 } else {
2469 sp->param_flags =
2470 (sp->param_flags & ~SPP_HB) | hb_change;
2474 /* When Path MTU discovery is disabled the value specified here will
2475 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2476 * include the flag SPP_PMTUD_DISABLE for this field to have any
2477 * effect).
2479 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2480 if (trans) {
2481 trans->pathmtu = params->spp_pathmtu;
2482 sctp_assoc_sync_pmtu(asoc);
2483 } else if (asoc) {
2484 asoc->pathmtu = params->spp_pathmtu;
2485 } else {
2486 sp->pathmtu = params->spp_pathmtu;
2490 if (pmtud_change) {
2491 if (trans) {
2492 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2493 (params->spp_flags & SPP_PMTUD_ENABLE);
2494 trans->param_flags =
2495 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2496 if (update) {
2497 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2498 sctp_assoc_sync_pmtu(asoc);
2500 } else if (asoc) {
2501 asoc->param_flags =
2502 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2503 } else {
2504 sp->param_flags =
2505 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2509 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2510 * value of this field is ignored. Note also that a value of zero
2511 * indicates the current setting should be left unchanged.
2513 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2514 if (trans) {
2515 trans->sackdelay =
2516 msecs_to_jiffies(params->spp_sackdelay);
2517 } else if (asoc) {
2518 asoc->sackdelay =
2519 msecs_to_jiffies(params->spp_sackdelay);
2520 } else {
2521 sp->sackdelay = params->spp_sackdelay;
2525 if (sackdelay_change) {
2526 if (trans) {
2527 trans->param_flags =
2528 (trans->param_flags & ~SPP_SACKDELAY) |
2529 sackdelay_change;
2530 } else if (asoc) {
2531 asoc->param_flags =
2532 (asoc->param_flags & ~SPP_SACKDELAY) |
2533 sackdelay_change;
2534 } else {
2535 sp->param_flags =
2536 (sp->param_flags & ~SPP_SACKDELAY) |
2537 sackdelay_change;
2541 /* Note that a value of zero indicates the current setting should be
2542 left unchanged.
2544 if (params->spp_pathmaxrxt) {
2545 if (trans) {
2546 trans->pathmaxrxt = params->spp_pathmaxrxt;
2547 } else if (asoc) {
2548 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2549 } else {
2550 sp->pathmaxrxt = params->spp_pathmaxrxt;
2554 return 0;
2557 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2558 char __user *optval,
2559 unsigned int optlen)
2561 struct sctp_paddrparams params;
2562 struct sctp_transport *trans = NULL;
2563 struct sctp_association *asoc = NULL;
2564 struct sctp_sock *sp = sctp_sk(sk);
2565 int error;
2566 int hb_change, pmtud_change, sackdelay_change;
2568 if (optlen != sizeof(struct sctp_paddrparams))
2569 return -EINVAL;
2571 if (copy_from_user(&params, optval, optlen))
2572 return -EFAULT;
2574 /* Validate flags and value parameters. */
2575 hb_change = params.spp_flags & SPP_HB;
2576 pmtud_change = params.spp_flags & SPP_PMTUD;
2577 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2579 if (hb_change == SPP_HB ||
2580 pmtud_change == SPP_PMTUD ||
2581 sackdelay_change == SPP_SACKDELAY ||
2582 params.spp_sackdelay > 500 ||
2583 (params.spp_pathmtu &&
2584 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2585 return -EINVAL;
2587 /* If an address other than INADDR_ANY is specified, and
2588 * no transport is found, then the request is invalid.
2590 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2591 trans = sctp_addr_id2transport(sk, &params.spp_address,
2592 params.spp_assoc_id);
2593 if (!trans)
2594 return -EINVAL;
2597 /* Get association, if assoc_id != 0 and the socket is a one
2598 * to many style socket, and an association was not found, then
2599 * the id was invalid.
2601 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2602 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2603 return -EINVAL;
2605 /* Heartbeat demand can only be sent on a transport or
2606 * association, but not a socket.
2608 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2609 return -EINVAL;
2611 /* Process parameters. */
2612 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2613 hb_change, pmtud_change,
2614 sackdelay_change);
2616 if (error)
2617 return error;
2619 /* If changes are for association, also apply parameters to each
2620 * transport.
2622 if (!trans && asoc) {
2623 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2624 transports) {
2625 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2626 hb_change, pmtud_change,
2627 sackdelay_change);
2631 return 0;
2634 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2636 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2639 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2641 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2645 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2647 * This option will effect the way delayed acks are performed. This
2648 * option allows you to get or set the delayed ack time, in
2649 * milliseconds. It also allows changing the delayed ack frequency.
2650 * Changing the frequency to 1 disables the delayed sack algorithm. If
2651 * the assoc_id is 0, then this sets or gets the endpoints default
2652 * values. If the assoc_id field is non-zero, then the set or get
2653 * effects the specified association for the one to many model (the
2654 * assoc_id field is ignored by the one to one model). Note that if
2655 * sack_delay or sack_freq are 0 when setting this option, then the
2656 * current values will remain unchanged.
2658 * struct sctp_sack_info {
2659 * sctp_assoc_t sack_assoc_id;
2660 * uint32_t sack_delay;
2661 * uint32_t sack_freq;
2662 * };
2664 * sack_assoc_id - This parameter, indicates which association the user
2665 * is performing an action upon. Note that if this field's value is
2666 * zero then the endpoints default value is changed (effecting future
2667 * associations only).
2669 * sack_delay - This parameter contains the number of milliseconds that
2670 * the user is requesting the delayed ACK timer be set to. Note that
2671 * this value is defined in the standard to be between 200 and 500
2672 * milliseconds.
2674 * sack_freq - This parameter contains the number of packets that must
2675 * be received before a sack is sent without waiting for the delay
2676 * timer to expire. The default value for this is 2, setting this
2677 * value to 1 will disable the delayed sack algorithm.
2680 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2681 char __user *optval, unsigned int optlen)
2683 struct sctp_sack_info params;
2684 struct sctp_transport *trans = NULL;
2685 struct sctp_association *asoc = NULL;
2686 struct sctp_sock *sp = sctp_sk(sk);
2688 if (optlen == sizeof(struct sctp_sack_info)) {
2689 if (copy_from_user(&params, optval, optlen))
2690 return -EFAULT;
2692 if (params.sack_delay == 0 && params.sack_freq == 0)
2693 return 0;
2694 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2695 pr_warn_ratelimited(DEPRECATED
2696 "%s (pid %d) "
2697 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2698 "Use struct sctp_sack_info instead\n",
2699 current->comm, task_pid_nr(current));
2700 if (copy_from_user(&params, optval, optlen))
2701 return -EFAULT;
2703 if (params.sack_delay == 0)
2704 params.sack_freq = 1;
2705 else
2706 params.sack_freq = 0;
2707 } else
2708 return -EINVAL;
2710 /* Validate value parameter. */
2711 if (params.sack_delay > 500)
2712 return -EINVAL;
2714 /* Get association, if sack_assoc_id != 0 and the socket is a one
2715 * to many style socket, and an association was not found, then
2716 * the id was invalid.
2718 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2719 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2720 return -EINVAL;
2722 if (params.sack_delay) {
2723 if (asoc) {
2724 asoc->sackdelay =
2725 msecs_to_jiffies(params.sack_delay);
2726 asoc->param_flags =
2727 sctp_spp_sackdelay_enable(asoc->param_flags);
2728 } else {
2729 sp->sackdelay = params.sack_delay;
2730 sp->param_flags =
2731 sctp_spp_sackdelay_enable(sp->param_flags);
2735 if (params.sack_freq == 1) {
2736 if (asoc) {
2737 asoc->param_flags =
2738 sctp_spp_sackdelay_disable(asoc->param_flags);
2739 } else {
2740 sp->param_flags =
2741 sctp_spp_sackdelay_disable(sp->param_flags);
2743 } else if (params.sack_freq > 1) {
2744 if (asoc) {
2745 asoc->sackfreq = params.sack_freq;
2746 asoc->param_flags =
2747 sctp_spp_sackdelay_enable(asoc->param_flags);
2748 } else {
2749 sp->sackfreq = params.sack_freq;
2750 sp->param_flags =
2751 sctp_spp_sackdelay_enable(sp->param_flags);
2755 /* If change is for association, also apply to each transport. */
2756 if (asoc) {
2757 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2758 transports) {
2759 if (params.sack_delay) {
2760 trans->sackdelay =
2761 msecs_to_jiffies(params.sack_delay);
2762 trans->param_flags =
2763 sctp_spp_sackdelay_enable(trans->param_flags);
2765 if (params.sack_freq == 1) {
2766 trans->param_flags =
2767 sctp_spp_sackdelay_disable(trans->param_flags);
2768 } else if (params.sack_freq > 1) {
2769 trans->sackfreq = params.sack_freq;
2770 trans->param_flags =
2771 sctp_spp_sackdelay_enable(trans->param_flags);
2776 return 0;
2779 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2781 * Applications can specify protocol parameters for the default association
2782 * initialization. The option name argument to setsockopt() and getsockopt()
2783 * is SCTP_INITMSG.
2785 * Setting initialization parameters is effective only on an unconnected
2786 * socket (for UDP-style sockets only future associations are effected
2787 * by the change). With TCP-style sockets, this option is inherited by
2788 * sockets derived from a listener socket.
2790 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2792 struct sctp_initmsg sinit;
2793 struct sctp_sock *sp = sctp_sk(sk);
2795 if (optlen != sizeof(struct sctp_initmsg))
2796 return -EINVAL;
2797 if (copy_from_user(&sinit, optval, optlen))
2798 return -EFAULT;
2800 if (sinit.sinit_num_ostreams)
2801 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2802 if (sinit.sinit_max_instreams)
2803 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2804 if (sinit.sinit_max_attempts)
2805 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2806 if (sinit.sinit_max_init_timeo)
2807 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2809 return 0;
2813 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2815 * Applications that wish to use the sendto() system call may wish to
2816 * specify a default set of parameters that would normally be supplied
2817 * through the inclusion of ancillary data. This socket option allows
2818 * such an application to set the default sctp_sndrcvinfo structure.
2819 * The application that wishes to use this socket option simply passes
2820 * in to this call the sctp_sndrcvinfo structure defined in Section
2821 * 5.2.2) The input parameters accepted by this call include
2822 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2823 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2824 * to this call if the caller is using the UDP model.
2826 static int sctp_setsockopt_default_send_param(struct sock *sk,
2827 char __user *optval,
2828 unsigned int optlen)
2830 struct sctp_sock *sp = sctp_sk(sk);
2831 struct sctp_association *asoc;
2832 struct sctp_sndrcvinfo info;
2834 if (optlen != sizeof(info))
2835 return -EINVAL;
2836 if (copy_from_user(&info, optval, optlen))
2837 return -EFAULT;
2838 if (info.sinfo_flags &
2839 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2840 SCTP_ABORT | SCTP_EOF))
2841 return -EINVAL;
2843 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2844 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2845 return -EINVAL;
2846 if (asoc) {
2847 asoc->default_stream = info.sinfo_stream;
2848 asoc->default_flags = info.sinfo_flags;
2849 asoc->default_ppid = info.sinfo_ppid;
2850 asoc->default_context = info.sinfo_context;
2851 asoc->default_timetolive = info.sinfo_timetolive;
2852 } else {
2853 sp->default_stream = info.sinfo_stream;
2854 sp->default_flags = info.sinfo_flags;
2855 sp->default_ppid = info.sinfo_ppid;
2856 sp->default_context = info.sinfo_context;
2857 sp->default_timetolive = info.sinfo_timetolive;
2860 return 0;
2863 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2864 * (SCTP_DEFAULT_SNDINFO)
2866 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2867 char __user *optval,
2868 unsigned int optlen)
2870 struct sctp_sock *sp = sctp_sk(sk);
2871 struct sctp_association *asoc;
2872 struct sctp_sndinfo info;
2874 if (optlen != sizeof(info))
2875 return -EINVAL;
2876 if (copy_from_user(&info, optval, optlen))
2877 return -EFAULT;
2878 if (info.snd_flags &
2879 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2880 SCTP_ABORT | SCTP_EOF))
2881 return -EINVAL;
2883 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2884 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2885 return -EINVAL;
2886 if (asoc) {
2887 asoc->default_stream = info.snd_sid;
2888 asoc->default_flags = info.snd_flags;
2889 asoc->default_ppid = info.snd_ppid;
2890 asoc->default_context = info.snd_context;
2891 } else {
2892 sp->default_stream = info.snd_sid;
2893 sp->default_flags = info.snd_flags;
2894 sp->default_ppid = info.snd_ppid;
2895 sp->default_context = info.snd_context;
2898 return 0;
2901 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2903 * Requests that the local SCTP stack use the enclosed peer address as
2904 * the association primary. The enclosed address must be one of the
2905 * association peer's addresses.
2907 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2908 unsigned int optlen)
2910 struct sctp_prim prim;
2911 struct sctp_transport *trans;
2913 if (optlen != sizeof(struct sctp_prim))
2914 return -EINVAL;
2916 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2917 return -EFAULT;
2919 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2920 if (!trans)
2921 return -EINVAL;
2923 sctp_assoc_set_primary(trans->asoc, trans);
2925 return 0;
2929 * 7.1.5 SCTP_NODELAY
2931 * Turn on/off any Nagle-like algorithm. This means that packets are
2932 * generally sent as soon as possible and no unnecessary delays are
2933 * introduced, at the cost of more packets in the network. Expects an
2934 * integer boolean flag.
2936 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2937 unsigned int optlen)
2939 int val;
2941 if (optlen < sizeof(int))
2942 return -EINVAL;
2943 if (get_user(val, (int __user *)optval))
2944 return -EFAULT;
2946 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2947 return 0;
2952 * 7.1.1 SCTP_RTOINFO
2954 * The protocol parameters used to initialize and bound retransmission
2955 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2956 * and modify these parameters.
2957 * All parameters are time values, in milliseconds. A value of 0, when
2958 * modifying the parameters, indicates that the current value should not
2959 * be changed.
2962 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2964 struct sctp_rtoinfo rtoinfo;
2965 struct sctp_association *asoc;
2966 unsigned long rto_min, rto_max;
2967 struct sctp_sock *sp = sctp_sk(sk);
2969 if (optlen != sizeof (struct sctp_rtoinfo))
2970 return -EINVAL;
2972 if (copy_from_user(&rtoinfo, optval, optlen))
2973 return -EFAULT;
2975 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2977 /* Set the values to the specific association */
2978 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2979 return -EINVAL;
2981 rto_max = rtoinfo.srto_max;
2982 rto_min = rtoinfo.srto_min;
2984 if (rto_max)
2985 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2986 else
2987 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2989 if (rto_min)
2990 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2991 else
2992 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2994 if (rto_min > rto_max)
2995 return -EINVAL;
2997 if (asoc) {
2998 if (rtoinfo.srto_initial != 0)
2999 asoc->rto_initial =
3000 msecs_to_jiffies(rtoinfo.srto_initial);
3001 asoc->rto_max = rto_max;
3002 asoc->rto_min = rto_min;
3003 } else {
3004 /* If there is no association or the association-id = 0
3005 * set the values to the endpoint.
3007 if (rtoinfo.srto_initial != 0)
3008 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3009 sp->rtoinfo.srto_max = rto_max;
3010 sp->rtoinfo.srto_min = rto_min;
3013 return 0;
3018 * 7.1.2 SCTP_ASSOCINFO
3020 * This option is used to tune the maximum retransmission attempts
3021 * of the association.
3022 * Returns an error if the new association retransmission value is
3023 * greater than the sum of the retransmission value of the peer.
3024 * See [SCTP] for more information.
3027 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3030 struct sctp_assocparams assocparams;
3031 struct sctp_association *asoc;
3033 if (optlen != sizeof(struct sctp_assocparams))
3034 return -EINVAL;
3035 if (copy_from_user(&assocparams, optval, optlen))
3036 return -EFAULT;
3038 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3040 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3041 return -EINVAL;
3043 /* Set the values to the specific association */
3044 if (asoc) {
3045 if (assocparams.sasoc_asocmaxrxt != 0) {
3046 __u32 path_sum = 0;
3047 int paths = 0;
3048 struct sctp_transport *peer_addr;
3050 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3051 transports) {
3052 path_sum += peer_addr->pathmaxrxt;
3053 paths++;
3056 /* Only validate asocmaxrxt if we have more than
3057 * one path/transport. We do this because path
3058 * retransmissions are only counted when we have more
3059 * then one path.
3061 if (paths > 1 &&
3062 assocparams.sasoc_asocmaxrxt > path_sum)
3063 return -EINVAL;
3065 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3068 if (assocparams.sasoc_cookie_life != 0)
3069 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3070 } else {
3071 /* Set the values to the endpoint */
3072 struct sctp_sock *sp = sctp_sk(sk);
3074 if (assocparams.sasoc_asocmaxrxt != 0)
3075 sp->assocparams.sasoc_asocmaxrxt =
3076 assocparams.sasoc_asocmaxrxt;
3077 if (assocparams.sasoc_cookie_life != 0)
3078 sp->assocparams.sasoc_cookie_life =
3079 assocparams.sasoc_cookie_life;
3081 return 0;
3085 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3087 * This socket option is a boolean flag which turns on or off mapped V4
3088 * addresses. If this option is turned on and the socket is type
3089 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3090 * If this option is turned off, then no mapping will be done of V4
3091 * addresses and a user will receive both PF_INET6 and PF_INET type
3092 * addresses on the socket.
3094 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3096 int val;
3097 struct sctp_sock *sp = sctp_sk(sk);
3099 if (optlen < sizeof(int))
3100 return -EINVAL;
3101 if (get_user(val, (int __user *)optval))
3102 return -EFAULT;
3103 if (val)
3104 sp->v4mapped = 1;
3105 else
3106 sp->v4mapped = 0;
3108 return 0;
3112 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3113 * This option will get or set the maximum size to put in any outgoing
3114 * SCTP DATA chunk. If a message is larger than this size it will be
3115 * fragmented by SCTP into the specified size. Note that the underlying
3116 * SCTP implementation may fragment into smaller sized chunks when the
3117 * PMTU of the underlying association is smaller than the value set by
3118 * the user. The default value for this option is '0' which indicates
3119 * the user is NOT limiting fragmentation and only the PMTU will effect
3120 * SCTP's choice of DATA chunk size. Note also that values set larger
3121 * than the maximum size of an IP datagram will effectively let SCTP
3122 * control fragmentation (i.e. the same as setting this option to 0).
3124 * The following structure is used to access and modify this parameter:
3126 * struct sctp_assoc_value {
3127 * sctp_assoc_t assoc_id;
3128 * uint32_t assoc_value;
3129 * };
3131 * assoc_id: This parameter is ignored for one-to-one style sockets.
3132 * For one-to-many style sockets this parameter indicates which
3133 * association the user is performing an action upon. Note that if
3134 * this field's value is zero then the endpoints default value is
3135 * changed (effecting future associations only).
3136 * assoc_value: This parameter specifies the maximum size in bytes.
3138 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3140 struct sctp_sock *sp = sctp_sk(sk);
3141 struct sctp_assoc_value params;
3142 struct sctp_association *asoc;
3143 int val;
3145 if (optlen == sizeof(int)) {
3146 pr_warn_ratelimited(DEPRECATED
3147 "%s (pid %d) "
3148 "Use of int in maxseg socket option.\n"
3149 "Use struct sctp_assoc_value instead\n",
3150 current->comm, task_pid_nr(current));
3151 if (copy_from_user(&val, optval, optlen))
3152 return -EFAULT;
3153 params.assoc_id = 0;
3154 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3155 if (copy_from_user(&params, optval, optlen))
3156 return -EFAULT;
3157 val = params.assoc_value;
3158 } else {
3159 return -EINVAL;
3162 if (val) {
3163 int min_len, max_len;
3165 min_len = SCTP_DEFAULT_MINSEGMENT - sp->pf->af->net_header_len;
3166 min_len -= sizeof(struct sctphdr) +
3167 sizeof(struct sctp_data_chunk);
3169 max_len = SCTP_MAX_CHUNK_LEN - sizeof(struct sctp_data_chunk);
3171 if (val < min_len || val > max_len)
3172 return -EINVAL;
3175 asoc = sctp_id2assoc(sk, params.assoc_id);
3176 if (asoc) {
3177 if (val == 0) {
3178 val = asoc->pathmtu - sp->pf->af->net_header_len;
3179 val -= sizeof(struct sctphdr) +
3180 sctp_datachk_len(&asoc->stream);
3182 asoc->user_frag = val;
3183 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3184 } else {
3185 if (params.assoc_id && sctp_style(sk, UDP))
3186 return -EINVAL;
3187 sp->user_frag = val;
3190 return 0;
3195 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3197 * Requests that the peer mark the enclosed address as the association
3198 * primary. The enclosed address must be one of the association's
3199 * locally bound addresses. The following structure is used to make a
3200 * set primary request:
3202 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3203 unsigned int optlen)
3205 struct net *net = sock_net(sk);
3206 struct sctp_sock *sp;
3207 struct sctp_association *asoc = NULL;
3208 struct sctp_setpeerprim prim;
3209 struct sctp_chunk *chunk;
3210 struct sctp_af *af;
3211 int err;
3213 sp = sctp_sk(sk);
3215 if (!net->sctp.addip_enable)
3216 return -EPERM;
3218 if (optlen != sizeof(struct sctp_setpeerprim))
3219 return -EINVAL;
3221 if (copy_from_user(&prim, optval, optlen))
3222 return -EFAULT;
3224 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3225 if (!asoc)
3226 return -EINVAL;
3228 if (!asoc->peer.asconf_capable)
3229 return -EPERM;
3231 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3232 return -EPERM;
3234 if (!sctp_state(asoc, ESTABLISHED))
3235 return -ENOTCONN;
3237 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3238 if (!af)
3239 return -EINVAL;
3241 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3242 return -EADDRNOTAVAIL;
3244 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3245 return -EADDRNOTAVAIL;
3247 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3248 chunk = sctp_make_asconf_set_prim(asoc,
3249 (union sctp_addr *)&prim.sspp_addr);
3250 if (!chunk)
3251 return -ENOMEM;
3253 err = sctp_send_asconf(asoc, chunk);
3255 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3257 return err;
3260 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3261 unsigned int optlen)
3263 struct sctp_setadaptation adaptation;
3265 if (optlen != sizeof(struct sctp_setadaptation))
3266 return -EINVAL;
3267 if (copy_from_user(&adaptation, optval, optlen))
3268 return -EFAULT;
3270 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3272 return 0;
3276 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3278 * The context field in the sctp_sndrcvinfo structure is normally only
3279 * used when a failed message is retrieved holding the value that was
3280 * sent down on the actual send call. This option allows the setting of
3281 * a default context on an association basis that will be received on
3282 * reading messages from the peer. This is especially helpful in the
3283 * one-2-many model for an application to keep some reference to an
3284 * internal state machine that is processing messages on the
3285 * association. Note that the setting of this value only effects
3286 * received messages from the peer and does not effect the value that is
3287 * saved with outbound messages.
3289 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3290 unsigned int optlen)
3292 struct sctp_assoc_value params;
3293 struct sctp_sock *sp;
3294 struct sctp_association *asoc;
3296 if (optlen != sizeof(struct sctp_assoc_value))
3297 return -EINVAL;
3298 if (copy_from_user(&params, optval, optlen))
3299 return -EFAULT;
3301 sp = sctp_sk(sk);
3303 if (params.assoc_id != 0) {
3304 asoc = sctp_id2assoc(sk, params.assoc_id);
3305 if (!asoc)
3306 return -EINVAL;
3307 asoc->default_rcv_context = params.assoc_value;
3308 } else {
3309 sp->default_rcv_context = params.assoc_value;
3312 return 0;
3316 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3318 * This options will at a minimum specify if the implementation is doing
3319 * fragmented interleave. Fragmented interleave, for a one to many
3320 * socket, is when subsequent calls to receive a message may return
3321 * parts of messages from different associations. Some implementations
3322 * may allow you to turn this value on or off. If so, when turned off,
3323 * no fragment interleave will occur (which will cause a head of line
3324 * blocking amongst multiple associations sharing the same one to many
3325 * socket). When this option is turned on, then each receive call may
3326 * come from a different association (thus the user must receive data
3327 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3328 * association each receive belongs to.
3330 * This option takes a boolean value. A non-zero value indicates that
3331 * fragmented interleave is on. A value of zero indicates that
3332 * fragmented interleave is off.
3334 * Note that it is important that an implementation that allows this
3335 * option to be turned on, have it off by default. Otherwise an unaware
3336 * application using the one to many model may become confused and act
3337 * incorrectly.
3339 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3340 char __user *optval,
3341 unsigned int optlen)
3343 int val;
3345 if (optlen != sizeof(int))
3346 return -EINVAL;
3347 if (get_user(val, (int __user *)optval))
3348 return -EFAULT;
3350 sctp_sk(sk)->frag_interleave = !!val;
3352 if (!sctp_sk(sk)->frag_interleave)
3353 sctp_sk(sk)->strm_interleave = 0;
3355 return 0;
3359 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3360 * (SCTP_PARTIAL_DELIVERY_POINT)
3362 * This option will set or get the SCTP partial delivery point. This
3363 * point is the size of a message where the partial delivery API will be
3364 * invoked to help free up rwnd space for the peer. Setting this to a
3365 * lower value will cause partial deliveries to happen more often. The
3366 * calls argument is an integer that sets or gets the partial delivery
3367 * point. Note also that the call will fail if the user attempts to set
3368 * this value larger than the socket receive buffer size.
3370 * Note that any single message having a length smaller than or equal to
3371 * the SCTP partial delivery point will be delivered in one single read
3372 * call as long as the user provided buffer is large enough to hold the
3373 * message.
3375 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3376 char __user *optval,
3377 unsigned int optlen)
3379 u32 val;
3381 if (optlen != sizeof(u32))
3382 return -EINVAL;
3383 if (get_user(val, (int __user *)optval))
3384 return -EFAULT;
3386 /* Note: We double the receive buffer from what the user sets
3387 * it to be, also initial rwnd is based on rcvbuf/2.
3389 if (val > (sk->sk_rcvbuf >> 1))
3390 return -EINVAL;
3392 sctp_sk(sk)->pd_point = val;
3394 return 0; /* is this the right error code? */
3398 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3400 * This option will allow a user to change the maximum burst of packets
3401 * that can be emitted by this association. Note that the default value
3402 * is 4, and some implementations may restrict this setting so that it
3403 * can only be lowered.
3405 * NOTE: This text doesn't seem right. Do this on a socket basis with
3406 * future associations inheriting the socket value.
3408 static int sctp_setsockopt_maxburst(struct sock *sk,
3409 char __user *optval,
3410 unsigned int optlen)
3412 struct sctp_assoc_value params;
3413 struct sctp_sock *sp;
3414 struct sctp_association *asoc;
3415 int val;
3416 int assoc_id = 0;
3418 if (optlen == sizeof(int)) {
3419 pr_warn_ratelimited(DEPRECATED
3420 "%s (pid %d) "
3421 "Use of int in max_burst socket option deprecated.\n"
3422 "Use struct sctp_assoc_value instead\n",
3423 current->comm, task_pid_nr(current));
3424 if (copy_from_user(&val, optval, optlen))
3425 return -EFAULT;
3426 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3427 if (copy_from_user(&params, optval, optlen))
3428 return -EFAULT;
3429 val = params.assoc_value;
3430 assoc_id = params.assoc_id;
3431 } else
3432 return -EINVAL;
3434 sp = sctp_sk(sk);
3436 if (assoc_id != 0) {
3437 asoc = sctp_id2assoc(sk, assoc_id);
3438 if (!asoc)
3439 return -EINVAL;
3440 asoc->max_burst = val;
3441 } else
3442 sp->max_burst = val;
3444 return 0;
3448 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3450 * This set option adds a chunk type that the user is requesting to be
3451 * received only in an authenticated way. Changes to the list of chunks
3452 * will only effect future associations on the socket.
3454 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3455 char __user *optval,
3456 unsigned int optlen)
3458 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3459 struct sctp_authchunk val;
3461 if (!ep->auth_enable)
3462 return -EACCES;
3464 if (optlen != sizeof(struct sctp_authchunk))
3465 return -EINVAL;
3466 if (copy_from_user(&val, optval, optlen))
3467 return -EFAULT;
3469 switch (val.sauth_chunk) {
3470 case SCTP_CID_INIT:
3471 case SCTP_CID_INIT_ACK:
3472 case SCTP_CID_SHUTDOWN_COMPLETE:
3473 case SCTP_CID_AUTH:
3474 return -EINVAL;
3477 /* add this chunk id to the endpoint */
3478 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3482 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3484 * This option gets or sets the list of HMAC algorithms that the local
3485 * endpoint requires the peer to use.
3487 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3488 char __user *optval,
3489 unsigned int optlen)
3491 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3492 struct sctp_hmacalgo *hmacs;
3493 u32 idents;
3494 int err;
3496 if (!ep->auth_enable)
3497 return -EACCES;
3499 if (optlen < sizeof(struct sctp_hmacalgo))
3500 return -EINVAL;
3501 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3502 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3504 hmacs = memdup_user(optval, optlen);
3505 if (IS_ERR(hmacs))
3506 return PTR_ERR(hmacs);
3508 idents = hmacs->shmac_num_idents;
3509 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3510 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3511 err = -EINVAL;
3512 goto out;
3515 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3516 out:
3517 kfree(hmacs);
3518 return err;
3522 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3524 * This option will set a shared secret key which is used to build an
3525 * association shared key.
3527 static int sctp_setsockopt_auth_key(struct sock *sk,
3528 char __user *optval,
3529 unsigned int optlen)
3531 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3532 struct sctp_authkey *authkey;
3533 struct sctp_association *asoc;
3534 int ret;
3536 if (!ep->auth_enable)
3537 return -EACCES;
3539 if (optlen <= sizeof(struct sctp_authkey))
3540 return -EINVAL;
3541 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3542 * this.
3544 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3545 sizeof(struct sctp_authkey));
3547 authkey = memdup_user(optval, optlen);
3548 if (IS_ERR(authkey))
3549 return PTR_ERR(authkey);
3551 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3552 ret = -EINVAL;
3553 goto out;
3556 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3557 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3558 ret = -EINVAL;
3559 goto out;
3562 ret = sctp_auth_set_key(ep, asoc, authkey);
3563 out:
3564 kzfree(authkey);
3565 return ret;
3569 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3571 * This option will get or set the active shared key to be used to build
3572 * the association shared key.
3574 static int sctp_setsockopt_active_key(struct sock *sk,
3575 char __user *optval,
3576 unsigned int optlen)
3578 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3579 struct sctp_authkeyid val;
3580 struct sctp_association *asoc;
3582 if (!ep->auth_enable)
3583 return -EACCES;
3585 if (optlen != sizeof(struct sctp_authkeyid))
3586 return -EINVAL;
3587 if (copy_from_user(&val, optval, optlen))
3588 return -EFAULT;
3590 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3591 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3592 return -EINVAL;
3594 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3598 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3600 * This set option will delete a shared secret key from use.
3602 static int sctp_setsockopt_del_key(struct sock *sk,
3603 char __user *optval,
3604 unsigned int optlen)
3606 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3607 struct sctp_authkeyid val;
3608 struct sctp_association *asoc;
3610 if (!ep->auth_enable)
3611 return -EACCES;
3613 if (optlen != sizeof(struct sctp_authkeyid))
3614 return -EINVAL;
3615 if (copy_from_user(&val, optval, optlen))
3616 return -EFAULT;
3618 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3619 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3620 return -EINVAL;
3622 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3627 * 8.1.23 SCTP_AUTO_ASCONF
3629 * This option will enable or disable the use of the automatic generation of
3630 * ASCONF chunks to add and delete addresses to an existing association. Note
3631 * that this option has two caveats namely: a) it only affects sockets that
3632 * are bound to all addresses available to the SCTP stack, and b) the system
3633 * administrator may have an overriding control that turns the ASCONF feature
3634 * off no matter what setting the socket option may have.
3635 * This option expects an integer boolean flag, where a non-zero value turns on
3636 * the option, and a zero value turns off the option.
3637 * Note. In this implementation, socket operation overrides default parameter
3638 * being set by sysctl as well as FreeBSD implementation
3640 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3641 unsigned int optlen)
3643 int val;
3644 struct sctp_sock *sp = sctp_sk(sk);
3646 if (optlen < sizeof(int))
3647 return -EINVAL;
3648 if (get_user(val, (int __user *)optval))
3649 return -EFAULT;
3650 if (!sctp_is_ep_boundall(sk) && val)
3651 return -EINVAL;
3652 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3653 return 0;
3655 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3656 if (val == 0 && sp->do_auto_asconf) {
3657 list_del(&sp->auto_asconf_list);
3658 sp->do_auto_asconf = 0;
3659 } else if (val && !sp->do_auto_asconf) {
3660 list_add_tail(&sp->auto_asconf_list,
3661 &sock_net(sk)->sctp.auto_asconf_splist);
3662 sp->do_auto_asconf = 1;
3664 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3665 return 0;
3669 * SCTP_PEER_ADDR_THLDS
3671 * This option allows us to alter the partially failed threshold for one or all
3672 * transports in an association. See Section 6.1 of:
3673 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3675 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3676 char __user *optval,
3677 unsigned int optlen)
3679 struct sctp_paddrthlds val;
3680 struct sctp_transport *trans;
3681 struct sctp_association *asoc;
3683 if (optlen < sizeof(struct sctp_paddrthlds))
3684 return -EINVAL;
3685 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3686 sizeof(struct sctp_paddrthlds)))
3687 return -EFAULT;
3690 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3691 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3692 if (!asoc)
3693 return -ENOENT;
3694 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3695 transports) {
3696 if (val.spt_pathmaxrxt)
3697 trans->pathmaxrxt = val.spt_pathmaxrxt;
3698 trans->pf_retrans = val.spt_pathpfthld;
3701 if (val.spt_pathmaxrxt)
3702 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3703 asoc->pf_retrans = val.spt_pathpfthld;
3704 } else {
3705 trans = sctp_addr_id2transport(sk, &val.spt_address,
3706 val.spt_assoc_id);
3707 if (!trans)
3708 return -ENOENT;
3710 if (val.spt_pathmaxrxt)
3711 trans->pathmaxrxt = val.spt_pathmaxrxt;
3712 trans->pf_retrans = val.spt_pathpfthld;
3715 return 0;
3718 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3719 char __user *optval,
3720 unsigned int optlen)
3722 int val;
3724 if (optlen < sizeof(int))
3725 return -EINVAL;
3726 if (get_user(val, (int __user *) optval))
3727 return -EFAULT;
3729 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3731 return 0;
3734 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3735 char __user *optval,
3736 unsigned int optlen)
3738 int val;
3740 if (optlen < sizeof(int))
3741 return -EINVAL;
3742 if (get_user(val, (int __user *) optval))
3743 return -EFAULT;
3745 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3747 return 0;
3750 static int sctp_setsockopt_pr_supported(struct sock *sk,
3751 char __user *optval,
3752 unsigned int optlen)
3754 struct sctp_assoc_value params;
3755 struct sctp_association *asoc;
3756 int retval = -EINVAL;
3758 if (optlen != sizeof(params))
3759 goto out;
3761 if (copy_from_user(&params, optval, optlen)) {
3762 retval = -EFAULT;
3763 goto out;
3766 asoc = sctp_id2assoc(sk, params.assoc_id);
3767 if (asoc) {
3768 asoc->prsctp_enable = !!params.assoc_value;
3769 } else if (!params.assoc_id) {
3770 struct sctp_sock *sp = sctp_sk(sk);
3772 sp->ep->prsctp_enable = !!params.assoc_value;
3773 } else {
3774 goto out;
3777 retval = 0;
3779 out:
3780 return retval;
3783 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3784 char __user *optval,
3785 unsigned int optlen)
3787 struct sctp_default_prinfo info;
3788 struct sctp_association *asoc;
3789 int retval = -EINVAL;
3791 if (optlen != sizeof(info))
3792 goto out;
3794 if (copy_from_user(&info, optval, sizeof(info))) {
3795 retval = -EFAULT;
3796 goto out;
3799 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3800 goto out;
3802 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3803 info.pr_value = 0;
3805 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3806 if (asoc) {
3807 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3808 asoc->default_timetolive = info.pr_value;
3809 } else if (!info.pr_assoc_id) {
3810 struct sctp_sock *sp = sctp_sk(sk);
3812 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3813 sp->default_timetolive = info.pr_value;
3814 } else {
3815 goto out;
3818 retval = 0;
3820 out:
3821 return retval;
3824 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3825 char __user *optval,
3826 unsigned int optlen)
3828 struct sctp_assoc_value params;
3829 struct sctp_association *asoc;
3830 int retval = -EINVAL;
3832 if (optlen != sizeof(params))
3833 goto out;
3835 if (copy_from_user(&params, optval, optlen)) {
3836 retval = -EFAULT;
3837 goto out;
3840 asoc = sctp_id2assoc(sk, params.assoc_id);
3841 if (asoc) {
3842 asoc->reconf_enable = !!params.assoc_value;
3843 } else if (!params.assoc_id) {
3844 struct sctp_sock *sp = sctp_sk(sk);
3846 sp->ep->reconf_enable = !!params.assoc_value;
3847 } else {
3848 goto out;
3851 retval = 0;
3853 out:
3854 return retval;
3857 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3858 char __user *optval,
3859 unsigned int optlen)
3861 struct sctp_assoc_value params;
3862 struct sctp_association *asoc;
3863 int retval = -EINVAL;
3865 if (optlen != sizeof(params))
3866 goto out;
3868 if (copy_from_user(&params, optval, optlen)) {
3869 retval = -EFAULT;
3870 goto out;
3873 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3874 goto out;
3876 asoc = sctp_id2assoc(sk, params.assoc_id);
3877 if (asoc) {
3878 asoc->strreset_enable = params.assoc_value;
3879 } else if (!params.assoc_id) {
3880 struct sctp_sock *sp = sctp_sk(sk);
3882 sp->ep->strreset_enable = params.assoc_value;
3883 } else {
3884 goto out;
3887 retval = 0;
3889 out:
3890 return retval;
3893 static int sctp_setsockopt_reset_streams(struct sock *sk,
3894 char __user *optval,
3895 unsigned int optlen)
3897 struct sctp_reset_streams *params;
3898 struct sctp_association *asoc;
3899 int retval = -EINVAL;
3901 if (optlen < sizeof(*params))
3902 return -EINVAL;
3903 /* srs_number_streams is u16, so optlen can't be bigger than this. */
3904 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3905 sizeof(__u16) * sizeof(*params));
3907 params = memdup_user(optval, optlen);
3908 if (IS_ERR(params))
3909 return PTR_ERR(params);
3911 if (params->srs_number_streams * sizeof(__u16) >
3912 optlen - sizeof(*params))
3913 goto out;
3915 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3916 if (!asoc)
3917 goto out;
3919 retval = sctp_send_reset_streams(asoc, params);
3921 out:
3922 kfree(params);
3923 return retval;
3926 static int sctp_setsockopt_reset_assoc(struct sock *sk,
3927 char __user *optval,
3928 unsigned int optlen)
3930 struct sctp_association *asoc;
3931 sctp_assoc_t associd;
3932 int retval = -EINVAL;
3934 if (optlen != sizeof(associd))
3935 goto out;
3937 if (copy_from_user(&associd, optval, optlen)) {
3938 retval = -EFAULT;
3939 goto out;
3942 asoc = sctp_id2assoc(sk, associd);
3943 if (!asoc)
3944 goto out;
3946 retval = sctp_send_reset_assoc(asoc);
3948 out:
3949 return retval;
3952 static int sctp_setsockopt_add_streams(struct sock *sk,
3953 char __user *optval,
3954 unsigned int optlen)
3956 struct sctp_association *asoc;
3957 struct sctp_add_streams params;
3958 int retval = -EINVAL;
3960 if (optlen != sizeof(params))
3961 goto out;
3963 if (copy_from_user(&params, optval, optlen)) {
3964 retval = -EFAULT;
3965 goto out;
3968 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
3969 if (!asoc)
3970 goto out;
3972 retval = sctp_send_add_streams(asoc, &params);
3974 out:
3975 return retval;
3978 static int sctp_setsockopt_scheduler(struct sock *sk,
3979 char __user *optval,
3980 unsigned int optlen)
3982 struct sctp_association *asoc;
3983 struct sctp_assoc_value params;
3984 int retval = -EINVAL;
3986 if (optlen < sizeof(params))
3987 goto out;
3989 optlen = sizeof(params);
3990 if (copy_from_user(&params, optval, optlen)) {
3991 retval = -EFAULT;
3992 goto out;
3995 if (params.assoc_value > SCTP_SS_MAX)
3996 goto out;
3998 asoc = sctp_id2assoc(sk, params.assoc_id);
3999 if (!asoc)
4000 goto out;
4002 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4004 out:
4005 return retval;
4008 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4009 char __user *optval,
4010 unsigned int optlen)
4012 struct sctp_association *asoc;
4013 struct sctp_stream_value params;
4014 int retval = -EINVAL;
4016 if (optlen < sizeof(params))
4017 goto out;
4019 optlen = sizeof(params);
4020 if (copy_from_user(&params, optval, optlen)) {
4021 retval = -EFAULT;
4022 goto out;
4025 asoc = sctp_id2assoc(sk, params.assoc_id);
4026 if (!asoc)
4027 goto out;
4029 retval = sctp_sched_set_value(asoc, params.stream_id,
4030 params.stream_value, GFP_KERNEL);
4032 out:
4033 return retval;
4036 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4037 char __user *optval,
4038 unsigned int optlen)
4040 struct sctp_sock *sp = sctp_sk(sk);
4041 struct net *net = sock_net(sk);
4042 struct sctp_assoc_value params;
4043 int retval = -EINVAL;
4045 if (optlen < sizeof(params))
4046 goto out;
4048 optlen = sizeof(params);
4049 if (copy_from_user(&params, optval, optlen)) {
4050 retval = -EFAULT;
4051 goto out;
4054 if (params.assoc_id)
4055 goto out;
4057 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4058 retval = -EPERM;
4059 goto out;
4062 sp->strm_interleave = !!params.assoc_value;
4064 retval = 0;
4066 out:
4067 return retval;
4070 /* API 6.2 setsockopt(), getsockopt()
4072 * Applications use setsockopt() and getsockopt() to set or retrieve
4073 * socket options. Socket options are used to change the default
4074 * behavior of sockets calls. They are described in Section 7.
4076 * The syntax is:
4078 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4079 * int __user *optlen);
4080 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4081 * int optlen);
4083 * sd - the socket descript.
4084 * level - set to IPPROTO_SCTP for all SCTP options.
4085 * optname - the option name.
4086 * optval - the buffer to store the value of the option.
4087 * optlen - the size of the buffer.
4089 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4090 char __user *optval, unsigned int optlen)
4092 int retval = 0;
4094 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4096 /* I can hardly begin to describe how wrong this is. This is
4097 * so broken as to be worse than useless. The API draft
4098 * REALLY is NOT helpful here... I am not convinced that the
4099 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4100 * are at all well-founded.
4102 if (level != SOL_SCTP) {
4103 struct sctp_af *af = sctp_sk(sk)->pf->af;
4104 retval = af->setsockopt(sk, level, optname, optval, optlen);
4105 goto out_nounlock;
4108 lock_sock(sk);
4110 switch (optname) {
4111 case SCTP_SOCKOPT_BINDX_ADD:
4112 /* 'optlen' is the size of the addresses buffer. */
4113 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4114 optlen, SCTP_BINDX_ADD_ADDR);
4115 break;
4117 case SCTP_SOCKOPT_BINDX_REM:
4118 /* 'optlen' is the size of the addresses buffer. */
4119 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4120 optlen, SCTP_BINDX_REM_ADDR);
4121 break;
4123 case SCTP_SOCKOPT_CONNECTX_OLD:
4124 /* 'optlen' is the size of the addresses buffer. */
4125 retval = sctp_setsockopt_connectx_old(sk,
4126 (struct sockaddr __user *)optval,
4127 optlen);
4128 break;
4130 case SCTP_SOCKOPT_CONNECTX:
4131 /* 'optlen' is the size of the addresses buffer. */
4132 retval = sctp_setsockopt_connectx(sk,
4133 (struct sockaddr __user *)optval,
4134 optlen);
4135 break;
4137 case SCTP_DISABLE_FRAGMENTS:
4138 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4139 break;
4141 case SCTP_EVENTS:
4142 retval = sctp_setsockopt_events(sk, optval, optlen);
4143 break;
4145 case SCTP_AUTOCLOSE:
4146 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4147 break;
4149 case SCTP_PEER_ADDR_PARAMS:
4150 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4151 break;
4153 case SCTP_DELAYED_SACK:
4154 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4155 break;
4156 case SCTP_PARTIAL_DELIVERY_POINT:
4157 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4158 break;
4160 case SCTP_INITMSG:
4161 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4162 break;
4163 case SCTP_DEFAULT_SEND_PARAM:
4164 retval = sctp_setsockopt_default_send_param(sk, optval,
4165 optlen);
4166 break;
4167 case SCTP_DEFAULT_SNDINFO:
4168 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4169 break;
4170 case SCTP_PRIMARY_ADDR:
4171 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4172 break;
4173 case SCTP_SET_PEER_PRIMARY_ADDR:
4174 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4175 break;
4176 case SCTP_NODELAY:
4177 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4178 break;
4179 case SCTP_RTOINFO:
4180 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4181 break;
4182 case SCTP_ASSOCINFO:
4183 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4184 break;
4185 case SCTP_I_WANT_MAPPED_V4_ADDR:
4186 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4187 break;
4188 case SCTP_MAXSEG:
4189 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4190 break;
4191 case SCTP_ADAPTATION_LAYER:
4192 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4193 break;
4194 case SCTP_CONTEXT:
4195 retval = sctp_setsockopt_context(sk, optval, optlen);
4196 break;
4197 case SCTP_FRAGMENT_INTERLEAVE:
4198 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4199 break;
4200 case SCTP_MAX_BURST:
4201 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4202 break;
4203 case SCTP_AUTH_CHUNK:
4204 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4205 break;
4206 case SCTP_HMAC_IDENT:
4207 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4208 break;
4209 case SCTP_AUTH_KEY:
4210 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4211 break;
4212 case SCTP_AUTH_ACTIVE_KEY:
4213 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4214 break;
4215 case SCTP_AUTH_DELETE_KEY:
4216 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4217 break;
4218 case SCTP_AUTO_ASCONF:
4219 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4220 break;
4221 case SCTP_PEER_ADDR_THLDS:
4222 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4223 break;
4224 case SCTP_RECVRCVINFO:
4225 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4226 break;
4227 case SCTP_RECVNXTINFO:
4228 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4229 break;
4230 case SCTP_PR_SUPPORTED:
4231 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4232 break;
4233 case SCTP_DEFAULT_PRINFO:
4234 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4235 break;
4236 case SCTP_RECONFIG_SUPPORTED:
4237 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4238 break;
4239 case SCTP_ENABLE_STREAM_RESET:
4240 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4241 break;
4242 case SCTP_RESET_STREAMS:
4243 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4244 break;
4245 case SCTP_RESET_ASSOC:
4246 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4247 break;
4248 case SCTP_ADD_STREAMS:
4249 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4250 break;
4251 case SCTP_STREAM_SCHEDULER:
4252 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4253 break;
4254 case SCTP_STREAM_SCHEDULER_VALUE:
4255 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4256 break;
4257 case SCTP_INTERLEAVING_SUPPORTED:
4258 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4259 optlen);
4260 break;
4261 default:
4262 retval = -ENOPROTOOPT;
4263 break;
4266 release_sock(sk);
4268 out_nounlock:
4269 return retval;
4272 /* API 3.1.6 connect() - UDP Style Syntax
4274 * An application may use the connect() call in the UDP model to initiate an
4275 * association without sending data.
4277 * The syntax is:
4279 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4281 * sd: the socket descriptor to have a new association added to.
4283 * nam: the address structure (either struct sockaddr_in or struct
4284 * sockaddr_in6 defined in RFC2553 [7]).
4286 * len: the size of the address.
4288 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4289 int addr_len)
4291 int err = 0;
4292 struct sctp_af *af;
4294 lock_sock(sk);
4296 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4297 addr, addr_len);
4299 /* Validate addr_len before calling common connect/connectx routine. */
4300 af = sctp_get_af_specific(addr->sa_family);
4301 if (!af || addr_len < af->sockaddr_len) {
4302 err = -EINVAL;
4303 } else {
4304 /* Pass correct addr len to common routine (so it knows there
4305 * is only one address being passed.
4307 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4310 release_sock(sk);
4311 return err;
4314 /* FIXME: Write comments. */
4315 static int sctp_disconnect(struct sock *sk, int flags)
4317 return -EOPNOTSUPP; /* STUB */
4320 /* 4.1.4 accept() - TCP Style Syntax
4322 * Applications use accept() call to remove an established SCTP
4323 * association from the accept queue of the endpoint. A new socket
4324 * descriptor will be returned from accept() to represent the newly
4325 * formed association.
4327 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4329 struct sctp_sock *sp;
4330 struct sctp_endpoint *ep;
4331 struct sock *newsk = NULL;
4332 struct sctp_association *asoc;
4333 long timeo;
4334 int error = 0;
4336 lock_sock(sk);
4338 sp = sctp_sk(sk);
4339 ep = sp->ep;
4341 if (!sctp_style(sk, TCP)) {
4342 error = -EOPNOTSUPP;
4343 goto out;
4346 if (!sctp_sstate(sk, LISTENING)) {
4347 error = -EINVAL;
4348 goto out;
4351 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4353 error = sctp_wait_for_accept(sk, timeo);
4354 if (error)
4355 goto out;
4357 /* We treat the list of associations on the endpoint as the accept
4358 * queue and pick the first association on the list.
4360 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4362 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4363 if (!newsk) {
4364 error = -ENOMEM;
4365 goto out;
4368 /* Populate the fields of the newsk from the oldsk and migrate the
4369 * asoc to the newsk.
4371 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4373 out:
4374 release_sock(sk);
4375 *err = error;
4376 return newsk;
4379 /* The SCTP ioctl handler. */
4380 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4382 int rc = -ENOTCONN;
4384 lock_sock(sk);
4387 * SEQPACKET-style sockets in LISTENING state are valid, for
4388 * SCTP, so only discard TCP-style sockets in LISTENING state.
4390 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4391 goto out;
4393 switch (cmd) {
4394 case SIOCINQ: {
4395 struct sk_buff *skb;
4396 unsigned int amount = 0;
4398 skb = skb_peek(&sk->sk_receive_queue);
4399 if (skb != NULL) {
4401 * We will only return the amount of this packet since
4402 * that is all that will be read.
4404 amount = skb->len;
4406 rc = put_user(amount, (int __user *)arg);
4407 break;
4409 default:
4410 rc = -ENOIOCTLCMD;
4411 break;
4413 out:
4414 release_sock(sk);
4415 return rc;
4418 /* This is the function which gets called during socket creation to
4419 * initialized the SCTP-specific portion of the sock.
4420 * The sock structure should already be zero-filled memory.
4422 static int sctp_init_sock(struct sock *sk)
4424 struct net *net = sock_net(sk);
4425 struct sctp_sock *sp;
4427 pr_debug("%s: sk:%p\n", __func__, sk);
4429 sp = sctp_sk(sk);
4431 /* Initialize the SCTP per socket area. */
4432 switch (sk->sk_type) {
4433 case SOCK_SEQPACKET:
4434 sp->type = SCTP_SOCKET_UDP;
4435 break;
4436 case SOCK_STREAM:
4437 sp->type = SCTP_SOCKET_TCP;
4438 break;
4439 default:
4440 return -ESOCKTNOSUPPORT;
4443 sk->sk_gso_type = SKB_GSO_SCTP;
4445 /* Initialize default send parameters. These parameters can be
4446 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4448 sp->default_stream = 0;
4449 sp->default_ppid = 0;
4450 sp->default_flags = 0;
4451 sp->default_context = 0;
4452 sp->default_timetolive = 0;
4454 sp->default_rcv_context = 0;
4455 sp->max_burst = net->sctp.max_burst;
4457 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4459 /* Initialize default setup parameters. These parameters
4460 * can be modified with the SCTP_INITMSG socket option or
4461 * overridden by the SCTP_INIT CMSG.
4463 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4464 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4465 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4466 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4468 /* Initialize default RTO related parameters. These parameters can
4469 * be modified for with the SCTP_RTOINFO socket option.
4471 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4472 sp->rtoinfo.srto_max = net->sctp.rto_max;
4473 sp->rtoinfo.srto_min = net->sctp.rto_min;
4475 /* Initialize default association related parameters. These parameters
4476 * can be modified with the SCTP_ASSOCINFO socket option.
4478 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4479 sp->assocparams.sasoc_number_peer_destinations = 0;
4480 sp->assocparams.sasoc_peer_rwnd = 0;
4481 sp->assocparams.sasoc_local_rwnd = 0;
4482 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4484 /* Initialize default event subscriptions. By default, all the
4485 * options are off.
4487 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4489 /* Default Peer Address Parameters. These defaults can
4490 * be modified via SCTP_PEER_ADDR_PARAMS
4492 sp->hbinterval = net->sctp.hb_interval;
4493 sp->pathmaxrxt = net->sctp.max_retrans_path;
4494 sp->pathmtu = 0; /* allow default discovery */
4495 sp->sackdelay = net->sctp.sack_timeout;
4496 sp->sackfreq = 2;
4497 sp->param_flags = SPP_HB_ENABLE |
4498 SPP_PMTUD_ENABLE |
4499 SPP_SACKDELAY_ENABLE;
4501 /* If enabled no SCTP message fragmentation will be performed.
4502 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4504 sp->disable_fragments = 0;
4506 /* Enable Nagle algorithm by default. */
4507 sp->nodelay = 0;
4509 sp->recvrcvinfo = 0;
4510 sp->recvnxtinfo = 0;
4512 /* Enable by default. */
4513 sp->v4mapped = 1;
4515 /* Auto-close idle associations after the configured
4516 * number of seconds. A value of 0 disables this
4517 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4518 * for UDP-style sockets only.
4520 sp->autoclose = 0;
4522 /* User specified fragmentation limit. */
4523 sp->user_frag = 0;
4525 sp->adaptation_ind = 0;
4527 sp->pf = sctp_get_pf_specific(sk->sk_family);
4529 /* Control variables for partial data delivery. */
4530 atomic_set(&sp->pd_mode, 0);
4531 skb_queue_head_init(&sp->pd_lobby);
4532 sp->frag_interleave = 0;
4534 /* Create a per socket endpoint structure. Even if we
4535 * change the data structure relationships, this may still
4536 * be useful for storing pre-connect address information.
4538 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4539 if (!sp->ep)
4540 return -ENOMEM;
4542 sp->hmac = NULL;
4544 sk->sk_destruct = sctp_destruct_sock;
4546 SCTP_DBG_OBJCNT_INC(sock);
4548 local_bh_disable();
4549 sk_sockets_allocated_inc(sk);
4550 sock_prot_inuse_add(net, sk->sk_prot, 1);
4552 /* Nothing can fail after this block, otherwise
4553 * sctp_destroy_sock() will be called without addr_wq_lock held
4555 if (net->sctp.default_auto_asconf) {
4556 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4557 list_add_tail(&sp->auto_asconf_list,
4558 &net->sctp.auto_asconf_splist);
4559 sp->do_auto_asconf = 1;
4560 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4561 } else {
4562 sp->do_auto_asconf = 0;
4565 local_bh_enable();
4567 return 0;
4570 /* Cleanup any SCTP per socket resources. Must be called with
4571 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4573 static void sctp_destroy_sock(struct sock *sk)
4575 struct sctp_sock *sp;
4577 pr_debug("%s: sk:%p\n", __func__, sk);
4579 /* Release our hold on the endpoint. */
4580 sp = sctp_sk(sk);
4581 /* This could happen during socket init, thus we bail out
4582 * early, since the rest of the below is not setup either.
4584 if (sp->ep == NULL)
4585 return;
4587 if (sp->do_auto_asconf) {
4588 sp->do_auto_asconf = 0;
4589 list_del(&sp->auto_asconf_list);
4591 sctp_endpoint_free(sp->ep);
4592 local_bh_disable();
4593 sk_sockets_allocated_dec(sk);
4594 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4595 local_bh_enable();
4598 /* Triggered when there are no references on the socket anymore */
4599 static void sctp_destruct_sock(struct sock *sk)
4601 struct sctp_sock *sp = sctp_sk(sk);
4603 /* Free up the HMAC transform. */
4604 crypto_free_shash(sp->hmac);
4606 inet_sock_destruct(sk);
4609 /* API 4.1.7 shutdown() - TCP Style Syntax
4610 * int shutdown(int socket, int how);
4612 * sd - the socket descriptor of the association to be closed.
4613 * how - Specifies the type of shutdown. The values are
4614 * as follows:
4615 * SHUT_RD
4616 * Disables further receive operations. No SCTP
4617 * protocol action is taken.
4618 * SHUT_WR
4619 * Disables further send operations, and initiates
4620 * the SCTP shutdown sequence.
4621 * SHUT_RDWR
4622 * Disables further send and receive operations
4623 * and initiates the SCTP shutdown sequence.
4625 static void sctp_shutdown(struct sock *sk, int how)
4627 struct net *net = sock_net(sk);
4628 struct sctp_endpoint *ep;
4630 if (!sctp_style(sk, TCP))
4631 return;
4633 ep = sctp_sk(sk)->ep;
4634 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4635 struct sctp_association *asoc;
4637 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4638 asoc = list_entry(ep->asocs.next,
4639 struct sctp_association, asocs);
4640 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4644 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4645 struct sctp_info *info)
4647 struct sctp_transport *prim;
4648 struct list_head *pos;
4649 int mask;
4651 memset(info, 0, sizeof(*info));
4652 if (!asoc) {
4653 struct sctp_sock *sp = sctp_sk(sk);
4655 info->sctpi_s_autoclose = sp->autoclose;
4656 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4657 info->sctpi_s_pd_point = sp->pd_point;
4658 info->sctpi_s_nodelay = sp->nodelay;
4659 info->sctpi_s_disable_fragments = sp->disable_fragments;
4660 info->sctpi_s_v4mapped = sp->v4mapped;
4661 info->sctpi_s_frag_interleave = sp->frag_interleave;
4662 info->sctpi_s_type = sp->type;
4664 return 0;
4667 info->sctpi_tag = asoc->c.my_vtag;
4668 info->sctpi_state = asoc->state;
4669 info->sctpi_rwnd = asoc->a_rwnd;
4670 info->sctpi_unackdata = asoc->unack_data;
4671 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4672 info->sctpi_instrms = asoc->stream.incnt;
4673 info->sctpi_outstrms = asoc->stream.outcnt;
4674 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4675 info->sctpi_inqueue++;
4676 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4677 info->sctpi_outqueue++;
4678 info->sctpi_overall_error = asoc->overall_error_count;
4679 info->sctpi_max_burst = asoc->max_burst;
4680 info->sctpi_maxseg = asoc->frag_point;
4681 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4682 info->sctpi_peer_tag = asoc->c.peer_vtag;
4684 mask = asoc->peer.ecn_capable << 1;
4685 mask = (mask | asoc->peer.ipv4_address) << 1;
4686 mask = (mask | asoc->peer.ipv6_address) << 1;
4687 mask = (mask | asoc->peer.hostname_address) << 1;
4688 mask = (mask | asoc->peer.asconf_capable) << 1;
4689 mask = (mask | asoc->peer.prsctp_capable) << 1;
4690 mask = (mask | asoc->peer.auth_capable);
4691 info->sctpi_peer_capable = mask;
4692 mask = asoc->peer.sack_needed << 1;
4693 mask = (mask | asoc->peer.sack_generation) << 1;
4694 mask = (mask | asoc->peer.zero_window_announced);
4695 info->sctpi_peer_sack = mask;
4697 info->sctpi_isacks = asoc->stats.isacks;
4698 info->sctpi_osacks = asoc->stats.osacks;
4699 info->sctpi_opackets = asoc->stats.opackets;
4700 info->sctpi_ipackets = asoc->stats.ipackets;
4701 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4702 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4703 info->sctpi_idupchunks = asoc->stats.idupchunks;
4704 info->sctpi_gapcnt = asoc->stats.gapcnt;
4705 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4706 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4707 info->sctpi_oodchunks = asoc->stats.oodchunks;
4708 info->sctpi_iodchunks = asoc->stats.iodchunks;
4709 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4710 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4712 prim = asoc->peer.primary_path;
4713 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4714 info->sctpi_p_state = prim->state;
4715 info->sctpi_p_cwnd = prim->cwnd;
4716 info->sctpi_p_srtt = prim->srtt;
4717 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4718 info->sctpi_p_hbinterval = prim->hbinterval;
4719 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4720 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4721 info->sctpi_p_ssthresh = prim->ssthresh;
4722 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4723 info->sctpi_p_flight_size = prim->flight_size;
4724 info->sctpi_p_error = prim->error_count;
4726 return 0;
4728 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4730 /* use callback to avoid exporting the core structure */
4731 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4733 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4735 rhashtable_walk_start(iter);
4738 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4740 rhashtable_walk_stop(iter);
4741 rhashtable_walk_exit(iter);
4744 struct sctp_transport *sctp_transport_get_next(struct net *net,
4745 struct rhashtable_iter *iter)
4747 struct sctp_transport *t;
4749 t = rhashtable_walk_next(iter);
4750 for (; t; t = rhashtable_walk_next(iter)) {
4751 if (IS_ERR(t)) {
4752 if (PTR_ERR(t) == -EAGAIN)
4753 continue;
4754 break;
4757 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4758 t->asoc->peer.primary_path == t)
4759 break;
4762 return t;
4765 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4766 struct rhashtable_iter *iter,
4767 int pos)
4769 void *obj = SEQ_START_TOKEN;
4771 while (pos && (obj = sctp_transport_get_next(net, iter)) &&
4772 !IS_ERR(obj))
4773 pos--;
4775 return obj;
4778 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4779 void *p) {
4780 int err = 0;
4781 int hash = 0;
4782 struct sctp_ep_common *epb;
4783 struct sctp_hashbucket *head;
4785 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4786 hash++, head++) {
4787 read_lock_bh(&head->lock);
4788 sctp_for_each_hentry(epb, &head->chain) {
4789 err = cb(sctp_ep(epb), p);
4790 if (err)
4791 break;
4793 read_unlock_bh(&head->lock);
4796 return err;
4798 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4800 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4801 struct net *net,
4802 const union sctp_addr *laddr,
4803 const union sctp_addr *paddr, void *p)
4805 struct sctp_transport *transport;
4806 int err;
4808 rcu_read_lock();
4809 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4810 rcu_read_unlock();
4811 if (!transport)
4812 return -ENOENT;
4814 err = cb(transport, p);
4815 sctp_transport_put(transport);
4817 return err;
4819 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4821 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4822 int (*cb_done)(struct sctp_transport *, void *),
4823 struct net *net, int *pos, void *p) {
4824 struct rhashtable_iter hti;
4825 struct sctp_transport *tsp;
4826 int ret;
4828 again:
4829 ret = 0;
4830 sctp_transport_walk_start(&hti);
4832 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
4833 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
4834 if (!sctp_transport_hold(tsp))
4835 continue;
4836 ret = cb(tsp, p);
4837 if (ret)
4838 break;
4839 (*pos)++;
4840 sctp_transport_put(tsp);
4842 sctp_transport_walk_stop(&hti);
4844 if (ret) {
4845 if (cb_done && !cb_done(tsp, p)) {
4846 (*pos)++;
4847 sctp_transport_put(tsp);
4848 goto again;
4850 sctp_transport_put(tsp);
4853 return ret;
4855 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4857 /* 7.2.1 Association Status (SCTP_STATUS)
4859 * Applications can retrieve current status information about an
4860 * association, including association state, peer receiver window size,
4861 * number of unacked data chunks, and number of data chunks pending
4862 * receipt. This information is read-only.
4864 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4865 char __user *optval,
4866 int __user *optlen)
4868 struct sctp_status status;
4869 struct sctp_association *asoc = NULL;
4870 struct sctp_transport *transport;
4871 sctp_assoc_t associd;
4872 int retval = 0;
4874 if (len < sizeof(status)) {
4875 retval = -EINVAL;
4876 goto out;
4879 len = sizeof(status);
4880 if (copy_from_user(&status, optval, len)) {
4881 retval = -EFAULT;
4882 goto out;
4885 associd = status.sstat_assoc_id;
4886 asoc = sctp_id2assoc(sk, associd);
4887 if (!asoc) {
4888 retval = -EINVAL;
4889 goto out;
4892 transport = asoc->peer.primary_path;
4894 status.sstat_assoc_id = sctp_assoc2id(asoc);
4895 status.sstat_state = sctp_assoc_to_state(asoc);
4896 status.sstat_rwnd = asoc->peer.rwnd;
4897 status.sstat_unackdata = asoc->unack_data;
4899 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4900 status.sstat_instrms = asoc->stream.incnt;
4901 status.sstat_outstrms = asoc->stream.outcnt;
4902 status.sstat_fragmentation_point = asoc->frag_point;
4903 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4904 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4905 transport->af_specific->sockaddr_len);
4906 /* Map ipv4 address into v4-mapped-on-v6 address. */
4907 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4908 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4909 status.sstat_primary.spinfo_state = transport->state;
4910 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4911 status.sstat_primary.spinfo_srtt = transport->srtt;
4912 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4913 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4915 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4916 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4918 if (put_user(len, optlen)) {
4919 retval = -EFAULT;
4920 goto out;
4923 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4924 __func__, len, status.sstat_state, status.sstat_rwnd,
4925 status.sstat_assoc_id);
4927 if (copy_to_user(optval, &status, len)) {
4928 retval = -EFAULT;
4929 goto out;
4932 out:
4933 return retval;
4937 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4939 * Applications can retrieve information about a specific peer address
4940 * of an association, including its reachability state, congestion
4941 * window, and retransmission timer values. This information is
4942 * read-only.
4944 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4945 char __user *optval,
4946 int __user *optlen)
4948 struct sctp_paddrinfo pinfo;
4949 struct sctp_transport *transport;
4950 int retval = 0;
4952 if (len < sizeof(pinfo)) {
4953 retval = -EINVAL;
4954 goto out;
4957 len = sizeof(pinfo);
4958 if (copy_from_user(&pinfo, optval, len)) {
4959 retval = -EFAULT;
4960 goto out;
4963 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4964 pinfo.spinfo_assoc_id);
4965 if (!transport)
4966 return -EINVAL;
4968 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4969 pinfo.spinfo_state = transport->state;
4970 pinfo.spinfo_cwnd = transport->cwnd;
4971 pinfo.spinfo_srtt = transport->srtt;
4972 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4973 pinfo.spinfo_mtu = transport->pathmtu;
4975 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4976 pinfo.spinfo_state = SCTP_ACTIVE;
4978 if (put_user(len, optlen)) {
4979 retval = -EFAULT;
4980 goto out;
4983 if (copy_to_user(optval, &pinfo, len)) {
4984 retval = -EFAULT;
4985 goto out;
4988 out:
4989 return retval;
4992 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4994 * This option is a on/off flag. If enabled no SCTP message
4995 * fragmentation will be performed. Instead if a message being sent
4996 * exceeds the current PMTU size, the message will NOT be sent and
4997 * instead a error will be indicated to the user.
4999 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5000 char __user *optval, int __user *optlen)
5002 int val;
5004 if (len < sizeof(int))
5005 return -EINVAL;
5007 len = sizeof(int);
5008 val = (sctp_sk(sk)->disable_fragments == 1);
5009 if (put_user(len, optlen))
5010 return -EFAULT;
5011 if (copy_to_user(optval, &val, len))
5012 return -EFAULT;
5013 return 0;
5016 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5018 * This socket option is used to specify various notifications and
5019 * ancillary data the user wishes to receive.
5021 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5022 int __user *optlen)
5024 if (len == 0)
5025 return -EINVAL;
5026 if (len > sizeof(struct sctp_event_subscribe))
5027 len = sizeof(struct sctp_event_subscribe);
5028 if (put_user(len, optlen))
5029 return -EFAULT;
5030 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5031 return -EFAULT;
5032 return 0;
5035 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5037 * This socket option is applicable to the UDP-style socket only. When
5038 * set it will cause associations that are idle for more than the
5039 * specified number of seconds to automatically close. An association
5040 * being idle is defined an association that has NOT sent or received
5041 * user data. The special value of '0' indicates that no automatic
5042 * close of any associations should be performed. The option expects an
5043 * integer defining the number of seconds of idle time before an
5044 * association is closed.
5046 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5048 /* Applicable to UDP-style socket only */
5049 if (sctp_style(sk, TCP))
5050 return -EOPNOTSUPP;
5051 if (len < sizeof(int))
5052 return -EINVAL;
5053 len = sizeof(int);
5054 if (put_user(len, optlen))
5055 return -EFAULT;
5056 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5057 return -EFAULT;
5058 return 0;
5061 /* Helper routine to branch off an association to a new socket. */
5062 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5064 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5065 struct sctp_sock *sp = sctp_sk(sk);
5066 struct socket *sock;
5067 int err = 0;
5069 /* Do not peel off from one netns to another one. */
5070 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5071 return -EINVAL;
5073 if (!asoc)
5074 return -EINVAL;
5076 /* An association cannot be branched off from an already peeled-off
5077 * socket, nor is this supported for tcp style sockets.
5079 if (!sctp_style(sk, UDP))
5080 return -EINVAL;
5082 /* Create a new socket. */
5083 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5084 if (err < 0)
5085 return err;
5087 sctp_copy_sock(sock->sk, sk, asoc);
5089 /* Make peeled-off sockets more like 1-1 accepted sockets.
5090 * Set the daddr and initialize id to something more random
5092 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5094 /* Populate the fields of the newsk from the oldsk and migrate the
5095 * asoc to the newsk.
5097 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5099 *sockp = sock;
5101 return err;
5103 EXPORT_SYMBOL(sctp_do_peeloff);
5105 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5106 struct file **newfile, unsigned flags)
5108 struct socket *newsock;
5109 int retval;
5111 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5112 if (retval < 0)
5113 goto out;
5115 /* Map the socket to an unused fd that can be returned to the user. */
5116 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5117 if (retval < 0) {
5118 sock_release(newsock);
5119 goto out;
5122 *newfile = sock_alloc_file(newsock, 0, NULL);
5123 if (IS_ERR(*newfile)) {
5124 put_unused_fd(retval);
5125 retval = PTR_ERR(*newfile);
5126 *newfile = NULL;
5127 return retval;
5130 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5131 retval);
5133 peeloff->sd = retval;
5135 if (flags & SOCK_NONBLOCK)
5136 (*newfile)->f_flags |= O_NONBLOCK;
5137 out:
5138 return retval;
5141 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5143 sctp_peeloff_arg_t peeloff;
5144 struct file *newfile = NULL;
5145 int retval = 0;
5147 if (len < sizeof(sctp_peeloff_arg_t))
5148 return -EINVAL;
5149 len = sizeof(sctp_peeloff_arg_t);
5150 if (copy_from_user(&peeloff, optval, len))
5151 return -EFAULT;
5153 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5154 if (retval < 0)
5155 goto out;
5157 /* Return the fd mapped to the new socket. */
5158 if (put_user(len, optlen)) {
5159 fput(newfile);
5160 put_unused_fd(retval);
5161 return -EFAULT;
5164 if (copy_to_user(optval, &peeloff, len)) {
5165 fput(newfile);
5166 put_unused_fd(retval);
5167 return -EFAULT;
5169 fd_install(retval, newfile);
5170 out:
5171 return retval;
5174 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5175 char __user *optval, int __user *optlen)
5177 sctp_peeloff_flags_arg_t peeloff;
5178 struct file *newfile = NULL;
5179 int retval = 0;
5181 if (len < sizeof(sctp_peeloff_flags_arg_t))
5182 return -EINVAL;
5183 len = sizeof(sctp_peeloff_flags_arg_t);
5184 if (copy_from_user(&peeloff, optval, len))
5185 return -EFAULT;
5187 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5188 &newfile, peeloff.flags);
5189 if (retval < 0)
5190 goto out;
5192 /* Return the fd mapped to the new socket. */
5193 if (put_user(len, optlen)) {
5194 fput(newfile);
5195 put_unused_fd(retval);
5196 return -EFAULT;
5199 if (copy_to_user(optval, &peeloff, len)) {
5200 fput(newfile);
5201 put_unused_fd(retval);
5202 return -EFAULT;
5204 fd_install(retval, newfile);
5205 out:
5206 return retval;
5209 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5211 * Applications can enable or disable heartbeats for any peer address of
5212 * an association, modify an address's heartbeat interval, force a
5213 * heartbeat to be sent immediately, and adjust the address's maximum
5214 * number of retransmissions sent before an address is considered
5215 * unreachable. The following structure is used to access and modify an
5216 * address's parameters:
5218 * struct sctp_paddrparams {
5219 * sctp_assoc_t spp_assoc_id;
5220 * struct sockaddr_storage spp_address;
5221 * uint32_t spp_hbinterval;
5222 * uint16_t spp_pathmaxrxt;
5223 * uint32_t spp_pathmtu;
5224 * uint32_t spp_sackdelay;
5225 * uint32_t spp_flags;
5226 * };
5228 * spp_assoc_id - (one-to-many style socket) This is filled in the
5229 * application, and identifies the association for
5230 * this query.
5231 * spp_address - This specifies which address is of interest.
5232 * spp_hbinterval - This contains the value of the heartbeat interval,
5233 * in milliseconds. If a value of zero
5234 * is present in this field then no changes are to
5235 * be made to this parameter.
5236 * spp_pathmaxrxt - This contains the maximum number of
5237 * retransmissions before this address shall be
5238 * considered unreachable. If a value of zero
5239 * is present in this field then no changes are to
5240 * be made to this parameter.
5241 * spp_pathmtu - When Path MTU discovery is disabled the value
5242 * specified here will be the "fixed" path mtu.
5243 * Note that if the spp_address field is empty
5244 * then all associations on this address will
5245 * have this fixed path mtu set upon them.
5247 * spp_sackdelay - When delayed sack is enabled, this value specifies
5248 * the number of milliseconds that sacks will be delayed
5249 * for. This value will apply to all addresses of an
5250 * association if the spp_address field is empty. Note
5251 * also, that if delayed sack is enabled and this
5252 * value is set to 0, no change is made to the last
5253 * recorded delayed sack timer value.
5255 * spp_flags - These flags are used to control various features
5256 * on an association. The flag field may contain
5257 * zero or more of the following options.
5259 * SPP_HB_ENABLE - Enable heartbeats on the
5260 * specified address. Note that if the address
5261 * field is empty all addresses for the association
5262 * have heartbeats enabled upon them.
5264 * SPP_HB_DISABLE - Disable heartbeats on the
5265 * speicifed address. Note that if the address
5266 * field is empty all addresses for the association
5267 * will have their heartbeats disabled. Note also
5268 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5269 * mutually exclusive, only one of these two should
5270 * be specified. Enabling both fields will have
5271 * undetermined results.
5273 * SPP_HB_DEMAND - Request a user initiated heartbeat
5274 * to be made immediately.
5276 * SPP_PMTUD_ENABLE - This field will enable PMTU
5277 * discovery upon the specified address. Note that
5278 * if the address feild is empty then all addresses
5279 * on the association are effected.
5281 * SPP_PMTUD_DISABLE - This field will disable PMTU
5282 * discovery upon the specified address. Note that
5283 * if the address feild is empty then all addresses
5284 * on the association are effected. Not also that
5285 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5286 * exclusive. Enabling both will have undetermined
5287 * results.
5289 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5290 * on delayed sack. The time specified in spp_sackdelay
5291 * is used to specify the sack delay for this address. Note
5292 * that if spp_address is empty then all addresses will
5293 * enable delayed sack and take on the sack delay
5294 * value specified in spp_sackdelay.
5295 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5296 * off delayed sack. If the spp_address field is blank then
5297 * delayed sack is disabled for the entire association. Note
5298 * also that this field is mutually exclusive to
5299 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5300 * results.
5302 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5303 char __user *optval, int __user *optlen)
5305 struct sctp_paddrparams params;
5306 struct sctp_transport *trans = NULL;
5307 struct sctp_association *asoc = NULL;
5308 struct sctp_sock *sp = sctp_sk(sk);
5310 if (len < sizeof(struct sctp_paddrparams))
5311 return -EINVAL;
5312 len = sizeof(struct sctp_paddrparams);
5313 if (copy_from_user(&params, optval, len))
5314 return -EFAULT;
5316 /* If an address other than INADDR_ANY is specified, and
5317 * no transport is found, then the request is invalid.
5319 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
5320 trans = sctp_addr_id2transport(sk, &params.spp_address,
5321 params.spp_assoc_id);
5322 if (!trans) {
5323 pr_debug("%s: failed no transport\n", __func__);
5324 return -EINVAL;
5328 /* Get association, if assoc_id != 0 and the socket is a one
5329 * to many style socket, and an association was not found, then
5330 * the id was invalid.
5332 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5333 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5334 pr_debug("%s: failed no association\n", __func__);
5335 return -EINVAL;
5338 if (trans) {
5339 /* Fetch transport values. */
5340 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5341 params.spp_pathmtu = trans->pathmtu;
5342 params.spp_pathmaxrxt = trans->pathmaxrxt;
5343 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5345 /*draft-11 doesn't say what to return in spp_flags*/
5346 params.spp_flags = trans->param_flags;
5347 } else if (asoc) {
5348 /* Fetch association values. */
5349 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5350 params.spp_pathmtu = asoc->pathmtu;
5351 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5352 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5354 /*draft-11 doesn't say what to return in spp_flags*/
5355 params.spp_flags = asoc->param_flags;
5356 } else {
5357 /* Fetch socket values. */
5358 params.spp_hbinterval = sp->hbinterval;
5359 params.spp_pathmtu = sp->pathmtu;
5360 params.spp_sackdelay = sp->sackdelay;
5361 params.spp_pathmaxrxt = sp->pathmaxrxt;
5363 /*draft-11 doesn't say what to return in spp_flags*/
5364 params.spp_flags = sp->param_flags;
5367 if (copy_to_user(optval, &params, len))
5368 return -EFAULT;
5370 if (put_user(len, optlen))
5371 return -EFAULT;
5373 return 0;
5377 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5379 * This option will effect the way delayed acks are performed. This
5380 * option allows you to get or set the delayed ack time, in
5381 * milliseconds. It also allows changing the delayed ack frequency.
5382 * Changing the frequency to 1 disables the delayed sack algorithm. If
5383 * the assoc_id is 0, then this sets or gets the endpoints default
5384 * values. If the assoc_id field is non-zero, then the set or get
5385 * effects the specified association for the one to many model (the
5386 * assoc_id field is ignored by the one to one model). Note that if
5387 * sack_delay or sack_freq are 0 when setting this option, then the
5388 * current values will remain unchanged.
5390 * struct sctp_sack_info {
5391 * sctp_assoc_t sack_assoc_id;
5392 * uint32_t sack_delay;
5393 * uint32_t sack_freq;
5394 * };
5396 * sack_assoc_id - This parameter, indicates which association the user
5397 * is performing an action upon. Note that if this field's value is
5398 * zero then the endpoints default value is changed (effecting future
5399 * associations only).
5401 * sack_delay - This parameter contains the number of milliseconds that
5402 * the user is requesting the delayed ACK timer be set to. Note that
5403 * this value is defined in the standard to be between 200 and 500
5404 * milliseconds.
5406 * sack_freq - This parameter contains the number of packets that must
5407 * be received before a sack is sent without waiting for the delay
5408 * timer to expire. The default value for this is 2, setting this
5409 * value to 1 will disable the delayed sack algorithm.
5411 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5412 char __user *optval,
5413 int __user *optlen)
5415 struct sctp_sack_info params;
5416 struct sctp_association *asoc = NULL;
5417 struct sctp_sock *sp = sctp_sk(sk);
5419 if (len >= sizeof(struct sctp_sack_info)) {
5420 len = sizeof(struct sctp_sack_info);
5422 if (copy_from_user(&params, optval, len))
5423 return -EFAULT;
5424 } else if (len == sizeof(struct sctp_assoc_value)) {
5425 pr_warn_ratelimited(DEPRECATED
5426 "%s (pid %d) "
5427 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5428 "Use struct sctp_sack_info instead\n",
5429 current->comm, task_pid_nr(current));
5430 if (copy_from_user(&params, optval, len))
5431 return -EFAULT;
5432 } else
5433 return -EINVAL;
5435 /* Get association, if sack_assoc_id != 0 and the socket is a one
5436 * to many style socket, and an association was not found, then
5437 * the id was invalid.
5439 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5440 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5441 return -EINVAL;
5443 if (asoc) {
5444 /* Fetch association values. */
5445 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5446 params.sack_delay = jiffies_to_msecs(
5447 asoc->sackdelay);
5448 params.sack_freq = asoc->sackfreq;
5450 } else {
5451 params.sack_delay = 0;
5452 params.sack_freq = 1;
5454 } else {
5455 /* Fetch socket values. */
5456 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5457 params.sack_delay = sp->sackdelay;
5458 params.sack_freq = sp->sackfreq;
5459 } else {
5460 params.sack_delay = 0;
5461 params.sack_freq = 1;
5465 if (copy_to_user(optval, &params, len))
5466 return -EFAULT;
5468 if (put_user(len, optlen))
5469 return -EFAULT;
5471 return 0;
5474 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5476 * Applications can specify protocol parameters for the default association
5477 * initialization. The option name argument to setsockopt() and getsockopt()
5478 * is SCTP_INITMSG.
5480 * Setting initialization parameters is effective only on an unconnected
5481 * socket (for UDP-style sockets only future associations are effected
5482 * by the change). With TCP-style sockets, this option is inherited by
5483 * sockets derived from a listener socket.
5485 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5487 if (len < sizeof(struct sctp_initmsg))
5488 return -EINVAL;
5489 len = sizeof(struct sctp_initmsg);
5490 if (put_user(len, optlen))
5491 return -EFAULT;
5492 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5493 return -EFAULT;
5494 return 0;
5498 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5499 char __user *optval, int __user *optlen)
5501 struct sctp_association *asoc;
5502 int cnt = 0;
5503 struct sctp_getaddrs getaddrs;
5504 struct sctp_transport *from;
5505 void __user *to;
5506 union sctp_addr temp;
5507 struct sctp_sock *sp = sctp_sk(sk);
5508 int addrlen;
5509 size_t space_left;
5510 int bytes_copied;
5512 if (len < sizeof(struct sctp_getaddrs))
5513 return -EINVAL;
5515 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5516 return -EFAULT;
5518 /* For UDP-style sockets, id specifies the association to query. */
5519 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5520 if (!asoc)
5521 return -EINVAL;
5523 to = optval + offsetof(struct sctp_getaddrs, addrs);
5524 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5526 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5527 transports) {
5528 memcpy(&temp, &from->ipaddr, sizeof(temp));
5529 addrlen = sctp_get_pf_specific(sk->sk_family)
5530 ->addr_to_user(sp, &temp);
5531 if (space_left < addrlen)
5532 return -ENOMEM;
5533 if (copy_to_user(to, &temp, addrlen))
5534 return -EFAULT;
5535 to += addrlen;
5536 cnt++;
5537 space_left -= addrlen;
5540 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5541 return -EFAULT;
5542 bytes_copied = ((char __user *)to) - optval;
5543 if (put_user(bytes_copied, optlen))
5544 return -EFAULT;
5546 return 0;
5549 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5550 size_t space_left, int *bytes_copied)
5552 struct sctp_sockaddr_entry *addr;
5553 union sctp_addr temp;
5554 int cnt = 0;
5555 int addrlen;
5556 struct net *net = sock_net(sk);
5558 rcu_read_lock();
5559 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5560 if (!addr->valid)
5561 continue;
5563 if ((PF_INET == sk->sk_family) &&
5564 (AF_INET6 == addr->a.sa.sa_family))
5565 continue;
5566 if ((PF_INET6 == sk->sk_family) &&
5567 inet_v6_ipv6only(sk) &&
5568 (AF_INET == addr->a.sa.sa_family))
5569 continue;
5570 memcpy(&temp, &addr->a, sizeof(temp));
5571 if (!temp.v4.sin_port)
5572 temp.v4.sin_port = htons(port);
5574 addrlen = sctp_get_pf_specific(sk->sk_family)
5575 ->addr_to_user(sctp_sk(sk), &temp);
5577 if (space_left < addrlen) {
5578 cnt = -ENOMEM;
5579 break;
5581 memcpy(to, &temp, addrlen);
5583 to += addrlen;
5584 cnt++;
5585 space_left -= addrlen;
5586 *bytes_copied += addrlen;
5588 rcu_read_unlock();
5590 return cnt;
5594 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5595 char __user *optval, int __user *optlen)
5597 struct sctp_bind_addr *bp;
5598 struct sctp_association *asoc;
5599 int cnt = 0;
5600 struct sctp_getaddrs getaddrs;
5601 struct sctp_sockaddr_entry *addr;
5602 void __user *to;
5603 union sctp_addr temp;
5604 struct sctp_sock *sp = sctp_sk(sk);
5605 int addrlen;
5606 int err = 0;
5607 size_t space_left;
5608 int bytes_copied = 0;
5609 void *addrs;
5610 void *buf;
5612 if (len < sizeof(struct sctp_getaddrs))
5613 return -EINVAL;
5615 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5616 return -EFAULT;
5619 * For UDP-style sockets, id specifies the association to query.
5620 * If the id field is set to the value '0' then the locally bound
5621 * addresses are returned without regard to any particular
5622 * association.
5624 if (0 == getaddrs.assoc_id) {
5625 bp = &sctp_sk(sk)->ep->base.bind_addr;
5626 } else {
5627 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5628 if (!asoc)
5629 return -EINVAL;
5630 bp = &asoc->base.bind_addr;
5633 to = optval + offsetof(struct sctp_getaddrs, addrs);
5634 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5636 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5637 if (!addrs)
5638 return -ENOMEM;
5640 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5641 * addresses from the global local address list.
5643 if (sctp_list_single_entry(&bp->address_list)) {
5644 addr = list_entry(bp->address_list.next,
5645 struct sctp_sockaddr_entry, list);
5646 if (sctp_is_any(sk, &addr->a)) {
5647 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5648 space_left, &bytes_copied);
5649 if (cnt < 0) {
5650 err = cnt;
5651 goto out;
5653 goto copy_getaddrs;
5657 buf = addrs;
5658 /* Protection on the bound address list is not needed since
5659 * in the socket option context we hold a socket lock and
5660 * thus the bound address list can't change.
5662 list_for_each_entry(addr, &bp->address_list, list) {
5663 memcpy(&temp, &addr->a, sizeof(temp));
5664 addrlen = sctp_get_pf_specific(sk->sk_family)
5665 ->addr_to_user(sp, &temp);
5666 if (space_left < addrlen) {
5667 err = -ENOMEM; /*fixme: right error?*/
5668 goto out;
5670 memcpy(buf, &temp, addrlen);
5671 buf += addrlen;
5672 bytes_copied += addrlen;
5673 cnt++;
5674 space_left -= addrlen;
5677 copy_getaddrs:
5678 if (copy_to_user(to, addrs, bytes_copied)) {
5679 err = -EFAULT;
5680 goto out;
5682 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5683 err = -EFAULT;
5684 goto out;
5686 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5687 * but we can't change it anymore.
5689 if (put_user(bytes_copied, optlen))
5690 err = -EFAULT;
5691 out:
5692 kfree(addrs);
5693 return err;
5696 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5698 * Requests that the local SCTP stack use the enclosed peer address as
5699 * the association primary. The enclosed address must be one of the
5700 * association peer's addresses.
5702 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5703 char __user *optval, int __user *optlen)
5705 struct sctp_prim prim;
5706 struct sctp_association *asoc;
5707 struct sctp_sock *sp = sctp_sk(sk);
5709 if (len < sizeof(struct sctp_prim))
5710 return -EINVAL;
5712 len = sizeof(struct sctp_prim);
5714 if (copy_from_user(&prim, optval, len))
5715 return -EFAULT;
5717 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5718 if (!asoc)
5719 return -EINVAL;
5721 if (!asoc->peer.primary_path)
5722 return -ENOTCONN;
5724 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5725 asoc->peer.primary_path->af_specific->sockaddr_len);
5727 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5728 (union sctp_addr *)&prim.ssp_addr);
5730 if (put_user(len, optlen))
5731 return -EFAULT;
5732 if (copy_to_user(optval, &prim, len))
5733 return -EFAULT;
5735 return 0;
5739 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5741 * Requests that the local endpoint set the specified Adaptation Layer
5742 * Indication parameter for all future INIT and INIT-ACK exchanges.
5744 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5745 char __user *optval, int __user *optlen)
5747 struct sctp_setadaptation adaptation;
5749 if (len < sizeof(struct sctp_setadaptation))
5750 return -EINVAL;
5752 len = sizeof(struct sctp_setadaptation);
5754 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5756 if (put_user(len, optlen))
5757 return -EFAULT;
5758 if (copy_to_user(optval, &adaptation, len))
5759 return -EFAULT;
5761 return 0;
5766 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5768 * Applications that wish to use the sendto() system call may wish to
5769 * specify a default set of parameters that would normally be supplied
5770 * through the inclusion of ancillary data. This socket option allows
5771 * such an application to set the default sctp_sndrcvinfo structure.
5774 * The application that wishes to use this socket option simply passes
5775 * in to this call the sctp_sndrcvinfo structure defined in Section
5776 * 5.2.2) The input parameters accepted by this call include
5777 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5778 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5779 * to this call if the caller is using the UDP model.
5781 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5783 static int sctp_getsockopt_default_send_param(struct sock *sk,
5784 int len, char __user *optval,
5785 int __user *optlen)
5787 struct sctp_sock *sp = sctp_sk(sk);
5788 struct sctp_association *asoc;
5789 struct sctp_sndrcvinfo info;
5791 if (len < sizeof(info))
5792 return -EINVAL;
5794 len = sizeof(info);
5796 if (copy_from_user(&info, optval, len))
5797 return -EFAULT;
5799 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5800 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5801 return -EINVAL;
5802 if (asoc) {
5803 info.sinfo_stream = asoc->default_stream;
5804 info.sinfo_flags = asoc->default_flags;
5805 info.sinfo_ppid = asoc->default_ppid;
5806 info.sinfo_context = asoc->default_context;
5807 info.sinfo_timetolive = asoc->default_timetolive;
5808 } else {
5809 info.sinfo_stream = sp->default_stream;
5810 info.sinfo_flags = sp->default_flags;
5811 info.sinfo_ppid = sp->default_ppid;
5812 info.sinfo_context = sp->default_context;
5813 info.sinfo_timetolive = sp->default_timetolive;
5816 if (put_user(len, optlen))
5817 return -EFAULT;
5818 if (copy_to_user(optval, &info, len))
5819 return -EFAULT;
5821 return 0;
5824 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5825 * (SCTP_DEFAULT_SNDINFO)
5827 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5828 char __user *optval,
5829 int __user *optlen)
5831 struct sctp_sock *sp = sctp_sk(sk);
5832 struct sctp_association *asoc;
5833 struct sctp_sndinfo info;
5835 if (len < sizeof(info))
5836 return -EINVAL;
5838 len = sizeof(info);
5840 if (copy_from_user(&info, optval, len))
5841 return -EFAULT;
5843 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5844 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5845 return -EINVAL;
5846 if (asoc) {
5847 info.snd_sid = asoc->default_stream;
5848 info.snd_flags = asoc->default_flags;
5849 info.snd_ppid = asoc->default_ppid;
5850 info.snd_context = asoc->default_context;
5851 } else {
5852 info.snd_sid = sp->default_stream;
5853 info.snd_flags = sp->default_flags;
5854 info.snd_ppid = sp->default_ppid;
5855 info.snd_context = sp->default_context;
5858 if (put_user(len, optlen))
5859 return -EFAULT;
5860 if (copy_to_user(optval, &info, len))
5861 return -EFAULT;
5863 return 0;
5868 * 7.1.5 SCTP_NODELAY
5870 * Turn on/off any Nagle-like algorithm. This means that packets are
5871 * generally sent as soon as possible and no unnecessary delays are
5872 * introduced, at the cost of more packets in the network. Expects an
5873 * integer boolean flag.
5876 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5877 char __user *optval, int __user *optlen)
5879 int val;
5881 if (len < sizeof(int))
5882 return -EINVAL;
5884 len = sizeof(int);
5885 val = (sctp_sk(sk)->nodelay == 1);
5886 if (put_user(len, optlen))
5887 return -EFAULT;
5888 if (copy_to_user(optval, &val, len))
5889 return -EFAULT;
5890 return 0;
5895 * 7.1.1 SCTP_RTOINFO
5897 * The protocol parameters used to initialize and bound retransmission
5898 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5899 * and modify these parameters.
5900 * All parameters are time values, in milliseconds. A value of 0, when
5901 * modifying the parameters, indicates that the current value should not
5902 * be changed.
5905 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5906 char __user *optval,
5907 int __user *optlen) {
5908 struct sctp_rtoinfo rtoinfo;
5909 struct sctp_association *asoc;
5911 if (len < sizeof (struct sctp_rtoinfo))
5912 return -EINVAL;
5914 len = sizeof(struct sctp_rtoinfo);
5916 if (copy_from_user(&rtoinfo, optval, len))
5917 return -EFAULT;
5919 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5921 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5922 return -EINVAL;
5924 /* Values corresponding to the specific association. */
5925 if (asoc) {
5926 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5927 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5928 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5929 } else {
5930 /* Values corresponding to the endpoint. */
5931 struct sctp_sock *sp = sctp_sk(sk);
5933 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5934 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5935 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5938 if (put_user(len, optlen))
5939 return -EFAULT;
5941 if (copy_to_user(optval, &rtoinfo, len))
5942 return -EFAULT;
5944 return 0;
5949 * 7.1.2 SCTP_ASSOCINFO
5951 * This option is used to tune the maximum retransmission attempts
5952 * of the association.
5953 * Returns an error if the new association retransmission value is
5954 * greater than the sum of the retransmission value of the peer.
5955 * See [SCTP] for more information.
5958 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5959 char __user *optval,
5960 int __user *optlen)
5963 struct sctp_assocparams assocparams;
5964 struct sctp_association *asoc;
5965 struct list_head *pos;
5966 int cnt = 0;
5968 if (len < sizeof (struct sctp_assocparams))
5969 return -EINVAL;
5971 len = sizeof(struct sctp_assocparams);
5973 if (copy_from_user(&assocparams, optval, len))
5974 return -EFAULT;
5976 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5978 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5979 return -EINVAL;
5981 /* Values correspoinding to the specific association */
5982 if (asoc) {
5983 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5984 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5985 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5986 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5988 list_for_each(pos, &asoc->peer.transport_addr_list) {
5989 cnt++;
5992 assocparams.sasoc_number_peer_destinations = cnt;
5993 } else {
5994 /* Values corresponding to the endpoint */
5995 struct sctp_sock *sp = sctp_sk(sk);
5997 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5998 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5999 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6000 assocparams.sasoc_cookie_life =
6001 sp->assocparams.sasoc_cookie_life;
6002 assocparams.sasoc_number_peer_destinations =
6003 sp->assocparams.
6004 sasoc_number_peer_destinations;
6007 if (put_user(len, optlen))
6008 return -EFAULT;
6010 if (copy_to_user(optval, &assocparams, len))
6011 return -EFAULT;
6013 return 0;
6017 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6019 * This socket option is a boolean flag which turns on or off mapped V4
6020 * addresses. If this option is turned on and the socket is type
6021 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6022 * If this option is turned off, then no mapping will be done of V4
6023 * addresses and a user will receive both PF_INET6 and PF_INET type
6024 * addresses on the socket.
6026 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6027 char __user *optval, int __user *optlen)
6029 int val;
6030 struct sctp_sock *sp = sctp_sk(sk);
6032 if (len < sizeof(int))
6033 return -EINVAL;
6035 len = sizeof(int);
6036 val = sp->v4mapped;
6037 if (put_user(len, optlen))
6038 return -EFAULT;
6039 if (copy_to_user(optval, &val, len))
6040 return -EFAULT;
6042 return 0;
6046 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6047 * (chapter and verse is quoted at sctp_setsockopt_context())
6049 static int sctp_getsockopt_context(struct sock *sk, int len,
6050 char __user *optval, int __user *optlen)
6052 struct sctp_assoc_value params;
6053 struct sctp_sock *sp;
6054 struct sctp_association *asoc;
6056 if (len < sizeof(struct sctp_assoc_value))
6057 return -EINVAL;
6059 len = sizeof(struct sctp_assoc_value);
6061 if (copy_from_user(&params, optval, len))
6062 return -EFAULT;
6064 sp = sctp_sk(sk);
6066 if (params.assoc_id != 0) {
6067 asoc = sctp_id2assoc(sk, params.assoc_id);
6068 if (!asoc)
6069 return -EINVAL;
6070 params.assoc_value = asoc->default_rcv_context;
6071 } else {
6072 params.assoc_value = sp->default_rcv_context;
6075 if (put_user(len, optlen))
6076 return -EFAULT;
6077 if (copy_to_user(optval, &params, len))
6078 return -EFAULT;
6080 return 0;
6084 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6085 * This option will get or set the maximum size to put in any outgoing
6086 * SCTP DATA chunk. If a message is larger than this size it will be
6087 * fragmented by SCTP into the specified size. Note that the underlying
6088 * SCTP implementation may fragment into smaller sized chunks when the
6089 * PMTU of the underlying association is smaller than the value set by
6090 * the user. The default value for this option is '0' which indicates
6091 * the user is NOT limiting fragmentation and only the PMTU will effect
6092 * SCTP's choice of DATA chunk size. Note also that values set larger
6093 * than the maximum size of an IP datagram will effectively let SCTP
6094 * control fragmentation (i.e. the same as setting this option to 0).
6096 * The following structure is used to access and modify this parameter:
6098 * struct sctp_assoc_value {
6099 * sctp_assoc_t assoc_id;
6100 * uint32_t assoc_value;
6101 * };
6103 * assoc_id: This parameter is ignored for one-to-one style sockets.
6104 * For one-to-many style sockets this parameter indicates which
6105 * association the user is performing an action upon. Note that if
6106 * this field's value is zero then the endpoints default value is
6107 * changed (effecting future associations only).
6108 * assoc_value: This parameter specifies the maximum size in bytes.
6110 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6111 char __user *optval, int __user *optlen)
6113 struct sctp_assoc_value params;
6114 struct sctp_association *asoc;
6116 if (len == sizeof(int)) {
6117 pr_warn_ratelimited(DEPRECATED
6118 "%s (pid %d) "
6119 "Use of int in maxseg socket option.\n"
6120 "Use struct sctp_assoc_value instead\n",
6121 current->comm, task_pid_nr(current));
6122 params.assoc_id = 0;
6123 } else if (len >= sizeof(struct sctp_assoc_value)) {
6124 len = sizeof(struct sctp_assoc_value);
6125 if (copy_from_user(&params, optval, len))
6126 return -EFAULT;
6127 } else
6128 return -EINVAL;
6130 asoc = sctp_id2assoc(sk, params.assoc_id);
6131 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6132 return -EINVAL;
6134 if (asoc)
6135 params.assoc_value = asoc->frag_point;
6136 else
6137 params.assoc_value = sctp_sk(sk)->user_frag;
6139 if (put_user(len, optlen))
6140 return -EFAULT;
6141 if (len == sizeof(int)) {
6142 if (copy_to_user(optval, &params.assoc_value, len))
6143 return -EFAULT;
6144 } else {
6145 if (copy_to_user(optval, &params, len))
6146 return -EFAULT;
6149 return 0;
6153 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6154 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6156 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6157 char __user *optval, int __user *optlen)
6159 int val;
6161 if (len < sizeof(int))
6162 return -EINVAL;
6164 len = sizeof(int);
6166 val = sctp_sk(sk)->frag_interleave;
6167 if (put_user(len, optlen))
6168 return -EFAULT;
6169 if (copy_to_user(optval, &val, len))
6170 return -EFAULT;
6172 return 0;
6176 * 7.1.25. Set or Get the sctp partial delivery point
6177 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6179 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6180 char __user *optval,
6181 int __user *optlen)
6183 u32 val;
6185 if (len < sizeof(u32))
6186 return -EINVAL;
6188 len = sizeof(u32);
6190 val = sctp_sk(sk)->pd_point;
6191 if (put_user(len, optlen))
6192 return -EFAULT;
6193 if (copy_to_user(optval, &val, len))
6194 return -EFAULT;
6196 return 0;
6200 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6201 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6203 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6204 char __user *optval,
6205 int __user *optlen)
6207 struct sctp_assoc_value params;
6208 struct sctp_sock *sp;
6209 struct sctp_association *asoc;
6211 if (len == sizeof(int)) {
6212 pr_warn_ratelimited(DEPRECATED
6213 "%s (pid %d) "
6214 "Use of int in max_burst socket option.\n"
6215 "Use struct sctp_assoc_value instead\n",
6216 current->comm, task_pid_nr(current));
6217 params.assoc_id = 0;
6218 } else if (len >= sizeof(struct sctp_assoc_value)) {
6219 len = sizeof(struct sctp_assoc_value);
6220 if (copy_from_user(&params, optval, len))
6221 return -EFAULT;
6222 } else
6223 return -EINVAL;
6225 sp = sctp_sk(sk);
6227 if (params.assoc_id != 0) {
6228 asoc = sctp_id2assoc(sk, params.assoc_id);
6229 if (!asoc)
6230 return -EINVAL;
6231 params.assoc_value = asoc->max_burst;
6232 } else
6233 params.assoc_value = sp->max_burst;
6235 if (len == sizeof(int)) {
6236 if (copy_to_user(optval, &params.assoc_value, len))
6237 return -EFAULT;
6238 } else {
6239 if (copy_to_user(optval, &params, len))
6240 return -EFAULT;
6243 return 0;
6247 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6248 char __user *optval, int __user *optlen)
6250 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6251 struct sctp_hmacalgo __user *p = (void __user *)optval;
6252 struct sctp_hmac_algo_param *hmacs;
6253 __u16 data_len = 0;
6254 u32 num_idents;
6255 int i;
6257 if (!ep->auth_enable)
6258 return -EACCES;
6260 hmacs = ep->auth_hmacs_list;
6261 data_len = ntohs(hmacs->param_hdr.length) -
6262 sizeof(struct sctp_paramhdr);
6264 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6265 return -EINVAL;
6267 len = sizeof(struct sctp_hmacalgo) + data_len;
6268 num_idents = data_len / sizeof(u16);
6270 if (put_user(len, optlen))
6271 return -EFAULT;
6272 if (put_user(num_idents, &p->shmac_num_idents))
6273 return -EFAULT;
6274 for (i = 0; i < num_idents; i++) {
6275 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6277 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6278 return -EFAULT;
6280 return 0;
6283 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6284 char __user *optval, int __user *optlen)
6286 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6287 struct sctp_authkeyid val;
6288 struct sctp_association *asoc;
6290 if (!ep->auth_enable)
6291 return -EACCES;
6293 if (len < sizeof(struct sctp_authkeyid))
6294 return -EINVAL;
6296 len = sizeof(struct sctp_authkeyid);
6297 if (copy_from_user(&val, optval, len))
6298 return -EFAULT;
6300 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6301 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6302 return -EINVAL;
6304 if (asoc)
6305 val.scact_keynumber = asoc->active_key_id;
6306 else
6307 val.scact_keynumber = ep->active_key_id;
6309 if (put_user(len, optlen))
6310 return -EFAULT;
6311 if (copy_to_user(optval, &val, len))
6312 return -EFAULT;
6314 return 0;
6317 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6318 char __user *optval, int __user *optlen)
6320 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6321 struct sctp_authchunks __user *p = (void __user *)optval;
6322 struct sctp_authchunks val;
6323 struct sctp_association *asoc;
6324 struct sctp_chunks_param *ch;
6325 u32 num_chunks = 0;
6326 char __user *to;
6328 if (!ep->auth_enable)
6329 return -EACCES;
6331 if (len < sizeof(struct sctp_authchunks))
6332 return -EINVAL;
6334 if (copy_from_user(&val, optval, sizeof(val)))
6335 return -EFAULT;
6337 to = p->gauth_chunks;
6338 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6339 if (!asoc)
6340 return -EINVAL;
6342 ch = asoc->peer.peer_chunks;
6343 if (!ch)
6344 goto num;
6346 /* See if the user provided enough room for all the data */
6347 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6348 if (len < num_chunks)
6349 return -EINVAL;
6351 if (copy_to_user(to, ch->chunks, num_chunks))
6352 return -EFAULT;
6353 num:
6354 len = sizeof(struct sctp_authchunks) + num_chunks;
6355 if (put_user(len, optlen))
6356 return -EFAULT;
6357 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6358 return -EFAULT;
6359 return 0;
6362 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6363 char __user *optval, int __user *optlen)
6365 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6366 struct sctp_authchunks __user *p = (void __user *)optval;
6367 struct sctp_authchunks val;
6368 struct sctp_association *asoc;
6369 struct sctp_chunks_param *ch;
6370 u32 num_chunks = 0;
6371 char __user *to;
6373 if (!ep->auth_enable)
6374 return -EACCES;
6376 if (len < sizeof(struct sctp_authchunks))
6377 return -EINVAL;
6379 if (copy_from_user(&val, optval, sizeof(val)))
6380 return -EFAULT;
6382 to = p->gauth_chunks;
6383 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6384 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6385 return -EINVAL;
6387 if (asoc)
6388 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6389 else
6390 ch = ep->auth_chunk_list;
6392 if (!ch)
6393 goto num;
6395 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6396 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6397 return -EINVAL;
6399 if (copy_to_user(to, ch->chunks, num_chunks))
6400 return -EFAULT;
6401 num:
6402 len = sizeof(struct sctp_authchunks) + num_chunks;
6403 if (put_user(len, optlen))
6404 return -EFAULT;
6405 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6406 return -EFAULT;
6408 return 0;
6412 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6413 * This option gets the current number of associations that are attached
6414 * to a one-to-many style socket. The option value is an uint32_t.
6416 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6417 char __user *optval, int __user *optlen)
6419 struct sctp_sock *sp = sctp_sk(sk);
6420 struct sctp_association *asoc;
6421 u32 val = 0;
6423 if (sctp_style(sk, TCP))
6424 return -EOPNOTSUPP;
6426 if (len < sizeof(u32))
6427 return -EINVAL;
6429 len = sizeof(u32);
6431 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6432 val++;
6435 if (put_user(len, optlen))
6436 return -EFAULT;
6437 if (copy_to_user(optval, &val, len))
6438 return -EFAULT;
6440 return 0;
6444 * 8.1.23 SCTP_AUTO_ASCONF
6445 * See the corresponding setsockopt entry as description
6447 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6448 char __user *optval, int __user *optlen)
6450 int val = 0;
6452 if (len < sizeof(int))
6453 return -EINVAL;
6455 len = sizeof(int);
6456 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6457 val = 1;
6458 if (put_user(len, optlen))
6459 return -EFAULT;
6460 if (copy_to_user(optval, &val, len))
6461 return -EFAULT;
6462 return 0;
6466 * 8.2.6. Get the Current Identifiers of Associations
6467 * (SCTP_GET_ASSOC_ID_LIST)
6469 * This option gets the current list of SCTP association identifiers of
6470 * the SCTP associations handled by a one-to-many style socket.
6472 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6473 char __user *optval, int __user *optlen)
6475 struct sctp_sock *sp = sctp_sk(sk);
6476 struct sctp_association *asoc;
6477 struct sctp_assoc_ids *ids;
6478 u32 num = 0;
6480 if (sctp_style(sk, TCP))
6481 return -EOPNOTSUPP;
6483 if (len < sizeof(struct sctp_assoc_ids))
6484 return -EINVAL;
6486 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6487 num++;
6490 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6491 return -EINVAL;
6493 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6495 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6496 if (unlikely(!ids))
6497 return -ENOMEM;
6499 ids->gaids_number_of_ids = num;
6500 num = 0;
6501 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6502 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6505 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6506 kfree(ids);
6507 return -EFAULT;
6510 kfree(ids);
6511 return 0;
6515 * SCTP_PEER_ADDR_THLDS
6517 * This option allows us to fetch the partially failed threshold for one or all
6518 * transports in an association. See Section 6.1 of:
6519 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6521 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6522 char __user *optval,
6523 int len,
6524 int __user *optlen)
6526 struct sctp_paddrthlds val;
6527 struct sctp_transport *trans;
6528 struct sctp_association *asoc;
6530 if (len < sizeof(struct sctp_paddrthlds))
6531 return -EINVAL;
6532 len = sizeof(struct sctp_paddrthlds);
6533 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6534 return -EFAULT;
6536 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6537 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6538 if (!asoc)
6539 return -ENOENT;
6541 val.spt_pathpfthld = asoc->pf_retrans;
6542 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6543 } else {
6544 trans = sctp_addr_id2transport(sk, &val.spt_address,
6545 val.spt_assoc_id);
6546 if (!trans)
6547 return -ENOENT;
6549 val.spt_pathmaxrxt = trans->pathmaxrxt;
6550 val.spt_pathpfthld = trans->pf_retrans;
6553 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6554 return -EFAULT;
6556 return 0;
6560 * SCTP_GET_ASSOC_STATS
6562 * This option retrieves local per endpoint statistics. It is modeled
6563 * after OpenSolaris' implementation
6565 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6566 char __user *optval,
6567 int __user *optlen)
6569 struct sctp_assoc_stats sas;
6570 struct sctp_association *asoc = NULL;
6572 /* User must provide at least the assoc id */
6573 if (len < sizeof(sctp_assoc_t))
6574 return -EINVAL;
6576 /* Allow the struct to grow and fill in as much as possible */
6577 len = min_t(size_t, len, sizeof(sas));
6579 if (copy_from_user(&sas, optval, len))
6580 return -EFAULT;
6582 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6583 if (!asoc)
6584 return -EINVAL;
6586 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6587 sas.sas_gapcnt = asoc->stats.gapcnt;
6588 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6589 sas.sas_osacks = asoc->stats.osacks;
6590 sas.sas_isacks = asoc->stats.isacks;
6591 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6592 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6593 sas.sas_oodchunks = asoc->stats.oodchunks;
6594 sas.sas_iodchunks = asoc->stats.iodchunks;
6595 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6596 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6597 sas.sas_idupchunks = asoc->stats.idupchunks;
6598 sas.sas_opackets = asoc->stats.opackets;
6599 sas.sas_ipackets = asoc->stats.ipackets;
6601 /* New high max rto observed, will return 0 if not a single
6602 * RTO update took place. obs_rto_ipaddr will be bogus
6603 * in such a case
6605 sas.sas_maxrto = asoc->stats.max_obs_rto;
6606 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6607 sizeof(struct sockaddr_storage));
6609 /* Mark beginning of a new observation period */
6610 asoc->stats.max_obs_rto = asoc->rto_min;
6612 if (put_user(len, optlen))
6613 return -EFAULT;
6615 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6617 if (copy_to_user(optval, &sas, len))
6618 return -EFAULT;
6620 return 0;
6623 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6624 char __user *optval,
6625 int __user *optlen)
6627 int val = 0;
6629 if (len < sizeof(int))
6630 return -EINVAL;
6632 len = sizeof(int);
6633 if (sctp_sk(sk)->recvrcvinfo)
6634 val = 1;
6635 if (put_user(len, optlen))
6636 return -EFAULT;
6637 if (copy_to_user(optval, &val, len))
6638 return -EFAULT;
6640 return 0;
6643 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6644 char __user *optval,
6645 int __user *optlen)
6647 int val = 0;
6649 if (len < sizeof(int))
6650 return -EINVAL;
6652 len = sizeof(int);
6653 if (sctp_sk(sk)->recvnxtinfo)
6654 val = 1;
6655 if (put_user(len, optlen))
6656 return -EFAULT;
6657 if (copy_to_user(optval, &val, len))
6658 return -EFAULT;
6660 return 0;
6663 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6664 char __user *optval,
6665 int __user *optlen)
6667 struct sctp_assoc_value params;
6668 struct sctp_association *asoc;
6669 int retval = -EFAULT;
6671 if (len < sizeof(params)) {
6672 retval = -EINVAL;
6673 goto out;
6676 len = sizeof(params);
6677 if (copy_from_user(&params, optval, len))
6678 goto out;
6680 asoc = sctp_id2assoc(sk, params.assoc_id);
6681 if (asoc) {
6682 params.assoc_value = asoc->prsctp_enable;
6683 } else if (!params.assoc_id) {
6684 struct sctp_sock *sp = sctp_sk(sk);
6686 params.assoc_value = sp->ep->prsctp_enable;
6687 } else {
6688 retval = -EINVAL;
6689 goto out;
6692 if (put_user(len, optlen))
6693 goto out;
6695 if (copy_to_user(optval, &params, len))
6696 goto out;
6698 retval = 0;
6700 out:
6701 return retval;
6704 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6705 char __user *optval,
6706 int __user *optlen)
6708 struct sctp_default_prinfo info;
6709 struct sctp_association *asoc;
6710 int retval = -EFAULT;
6712 if (len < sizeof(info)) {
6713 retval = -EINVAL;
6714 goto out;
6717 len = sizeof(info);
6718 if (copy_from_user(&info, optval, len))
6719 goto out;
6721 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6722 if (asoc) {
6723 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6724 info.pr_value = asoc->default_timetolive;
6725 } else if (!info.pr_assoc_id) {
6726 struct sctp_sock *sp = sctp_sk(sk);
6728 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6729 info.pr_value = sp->default_timetolive;
6730 } else {
6731 retval = -EINVAL;
6732 goto out;
6735 if (put_user(len, optlen))
6736 goto out;
6738 if (copy_to_user(optval, &info, len))
6739 goto out;
6741 retval = 0;
6743 out:
6744 return retval;
6747 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6748 char __user *optval,
6749 int __user *optlen)
6751 struct sctp_prstatus params;
6752 struct sctp_association *asoc;
6753 int policy;
6754 int retval = -EINVAL;
6756 if (len < sizeof(params))
6757 goto out;
6759 len = sizeof(params);
6760 if (copy_from_user(&params, optval, len)) {
6761 retval = -EFAULT;
6762 goto out;
6765 policy = params.sprstat_policy;
6766 if (policy & ~SCTP_PR_SCTP_MASK)
6767 goto out;
6769 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6770 if (!asoc)
6771 goto out;
6773 if (policy == SCTP_PR_SCTP_NONE) {
6774 params.sprstat_abandoned_unsent = 0;
6775 params.sprstat_abandoned_sent = 0;
6776 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6777 params.sprstat_abandoned_unsent +=
6778 asoc->abandoned_unsent[policy];
6779 params.sprstat_abandoned_sent +=
6780 asoc->abandoned_sent[policy];
6782 } else {
6783 params.sprstat_abandoned_unsent =
6784 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6785 params.sprstat_abandoned_sent =
6786 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6789 if (put_user(len, optlen)) {
6790 retval = -EFAULT;
6791 goto out;
6794 if (copy_to_user(optval, &params, len)) {
6795 retval = -EFAULT;
6796 goto out;
6799 retval = 0;
6801 out:
6802 return retval;
6805 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6806 char __user *optval,
6807 int __user *optlen)
6809 struct sctp_stream_out_ext *streamoute;
6810 struct sctp_association *asoc;
6811 struct sctp_prstatus params;
6812 int retval = -EINVAL;
6813 int policy;
6815 if (len < sizeof(params))
6816 goto out;
6818 len = sizeof(params);
6819 if (copy_from_user(&params, optval, len)) {
6820 retval = -EFAULT;
6821 goto out;
6824 policy = params.sprstat_policy;
6825 if (policy & ~SCTP_PR_SCTP_MASK)
6826 goto out;
6828 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6829 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6830 goto out;
6832 streamoute = asoc->stream.out[params.sprstat_sid].ext;
6833 if (!streamoute) {
6834 /* Not allocated yet, means all stats are 0 */
6835 params.sprstat_abandoned_unsent = 0;
6836 params.sprstat_abandoned_sent = 0;
6837 retval = 0;
6838 goto out;
6841 if (policy == SCTP_PR_SCTP_NONE) {
6842 params.sprstat_abandoned_unsent = 0;
6843 params.sprstat_abandoned_sent = 0;
6844 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6845 params.sprstat_abandoned_unsent +=
6846 streamoute->abandoned_unsent[policy];
6847 params.sprstat_abandoned_sent +=
6848 streamoute->abandoned_sent[policy];
6850 } else {
6851 params.sprstat_abandoned_unsent =
6852 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6853 params.sprstat_abandoned_sent =
6854 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
6857 if (put_user(len, optlen) || copy_to_user(optval, &params, len)) {
6858 retval = -EFAULT;
6859 goto out;
6862 retval = 0;
6864 out:
6865 return retval;
6868 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6869 char __user *optval,
6870 int __user *optlen)
6872 struct sctp_assoc_value params;
6873 struct sctp_association *asoc;
6874 int retval = -EFAULT;
6876 if (len < sizeof(params)) {
6877 retval = -EINVAL;
6878 goto out;
6881 len = sizeof(params);
6882 if (copy_from_user(&params, optval, len))
6883 goto out;
6885 asoc = sctp_id2assoc(sk, params.assoc_id);
6886 if (asoc) {
6887 params.assoc_value = asoc->reconf_enable;
6888 } else if (!params.assoc_id) {
6889 struct sctp_sock *sp = sctp_sk(sk);
6891 params.assoc_value = sp->ep->reconf_enable;
6892 } else {
6893 retval = -EINVAL;
6894 goto out;
6897 if (put_user(len, optlen))
6898 goto out;
6900 if (copy_to_user(optval, &params, len))
6901 goto out;
6903 retval = 0;
6905 out:
6906 return retval;
6909 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6910 char __user *optval,
6911 int __user *optlen)
6913 struct sctp_assoc_value params;
6914 struct sctp_association *asoc;
6915 int retval = -EFAULT;
6917 if (len < sizeof(params)) {
6918 retval = -EINVAL;
6919 goto out;
6922 len = sizeof(params);
6923 if (copy_from_user(&params, optval, len))
6924 goto out;
6926 asoc = sctp_id2assoc(sk, params.assoc_id);
6927 if (asoc) {
6928 params.assoc_value = asoc->strreset_enable;
6929 } else if (!params.assoc_id) {
6930 struct sctp_sock *sp = sctp_sk(sk);
6932 params.assoc_value = sp->ep->strreset_enable;
6933 } else {
6934 retval = -EINVAL;
6935 goto out;
6938 if (put_user(len, optlen))
6939 goto out;
6941 if (copy_to_user(optval, &params, len))
6942 goto out;
6944 retval = 0;
6946 out:
6947 return retval;
6950 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
6951 char __user *optval,
6952 int __user *optlen)
6954 struct sctp_assoc_value params;
6955 struct sctp_association *asoc;
6956 int retval = -EFAULT;
6958 if (len < sizeof(params)) {
6959 retval = -EINVAL;
6960 goto out;
6963 len = sizeof(params);
6964 if (copy_from_user(&params, optval, len))
6965 goto out;
6967 asoc = sctp_id2assoc(sk, params.assoc_id);
6968 if (!asoc) {
6969 retval = -EINVAL;
6970 goto out;
6973 params.assoc_value = sctp_sched_get_sched(asoc);
6975 if (put_user(len, optlen))
6976 goto out;
6978 if (copy_to_user(optval, &params, len))
6979 goto out;
6981 retval = 0;
6983 out:
6984 return retval;
6987 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
6988 char __user *optval,
6989 int __user *optlen)
6991 struct sctp_stream_value params;
6992 struct sctp_association *asoc;
6993 int retval = -EFAULT;
6995 if (len < sizeof(params)) {
6996 retval = -EINVAL;
6997 goto out;
7000 len = sizeof(params);
7001 if (copy_from_user(&params, optval, len))
7002 goto out;
7004 asoc = sctp_id2assoc(sk, params.assoc_id);
7005 if (!asoc) {
7006 retval = -EINVAL;
7007 goto out;
7010 retval = sctp_sched_get_value(asoc, params.stream_id,
7011 &params.stream_value);
7012 if (retval)
7013 goto out;
7015 if (put_user(len, optlen)) {
7016 retval = -EFAULT;
7017 goto out;
7020 if (copy_to_user(optval, &params, len)) {
7021 retval = -EFAULT;
7022 goto out;
7025 out:
7026 return retval;
7029 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7030 char __user *optval,
7031 int __user *optlen)
7033 struct sctp_assoc_value params;
7034 struct sctp_association *asoc;
7035 int retval = -EFAULT;
7037 if (len < sizeof(params)) {
7038 retval = -EINVAL;
7039 goto out;
7042 len = sizeof(params);
7043 if (copy_from_user(&params, optval, len))
7044 goto out;
7046 asoc = sctp_id2assoc(sk, params.assoc_id);
7047 if (asoc) {
7048 params.assoc_value = asoc->intl_enable;
7049 } else if (!params.assoc_id) {
7050 struct sctp_sock *sp = sctp_sk(sk);
7052 params.assoc_value = sp->strm_interleave;
7053 } else {
7054 retval = -EINVAL;
7055 goto out;
7058 if (put_user(len, optlen))
7059 goto out;
7061 if (copy_to_user(optval, &params, len))
7062 goto out;
7064 retval = 0;
7066 out:
7067 return retval;
7070 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7071 char __user *optval, int __user *optlen)
7073 int retval = 0;
7074 int len;
7076 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7078 /* I can hardly begin to describe how wrong this is. This is
7079 * so broken as to be worse than useless. The API draft
7080 * REALLY is NOT helpful here... I am not convinced that the
7081 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7082 * are at all well-founded.
7084 if (level != SOL_SCTP) {
7085 struct sctp_af *af = sctp_sk(sk)->pf->af;
7087 retval = af->getsockopt(sk, level, optname, optval, optlen);
7088 return retval;
7091 if (get_user(len, optlen))
7092 return -EFAULT;
7094 if (len < 0)
7095 return -EINVAL;
7097 lock_sock(sk);
7099 switch (optname) {
7100 case SCTP_STATUS:
7101 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7102 break;
7103 case SCTP_DISABLE_FRAGMENTS:
7104 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7105 optlen);
7106 break;
7107 case SCTP_EVENTS:
7108 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7109 break;
7110 case SCTP_AUTOCLOSE:
7111 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7112 break;
7113 case SCTP_SOCKOPT_PEELOFF:
7114 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7115 break;
7116 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7117 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7118 break;
7119 case SCTP_PEER_ADDR_PARAMS:
7120 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7121 optlen);
7122 break;
7123 case SCTP_DELAYED_SACK:
7124 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7125 optlen);
7126 break;
7127 case SCTP_INITMSG:
7128 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7129 break;
7130 case SCTP_GET_PEER_ADDRS:
7131 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7132 optlen);
7133 break;
7134 case SCTP_GET_LOCAL_ADDRS:
7135 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7136 optlen);
7137 break;
7138 case SCTP_SOCKOPT_CONNECTX3:
7139 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7140 break;
7141 case SCTP_DEFAULT_SEND_PARAM:
7142 retval = sctp_getsockopt_default_send_param(sk, len,
7143 optval, optlen);
7144 break;
7145 case SCTP_DEFAULT_SNDINFO:
7146 retval = sctp_getsockopt_default_sndinfo(sk, len,
7147 optval, optlen);
7148 break;
7149 case SCTP_PRIMARY_ADDR:
7150 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7151 break;
7152 case SCTP_NODELAY:
7153 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7154 break;
7155 case SCTP_RTOINFO:
7156 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7157 break;
7158 case SCTP_ASSOCINFO:
7159 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7160 break;
7161 case SCTP_I_WANT_MAPPED_V4_ADDR:
7162 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7163 break;
7164 case SCTP_MAXSEG:
7165 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7166 break;
7167 case SCTP_GET_PEER_ADDR_INFO:
7168 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7169 optlen);
7170 break;
7171 case SCTP_ADAPTATION_LAYER:
7172 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7173 optlen);
7174 break;
7175 case SCTP_CONTEXT:
7176 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7177 break;
7178 case SCTP_FRAGMENT_INTERLEAVE:
7179 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7180 optlen);
7181 break;
7182 case SCTP_PARTIAL_DELIVERY_POINT:
7183 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7184 optlen);
7185 break;
7186 case SCTP_MAX_BURST:
7187 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7188 break;
7189 case SCTP_AUTH_KEY:
7190 case SCTP_AUTH_CHUNK:
7191 case SCTP_AUTH_DELETE_KEY:
7192 retval = -EOPNOTSUPP;
7193 break;
7194 case SCTP_HMAC_IDENT:
7195 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7196 break;
7197 case SCTP_AUTH_ACTIVE_KEY:
7198 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7199 break;
7200 case SCTP_PEER_AUTH_CHUNKS:
7201 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7202 optlen);
7203 break;
7204 case SCTP_LOCAL_AUTH_CHUNKS:
7205 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7206 optlen);
7207 break;
7208 case SCTP_GET_ASSOC_NUMBER:
7209 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7210 break;
7211 case SCTP_GET_ASSOC_ID_LIST:
7212 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7213 break;
7214 case SCTP_AUTO_ASCONF:
7215 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7216 break;
7217 case SCTP_PEER_ADDR_THLDS:
7218 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7219 break;
7220 case SCTP_GET_ASSOC_STATS:
7221 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7222 break;
7223 case SCTP_RECVRCVINFO:
7224 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7225 break;
7226 case SCTP_RECVNXTINFO:
7227 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7228 break;
7229 case SCTP_PR_SUPPORTED:
7230 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7231 break;
7232 case SCTP_DEFAULT_PRINFO:
7233 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7234 optlen);
7235 break;
7236 case SCTP_PR_ASSOC_STATUS:
7237 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7238 optlen);
7239 break;
7240 case SCTP_PR_STREAM_STATUS:
7241 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7242 optlen);
7243 break;
7244 case SCTP_RECONFIG_SUPPORTED:
7245 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7246 optlen);
7247 break;
7248 case SCTP_ENABLE_STREAM_RESET:
7249 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7250 optlen);
7251 break;
7252 case SCTP_STREAM_SCHEDULER:
7253 retval = sctp_getsockopt_scheduler(sk, len, optval,
7254 optlen);
7255 break;
7256 case SCTP_STREAM_SCHEDULER_VALUE:
7257 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7258 optlen);
7259 break;
7260 case SCTP_INTERLEAVING_SUPPORTED:
7261 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7262 optlen);
7263 break;
7264 default:
7265 retval = -ENOPROTOOPT;
7266 break;
7269 release_sock(sk);
7270 return retval;
7273 static int sctp_hash(struct sock *sk)
7275 /* STUB */
7276 return 0;
7279 static void sctp_unhash(struct sock *sk)
7281 /* STUB */
7284 /* Check if port is acceptable. Possibly find first available port.
7286 * The port hash table (contained in the 'global' SCTP protocol storage
7287 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7288 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7289 * list (the list number is the port number hashed out, so as you
7290 * would expect from a hash function, all the ports in a given list have
7291 * such a number that hashes out to the same list number; you were
7292 * expecting that, right?); so each list has a set of ports, with a
7293 * link to the socket (struct sock) that uses it, the port number and
7294 * a fastreuse flag (FIXME: NPI ipg).
7296 static struct sctp_bind_bucket *sctp_bucket_create(
7297 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7299 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7301 struct sctp_bind_hashbucket *head; /* hash list */
7302 struct sctp_bind_bucket *pp;
7303 unsigned short snum;
7304 int ret;
7306 snum = ntohs(addr->v4.sin_port);
7308 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7310 local_bh_disable();
7312 if (snum == 0) {
7313 /* Search for an available port. */
7314 int low, high, remaining, index;
7315 unsigned int rover;
7316 struct net *net = sock_net(sk);
7318 inet_get_local_port_range(net, &low, &high);
7319 remaining = (high - low) + 1;
7320 rover = prandom_u32() % remaining + low;
7322 do {
7323 rover++;
7324 if ((rover < low) || (rover > high))
7325 rover = low;
7326 if (inet_is_local_reserved_port(net, rover))
7327 continue;
7328 index = sctp_phashfn(sock_net(sk), rover);
7329 head = &sctp_port_hashtable[index];
7330 spin_lock(&head->lock);
7331 sctp_for_each_hentry(pp, &head->chain)
7332 if ((pp->port == rover) &&
7333 net_eq(sock_net(sk), pp->net))
7334 goto next;
7335 break;
7336 next:
7337 spin_unlock(&head->lock);
7338 } while (--remaining > 0);
7340 /* Exhausted local port range during search? */
7341 ret = 1;
7342 if (remaining <= 0)
7343 goto fail;
7345 /* OK, here is the one we will use. HEAD (the port
7346 * hash table list entry) is non-NULL and we hold it's
7347 * mutex.
7349 snum = rover;
7350 } else {
7351 /* We are given an specific port number; we verify
7352 * that it is not being used. If it is used, we will
7353 * exahust the search in the hash list corresponding
7354 * to the port number (snum) - we detect that with the
7355 * port iterator, pp being NULL.
7357 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7358 spin_lock(&head->lock);
7359 sctp_for_each_hentry(pp, &head->chain) {
7360 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7361 goto pp_found;
7364 pp = NULL;
7365 goto pp_not_found;
7366 pp_found:
7367 if (!hlist_empty(&pp->owner)) {
7368 /* We had a port hash table hit - there is an
7369 * available port (pp != NULL) and it is being
7370 * used by other socket (pp->owner not empty); that other
7371 * socket is going to be sk2.
7373 int reuse = sk->sk_reuse;
7374 struct sock *sk2;
7376 pr_debug("%s: found a possible match\n", __func__);
7378 if (pp->fastreuse && sk->sk_reuse &&
7379 sk->sk_state != SCTP_SS_LISTENING)
7380 goto success;
7382 /* Run through the list of sockets bound to the port
7383 * (pp->port) [via the pointers bind_next and
7384 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7385 * we get the endpoint they describe and run through
7386 * the endpoint's list of IP (v4 or v6) addresses,
7387 * comparing each of the addresses with the address of
7388 * the socket sk. If we find a match, then that means
7389 * that this port/socket (sk) combination are already
7390 * in an endpoint.
7392 sk_for_each_bound(sk2, &pp->owner) {
7393 struct sctp_endpoint *ep2;
7394 ep2 = sctp_sk(sk2)->ep;
7396 if (sk == sk2 ||
7397 (reuse && sk2->sk_reuse &&
7398 sk2->sk_state != SCTP_SS_LISTENING))
7399 continue;
7401 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7402 sctp_sk(sk2), sctp_sk(sk))) {
7403 ret = (long)sk2;
7404 goto fail_unlock;
7408 pr_debug("%s: found a match\n", __func__);
7410 pp_not_found:
7411 /* If there was a hash table miss, create a new port. */
7412 ret = 1;
7413 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7414 goto fail_unlock;
7416 /* In either case (hit or miss), make sure fastreuse is 1 only
7417 * if sk->sk_reuse is too (that is, if the caller requested
7418 * SO_REUSEADDR on this socket -sk-).
7420 if (hlist_empty(&pp->owner)) {
7421 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7422 pp->fastreuse = 1;
7423 else
7424 pp->fastreuse = 0;
7425 } else if (pp->fastreuse &&
7426 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7427 pp->fastreuse = 0;
7429 /* We are set, so fill up all the data in the hash table
7430 * entry, tie the socket list information with the rest of the
7431 * sockets FIXME: Blurry, NPI (ipg).
7433 success:
7434 if (!sctp_sk(sk)->bind_hash) {
7435 inet_sk(sk)->inet_num = snum;
7436 sk_add_bind_node(sk, &pp->owner);
7437 sctp_sk(sk)->bind_hash = pp;
7439 ret = 0;
7441 fail_unlock:
7442 spin_unlock(&head->lock);
7444 fail:
7445 local_bh_enable();
7446 return ret;
7449 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7450 * port is requested.
7452 static int sctp_get_port(struct sock *sk, unsigned short snum)
7454 union sctp_addr addr;
7455 struct sctp_af *af = sctp_sk(sk)->pf->af;
7457 /* Set up a dummy address struct from the sk. */
7458 af->from_sk(&addr, sk);
7459 addr.v4.sin_port = htons(snum);
7461 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7462 return !!sctp_get_port_local(sk, &addr);
7466 * Move a socket to LISTENING state.
7468 static int sctp_listen_start(struct sock *sk, int backlog)
7470 struct sctp_sock *sp = sctp_sk(sk);
7471 struct sctp_endpoint *ep = sp->ep;
7472 struct crypto_shash *tfm = NULL;
7473 char alg[32];
7475 /* Allocate HMAC for generating cookie. */
7476 if (!sp->hmac && sp->sctp_hmac_alg) {
7477 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7478 tfm = crypto_alloc_shash(alg, 0, 0);
7479 if (IS_ERR(tfm)) {
7480 net_info_ratelimited("failed to load transform for %s: %ld\n",
7481 sp->sctp_hmac_alg, PTR_ERR(tfm));
7482 return -ENOSYS;
7484 sctp_sk(sk)->hmac = tfm;
7488 * If a bind() or sctp_bindx() is not called prior to a listen()
7489 * call that allows new associations to be accepted, the system
7490 * picks an ephemeral port and will choose an address set equivalent
7491 * to binding with a wildcard address.
7493 * This is not currently spelled out in the SCTP sockets
7494 * extensions draft, but follows the practice as seen in TCP
7495 * sockets.
7498 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7499 if (!ep->base.bind_addr.port) {
7500 if (sctp_autobind(sk))
7501 return -EAGAIN;
7502 } else {
7503 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7504 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7505 return -EADDRINUSE;
7509 sk->sk_max_ack_backlog = backlog;
7510 sctp_hash_endpoint(ep);
7511 return 0;
7515 * 4.1.3 / 5.1.3 listen()
7517 * By default, new associations are not accepted for UDP style sockets.
7518 * An application uses listen() to mark a socket as being able to
7519 * accept new associations.
7521 * On TCP style sockets, applications use listen() to ready the SCTP
7522 * endpoint for accepting inbound associations.
7524 * On both types of endpoints a backlog of '0' disables listening.
7526 * Move a socket to LISTENING state.
7528 int sctp_inet_listen(struct socket *sock, int backlog)
7530 struct sock *sk = sock->sk;
7531 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7532 int err = -EINVAL;
7534 if (unlikely(backlog < 0))
7535 return err;
7537 lock_sock(sk);
7539 /* Peeled-off sockets are not allowed to listen(). */
7540 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7541 goto out;
7543 if (sock->state != SS_UNCONNECTED)
7544 goto out;
7546 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7547 goto out;
7549 /* If backlog is zero, disable listening. */
7550 if (!backlog) {
7551 if (sctp_sstate(sk, CLOSED))
7552 goto out;
7554 err = 0;
7555 sctp_unhash_endpoint(ep);
7556 sk->sk_state = SCTP_SS_CLOSED;
7557 if (sk->sk_reuse)
7558 sctp_sk(sk)->bind_hash->fastreuse = 1;
7559 goto out;
7562 /* If we are already listening, just update the backlog */
7563 if (sctp_sstate(sk, LISTENING))
7564 sk->sk_max_ack_backlog = backlog;
7565 else {
7566 err = sctp_listen_start(sk, backlog);
7567 if (err)
7568 goto out;
7571 err = 0;
7572 out:
7573 release_sock(sk);
7574 return err;
7578 * This function is done by modeling the current datagram_poll() and the
7579 * tcp_poll(). Note that, based on these implementations, we don't
7580 * lock the socket in this function, even though it seems that,
7581 * ideally, locking or some other mechanisms can be used to ensure
7582 * the integrity of the counters (sndbuf and wmem_alloc) used
7583 * in this place. We assume that we don't need locks either until proven
7584 * otherwise.
7586 * Another thing to note is that we include the Async I/O support
7587 * here, again, by modeling the current TCP/UDP code. We don't have
7588 * a good way to test with it yet.
7590 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7592 struct sock *sk = sock->sk;
7593 struct sctp_sock *sp = sctp_sk(sk);
7594 __poll_t mask;
7596 poll_wait(file, sk_sleep(sk), wait);
7598 sock_rps_record_flow(sk);
7600 /* A TCP-style listening socket becomes readable when the accept queue
7601 * is not empty.
7603 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7604 return (!list_empty(&sp->ep->asocs)) ?
7605 (EPOLLIN | EPOLLRDNORM) : 0;
7607 mask = 0;
7609 /* Is there any exceptional events? */
7610 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7611 mask |= EPOLLERR |
7612 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7613 if (sk->sk_shutdown & RCV_SHUTDOWN)
7614 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7615 if (sk->sk_shutdown == SHUTDOWN_MASK)
7616 mask |= EPOLLHUP;
7618 /* Is it readable? Reconsider this code with TCP-style support. */
7619 if (!skb_queue_empty(&sk->sk_receive_queue))
7620 mask |= EPOLLIN | EPOLLRDNORM;
7622 /* The association is either gone or not ready. */
7623 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7624 return mask;
7626 /* Is it writable? */
7627 if (sctp_writeable(sk)) {
7628 mask |= EPOLLOUT | EPOLLWRNORM;
7629 } else {
7630 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7632 * Since the socket is not locked, the buffer
7633 * might be made available after the writeable check and
7634 * before the bit is set. This could cause a lost I/O
7635 * signal. tcp_poll() has a race breaker for this race
7636 * condition. Based on their implementation, we put
7637 * in the following code to cover it as well.
7639 if (sctp_writeable(sk))
7640 mask |= EPOLLOUT | EPOLLWRNORM;
7642 return mask;
7645 /********************************************************************
7646 * 2nd Level Abstractions
7647 ********************************************************************/
7649 static struct sctp_bind_bucket *sctp_bucket_create(
7650 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7652 struct sctp_bind_bucket *pp;
7654 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7655 if (pp) {
7656 SCTP_DBG_OBJCNT_INC(bind_bucket);
7657 pp->port = snum;
7658 pp->fastreuse = 0;
7659 INIT_HLIST_HEAD(&pp->owner);
7660 pp->net = net;
7661 hlist_add_head(&pp->node, &head->chain);
7663 return pp;
7666 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7667 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7669 if (pp && hlist_empty(&pp->owner)) {
7670 __hlist_del(&pp->node);
7671 kmem_cache_free(sctp_bucket_cachep, pp);
7672 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7676 /* Release this socket's reference to a local port. */
7677 static inline void __sctp_put_port(struct sock *sk)
7679 struct sctp_bind_hashbucket *head =
7680 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7681 inet_sk(sk)->inet_num)];
7682 struct sctp_bind_bucket *pp;
7684 spin_lock(&head->lock);
7685 pp = sctp_sk(sk)->bind_hash;
7686 __sk_del_bind_node(sk);
7687 sctp_sk(sk)->bind_hash = NULL;
7688 inet_sk(sk)->inet_num = 0;
7689 sctp_bucket_destroy(pp);
7690 spin_unlock(&head->lock);
7693 void sctp_put_port(struct sock *sk)
7695 local_bh_disable();
7696 __sctp_put_port(sk);
7697 local_bh_enable();
7701 * The system picks an ephemeral port and choose an address set equivalent
7702 * to binding with a wildcard address.
7703 * One of those addresses will be the primary address for the association.
7704 * This automatically enables the multihoming capability of SCTP.
7706 static int sctp_autobind(struct sock *sk)
7708 union sctp_addr autoaddr;
7709 struct sctp_af *af;
7710 __be16 port;
7712 /* Initialize a local sockaddr structure to INADDR_ANY. */
7713 af = sctp_sk(sk)->pf->af;
7715 port = htons(inet_sk(sk)->inet_num);
7716 af->inaddr_any(&autoaddr, port);
7718 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7721 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7723 * From RFC 2292
7724 * 4.2 The cmsghdr Structure *
7726 * When ancillary data is sent or received, any number of ancillary data
7727 * objects can be specified by the msg_control and msg_controllen members of
7728 * the msghdr structure, because each object is preceded by
7729 * a cmsghdr structure defining the object's length (the cmsg_len member).
7730 * Historically Berkeley-derived implementations have passed only one object
7731 * at a time, but this API allows multiple objects to be
7732 * passed in a single call to sendmsg() or recvmsg(). The following example
7733 * shows two ancillary data objects in a control buffer.
7735 * |<--------------------------- msg_controllen -------------------------->|
7736 * | |
7738 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7740 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7741 * | | |
7743 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7745 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7746 * | | | | |
7748 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7749 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7751 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7753 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7757 * msg_control
7758 * points here
7760 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
7762 struct msghdr *my_msg = (struct msghdr *)msg;
7763 struct cmsghdr *cmsg;
7765 for_each_cmsghdr(cmsg, my_msg) {
7766 if (!CMSG_OK(my_msg, cmsg))
7767 return -EINVAL;
7769 /* Should we parse this header or ignore? */
7770 if (cmsg->cmsg_level != IPPROTO_SCTP)
7771 continue;
7773 /* Strictly check lengths following example in SCM code. */
7774 switch (cmsg->cmsg_type) {
7775 case SCTP_INIT:
7776 /* SCTP Socket API Extension
7777 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7779 * This cmsghdr structure provides information for
7780 * initializing new SCTP associations with sendmsg().
7781 * The SCTP_INITMSG socket option uses this same data
7782 * structure. This structure is not used for
7783 * recvmsg().
7785 * cmsg_level cmsg_type cmsg_data[]
7786 * ------------ ------------ ----------------------
7787 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7789 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7790 return -EINVAL;
7792 cmsgs->init = CMSG_DATA(cmsg);
7793 break;
7795 case SCTP_SNDRCV:
7796 /* SCTP Socket API Extension
7797 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7799 * This cmsghdr structure specifies SCTP options for
7800 * sendmsg() and describes SCTP header information
7801 * about a received message through recvmsg().
7803 * cmsg_level cmsg_type cmsg_data[]
7804 * ------------ ------------ ----------------------
7805 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7807 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7808 return -EINVAL;
7810 cmsgs->srinfo = CMSG_DATA(cmsg);
7812 if (cmsgs->srinfo->sinfo_flags &
7813 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7814 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7815 SCTP_ABORT | SCTP_EOF))
7816 return -EINVAL;
7817 break;
7819 case SCTP_SNDINFO:
7820 /* SCTP Socket API Extension
7821 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7823 * This cmsghdr structure specifies SCTP options for
7824 * sendmsg(). This structure and SCTP_RCVINFO replaces
7825 * SCTP_SNDRCV which has been deprecated.
7827 * cmsg_level cmsg_type cmsg_data[]
7828 * ------------ ------------ ---------------------
7829 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7831 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7832 return -EINVAL;
7834 cmsgs->sinfo = CMSG_DATA(cmsg);
7836 if (cmsgs->sinfo->snd_flags &
7837 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7838 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7839 SCTP_ABORT | SCTP_EOF))
7840 return -EINVAL;
7841 break;
7842 default:
7843 return -EINVAL;
7847 return 0;
7851 * Wait for a packet..
7852 * Note: This function is the same function as in core/datagram.c
7853 * with a few modifications to make lksctp work.
7855 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7857 int error;
7858 DEFINE_WAIT(wait);
7860 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7862 /* Socket errors? */
7863 error = sock_error(sk);
7864 if (error)
7865 goto out;
7867 if (!skb_queue_empty(&sk->sk_receive_queue))
7868 goto ready;
7870 /* Socket shut down? */
7871 if (sk->sk_shutdown & RCV_SHUTDOWN)
7872 goto out;
7874 /* Sequenced packets can come disconnected. If so we report the
7875 * problem.
7877 error = -ENOTCONN;
7879 /* Is there a good reason to think that we may receive some data? */
7880 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7881 goto out;
7883 /* Handle signals. */
7884 if (signal_pending(current))
7885 goto interrupted;
7887 /* Let another process have a go. Since we are going to sleep
7888 * anyway. Note: This may cause odd behaviors if the message
7889 * does not fit in the user's buffer, but this seems to be the
7890 * only way to honor MSG_DONTWAIT realistically.
7892 release_sock(sk);
7893 *timeo_p = schedule_timeout(*timeo_p);
7894 lock_sock(sk);
7896 ready:
7897 finish_wait(sk_sleep(sk), &wait);
7898 return 0;
7900 interrupted:
7901 error = sock_intr_errno(*timeo_p);
7903 out:
7904 finish_wait(sk_sleep(sk), &wait);
7905 *err = error;
7906 return error;
7909 /* Receive a datagram.
7910 * Note: This is pretty much the same routine as in core/datagram.c
7911 * with a few changes to make lksctp work.
7913 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7914 int noblock, int *err)
7916 int error;
7917 struct sk_buff *skb;
7918 long timeo;
7920 timeo = sock_rcvtimeo(sk, noblock);
7922 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7923 MAX_SCHEDULE_TIMEOUT);
7925 do {
7926 /* Again only user level code calls this function,
7927 * so nothing interrupt level
7928 * will suddenly eat the receive_queue.
7930 * Look at current nfs client by the way...
7931 * However, this function was correct in any case. 8)
7933 if (flags & MSG_PEEK) {
7934 skb = skb_peek(&sk->sk_receive_queue);
7935 if (skb)
7936 refcount_inc(&skb->users);
7937 } else {
7938 skb = __skb_dequeue(&sk->sk_receive_queue);
7941 if (skb)
7942 return skb;
7944 /* Caller is allowed not to check sk->sk_err before calling. */
7945 error = sock_error(sk);
7946 if (error)
7947 goto no_packet;
7949 if (sk->sk_shutdown & RCV_SHUTDOWN)
7950 break;
7952 if (sk_can_busy_loop(sk)) {
7953 sk_busy_loop(sk, noblock);
7955 if (!skb_queue_empty(&sk->sk_receive_queue))
7956 continue;
7959 /* User doesn't want to wait. */
7960 error = -EAGAIN;
7961 if (!timeo)
7962 goto no_packet;
7963 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7965 return NULL;
7967 no_packet:
7968 *err = error;
7969 return NULL;
7972 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7973 static void __sctp_write_space(struct sctp_association *asoc)
7975 struct sock *sk = asoc->base.sk;
7977 if (sctp_wspace(asoc) <= 0)
7978 return;
7980 if (waitqueue_active(&asoc->wait))
7981 wake_up_interruptible(&asoc->wait);
7983 if (sctp_writeable(sk)) {
7984 struct socket_wq *wq;
7986 rcu_read_lock();
7987 wq = rcu_dereference(sk->sk_wq);
7988 if (wq) {
7989 if (waitqueue_active(&wq->wait))
7990 wake_up_interruptible(&wq->wait);
7992 /* Note that we try to include the Async I/O support
7993 * here by modeling from the current TCP/UDP code.
7994 * We have not tested with it yet.
7996 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7997 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7999 rcu_read_unlock();
8003 static void sctp_wake_up_waiters(struct sock *sk,
8004 struct sctp_association *asoc)
8006 struct sctp_association *tmp = asoc;
8008 /* We do accounting for the sndbuf space per association,
8009 * so we only need to wake our own association.
8011 if (asoc->ep->sndbuf_policy)
8012 return __sctp_write_space(asoc);
8014 /* If association goes down and is just flushing its
8015 * outq, then just normally notify others.
8017 if (asoc->base.dead)
8018 return sctp_write_space(sk);
8020 /* Accounting for the sndbuf space is per socket, so we
8021 * need to wake up others, try to be fair and in case of
8022 * other associations, let them have a go first instead
8023 * of just doing a sctp_write_space() call.
8025 * Note that we reach sctp_wake_up_waiters() only when
8026 * associations free up queued chunks, thus we are under
8027 * lock and the list of associations on a socket is
8028 * guaranteed not to change.
8030 for (tmp = list_next_entry(tmp, asocs); 1;
8031 tmp = list_next_entry(tmp, asocs)) {
8032 /* Manually skip the head element. */
8033 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8034 continue;
8035 /* Wake up association. */
8036 __sctp_write_space(tmp);
8037 /* We've reached the end. */
8038 if (tmp == asoc)
8039 break;
8043 /* Do accounting for the sndbuf space.
8044 * Decrement the used sndbuf space of the corresponding association by the
8045 * data size which was just transmitted(freed).
8047 static void sctp_wfree(struct sk_buff *skb)
8049 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8050 struct sctp_association *asoc = chunk->asoc;
8051 struct sock *sk = asoc->base.sk;
8053 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
8054 sizeof(struct sk_buff) +
8055 sizeof(struct sctp_chunk);
8057 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
8060 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
8062 sk->sk_wmem_queued -= skb->truesize;
8063 sk_mem_uncharge(sk, skb->truesize);
8065 sock_wfree(skb);
8066 sctp_wake_up_waiters(sk, asoc);
8068 sctp_association_put(asoc);
8071 /* Do accounting for the receive space on the socket.
8072 * Accounting for the association is done in ulpevent.c
8073 * We set this as a destructor for the cloned data skbs so that
8074 * accounting is done at the correct time.
8076 void sctp_sock_rfree(struct sk_buff *skb)
8078 struct sock *sk = skb->sk;
8079 struct sctp_ulpevent *event = sctp_skb2event(skb);
8081 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8084 * Mimic the behavior of sock_rfree
8086 sk_mem_uncharge(sk, event->rmem_len);
8090 /* Helper function to wait for space in the sndbuf. */
8091 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8092 size_t msg_len)
8094 struct sock *sk = asoc->base.sk;
8095 long current_timeo = *timeo_p;
8096 DEFINE_WAIT(wait);
8097 int err = 0;
8099 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8100 *timeo_p, msg_len);
8102 /* Increment the association's refcnt. */
8103 sctp_association_hold(asoc);
8105 /* Wait on the association specific sndbuf space. */
8106 for (;;) {
8107 prepare_to_wait_exclusive(&asoc->wait, &wait,
8108 TASK_INTERRUPTIBLE);
8109 if (asoc->base.dead)
8110 goto do_dead;
8111 if (!*timeo_p)
8112 goto do_nonblock;
8113 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8114 goto do_error;
8115 if (signal_pending(current))
8116 goto do_interrupted;
8117 if (msg_len <= sctp_wspace(asoc))
8118 break;
8120 /* Let another process have a go. Since we are going
8121 * to sleep anyway.
8123 release_sock(sk);
8124 current_timeo = schedule_timeout(current_timeo);
8125 lock_sock(sk);
8126 if (sk != asoc->base.sk)
8127 goto do_error;
8129 *timeo_p = current_timeo;
8132 out:
8133 finish_wait(&asoc->wait, &wait);
8135 /* Release the association's refcnt. */
8136 sctp_association_put(asoc);
8138 return err;
8140 do_dead:
8141 err = -ESRCH;
8142 goto out;
8144 do_error:
8145 err = -EPIPE;
8146 goto out;
8148 do_interrupted:
8149 err = sock_intr_errno(*timeo_p);
8150 goto out;
8152 do_nonblock:
8153 err = -EAGAIN;
8154 goto out;
8157 void sctp_data_ready(struct sock *sk)
8159 struct socket_wq *wq;
8161 rcu_read_lock();
8162 wq = rcu_dereference(sk->sk_wq);
8163 if (skwq_has_sleeper(wq))
8164 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8165 EPOLLRDNORM | EPOLLRDBAND);
8166 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8167 rcu_read_unlock();
8170 /* If socket sndbuf has changed, wake up all per association waiters. */
8171 void sctp_write_space(struct sock *sk)
8173 struct sctp_association *asoc;
8175 /* Wake up the tasks in each wait queue. */
8176 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8177 __sctp_write_space(asoc);
8181 /* Is there any sndbuf space available on the socket?
8183 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8184 * associations on the same socket. For a UDP-style socket with
8185 * multiple associations, it is possible for it to be "unwriteable"
8186 * prematurely. I assume that this is acceptable because
8187 * a premature "unwriteable" is better than an accidental "writeable" which
8188 * would cause an unwanted block under certain circumstances. For the 1-1
8189 * UDP-style sockets or TCP-style sockets, this code should work.
8190 * - Daisy
8192 static int sctp_writeable(struct sock *sk)
8194 int amt = 0;
8196 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
8197 if (amt < 0)
8198 amt = 0;
8199 return amt;
8202 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8203 * returns immediately with EINPROGRESS.
8205 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8207 struct sock *sk = asoc->base.sk;
8208 int err = 0;
8209 long current_timeo = *timeo_p;
8210 DEFINE_WAIT(wait);
8212 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8214 /* Increment the association's refcnt. */
8215 sctp_association_hold(asoc);
8217 for (;;) {
8218 prepare_to_wait_exclusive(&asoc->wait, &wait,
8219 TASK_INTERRUPTIBLE);
8220 if (!*timeo_p)
8221 goto do_nonblock;
8222 if (sk->sk_shutdown & RCV_SHUTDOWN)
8223 break;
8224 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8225 asoc->base.dead)
8226 goto do_error;
8227 if (signal_pending(current))
8228 goto do_interrupted;
8230 if (sctp_state(asoc, ESTABLISHED))
8231 break;
8233 /* Let another process have a go. Since we are going
8234 * to sleep anyway.
8236 release_sock(sk);
8237 current_timeo = schedule_timeout(current_timeo);
8238 lock_sock(sk);
8240 *timeo_p = current_timeo;
8243 out:
8244 finish_wait(&asoc->wait, &wait);
8246 /* Release the association's refcnt. */
8247 sctp_association_put(asoc);
8249 return err;
8251 do_error:
8252 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8253 err = -ETIMEDOUT;
8254 else
8255 err = -ECONNREFUSED;
8256 goto out;
8258 do_interrupted:
8259 err = sock_intr_errno(*timeo_p);
8260 goto out;
8262 do_nonblock:
8263 err = -EINPROGRESS;
8264 goto out;
8267 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8269 struct sctp_endpoint *ep;
8270 int err = 0;
8271 DEFINE_WAIT(wait);
8273 ep = sctp_sk(sk)->ep;
8276 for (;;) {
8277 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8278 TASK_INTERRUPTIBLE);
8280 if (list_empty(&ep->asocs)) {
8281 release_sock(sk);
8282 timeo = schedule_timeout(timeo);
8283 lock_sock(sk);
8286 err = -EINVAL;
8287 if (!sctp_sstate(sk, LISTENING))
8288 break;
8290 err = 0;
8291 if (!list_empty(&ep->asocs))
8292 break;
8294 err = sock_intr_errno(timeo);
8295 if (signal_pending(current))
8296 break;
8298 err = -EAGAIN;
8299 if (!timeo)
8300 break;
8303 finish_wait(sk_sleep(sk), &wait);
8305 return err;
8308 static void sctp_wait_for_close(struct sock *sk, long timeout)
8310 DEFINE_WAIT(wait);
8312 do {
8313 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8314 if (list_empty(&sctp_sk(sk)->ep->asocs))
8315 break;
8316 release_sock(sk);
8317 timeout = schedule_timeout(timeout);
8318 lock_sock(sk);
8319 } while (!signal_pending(current) && timeout);
8321 finish_wait(sk_sleep(sk), &wait);
8324 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8326 struct sk_buff *frag;
8328 if (!skb->data_len)
8329 goto done;
8331 /* Don't forget the fragments. */
8332 skb_walk_frags(skb, frag)
8333 sctp_skb_set_owner_r_frag(frag, sk);
8335 done:
8336 sctp_skb_set_owner_r(skb, sk);
8339 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8340 struct sctp_association *asoc)
8342 struct inet_sock *inet = inet_sk(sk);
8343 struct inet_sock *newinet;
8345 newsk->sk_type = sk->sk_type;
8346 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8347 newsk->sk_flags = sk->sk_flags;
8348 newsk->sk_tsflags = sk->sk_tsflags;
8349 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8350 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8351 newsk->sk_reuse = sk->sk_reuse;
8353 newsk->sk_shutdown = sk->sk_shutdown;
8354 newsk->sk_destruct = sctp_destruct_sock;
8355 newsk->sk_family = sk->sk_family;
8356 newsk->sk_protocol = IPPROTO_SCTP;
8357 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8358 newsk->sk_sndbuf = sk->sk_sndbuf;
8359 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8360 newsk->sk_lingertime = sk->sk_lingertime;
8361 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8362 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8363 newsk->sk_rxhash = sk->sk_rxhash;
8365 newinet = inet_sk(newsk);
8367 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8368 * getsockname() and getpeername()
8370 newinet->inet_sport = inet->inet_sport;
8371 newinet->inet_saddr = inet->inet_saddr;
8372 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8373 newinet->inet_dport = htons(asoc->peer.port);
8374 newinet->pmtudisc = inet->pmtudisc;
8375 newinet->inet_id = asoc->next_tsn ^ jiffies;
8377 newinet->uc_ttl = inet->uc_ttl;
8378 newinet->mc_loop = 1;
8379 newinet->mc_ttl = 1;
8380 newinet->mc_index = 0;
8381 newinet->mc_list = NULL;
8383 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8384 net_enable_timestamp();
8386 security_sk_clone(sk, newsk);
8389 static inline void sctp_copy_descendant(struct sock *sk_to,
8390 const struct sock *sk_from)
8392 int ancestor_size = sizeof(struct inet_sock) +
8393 sizeof(struct sctp_sock) -
8394 offsetof(struct sctp_sock, auto_asconf_list);
8396 if (sk_from->sk_family == PF_INET6)
8397 ancestor_size += sizeof(struct ipv6_pinfo);
8399 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8402 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8403 * and its messages to the newsk.
8405 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8406 struct sctp_association *assoc,
8407 enum sctp_socket_type type)
8409 struct sctp_sock *oldsp = sctp_sk(oldsk);
8410 struct sctp_sock *newsp = sctp_sk(newsk);
8411 struct sctp_bind_bucket *pp; /* hash list port iterator */
8412 struct sctp_endpoint *newep = newsp->ep;
8413 struct sk_buff *skb, *tmp;
8414 struct sctp_ulpevent *event;
8415 struct sctp_bind_hashbucket *head;
8417 /* Migrate socket buffer sizes and all the socket level options to the
8418 * new socket.
8420 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8421 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8422 /* Brute force copy old sctp opt. */
8423 sctp_copy_descendant(newsk, oldsk);
8425 /* Restore the ep value that was overwritten with the above structure
8426 * copy.
8428 newsp->ep = newep;
8429 newsp->hmac = NULL;
8431 /* Hook this new socket in to the bind_hash list. */
8432 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8433 inet_sk(oldsk)->inet_num)];
8434 spin_lock_bh(&head->lock);
8435 pp = sctp_sk(oldsk)->bind_hash;
8436 sk_add_bind_node(newsk, &pp->owner);
8437 sctp_sk(newsk)->bind_hash = pp;
8438 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8439 spin_unlock_bh(&head->lock);
8441 /* Copy the bind_addr list from the original endpoint to the new
8442 * endpoint so that we can handle restarts properly
8444 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8445 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8447 /* Move any messages in the old socket's receive queue that are for the
8448 * peeled off association to the new socket's receive queue.
8450 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8451 event = sctp_skb2event(skb);
8452 if (event->asoc == assoc) {
8453 __skb_unlink(skb, &oldsk->sk_receive_queue);
8454 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8455 sctp_skb_set_owner_r_frag(skb, newsk);
8459 /* Clean up any messages pending delivery due to partial
8460 * delivery. Three cases:
8461 * 1) No partial deliver; no work.
8462 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8463 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8465 skb_queue_head_init(&newsp->pd_lobby);
8466 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8468 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8469 struct sk_buff_head *queue;
8471 /* Decide which queue to move pd_lobby skbs to. */
8472 if (assoc->ulpq.pd_mode) {
8473 queue = &newsp->pd_lobby;
8474 } else
8475 queue = &newsk->sk_receive_queue;
8477 /* Walk through the pd_lobby, looking for skbs that
8478 * need moved to the new socket.
8480 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8481 event = sctp_skb2event(skb);
8482 if (event->asoc == assoc) {
8483 __skb_unlink(skb, &oldsp->pd_lobby);
8484 __skb_queue_tail(queue, skb);
8485 sctp_skb_set_owner_r_frag(skb, newsk);
8489 /* Clear up any skbs waiting for the partial
8490 * delivery to finish.
8492 if (assoc->ulpq.pd_mode)
8493 sctp_clear_pd(oldsk, NULL);
8497 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8499 /* Set the type of socket to indicate that it is peeled off from the
8500 * original UDP-style socket or created with the accept() call on a
8501 * TCP-style socket..
8503 newsp->type = type;
8505 /* Mark the new socket "in-use" by the user so that any packets
8506 * that may arrive on the association after we've moved it are
8507 * queued to the backlog. This prevents a potential race between
8508 * backlog processing on the old socket and new-packet processing
8509 * on the new socket.
8511 * The caller has just allocated newsk so we can guarantee that other
8512 * paths won't try to lock it and then oldsk.
8514 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8515 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8516 sctp_assoc_migrate(assoc, newsk);
8517 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8519 /* If the association on the newsk is already closed before accept()
8520 * is called, set RCV_SHUTDOWN flag.
8522 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8523 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8524 newsk->sk_shutdown |= RCV_SHUTDOWN;
8525 } else {
8526 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8529 release_sock(newsk);
8533 /* This proto struct describes the ULP interface for SCTP. */
8534 struct proto sctp_prot = {
8535 .name = "SCTP",
8536 .owner = THIS_MODULE,
8537 .close = sctp_close,
8538 .connect = sctp_connect,
8539 .disconnect = sctp_disconnect,
8540 .accept = sctp_accept,
8541 .ioctl = sctp_ioctl,
8542 .init = sctp_init_sock,
8543 .destroy = sctp_destroy_sock,
8544 .shutdown = sctp_shutdown,
8545 .setsockopt = sctp_setsockopt,
8546 .getsockopt = sctp_getsockopt,
8547 .sendmsg = sctp_sendmsg,
8548 .recvmsg = sctp_recvmsg,
8549 .bind = sctp_bind,
8550 .backlog_rcv = sctp_backlog_rcv,
8551 .hash = sctp_hash,
8552 .unhash = sctp_unhash,
8553 .get_port = sctp_get_port,
8554 .obj_size = sizeof(struct sctp_sock),
8555 .useroffset = offsetof(struct sctp_sock, subscribe),
8556 .usersize = offsetof(struct sctp_sock, initmsg) -
8557 offsetof(struct sctp_sock, subscribe) +
8558 sizeof_field(struct sctp_sock, initmsg),
8559 .sysctl_mem = sysctl_sctp_mem,
8560 .sysctl_rmem = sysctl_sctp_rmem,
8561 .sysctl_wmem = sysctl_sctp_wmem,
8562 .memory_pressure = &sctp_memory_pressure,
8563 .enter_memory_pressure = sctp_enter_memory_pressure,
8564 .memory_allocated = &sctp_memory_allocated,
8565 .sockets_allocated = &sctp_sockets_allocated,
8568 #if IS_ENABLED(CONFIG_IPV6)
8570 #include <net/transp_v6.h>
8571 static void sctp_v6_destroy_sock(struct sock *sk)
8573 sctp_destroy_sock(sk);
8574 inet6_destroy_sock(sk);
8577 struct proto sctpv6_prot = {
8578 .name = "SCTPv6",
8579 .owner = THIS_MODULE,
8580 .close = sctp_close,
8581 .connect = sctp_connect,
8582 .disconnect = sctp_disconnect,
8583 .accept = sctp_accept,
8584 .ioctl = sctp_ioctl,
8585 .init = sctp_init_sock,
8586 .destroy = sctp_v6_destroy_sock,
8587 .shutdown = sctp_shutdown,
8588 .setsockopt = sctp_setsockopt,
8589 .getsockopt = sctp_getsockopt,
8590 .sendmsg = sctp_sendmsg,
8591 .recvmsg = sctp_recvmsg,
8592 .bind = sctp_bind,
8593 .backlog_rcv = sctp_backlog_rcv,
8594 .hash = sctp_hash,
8595 .unhash = sctp_unhash,
8596 .get_port = sctp_get_port,
8597 .obj_size = sizeof(struct sctp6_sock),
8598 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
8599 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
8600 offsetof(struct sctp6_sock, sctp.subscribe) +
8601 sizeof_field(struct sctp6_sock, sctp.initmsg),
8602 .sysctl_mem = sysctl_sctp_mem,
8603 .sysctl_rmem = sysctl_sctp_rmem,
8604 .sysctl_wmem = sysctl_sctp_wmem,
8605 .memory_pressure = &sctp_memory_pressure,
8606 .enter_memory_pressure = sctp_enter_memory_pressure,
8607 .memory_allocated = &sctp_memory_allocated,
8608 .sockets_allocated = &sctp_sockets_allocated,
8610 #endif /* IS_ENABLED(CONFIG_IPV6) */