cifs: update ctime and mtime during truncate
[linux/fpc-iii.git] / net / sctp / socket.c
blob95f39dde1e08e0b4f56a0b267f858b82aee396f9
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/ip.h>
61 #include <linux/capability.h>
62 #include <linux/fcntl.h>
63 #include <linux/poll.h>
64 #include <linux/init.h>
65 #include <linux/slab.h>
66 #include <linux/file.h>
67 #include <linux/compat.h>
69 #include <net/ip.h>
70 #include <net/icmp.h>
71 #include <net/route.h>
72 #include <net/ipv6.h>
73 #include <net/inet_common.h>
74 #include <net/busy_poll.h>
76 #include <linux/socket.h> /* for sa_family_t */
77 #include <linux/export.h>
78 #include <net/sock.h>
79 #include <net/sctp/sctp.h>
80 #include <net/sctp/sm.h>
82 /* Forward declarations for internal helper functions. */
83 static int sctp_writeable(struct sock *sk);
84 static void sctp_wfree(struct sk_buff *skb);
85 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
86 size_t msg_len);
87 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
88 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
89 static int sctp_wait_for_accept(struct sock *sk, long timeo);
90 static void sctp_wait_for_close(struct sock *sk, long timeo);
91 static void sctp_destruct_sock(struct sock *sk);
92 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
93 union sctp_addr *addr, int len);
94 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
95 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
96 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf(struct sctp_association *asoc,
99 struct sctp_chunk *chunk);
100 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
101 static int sctp_autobind(struct sock *sk);
102 static void sctp_sock_migrate(struct sock *, struct sock *,
103 struct sctp_association *, sctp_socket_type_t);
105 static int sctp_memory_pressure;
106 static atomic_long_t sctp_memory_allocated;
107 struct percpu_counter sctp_sockets_allocated;
109 static void sctp_enter_memory_pressure(struct sock *sk)
111 sctp_memory_pressure = 1;
115 /* Get the sndbuf space available at the time on the association. */
116 static inline int sctp_wspace(struct sctp_association *asoc)
118 int amt;
120 if (asoc->ep->sndbuf_policy)
121 amt = asoc->sndbuf_used;
122 else
123 amt = sk_wmem_alloc_get(asoc->base.sk);
125 if (amt >= asoc->base.sk->sk_sndbuf) {
126 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
127 amt = 0;
128 else {
129 amt = sk_stream_wspace(asoc->base.sk);
130 if (amt < 0)
131 amt = 0;
133 } else {
134 amt = asoc->base.sk->sk_sndbuf - amt;
136 return amt;
139 /* Increment the used sndbuf space count of the corresponding association by
140 * the size of the outgoing data chunk.
141 * Also, set the skb destructor for sndbuf accounting later.
143 * Since it is always 1-1 between chunk and skb, and also a new skb is always
144 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
145 * destructor in the data chunk skb for the purpose of the sndbuf space
146 * tracking.
148 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
150 struct sctp_association *asoc = chunk->asoc;
151 struct sock *sk = asoc->base.sk;
153 /* The sndbuf space is tracked per association. */
154 sctp_association_hold(asoc);
156 skb_set_owner_w(chunk->skb, sk);
158 chunk->skb->destructor = sctp_wfree;
159 /* Save the chunk pointer in skb for sctp_wfree to use later. */
160 skb_shinfo(chunk->skb)->destructor_arg = chunk;
162 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
163 sizeof(struct sk_buff) +
164 sizeof(struct sctp_chunk);
166 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
167 sk->sk_wmem_queued += chunk->skb->truesize;
168 sk_mem_charge(sk, chunk->skb->truesize);
171 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
173 skb_orphan(chunk->skb);
176 #define traverse_and_process() \
177 do { \
178 msg = chunk->msg; \
179 if (msg == prev_msg) \
180 continue; \
181 list_for_each_entry(c, &msg->chunks, frag_list) { \
182 if ((clear && asoc->base.sk == c->skb->sk) || \
183 (!clear && asoc->base.sk != c->skb->sk)) \
184 cb(c); \
186 prev_msg = msg; \
187 } while (0)
189 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
190 bool clear,
191 void (*cb)(struct sctp_chunk *))
194 struct sctp_datamsg *msg, *prev_msg = NULL;
195 struct sctp_outq *q = &asoc->outqueue;
196 struct sctp_chunk *chunk, *c;
197 struct sctp_transport *t;
199 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
200 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
201 traverse_and_process();
203 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
204 traverse_and_process();
206 list_for_each_entry(chunk, &q->sacked, transmitted_list)
207 traverse_and_process();
209 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
210 traverse_and_process();
212 list_for_each_entry(chunk, &q->out_chunk_list, list)
213 traverse_and_process();
216 /* Verify that this is a valid address. */
217 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
218 int len)
220 struct sctp_af *af;
222 /* Verify basic sockaddr. */
223 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
224 if (!af)
225 return -EINVAL;
227 /* Is this a valid SCTP address? */
228 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
229 return -EINVAL;
231 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
232 return -EINVAL;
234 return 0;
237 /* Look up the association by its id. If this is not a UDP-style
238 * socket, the ID field is always ignored.
240 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
242 struct sctp_association *asoc = NULL;
244 /* If this is not a UDP-style socket, assoc id should be ignored. */
245 if (!sctp_style(sk, UDP)) {
246 /* Return NULL if the socket state is not ESTABLISHED. It
247 * could be a TCP-style listening socket or a socket which
248 * hasn't yet called connect() to establish an association.
250 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
251 return NULL;
253 /* Get the first and the only association from the list. */
254 if (!list_empty(&sctp_sk(sk)->ep->asocs))
255 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
256 struct sctp_association, asocs);
257 return asoc;
260 /* Otherwise this is a UDP-style socket. */
261 if (!id || (id == (sctp_assoc_t)-1))
262 return NULL;
264 spin_lock_bh(&sctp_assocs_id_lock);
265 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
266 if (asoc && (asoc->base.sk != sk || asoc->base.dead))
267 asoc = NULL;
268 spin_unlock_bh(&sctp_assocs_id_lock);
270 return asoc;
273 /* Look up the transport from an address and an assoc id. If both address and
274 * id are specified, the associations matching the address and the id should be
275 * the same.
277 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
278 struct sockaddr_storage *addr,
279 sctp_assoc_t id)
281 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
282 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
283 union sctp_addr *laddr = (union sctp_addr *)addr;
284 struct sctp_transport *transport;
286 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
287 return NULL;
289 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
290 laddr,
291 &transport);
293 if (!addr_asoc)
294 return NULL;
296 id_asoc = sctp_id2assoc(sk, id);
297 if (id_asoc && (id_asoc != addr_asoc))
298 return NULL;
300 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
301 (union sctp_addr *)addr);
303 return transport;
306 /* API 3.1.2 bind() - UDP Style Syntax
307 * The syntax of bind() is,
309 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
311 * sd - the socket descriptor returned by socket().
312 * addr - the address structure (struct sockaddr_in or struct
313 * sockaddr_in6 [RFC 2553]),
314 * addr_len - the size of the address structure.
316 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
318 int retval = 0;
320 lock_sock(sk);
322 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
323 addr, addr_len);
325 /* Disallow binding twice. */
326 if (!sctp_sk(sk)->ep->base.bind_addr.port)
327 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
328 addr_len);
329 else
330 retval = -EINVAL;
332 release_sock(sk);
334 return retval;
337 static long sctp_get_port_local(struct sock *, union sctp_addr *);
339 /* Verify this is a valid sockaddr. */
340 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
341 union sctp_addr *addr, int len)
343 struct sctp_af *af;
345 /* Check minimum size. */
346 if (len < sizeof (struct sockaddr))
347 return NULL;
349 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
350 return NULL;
352 if (addr->sa.sa_family == AF_INET6) {
353 if (len < SIN6_LEN_RFC2133)
354 return NULL;
355 /* V4 mapped address are really of AF_INET family */
356 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
357 !opt->pf->af_supported(AF_INET, opt))
358 return NULL;
361 /* If we get this far, af is valid. */
362 af = sctp_get_af_specific(addr->sa.sa_family);
364 if (len < af->sockaddr_len)
365 return NULL;
367 return af;
370 /* Bind a local address either to an endpoint or to an association. */
371 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
373 struct net *net = sock_net(sk);
374 struct sctp_sock *sp = sctp_sk(sk);
375 struct sctp_endpoint *ep = sp->ep;
376 struct sctp_bind_addr *bp = &ep->base.bind_addr;
377 struct sctp_af *af;
378 unsigned short snum;
379 int ret = 0;
381 /* Common sockaddr verification. */
382 af = sctp_sockaddr_af(sp, addr, len);
383 if (!af) {
384 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
385 __func__, sk, addr, len);
386 return -EINVAL;
389 snum = ntohs(addr->v4.sin_port);
391 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
392 __func__, sk, &addr->sa, bp->port, snum, len);
394 /* PF specific bind() address verification. */
395 if (!sp->pf->bind_verify(sp, addr))
396 return -EADDRNOTAVAIL;
398 /* We must either be unbound, or bind to the same port.
399 * It's OK to allow 0 ports if we are already bound.
400 * We'll just inhert an already bound port in this case
402 if (bp->port) {
403 if (!snum)
404 snum = bp->port;
405 else if (snum != bp->port) {
406 pr_debug("%s: new port %d doesn't match existing port "
407 "%d\n", __func__, snum, bp->port);
408 return -EINVAL;
412 if (snum && snum < PROT_SOCK &&
413 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
414 return -EACCES;
416 /* See if the address matches any of the addresses we may have
417 * already bound before checking against other endpoints.
419 if (sctp_bind_addr_match(bp, addr, sp))
420 return -EINVAL;
422 /* Make sure we are allowed to bind here.
423 * The function sctp_get_port_local() does duplicate address
424 * detection.
426 addr->v4.sin_port = htons(snum);
427 if ((ret = sctp_get_port_local(sk, addr))) {
428 return -EADDRINUSE;
431 /* Refresh ephemeral port. */
432 if (!bp->port)
433 bp->port = inet_sk(sk)->inet_num;
435 /* Add the address to the bind address list.
436 * Use GFP_ATOMIC since BHs will be disabled.
438 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
439 SCTP_ADDR_SRC, GFP_ATOMIC);
441 /* Copy back into socket for getsockname() use. */
442 if (!ret) {
443 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
444 sp->pf->to_sk_saddr(addr, sk);
447 return ret;
450 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
452 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
453 * at any one time. If a sender, after sending an ASCONF chunk, decides
454 * it needs to transfer another ASCONF Chunk, it MUST wait until the
455 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
456 * subsequent ASCONF. Note this restriction binds each side, so at any
457 * time two ASCONF may be in-transit on any given association (one sent
458 * from each endpoint).
460 static int sctp_send_asconf(struct sctp_association *asoc,
461 struct sctp_chunk *chunk)
463 struct net *net = sock_net(asoc->base.sk);
464 int retval = 0;
466 /* If there is an outstanding ASCONF chunk, queue it for later
467 * transmission.
469 if (asoc->addip_last_asconf) {
470 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
471 goto out;
474 /* Hold the chunk until an ASCONF_ACK is received. */
475 sctp_chunk_hold(chunk);
476 retval = sctp_primitive_ASCONF(net, asoc, chunk);
477 if (retval)
478 sctp_chunk_free(chunk);
479 else
480 asoc->addip_last_asconf = chunk;
482 out:
483 return retval;
486 /* Add a list of addresses as bind addresses to local endpoint or
487 * association.
489 * Basically run through each address specified in the addrs/addrcnt
490 * array/length pair, determine if it is IPv6 or IPv4 and call
491 * sctp_do_bind() on it.
493 * If any of them fails, then the operation will be reversed and the
494 * ones that were added will be removed.
496 * Only sctp_setsockopt_bindx() is supposed to call this function.
498 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
500 int cnt;
501 int retval = 0;
502 void *addr_buf;
503 struct sockaddr *sa_addr;
504 struct sctp_af *af;
506 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
507 addrs, addrcnt);
509 addr_buf = addrs;
510 for (cnt = 0; cnt < addrcnt; cnt++) {
511 /* The list may contain either IPv4 or IPv6 address;
512 * determine the address length for walking thru the list.
514 sa_addr = addr_buf;
515 af = sctp_get_af_specific(sa_addr->sa_family);
516 if (!af) {
517 retval = -EINVAL;
518 goto err_bindx_add;
521 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
522 af->sockaddr_len);
524 addr_buf += af->sockaddr_len;
526 err_bindx_add:
527 if (retval < 0) {
528 /* Failed. Cleanup the ones that have been added */
529 if (cnt > 0)
530 sctp_bindx_rem(sk, addrs, cnt);
531 return retval;
535 return retval;
538 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
539 * associations that are part of the endpoint indicating that a list of local
540 * addresses are added to the endpoint.
542 * If any of the addresses is already in the bind address list of the
543 * association, we do not send the chunk for that association. But it will not
544 * affect other associations.
546 * Only sctp_setsockopt_bindx() is supposed to call this function.
548 static int sctp_send_asconf_add_ip(struct sock *sk,
549 struct sockaddr *addrs,
550 int addrcnt)
552 struct net *net = sock_net(sk);
553 struct sctp_sock *sp;
554 struct sctp_endpoint *ep;
555 struct sctp_association *asoc;
556 struct sctp_bind_addr *bp;
557 struct sctp_chunk *chunk;
558 struct sctp_sockaddr_entry *laddr;
559 union sctp_addr *addr;
560 union sctp_addr saveaddr;
561 void *addr_buf;
562 struct sctp_af *af;
563 struct list_head *p;
564 int i;
565 int retval = 0;
567 if (!net->sctp.addip_enable)
568 return retval;
570 sp = sctp_sk(sk);
571 ep = sp->ep;
573 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
574 __func__, sk, addrs, addrcnt);
576 list_for_each_entry(asoc, &ep->asocs, asocs) {
577 if (!asoc->peer.asconf_capable)
578 continue;
580 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
581 continue;
583 if (!sctp_state(asoc, ESTABLISHED))
584 continue;
586 /* Check if any address in the packed array of addresses is
587 * in the bind address list of the association. If so,
588 * do not send the asconf chunk to its peer, but continue with
589 * other associations.
591 addr_buf = addrs;
592 for (i = 0; i < addrcnt; i++) {
593 addr = addr_buf;
594 af = sctp_get_af_specific(addr->v4.sin_family);
595 if (!af) {
596 retval = -EINVAL;
597 goto out;
600 if (sctp_assoc_lookup_laddr(asoc, addr))
601 break;
603 addr_buf += af->sockaddr_len;
605 if (i < addrcnt)
606 continue;
608 /* Use the first valid address in bind addr list of
609 * association as Address Parameter of ASCONF CHUNK.
611 bp = &asoc->base.bind_addr;
612 p = bp->address_list.next;
613 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
614 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
615 addrcnt, SCTP_PARAM_ADD_IP);
616 if (!chunk) {
617 retval = -ENOMEM;
618 goto out;
621 /* Add the new addresses to the bind address list with
622 * use_as_src set to 0.
624 addr_buf = addrs;
625 for (i = 0; i < addrcnt; i++) {
626 addr = addr_buf;
627 af = sctp_get_af_specific(addr->v4.sin_family);
628 memcpy(&saveaddr, addr, af->sockaddr_len);
629 retval = sctp_add_bind_addr(bp, &saveaddr,
630 sizeof(saveaddr),
631 SCTP_ADDR_NEW, GFP_ATOMIC);
632 addr_buf += af->sockaddr_len;
634 if (asoc->src_out_of_asoc_ok) {
635 struct sctp_transport *trans;
637 list_for_each_entry(trans,
638 &asoc->peer.transport_addr_list, transports) {
639 /* Clear the source and route cache */
640 dst_release(trans->dst);
641 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
642 2*asoc->pathmtu, 4380));
643 trans->ssthresh = asoc->peer.i.a_rwnd;
644 trans->rto = asoc->rto_initial;
645 sctp_max_rto(asoc, trans);
646 trans->rtt = trans->srtt = trans->rttvar = 0;
647 sctp_transport_route(trans, NULL,
648 sctp_sk(asoc->base.sk));
651 retval = sctp_send_asconf(asoc, chunk);
654 out:
655 return retval;
658 /* Remove a list of addresses from bind addresses list. Do not remove the
659 * last address.
661 * Basically run through each address specified in the addrs/addrcnt
662 * array/length pair, determine if it is IPv6 or IPv4 and call
663 * sctp_del_bind() on it.
665 * If any of them fails, then the operation will be reversed and the
666 * ones that were removed will be added back.
668 * At least one address has to be left; if only one address is
669 * available, the operation will return -EBUSY.
671 * Only sctp_setsockopt_bindx() is supposed to call this function.
673 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
675 struct sctp_sock *sp = sctp_sk(sk);
676 struct sctp_endpoint *ep = sp->ep;
677 int cnt;
678 struct sctp_bind_addr *bp = &ep->base.bind_addr;
679 int retval = 0;
680 void *addr_buf;
681 union sctp_addr *sa_addr;
682 struct sctp_af *af;
684 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
685 __func__, sk, addrs, addrcnt);
687 addr_buf = addrs;
688 for (cnt = 0; cnt < addrcnt; cnt++) {
689 /* If the bind address list is empty or if there is only one
690 * bind address, there is nothing more to be removed (we need
691 * at least one address here).
693 if (list_empty(&bp->address_list) ||
694 (sctp_list_single_entry(&bp->address_list))) {
695 retval = -EBUSY;
696 goto err_bindx_rem;
699 sa_addr = addr_buf;
700 af = sctp_get_af_specific(sa_addr->sa.sa_family);
701 if (!af) {
702 retval = -EINVAL;
703 goto err_bindx_rem;
706 if (!af->addr_valid(sa_addr, sp, NULL)) {
707 retval = -EADDRNOTAVAIL;
708 goto err_bindx_rem;
711 if (sa_addr->v4.sin_port &&
712 sa_addr->v4.sin_port != htons(bp->port)) {
713 retval = -EINVAL;
714 goto err_bindx_rem;
717 if (!sa_addr->v4.sin_port)
718 sa_addr->v4.sin_port = htons(bp->port);
720 /* FIXME - There is probably a need to check if sk->sk_saddr and
721 * sk->sk_rcv_addr are currently set to one of the addresses to
722 * be removed. This is something which needs to be looked into
723 * when we are fixing the outstanding issues with multi-homing
724 * socket routing and failover schemes. Refer to comments in
725 * sctp_do_bind(). -daisy
727 retval = sctp_del_bind_addr(bp, sa_addr);
729 addr_buf += af->sockaddr_len;
730 err_bindx_rem:
731 if (retval < 0) {
732 /* Failed. Add the ones that has been removed back */
733 if (cnt > 0)
734 sctp_bindx_add(sk, addrs, cnt);
735 return retval;
739 return retval;
742 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
743 * the associations that are part of the endpoint indicating that a list of
744 * local addresses are removed from the endpoint.
746 * If any of the addresses is already in the bind address list of the
747 * association, we do not send the chunk for that association. But it will not
748 * affect other associations.
750 * Only sctp_setsockopt_bindx() is supposed to call this function.
752 static int sctp_send_asconf_del_ip(struct sock *sk,
753 struct sockaddr *addrs,
754 int addrcnt)
756 struct net *net = sock_net(sk);
757 struct sctp_sock *sp;
758 struct sctp_endpoint *ep;
759 struct sctp_association *asoc;
760 struct sctp_transport *transport;
761 struct sctp_bind_addr *bp;
762 struct sctp_chunk *chunk;
763 union sctp_addr *laddr;
764 void *addr_buf;
765 struct sctp_af *af;
766 struct sctp_sockaddr_entry *saddr;
767 int i;
768 int retval = 0;
769 int stored = 0;
771 chunk = NULL;
772 if (!net->sctp.addip_enable)
773 return retval;
775 sp = sctp_sk(sk);
776 ep = sp->ep;
778 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
779 __func__, sk, addrs, addrcnt);
781 list_for_each_entry(asoc, &ep->asocs, asocs) {
783 if (!asoc->peer.asconf_capable)
784 continue;
786 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
787 continue;
789 if (!sctp_state(asoc, ESTABLISHED))
790 continue;
792 /* Check if any address in the packed array of addresses is
793 * not present in the bind address list of the association.
794 * If so, do not send the asconf chunk to its peer, but
795 * continue with other associations.
797 addr_buf = addrs;
798 for (i = 0; i < addrcnt; i++) {
799 laddr = addr_buf;
800 af = sctp_get_af_specific(laddr->v4.sin_family);
801 if (!af) {
802 retval = -EINVAL;
803 goto out;
806 if (!sctp_assoc_lookup_laddr(asoc, laddr))
807 break;
809 addr_buf += af->sockaddr_len;
811 if (i < addrcnt)
812 continue;
814 /* Find one address in the association's bind address list
815 * that is not in the packed array of addresses. This is to
816 * make sure that we do not delete all the addresses in the
817 * association.
819 bp = &asoc->base.bind_addr;
820 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
821 addrcnt, sp);
822 if ((laddr == NULL) && (addrcnt == 1)) {
823 if (asoc->asconf_addr_del_pending)
824 continue;
825 asoc->asconf_addr_del_pending =
826 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
827 if (asoc->asconf_addr_del_pending == NULL) {
828 retval = -ENOMEM;
829 goto out;
831 asoc->asconf_addr_del_pending->sa.sa_family =
832 addrs->sa_family;
833 asoc->asconf_addr_del_pending->v4.sin_port =
834 htons(bp->port);
835 if (addrs->sa_family == AF_INET) {
836 struct sockaddr_in *sin;
838 sin = (struct sockaddr_in *)addrs;
839 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
840 } else if (addrs->sa_family == AF_INET6) {
841 struct sockaddr_in6 *sin6;
843 sin6 = (struct sockaddr_in6 *)addrs;
844 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
847 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
848 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
849 asoc->asconf_addr_del_pending);
851 asoc->src_out_of_asoc_ok = 1;
852 stored = 1;
853 goto skip_mkasconf;
856 if (laddr == NULL)
857 return -EINVAL;
859 /* We do not need RCU protection throughout this loop
860 * because this is done under a socket lock from the
861 * setsockopt call.
863 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
864 SCTP_PARAM_DEL_IP);
865 if (!chunk) {
866 retval = -ENOMEM;
867 goto out;
870 skip_mkasconf:
871 /* Reset use_as_src flag for the addresses in the bind address
872 * list that are to be deleted.
874 addr_buf = addrs;
875 for (i = 0; i < addrcnt; i++) {
876 laddr = addr_buf;
877 af = sctp_get_af_specific(laddr->v4.sin_family);
878 list_for_each_entry(saddr, &bp->address_list, list) {
879 if (sctp_cmp_addr_exact(&saddr->a, laddr))
880 saddr->state = SCTP_ADDR_DEL;
882 addr_buf += af->sockaddr_len;
885 /* Update the route and saddr entries for all the transports
886 * as some of the addresses in the bind address list are
887 * about to be deleted and cannot be used as source addresses.
889 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
890 transports) {
891 dst_release(transport->dst);
892 sctp_transport_route(transport, NULL,
893 sctp_sk(asoc->base.sk));
896 if (stored)
897 /* We don't need to transmit ASCONF */
898 continue;
899 retval = sctp_send_asconf(asoc, chunk);
901 out:
902 return retval;
905 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
906 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
908 struct sock *sk = sctp_opt2sk(sp);
909 union sctp_addr *addr;
910 struct sctp_af *af;
912 /* It is safe to write port space in caller. */
913 addr = &addrw->a;
914 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
915 af = sctp_get_af_specific(addr->sa.sa_family);
916 if (!af)
917 return -EINVAL;
918 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
919 return -EINVAL;
921 if (addrw->state == SCTP_ADDR_NEW)
922 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
923 else
924 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
927 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
929 * API 8.1
930 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
931 * int flags);
933 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
934 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
935 * or IPv6 addresses.
937 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
938 * Section 3.1.2 for this usage.
940 * addrs is a pointer to an array of one or more socket addresses. Each
941 * address is contained in its appropriate structure (i.e. struct
942 * sockaddr_in or struct sockaddr_in6) the family of the address type
943 * must be used to distinguish the address length (note that this
944 * representation is termed a "packed array" of addresses). The caller
945 * specifies the number of addresses in the array with addrcnt.
947 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
948 * -1, and sets errno to the appropriate error code.
950 * For SCTP, the port given in each socket address must be the same, or
951 * sctp_bindx() will fail, setting errno to EINVAL.
953 * The flags parameter is formed from the bitwise OR of zero or more of
954 * the following currently defined flags:
956 * SCTP_BINDX_ADD_ADDR
958 * SCTP_BINDX_REM_ADDR
960 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
961 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
962 * addresses from the association. The two flags are mutually exclusive;
963 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
964 * not remove all addresses from an association; sctp_bindx() will
965 * reject such an attempt with EINVAL.
967 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
968 * additional addresses with an endpoint after calling bind(). Or use
969 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
970 * socket is associated with so that no new association accepted will be
971 * associated with those addresses. If the endpoint supports dynamic
972 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
973 * endpoint to send the appropriate message to the peer to change the
974 * peers address lists.
976 * Adding and removing addresses from a connected association is
977 * optional functionality. Implementations that do not support this
978 * functionality should return EOPNOTSUPP.
980 * Basically do nothing but copying the addresses from user to kernel
981 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
982 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
983 * from userspace.
985 * We don't use copy_from_user() for optimization: we first do the
986 * sanity checks (buffer size -fast- and access check-healthy
987 * pointer); if all of those succeed, then we can alloc the memory
988 * (expensive operation) needed to copy the data to kernel. Then we do
989 * the copying without checking the user space area
990 * (__copy_from_user()).
992 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
993 * it.
995 * sk The sk of the socket
996 * addrs The pointer to the addresses in user land
997 * addrssize Size of the addrs buffer
998 * op Operation to perform (add or remove, see the flags of
999 * sctp_bindx)
1001 * Returns 0 if ok, <0 errno code on error.
1003 static int sctp_setsockopt_bindx(struct sock *sk,
1004 struct sockaddr __user *addrs,
1005 int addrs_size, int op)
1007 struct sockaddr *kaddrs;
1008 int err;
1009 int addrcnt = 0;
1010 int walk_size = 0;
1011 struct sockaddr *sa_addr;
1012 void *addr_buf;
1013 struct sctp_af *af;
1015 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1016 __func__, sk, addrs, addrs_size, op);
1018 if (unlikely(addrs_size <= 0))
1019 return -EINVAL;
1021 /* Check the user passed a healthy pointer. */
1022 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1023 return -EFAULT;
1025 /* Alloc space for the address array in kernel memory. */
1026 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
1027 if (unlikely(!kaddrs))
1028 return -ENOMEM;
1030 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1031 kfree(kaddrs);
1032 return -EFAULT;
1035 /* Walk through the addrs buffer and count the number of addresses. */
1036 addr_buf = kaddrs;
1037 while (walk_size < addrs_size) {
1038 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1039 kfree(kaddrs);
1040 return -EINVAL;
1043 sa_addr = addr_buf;
1044 af = sctp_get_af_specific(sa_addr->sa_family);
1046 /* If the address family is not supported or if this address
1047 * causes the address buffer to overflow return EINVAL.
1049 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1050 kfree(kaddrs);
1051 return -EINVAL;
1053 addrcnt++;
1054 addr_buf += af->sockaddr_len;
1055 walk_size += af->sockaddr_len;
1058 /* Do the work. */
1059 switch (op) {
1060 case SCTP_BINDX_ADD_ADDR:
1061 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1062 if (err)
1063 goto out;
1064 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1065 break;
1067 case SCTP_BINDX_REM_ADDR:
1068 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1069 if (err)
1070 goto out;
1071 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1072 break;
1074 default:
1075 err = -EINVAL;
1076 break;
1079 out:
1080 kfree(kaddrs);
1082 return err;
1085 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1087 * Common routine for handling connect() and sctp_connectx().
1088 * Connect will come in with just a single address.
1090 static int __sctp_connect(struct sock *sk,
1091 struct sockaddr *kaddrs,
1092 int addrs_size, int flags,
1093 sctp_assoc_t *assoc_id)
1095 struct net *net = sock_net(sk);
1096 struct sctp_sock *sp;
1097 struct sctp_endpoint *ep;
1098 struct sctp_association *asoc = NULL;
1099 struct sctp_association *asoc2;
1100 struct sctp_transport *transport;
1101 union sctp_addr to;
1102 sctp_scope_t scope;
1103 long timeo;
1104 int err = 0;
1105 int addrcnt = 0;
1106 int walk_size = 0;
1107 union sctp_addr *sa_addr = NULL;
1108 void *addr_buf;
1109 unsigned short port;
1111 sp = sctp_sk(sk);
1112 ep = sp->ep;
1114 /* connect() cannot be done on a socket that is already in ESTABLISHED
1115 * state - UDP-style peeled off socket or a TCP-style socket that
1116 * is already connected.
1117 * It cannot be done even on a TCP-style listening socket.
1119 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1120 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1121 err = -EISCONN;
1122 goto out_free;
1125 /* Walk through the addrs buffer and count the number of addresses. */
1126 addr_buf = kaddrs;
1127 while (walk_size < addrs_size) {
1128 struct sctp_af *af;
1130 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1131 err = -EINVAL;
1132 goto out_free;
1135 sa_addr = addr_buf;
1136 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1138 /* If the address family is not supported or if this address
1139 * causes the address buffer to overflow return EINVAL.
1141 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1142 err = -EINVAL;
1143 goto out_free;
1146 port = ntohs(sa_addr->v4.sin_port);
1148 /* Save current address so we can work with it */
1149 memcpy(&to, sa_addr, af->sockaddr_len);
1151 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1152 if (err)
1153 goto out_free;
1155 /* Make sure the destination port is correctly set
1156 * in all addresses.
1158 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1159 err = -EINVAL;
1160 goto out_free;
1163 /* Check if there already is a matching association on the
1164 * endpoint (other than the one created here).
1166 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1167 if (asoc2 && asoc2 != asoc) {
1168 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1169 err = -EISCONN;
1170 else
1171 err = -EALREADY;
1172 goto out_free;
1175 /* If we could not find a matching association on the endpoint,
1176 * make sure that there is no peeled-off association matching
1177 * the peer address even on another socket.
1179 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1180 err = -EADDRNOTAVAIL;
1181 goto out_free;
1184 if (!asoc) {
1185 /* If a bind() or sctp_bindx() is not called prior to
1186 * an sctp_connectx() call, the system picks an
1187 * ephemeral port and will choose an address set
1188 * equivalent to binding with a wildcard address.
1190 if (!ep->base.bind_addr.port) {
1191 if (sctp_autobind(sk)) {
1192 err = -EAGAIN;
1193 goto out_free;
1195 } else {
1197 * If an unprivileged user inherits a 1-many
1198 * style socket with open associations on a
1199 * privileged port, it MAY be permitted to
1200 * accept new associations, but it SHOULD NOT
1201 * be permitted to open new associations.
1203 if (ep->base.bind_addr.port < PROT_SOCK &&
1204 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1205 err = -EACCES;
1206 goto out_free;
1210 scope = sctp_scope(&to);
1211 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1212 if (!asoc) {
1213 err = -ENOMEM;
1214 goto out_free;
1217 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1218 GFP_KERNEL);
1219 if (err < 0) {
1220 goto out_free;
1225 /* Prime the peer's transport structures. */
1226 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1227 SCTP_UNKNOWN);
1228 if (!transport) {
1229 err = -ENOMEM;
1230 goto out_free;
1233 addrcnt++;
1234 addr_buf += af->sockaddr_len;
1235 walk_size += af->sockaddr_len;
1238 /* In case the user of sctp_connectx() wants an association
1239 * id back, assign one now.
1241 if (assoc_id) {
1242 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1243 if (err < 0)
1244 goto out_free;
1247 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1248 if (err < 0) {
1249 goto out_free;
1252 /* Initialize sk's dport and daddr for getpeername() */
1253 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1254 sp->pf->to_sk_daddr(sa_addr, sk);
1255 sk->sk_err = 0;
1257 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1259 if (assoc_id)
1260 *assoc_id = asoc->assoc_id;
1261 err = sctp_wait_for_connect(asoc, &timeo);
1262 /* Note: the asoc may be freed after the return of
1263 * sctp_wait_for_connect.
1266 /* Don't free association on exit. */
1267 asoc = NULL;
1269 out_free:
1270 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1271 __func__, asoc, kaddrs, err);
1273 if (asoc) {
1274 /* sctp_primitive_ASSOCIATE may have added this association
1275 * To the hash table, try to unhash it, just in case, its a noop
1276 * if it wasn't hashed so we're safe
1278 sctp_association_free(asoc);
1280 return err;
1283 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1285 * API 8.9
1286 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1287 * sctp_assoc_t *asoc);
1289 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1290 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1291 * or IPv6 addresses.
1293 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1294 * Section 3.1.2 for this usage.
1296 * addrs is a pointer to an array of one or more socket addresses. Each
1297 * address is contained in its appropriate structure (i.e. struct
1298 * sockaddr_in or struct sockaddr_in6) the family of the address type
1299 * must be used to distengish the address length (note that this
1300 * representation is termed a "packed array" of addresses). The caller
1301 * specifies the number of addresses in the array with addrcnt.
1303 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1304 * the association id of the new association. On failure, sctp_connectx()
1305 * returns -1, and sets errno to the appropriate error code. The assoc_id
1306 * is not touched by the kernel.
1308 * For SCTP, the port given in each socket address must be the same, or
1309 * sctp_connectx() will fail, setting errno to EINVAL.
1311 * An application can use sctp_connectx to initiate an association with
1312 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1313 * allows a caller to specify multiple addresses at which a peer can be
1314 * reached. The way the SCTP stack uses the list of addresses to set up
1315 * the association is implementation dependent. This function only
1316 * specifies that the stack will try to make use of all the addresses in
1317 * the list when needed.
1319 * Note that the list of addresses passed in is only used for setting up
1320 * the association. It does not necessarily equal the set of addresses
1321 * the peer uses for the resulting association. If the caller wants to
1322 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1323 * retrieve them after the association has been set up.
1325 * Basically do nothing but copying the addresses from user to kernel
1326 * land and invoking either sctp_connectx(). This is used for tunneling
1327 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1329 * We don't use copy_from_user() for optimization: we first do the
1330 * sanity checks (buffer size -fast- and access check-healthy
1331 * pointer); if all of those succeed, then we can alloc the memory
1332 * (expensive operation) needed to copy the data to kernel. Then we do
1333 * the copying without checking the user space area
1334 * (__copy_from_user()).
1336 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1337 * it.
1339 * sk The sk of the socket
1340 * addrs The pointer to the addresses in user land
1341 * addrssize Size of the addrs buffer
1343 * Returns >=0 if ok, <0 errno code on error.
1345 static int __sctp_setsockopt_connectx(struct sock *sk,
1346 struct sockaddr __user *addrs,
1347 int addrs_size,
1348 sctp_assoc_t *assoc_id)
1350 struct sockaddr *kaddrs;
1351 gfp_t gfp = GFP_KERNEL;
1352 int err = 0, flags = 0;
1354 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1355 __func__, sk, addrs, addrs_size);
1357 if (unlikely(addrs_size <= 0))
1358 return -EINVAL;
1360 /* Check the user passed a healthy pointer. */
1361 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1362 return -EFAULT;
1364 /* Alloc space for the address array in kernel memory. */
1365 if (sk->sk_socket->file)
1366 gfp = GFP_USER | __GFP_NOWARN;
1367 kaddrs = kmalloc(addrs_size, gfp);
1368 if (unlikely(!kaddrs))
1369 return -ENOMEM;
1371 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1372 kfree(kaddrs);
1373 return -EFAULT;
1376 /* in-kernel sockets don't generally have a file allocated to them
1377 * if all they do is call sock_create_kern().
1379 if (sk->sk_socket->file)
1380 flags = sk->sk_socket->file->f_flags;
1382 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1384 kfree(kaddrs);
1386 return err;
1390 * This is an older interface. It's kept for backward compatibility
1391 * to the option that doesn't provide association id.
1393 static int sctp_setsockopt_connectx_old(struct sock *sk,
1394 struct sockaddr __user *addrs,
1395 int addrs_size)
1397 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1401 * New interface for the API. The since the API is done with a socket
1402 * option, to make it simple we feed back the association id is as a return
1403 * indication to the call. Error is always negative and association id is
1404 * always positive.
1406 static int sctp_setsockopt_connectx(struct sock *sk,
1407 struct sockaddr __user *addrs,
1408 int addrs_size)
1410 sctp_assoc_t assoc_id = 0;
1411 int err = 0;
1413 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1415 if (err)
1416 return err;
1417 else
1418 return assoc_id;
1422 * New (hopefully final) interface for the API.
1423 * We use the sctp_getaddrs_old structure so that use-space library
1424 * can avoid any unnecessary allocations. The only different part
1425 * is that we store the actual length of the address buffer into the
1426 * addrs_num structure member. That way we can re-use the existing
1427 * code.
1429 #ifdef CONFIG_COMPAT
1430 struct compat_sctp_getaddrs_old {
1431 sctp_assoc_t assoc_id;
1432 s32 addr_num;
1433 compat_uptr_t addrs; /* struct sockaddr * */
1435 #endif
1437 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1438 char __user *optval,
1439 int __user *optlen)
1441 struct sctp_getaddrs_old param;
1442 sctp_assoc_t assoc_id = 0;
1443 int err = 0;
1445 #ifdef CONFIG_COMPAT
1446 if (in_compat_syscall()) {
1447 struct compat_sctp_getaddrs_old param32;
1449 if (len < sizeof(param32))
1450 return -EINVAL;
1451 if (copy_from_user(&param32, optval, sizeof(param32)))
1452 return -EFAULT;
1454 param.assoc_id = param32.assoc_id;
1455 param.addr_num = param32.addr_num;
1456 param.addrs = compat_ptr(param32.addrs);
1457 } else
1458 #endif
1460 if (len < sizeof(param))
1461 return -EINVAL;
1462 if (copy_from_user(&param, optval, sizeof(param)))
1463 return -EFAULT;
1466 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1467 param.addrs, param.addr_num,
1468 &assoc_id);
1469 if (err == 0 || err == -EINPROGRESS) {
1470 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1471 return -EFAULT;
1472 if (put_user(sizeof(assoc_id), optlen))
1473 return -EFAULT;
1476 return err;
1479 /* API 3.1.4 close() - UDP Style Syntax
1480 * Applications use close() to perform graceful shutdown (as described in
1481 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1482 * by a UDP-style socket.
1484 * The syntax is
1486 * ret = close(int sd);
1488 * sd - the socket descriptor of the associations to be closed.
1490 * To gracefully shutdown a specific association represented by the
1491 * UDP-style socket, an application should use the sendmsg() call,
1492 * passing no user data, but including the appropriate flag in the
1493 * ancillary data (see Section xxxx).
1495 * If sd in the close() call is a branched-off socket representing only
1496 * one association, the shutdown is performed on that association only.
1498 * 4.1.6 close() - TCP Style Syntax
1500 * Applications use close() to gracefully close down an association.
1502 * The syntax is:
1504 * int close(int sd);
1506 * sd - the socket descriptor of the association to be closed.
1508 * After an application calls close() on a socket descriptor, no further
1509 * socket operations will succeed on that descriptor.
1511 * API 7.1.4 SO_LINGER
1513 * An application using the TCP-style socket can use this option to
1514 * perform the SCTP ABORT primitive. The linger option structure is:
1516 * struct linger {
1517 * int l_onoff; // option on/off
1518 * int l_linger; // linger time
1519 * };
1521 * To enable the option, set l_onoff to 1. If the l_linger value is set
1522 * to 0, calling close() is the same as the ABORT primitive. If the
1523 * value is set to a negative value, the setsockopt() call will return
1524 * an error. If the value is set to a positive value linger_time, the
1525 * close() can be blocked for at most linger_time ms. If the graceful
1526 * shutdown phase does not finish during this period, close() will
1527 * return but the graceful shutdown phase continues in the system.
1529 static void sctp_close(struct sock *sk, long timeout)
1531 struct net *net = sock_net(sk);
1532 struct sctp_endpoint *ep;
1533 struct sctp_association *asoc;
1534 struct list_head *pos, *temp;
1535 unsigned int data_was_unread;
1537 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1539 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1540 sk->sk_shutdown = SHUTDOWN_MASK;
1541 sk->sk_state = SCTP_SS_CLOSING;
1543 ep = sctp_sk(sk)->ep;
1545 /* Clean up any skbs sitting on the receive queue. */
1546 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1547 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1549 /* Walk all associations on an endpoint. */
1550 list_for_each_safe(pos, temp, &ep->asocs) {
1551 asoc = list_entry(pos, struct sctp_association, asocs);
1553 if (sctp_style(sk, TCP)) {
1554 /* A closed association can still be in the list if
1555 * it belongs to a TCP-style listening socket that is
1556 * not yet accepted. If so, free it. If not, send an
1557 * ABORT or SHUTDOWN based on the linger options.
1559 if (sctp_state(asoc, CLOSED)) {
1560 sctp_association_free(asoc);
1561 continue;
1565 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1566 !skb_queue_empty(&asoc->ulpq.reasm) ||
1567 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1568 struct sctp_chunk *chunk;
1570 chunk = sctp_make_abort_user(asoc, NULL, 0);
1571 sctp_primitive_ABORT(net, asoc, chunk);
1572 } else
1573 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1576 /* On a TCP-style socket, block for at most linger_time if set. */
1577 if (sctp_style(sk, TCP) && timeout)
1578 sctp_wait_for_close(sk, timeout);
1580 /* This will run the backlog queue. */
1581 release_sock(sk);
1583 /* Supposedly, no process has access to the socket, but
1584 * the net layers still may.
1585 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1586 * held and that should be grabbed before socket lock.
1588 spin_lock_bh(&net->sctp.addr_wq_lock);
1589 bh_lock_sock_nested(sk);
1591 /* Hold the sock, since sk_common_release() will put sock_put()
1592 * and we have just a little more cleanup.
1594 sock_hold(sk);
1595 sk_common_release(sk);
1597 bh_unlock_sock(sk);
1598 spin_unlock_bh(&net->sctp.addr_wq_lock);
1600 sock_put(sk);
1602 SCTP_DBG_OBJCNT_DEC(sock);
1605 /* Handle EPIPE error. */
1606 static int sctp_error(struct sock *sk, int flags, int err)
1608 if (err == -EPIPE)
1609 err = sock_error(sk) ? : -EPIPE;
1610 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1611 send_sig(SIGPIPE, current, 0);
1612 return err;
1615 /* API 3.1.3 sendmsg() - UDP Style Syntax
1617 * An application uses sendmsg() and recvmsg() calls to transmit data to
1618 * and receive data from its peer.
1620 * ssize_t sendmsg(int socket, const struct msghdr *message,
1621 * int flags);
1623 * socket - the socket descriptor of the endpoint.
1624 * message - pointer to the msghdr structure which contains a single
1625 * user message and possibly some ancillary data.
1627 * See Section 5 for complete description of the data
1628 * structures.
1630 * flags - flags sent or received with the user message, see Section
1631 * 5 for complete description of the flags.
1633 * Note: This function could use a rewrite especially when explicit
1634 * connect support comes in.
1636 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1638 static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1640 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1642 struct net *net = sock_net(sk);
1643 struct sctp_sock *sp;
1644 struct sctp_endpoint *ep;
1645 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1646 struct sctp_transport *transport, *chunk_tp;
1647 struct sctp_chunk *chunk;
1648 union sctp_addr to;
1649 struct sockaddr *msg_name = NULL;
1650 struct sctp_sndrcvinfo default_sinfo;
1651 struct sctp_sndrcvinfo *sinfo;
1652 struct sctp_initmsg *sinit;
1653 sctp_assoc_t associd = 0;
1654 sctp_cmsgs_t cmsgs = { NULL };
1655 sctp_scope_t scope;
1656 bool fill_sinfo_ttl = false, wait_connect = false;
1657 struct sctp_datamsg *datamsg;
1658 int msg_flags = msg->msg_flags;
1659 __u16 sinfo_flags = 0;
1660 long timeo;
1661 int err;
1663 err = 0;
1664 sp = sctp_sk(sk);
1665 ep = sp->ep;
1667 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1668 msg, msg_len, ep);
1670 /* We cannot send a message over a TCP-style listening socket. */
1671 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1672 err = -EPIPE;
1673 goto out_nounlock;
1676 /* Parse out the SCTP CMSGs. */
1677 err = sctp_msghdr_parse(msg, &cmsgs);
1678 if (err) {
1679 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1680 goto out_nounlock;
1683 /* Fetch the destination address for this packet. This
1684 * address only selects the association--it is not necessarily
1685 * the address we will send to.
1686 * For a peeled-off socket, msg_name is ignored.
1688 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1689 int msg_namelen = msg->msg_namelen;
1691 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1692 msg_namelen);
1693 if (err)
1694 return err;
1696 if (msg_namelen > sizeof(to))
1697 msg_namelen = sizeof(to);
1698 memcpy(&to, msg->msg_name, msg_namelen);
1699 msg_name = msg->msg_name;
1702 sinit = cmsgs.init;
1703 if (cmsgs.sinfo != NULL) {
1704 memset(&default_sinfo, 0, sizeof(default_sinfo));
1705 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1706 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1707 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1708 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1709 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1711 sinfo = &default_sinfo;
1712 fill_sinfo_ttl = true;
1713 } else {
1714 sinfo = cmsgs.srinfo;
1716 /* Did the user specify SNDINFO/SNDRCVINFO? */
1717 if (sinfo) {
1718 sinfo_flags = sinfo->sinfo_flags;
1719 associd = sinfo->sinfo_assoc_id;
1722 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1723 msg_len, sinfo_flags);
1725 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1726 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1727 err = -EINVAL;
1728 goto out_nounlock;
1731 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1732 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1733 * If SCTP_ABORT is set, the message length could be non zero with
1734 * the msg_iov set to the user abort reason.
1736 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1737 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1738 err = -EINVAL;
1739 goto out_nounlock;
1742 /* If SCTP_ADDR_OVER is set, there must be an address
1743 * specified in msg_name.
1745 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1746 err = -EINVAL;
1747 goto out_nounlock;
1750 transport = NULL;
1752 pr_debug("%s: about to look up association\n", __func__);
1754 lock_sock(sk);
1756 /* If a msg_name has been specified, assume this is to be used. */
1757 if (msg_name) {
1758 /* Look for a matching association on the endpoint. */
1759 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1761 /* If we could not find a matching association on the
1762 * endpoint, make sure that it is not a TCP-style
1763 * socket that already has an association or there is
1764 * no peeled-off association on another socket.
1766 if (!asoc &&
1767 ((sctp_style(sk, TCP) &&
1768 (sctp_sstate(sk, ESTABLISHED) ||
1769 sctp_sstate(sk, CLOSING))) ||
1770 sctp_endpoint_is_peeled_off(ep, &to))) {
1771 err = -EADDRNOTAVAIL;
1772 goto out_unlock;
1774 } else {
1775 asoc = sctp_id2assoc(sk, associd);
1776 if (!asoc) {
1777 err = -EPIPE;
1778 goto out_unlock;
1782 if (asoc) {
1783 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1785 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1786 * socket that has an association in CLOSED state. This can
1787 * happen when an accepted socket has an association that is
1788 * already CLOSED.
1790 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1791 err = -EPIPE;
1792 goto out_unlock;
1795 if (sinfo_flags & SCTP_EOF) {
1796 pr_debug("%s: shutting down association:%p\n",
1797 __func__, asoc);
1799 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1800 err = 0;
1801 goto out_unlock;
1803 if (sinfo_flags & SCTP_ABORT) {
1805 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1806 if (!chunk) {
1807 err = -ENOMEM;
1808 goto out_unlock;
1811 pr_debug("%s: aborting association:%p\n",
1812 __func__, asoc);
1814 sctp_primitive_ABORT(net, asoc, chunk);
1815 err = 0;
1816 goto out_unlock;
1820 /* Do we need to create the association? */
1821 if (!asoc) {
1822 pr_debug("%s: there is no association yet\n", __func__);
1824 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1825 err = -EINVAL;
1826 goto out_unlock;
1829 /* Check for invalid stream against the stream counts,
1830 * either the default or the user specified stream counts.
1832 if (sinfo) {
1833 if (!sinit || !sinit->sinit_num_ostreams) {
1834 /* Check against the defaults. */
1835 if (sinfo->sinfo_stream >=
1836 sp->initmsg.sinit_num_ostreams) {
1837 err = -EINVAL;
1838 goto out_unlock;
1840 } else {
1841 /* Check against the requested. */
1842 if (sinfo->sinfo_stream >=
1843 sinit->sinit_num_ostreams) {
1844 err = -EINVAL;
1845 goto out_unlock;
1851 * API 3.1.2 bind() - UDP Style Syntax
1852 * If a bind() or sctp_bindx() is not called prior to a
1853 * sendmsg() call that initiates a new association, the
1854 * system picks an ephemeral port and will choose an address
1855 * set equivalent to binding with a wildcard address.
1857 if (!ep->base.bind_addr.port) {
1858 if (sctp_autobind(sk)) {
1859 err = -EAGAIN;
1860 goto out_unlock;
1862 } else {
1864 * If an unprivileged user inherits a one-to-many
1865 * style socket with open associations on a privileged
1866 * port, it MAY be permitted to accept new associations,
1867 * but it SHOULD NOT be permitted to open new
1868 * associations.
1870 if (ep->base.bind_addr.port < PROT_SOCK &&
1871 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1872 err = -EACCES;
1873 goto out_unlock;
1877 scope = sctp_scope(&to);
1878 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1879 if (!new_asoc) {
1880 err = -ENOMEM;
1881 goto out_unlock;
1883 asoc = new_asoc;
1884 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1885 if (err < 0) {
1886 err = -ENOMEM;
1887 goto out_free;
1890 /* If the SCTP_INIT ancillary data is specified, set all
1891 * the association init values accordingly.
1893 if (sinit) {
1894 if (sinit->sinit_num_ostreams) {
1895 asoc->c.sinit_num_ostreams =
1896 sinit->sinit_num_ostreams;
1898 if (sinit->sinit_max_instreams) {
1899 asoc->c.sinit_max_instreams =
1900 sinit->sinit_max_instreams;
1902 if (sinit->sinit_max_attempts) {
1903 asoc->max_init_attempts
1904 = sinit->sinit_max_attempts;
1906 if (sinit->sinit_max_init_timeo) {
1907 asoc->max_init_timeo =
1908 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1912 /* Prime the peer's transport structures. */
1913 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1914 if (!transport) {
1915 err = -ENOMEM;
1916 goto out_free;
1920 /* ASSERT: we have a valid association at this point. */
1921 pr_debug("%s: we have a valid association\n", __func__);
1923 if (!sinfo) {
1924 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1925 * one with some defaults.
1927 memset(&default_sinfo, 0, sizeof(default_sinfo));
1928 default_sinfo.sinfo_stream = asoc->default_stream;
1929 default_sinfo.sinfo_flags = asoc->default_flags;
1930 default_sinfo.sinfo_ppid = asoc->default_ppid;
1931 default_sinfo.sinfo_context = asoc->default_context;
1932 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1933 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1935 sinfo = &default_sinfo;
1936 } else if (fill_sinfo_ttl) {
1937 /* In case SNDINFO was specified, we still need to fill
1938 * it with a default ttl from the assoc here.
1940 sinfo->sinfo_timetolive = asoc->default_timetolive;
1943 /* API 7.1.7, the sndbuf size per association bounds the
1944 * maximum size of data that can be sent in a single send call.
1946 if (msg_len > sk->sk_sndbuf) {
1947 err = -EMSGSIZE;
1948 goto out_free;
1951 if (asoc->pmtu_pending)
1952 sctp_assoc_pending_pmtu(sk, asoc);
1954 /* If fragmentation is disabled and the message length exceeds the
1955 * association fragmentation point, return EMSGSIZE. The I-D
1956 * does not specify what this error is, but this looks like
1957 * a great fit.
1959 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1960 err = -EMSGSIZE;
1961 goto out_free;
1964 /* Check for invalid stream. */
1965 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1966 err = -EINVAL;
1967 goto out_free;
1970 if (sctp_wspace(asoc) < msg_len)
1971 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1973 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1974 if (!sctp_wspace(asoc)) {
1975 /* sk can be changed by peel off when waiting for buf. */
1976 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1977 if (err) {
1978 if (err == -ESRCH) {
1979 /* asoc is already dead. */
1980 new_asoc = NULL;
1981 err = -EPIPE;
1983 goto out_free;
1987 /* If an address is passed with the sendto/sendmsg call, it is used
1988 * to override the primary destination address in the TCP model, or
1989 * when SCTP_ADDR_OVER flag is set in the UDP model.
1991 if ((sctp_style(sk, TCP) && msg_name) ||
1992 (sinfo_flags & SCTP_ADDR_OVER)) {
1993 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1994 if (!chunk_tp) {
1995 err = -EINVAL;
1996 goto out_free;
1998 } else
1999 chunk_tp = NULL;
2001 /* Auto-connect, if we aren't connected already. */
2002 if (sctp_state(asoc, CLOSED)) {
2003 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
2004 if (err < 0)
2005 goto out_free;
2007 wait_connect = true;
2008 pr_debug("%s: we associated primitively\n", __func__);
2011 /* Break the message into multiple chunks of maximum size. */
2012 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
2013 if (IS_ERR(datamsg)) {
2014 err = PTR_ERR(datamsg);
2015 goto out_free;
2018 /* Now send the (possibly) fragmented message. */
2019 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
2020 sctp_chunk_hold(chunk);
2022 /* Do accounting for the write space. */
2023 sctp_set_owner_w(chunk);
2025 chunk->transport = chunk_tp;
2028 /* Send it to the lower layers. Note: all chunks
2029 * must either fail or succeed. The lower layer
2030 * works that way today. Keep it that way or this
2031 * breaks.
2033 err = sctp_primitive_SEND(net, asoc, datamsg);
2034 /* Did the lower layer accept the chunk? */
2035 if (err) {
2036 sctp_datamsg_free(datamsg);
2037 goto out_free;
2040 pr_debug("%s: we sent primitively\n", __func__);
2042 sctp_datamsg_put(datamsg);
2043 err = msg_len;
2045 if (unlikely(wait_connect)) {
2046 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2047 sctp_wait_for_connect(asoc, &timeo);
2050 /* If we are already past ASSOCIATE, the lower
2051 * layers are responsible for association cleanup.
2053 goto out_unlock;
2055 out_free:
2056 if (new_asoc)
2057 sctp_association_free(asoc);
2058 out_unlock:
2059 release_sock(sk);
2061 out_nounlock:
2062 return sctp_error(sk, msg_flags, err);
2064 #if 0
2065 do_sock_err:
2066 if (msg_len)
2067 err = msg_len;
2068 else
2069 err = sock_error(sk);
2070 goto out;
2072 do_interrupted:
2073 if (msg_len)
2074 err = msg_len;
2075 goto out;
2076 #endif /* 0 */
2079 /* This is an extended version of skb_pull() that removes the data from the
2080 * start of a skb even when data is spread across the list of skb's in the
2081 * frag_list. len specifies the total amount of data that needs to be removed.
2082 * when 'len' bytes could be removed from the skb, it returns 0.
2083 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2084 * could not be removed.
2086 static int sctp_skb_pull(struct sk_buff *skb, int len)
2088 struct sk_buff *list;
2089 int skb_len = skb_headlen(skb);
2090 int rlen;
2092 if (len <= skb_len) {
2093 __skb_pull(skb, len);
2094 return 0;
2096 len -= skb_len;
2097 __skb_pull(skb, skb_len);
2099 skb_walk_frags(skb, list) {
2100 rlen = sctp_skb_pull(list, len);
2101 skb->len -= (len-rlen);
2102 skb->data_len -= (len-rlen);
2104 if (!rlen)
2105 return 0;
2107 len = rlen;
2110 return len;
2113 /* API 3.1.3 recvmsg() - UDP Style Syntax
2115 * ssize_t recvmsg(int socket, struct msghdr *message,
2116 * int flags);
2118 * socket - the socket descriptor of the endpoint.
2119 * message - pointer to the msghdr structure which contains a single
2120 * user message and possibly some ancillary data.
2122 * See Section 5 for complete description of the data
2123 * structures.
2125 * flags - flags sent or received with the user message, see Section
2126 * 5 for complete description of the flags.
2128 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2129 int noblock, int flags, int *addr_len)
2131 struct sctp_ulpevent *event = NULL;
2132 struct sctp_sock *sp = sctp_sk(sk);
2133 struct sk_buff *skb, *head_skb;
2134 int copied;
2135 int err = 0;
2136 int skb_len;
2138 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2139 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2140 addr_len);
2142 lock_sock(sk);
2144 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2145 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2146 err = -ENOTCONN;
2147 goto out;
2150 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2151 if (!skb)
2152 goto out;
2154 /* Get the total length of the skb including any skb's in the
2155 * frag_list.
2157 skb_len = skb->len;
2159 copied = skb_len;
2160 if (copied > len)
2161 copied = len;
2163 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2165 event = sctp_skb2event(skb);
2167 if (err)
2168 goto out_free;
2170 if (event->chunk && event->chunk->head_skb)
2171 head_skb = event->chunk->head_skb;
2172 else
2173 head_skb = skb;
2174 sock_recv_ts_and_drops(msg, sk, head_skb);
2175 if (sctp_ulpevent_is_notification(event)) {
2176 msg->msg_flags |= MSG_NOTIFICATION;
2177 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2178 } else {
2179 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2182 /* Check if we allow SCTP_NXTINFO. */
2183 if (sp->recvnxtinfo)
2184 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2185 /* Check if we allow SCTP_RCVINFO. */
2186 if (sp->recvrcvinfo)
2187 sctp_ulpevent_read_rcvinfo(event, msg);
2188 /* Check if we allow SCTP_SNDRCVINFO. */
2189 if (sp->subscribe.sctp_data_io_event)
2190 sctp_ulpevent_read_sndrcvinfo(event, msg);
2192 err = copied;
2194 /* If skb's length exceeds the user's buffer, update the skb and
2195 * push it back to the receive_queue so that the next call to
2196 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2198 if (skb_len > copied) {
2199 msg->msg_flags &= ~MSG_EOR;
2200 if (flags & MSG_PEEK)
2201 goto out_free;
2202 sctp_skb_pull(skb, copied);
2203 skb_queue_head(&sk->sk_receive_queue, skb);
2205 /* When only partial message is copied to the user, increase
2206 * rwnd by that amount. If all the data in the skb is read,
2207 * rwnd is updated when the event is freed.
2209 if (!sctp_ulpevent_is_notification(event))
2210 sctp_assoc_rwnd_increase(event->asoc, copied);
2211 goto out;
2212 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2213 (event->msg_flags & MSG_EOR))
2214 msg->msg_flags |= MSG_EOR;
2215 else
2216 msg->msg_flags &= ~MSG_EOR;
2218 out_free:
2219 if (flags & MSG_PEEK) {
2220 /* Release the skb reference acquired after peeking the skb in
2221 * sctp_skb_recv_datagram().
2223 kfree_skb(skb);
2224 } else {
2225 /* Free the event which includes releasing the reference to
2226 * the owner of the skb, freeing the skb and updating the
2227 * rwnd.
2229 sctp_ulpevent_free(event);
2231 out:
2232 release_sock(sk);
2233 return err;
2236 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2238 * This option is a on/off flag. If enabled no SCTP message
2239 * fragmentation will be performed. Instead if a message being sent
2240 * exceeds the current PMTU size, the message will NOT be sent and
2241 * instead a error will be indicated to the user.
2243 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2244 char __user *optval,
2245 unsigned int optlen)
2247 int val;
2249 if (optlen < sizeof(int))
2250 return -EINVAL;
2252 if (get_user(val, (int __user *)optval))
2253 return -EFAULT;
2255 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2257 return 0;
2260 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2261 unsigned int optlen)
2263 struct sctp_association *asoc;
2264 struct sctp_ulpevent *event;
2266 if (optlen > sizeof(struct sctp_event_subscribe))
2267 return -EINVAL;
2268 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2269 return -EFAULT;
2271 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2272 * if there is no data to be sent or retransmit, the stack will
2273 * immediately send up this notification.
2275 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2276 &sctp_sk(sk)->subscribe)) {
2277 asoc = sctp_id2assoc(sk, 0);
2279 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2280 event = sctp_ulpevent_make_sender_dry_event(asoc,
2281 GFP_ATOMIC);
2282 if (!event)
2283 return -ENOMEM;
2285 sctp_ulpq_tail_event(&asoc->ulpq, event);
2289 return 0;
2292 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2294 * This socket option is applicable to the UDP-style socket only. When
2295 * set it will cause associations that are idle for more than the
2296 * specified number of seconds to automatically close. An association
2297 * being idle is defined an association that has NOT sent or received
2298 * user data. The special value of '0' indicates that no automatic
2299 * close of any associations should be performed. The option expects an
2300 * integer defining the number of seconds of idle time before an
2301 * association is closed.
2303 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2304 unsigned int optlen)
2306 struct sctp_sock *sp = sctp_sk(sk);
2307 struct net *net = sock_net(sk);
2309 /* Applicable to UDP-style socket only */
2310 if (sctp_style(sk, TCP))
2311 return -EOPNOTSUPP;
2312 if (optlen != sizeof(int))
2313 return -EINVAL;
2314 if (copy_from_user(&sp->autoclose, optval, optlen))
2315 return -EFAULT;
2317 if (sp->autoclose > net->sctp.max_autoclose)
2318 sp->autoclose = net->sctp.max_autoclose;
2320 return 0;
2323 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2325 * Applications can enable or disable heartbeats for any peer address of
2326 * an association, modify an address's heartbeat interval, force a
2327 * heartbeat to be sent immediately, and adjust the address's maximum
2328 * number of retransmissions sent before an address is considered
2329 * unreachable. The following structure is used to access and modify an
2330 * address's parameters:
2332 * struct sctp_paddrparams {
2333 * sctp_assoc_t spp_assoc_id;
2334 * struct sockaddr_storage spp_address;
2335 * uint32_t spp_hbinterval;
2336 * uint16_t spp_pathmaxrxt;
2337 * uint32_t spp_pathmtu;
2338 * uint32_t spp_sackdelay;
2339 * uint32_t spp_flags;
2340 * };
2342 * spp_assoc_id - (one-to-many style socket) This is filled in the
2343 * application, and identifies the association for
2344 * this query.
2345 * spp_address - This specifies which address is of interest.
2346 * spp_hbinterval - This contains the value of the heartbeat interval,
2347 * in milliseconds. If a value of zero
2348 * is present in this field then no changes are to
2349 * be made to this parameter.
2350 * spp_pathmaxrxt - This contains the maximum number of
2351 * retransmissions before this address shall be
2352 * considered unreachable. If a value of zero
2353 * is present in this field then no changes are to
2354 * be made to this parameter.
2355 * spp_pathmtu - When Path MTU discovery is disabled the value
2356 * specified here will be the "fixed" path mtu.
2357 * Note that if the spp_address field is empty
2358 * then all associations on this address will
2359 * have this fixed path mtu set upon them.
2361 * spp_sackdelay - When delayed sack is enabled, this value specifies
2362 * the number of milliseconds that sacks will be delayed
2363 * for. This value will apply to all addresses of an
2364 * association if the spp_address field is empty. Note
2365 * also, that if delayed sack is enabled and this
2366 * value is set to 0, no change is made to the last
2367 * recorded delayed sack timer value.
2369 * spp_flags - These flags are used to control various features
2370 * on an association. The flag field may contain
2371 * zero or more of the following options.
2373 * SPP_HB_ENABLE - Enable heartbeats on the
2374 * specified address. Note that if the address
2375 * field is empty all addresses for the association
2376 * have heartbeats enabled upon them.
2378 * SPP_HB_DISABLE - Disable heartbeats on the
2379 * speicifed address. Note that if the address
2380 * field is empty all addresses for the association
2381 * will have their heartbeats disabled. Note also
2382 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2383 * mutually exclusive, only one of these two should
2384 * be specified. Enabling both fields will have
2385 * undetermined results.
2387 * SPP_HB_DEMAND - Request a user initiated heartbeat
2388 * to be made immediately.
2390 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2391 * heartbeat delayis to be set to the value of 0
2392 * milliseconds.
2394 * SPP_PMTUD_ENABLE - This field will enable PMTU
2395 * discovery upon the specified address. Note that
2396 * if the address feild is empty then all addresses
2397 * on the association are effected.
2399 * SPP_PMTUD_DISABLE - This field will disable PMTU
2400 * discovery upon the specified address. Note that
2401 * if the address feild is empty then all addresses
2402 * on the association are effected. Not also that
2403 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2404 * exclusive. Enabling both will have undetermined
2405 * results.
2407 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2408 * on delayed sack. The time specified in spp_sackdelay
2409 * is used to specify the sack delay for this address. Note
2410 * that if spp_address is empty then all addresses will
2411 * enable delayed sack and take on the sack delay
2412 * value specified in spp_sackdelay.
2413 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2414 * off delayed sack. If the spp_address field is blank then
2415 * delayed sack is disabled for the entire association. Note
2416 * also that this field is mutually exclusive to
2417 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2418 * results.
2420 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2421 struct sctp_transport *trans,
2422 struct sctp_association *asoc,
2423 struct sctp_sock *sp,
2424 int hb_change,
2425 int pmtud_change,
2426 int sackdelay_change)
2428 int error;
2430 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2431 struct net *net = sock_net(trans->asoc->base.sk);
2433 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2434 if (error)
2435 return error;
2438 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2439 * this field is ignored. Note also that a value of zero indicates
2440 * the current setting should be left unchanged.
2442 if (params->spp_flags & SPP_HB_ENABLE) {
2444 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2445 * set. This lets us use 0 value when this flag
2446 * is set.
2448 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2449 params->spp_hbinterval = 0;
2451 if (params->spp_hbinterval ||
2452 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2453 if (trans) {
2454 trans->hbinterval =
2455 msecs_to_jiffies(params->spp_hbinterval);
2456 } else if (asoc) {
2457 asoc->hbinterval =
2458 msecs_to_jiffies(params->spp_hbinterval);
2459 } else {
2460 sp->hbinterval = params->spp_hbinterval;
2465 if (hb_change) {
2466 if (trans) {
2467 trans->param_flags =
2468 (trans->param_flags & ~SPP_HB) | hb_change;
2469 } else if (asoc) {
2470 asoc->param_flags =
2471 (asoc->param_flags & ~SPP_HB) | hb_change;
2472 } else {
2473 sp->param_flags =
2474 (sp->param_flags & ~SPP_HB) | hb_change;
2478 /* When Path MTU discovery is disabled the value specified here will
2479 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2480 * include the flag SPP_PMTUD_DISABLE for this field to have any
2481 * effect).
2483 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2484 if (trans) {
2485 trans->pathmtu = params->spp_pathmtu;
2486 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2487 } else if (asoc) {
2488 asoc->pathmtu = params->spp_pathmtu;
2489 sctp_frag_point(asoc, params->spp_pathmtu);
2490 } else {
2491 sp->pathmtu = params->spp_pathmtu;
2495 if (pmtud_change) {
2496 if (trans) {
2497 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2498 (params->spp_flags & SPP_PMTUD_ENABLE);
2499 trans->param_flags =
2500 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2501 if (update) {
2502 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2503 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2505 } else if (asoc) {
2506 asoc->param_flags =
2507 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2508 } else {
2509 sp->param_flags =
2510 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2514 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2515 * value of this field is ignored. Note also that a value of zero
2516 * indicates the current setting should be left unchanged.
2518 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2519 if (trans) {
2520 trans->sackdelay =
2521 msecs_to_jiffies(params->spp_sackdelay);
2522 } else if (asoc) {
2523 asoc->sackdelay =
2524 msecs_to_jiffies(params->spp_sackdelay);
2525 } else {
2526 sp->sackdelay = params->spp_sackdelay;
2530 if (sackdelay_change) {
2531 if (trans) {
2532 trans->param_flags =
2533 (trans->param_flags & ~SPP_SACKDELAY) |
2534 sackdelay_change;
2535 } else if (asoc) {
2536 asoc->param_flags =
2537 (asoc->param_flags & ~SPP_SACKDELAY) |
2538 sackdelay_change;
2539 } else {
2540 sp->param_flags =
2541 (sp->param_flags & ~SPP_SACKDELAY) |
2542 sackdelay_change;
2546 /* Note that a value of zero indicates the current setting should be
2547 left unchanged.
2549 if (params->spp_pathmaxrxt) {
2550 if (trans) {
2551 trans->pathmaxrxt = params->spp_pathmaxrxt;
2552 } else if (asoc) {
2553 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2554 } else {
2555 sp->pathmaxrxt = params->spp_pathmaxrxt;
2559 return 0;
2562 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2563 char __user *optval,
2564 unsigned int optlen)
2566 struct sctp_paddrparams params;
2567 struct sctp_transport *trans = NULL;
2568 struct sctp_association *asoc = NULL;
2569 struct sctp_sock *sp = sctp_sk(sk);
2570 int error;
2571 int hb_change, pmtud_change, sackdelay_change;
2573 if (optlen != sizeof(struct sctp_paddrparams))
2574 return -EINVAL;
2576 if (copy_from_user(&params, optval, optlen))
2577 return -EFAULT;
2579 /* Validate flags and value parameters. */
2580 hb_change = params.spp_flags & SPP_HB;
2581 pmtud_change = params.spp_flags & SPP_PMTUD;
2582 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2584 if (hb_change == SPP_HB ||
2585 pmtud_change == SPP_PMTUD ||
2586 sackdelay_change == SPP_SACKDELAY ||
2587 params.spp_sackdelay > 500 ||
2588 (params.spp_pathmtu &&
2589 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2590 return -EINVAL;
2592 /* If an address other than INADDR_ANY is specified, and
2593 * no transport is found, then the request is invalid.
2595 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2596 trans = sctp_addr_id2transport(sk, &params.spp_address,
2597 params.spp_assoc_id);
2598 if (!trans)
2599 return -EINVAL;
2602 /* Get association, if assoc_id != 0 and the socket is a one
2603 * to many style socket, and an association was not found, then
2604 * the id was invalid.
2606 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2607 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2608 return -EINVAL;
2610 /* Heartbeat demand can only be sent on a transport or
2611 * association, but not a socket.
2613 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2614 return -EINVAL;
2616 /* Process parameters. */
2617 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2618 hb_change, pmtud_change,
2619 sackdelay_change);
2621 if (error)
2622 return error;
2624 /* If changes are for association, also apply parameters to each
2625 * transport.
2627 if (!trans && asoc) {
2628 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2629 transports) {
2630 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2631 hb_change, pmtud_change,
2632 sackdelay_change);
2636 return 0;
2639 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2641 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2644 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2646 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2650 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2652 * This option will effect the way delayed acks are performed. This
2653 * option allows you to get or set the delayed ack time, in
2654 * milliseconds. It also allows changing the delayed ack frequency.
2655 * Changing the frequency to 1 disables the delayed sack algorithm. If
2656 * the assoc_id is 0, then this sets or gets the endpoints default
2657 * values. If the assoc_id field is non-zero, then the set or get
2658 * effects the specified association for the one to many model (the
2659 * assoc_id field is ignored by the one to one model). Note that if
2660 * sack_delay or sack_freq are 0 when setting this option, then the
2661 * current values will remain unchanged.
2663 * struct sctp_sack_info {
2664 * sctp_assoc_t sack_assoc_id;
2665 * uint32_t sack_delay;
2666 * uint32_t sack_freq;
2667 * };
2669 * sack_assoc_id - This parameter, indicates which association the user
2670 * is performing an action upon. Note that if this field's value is
2671 * zero then the endpoints default value is changed (effecting future
2672 * associations only).
2674 * sack_delay - This parameter contains the number of milliseconds that
2675 * the user is requesting the delayed ACK timer be set to. Note that
2676 * this value is defined in the standard to be between 200 and 500
2677 * milliseconds.
2679 * sack_freq - This parameter contains the number of packets that must
2680 * be received before a sack is sent without waiting for the delay
2681 * timer to expire. The default value for this is 2, setting this
2682 * value to 1 will disable the delayed sack algorithm.
2685 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2686 char __user *optval, unsigned int optlen)
2688 struct sctp_sack_info params;
2689 struct sctp_transport *trans = NULL;
2690 struct sctp_association *asoc = NULL;
2691 struct sctp_sock *sp = sctp_sk(sk);
2693 if (optlen == sizeof(struct sctp_sack_info)) {
2694 if (copy_from_user(&params, optval, optlen))
2695 return -EFAULT;
2697 if (params.sack_delay == 0 && params.sack_freq == 0)
2698 return 0;
2699 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2700 pr_warn_ratelimited(DEPRECATED
2701 "%s (pid %d) "
2702 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2703 "Use struct sctp_sack_info instead\n",
2704 current->comm, task_pid_nr(current));
2705 if (copy_from_user(&params, optval, optlen))
2706 return -EFAULT;
2708 if (params.sack_delay == 0)
2709 params.sack_freq = 1;
2710 else
2711 params.sack_freq = 0;
2712 } else
2713 return -EINVAL;
2715 /* Validate value parameter. */
2716 if (params.sack_delay > 500)
2717 return -EINVAL;
2719 /* Get association, if sack_assoc_id != 0 and the socket is a one
2720 * to many style socket, and an association was not found, then
2721 * the id was invalid.
2723 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2724 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2725 return -EINVAL;
2727 if (params.sack_delay) {
2728 if (asoc) {
2729 asoc->sackdelay =
2730 msecs_to_jiffies(params.sack_delay);
2731 asoc->param_flags =
2732 sctp_spp_sackdelay_enable(asoc->param_flags);
2733 } else {
2734 sp->sackdelay = params.sack_delay;
2735 sp->param_flags =
2736 sctp_spp_sackdelay_enable(sp->param_flags);
2740 if (params.sack_freq == 1) {
2741 if (asoc) {
2742 asoc->param_flags =
2743 sctp_spp_sackdelay_disable(asoc->param_flags);
2744 } else {
2745 sp->param_flags =
2746 sctp_spp_sackdelay_disable(sp->param_flags);
2748 } else if (params.sack_freq > 1) {
2749 if (asoc) {
2750 asoc->sackfreq = params.sack_freq;
2751 asoc->param_flags =
2752 sctp_spp_sackdelay_enable(asoc->param_flags);
2753 } else {
2754 sp->sackfreq = params.sack_freq;
2755 sp->param_flags =
2756 sctp_spp_sackdelay_enable(sp->param_flags);
2760 /* If change is for association, also apply to each transport. */
2761 if (asoc) {
2762 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2763 transports) {
2764 if (params.sack_delay) {
2765 trans->sackdelay =
2766 msecs_to_jiffies(params.sack_delay);
2767 trans->param_flags =
2768 sctp_spp_sackdelay_enable(trans->param_flags);
2770 if (params.sack_freq == 1) {
2771 trans->param_flags =
2772 sctp_spp_sackdelay_disable(trans->param_flags);
2773 } else if (params.sack_freq > 1) {
2774 trans->sackfreq = params.sack_freq;
2775 trans->param_flags =
2776 sctp_spp_sackdelay_enable(trans->param_flags);
2781 return 0;
2784 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2786 * Applications can specify protocol parameters for the default association
2787 * initialization. The option name argument to setsockopt() and getsockopt()
2788 * is SCTP_INITMSG.
2790 * Setting initialization parameters is effective only on an unconnected
2791 * socket (for UDP-style sockets only future associations are effected
2792 * by the change). With TCP-style sockets, this option is inherited by
2793 * sockets derived from a listener socket.
2795 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2797 struct sctp_initmsg sinit;
2798 struct sctp_sock *sp = sctp_sk(sk);
2800 if (optlen != sizeof(struct sctp_initmsg))
2801 return -EINVAL;
2802 if (copy_from_user(&sinit, optval, optlen))
2803 return -EFAULT;
2805 if (sinit.sinit_num_ostreams)
2806 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2807 if (sinit.sinit_max_instreams)
2808 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2809 if (sinit.sinit_max_attempts)
2810 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2811 if (sinit.sinit_max_init_timeo)
2812 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2814 return 0;
2818 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2820 * Applications that wish to use the sendto() system call may wish to
2821 * specify a default set of parameters that would normally be supplied
2822 * through the inclusion of ancillary data. This socket option allows
2823 * such an application to set the default sctp_sndrcvinfo structure.
2824 * The application that wishes to use this socket option simply passes
2825 * in to this call the sctp_sndrcvinfo structure defined in Section
2826 * 5.2.2) The input parameters accepted by this call include
2827 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2828 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2829 * to this call if the caller is using the UDP model.
2831 static int sctp_setsockopt_default_send_param(struct sock *sk,
2832 char __user *optval,
2833 unsigned int optlen)
2835 struct sctp_sock *sp = sctp_sk(sk);
2836 struct sctp_association *asoc;
2837 struct sctp_sndrcvinfo info;
2839 if (optlen != sizeof(info))
2840 return -EINVAL;
2841 if (copy_from_user(&info, optval, optlen))
2842 return -EFAULT;
2843 if (info.sinfo_flags &
2844 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2845 SCTP_ABORT | SCTP_EOF))
2846 return -EINVAL;
2848 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2849 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2850 return -EINVAL;
2851 if (asoc) {
2852 asoc->default_stream = info.sinfo_stream;
2853 asoc->default_flags = info.sinfo_flags;
2854 asoc->default_ppid = info.sinfo_ppid;
2855 asoc->default_context = info.sinfo_context;
2856 asoc->default_timetolive = info.sinfo_timetolive;
2857 } else {
2858 sp->default_stream = info.sinfo_stream;
2859 sp->default_flags = info.sinfo_flags;
2860 sp->default_ppid = info.sinfo_ppid;
2861 sp->default_context = info.sinfo_context;
2862 sp->default_timetolive = info.sinfo_timetolive;
2865 return 0;
2868 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2869 * (SCTP_DEFAULT_SNDINFO)
2871 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2872 char __user *optval,
2873 unsigned int optlen)
2875 struct sctp_sock *sp = sctp_sk(sk);
2876 struct sctp_association *asoc;
2877 struct sctp_sndinfo info;
2879 if (optlen != sizeof(info))
2880 return -EINVAL;
2881 if (copy_from_user(&info, optval, optlen))
2882 return -EFAULT;
2883 if (info.snd_flags &
2884 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2885 SCTP_ABORT | SCTP_EOF))
2886 return -EINVAL;
2888 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2889 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2890 return -EINVAL;
2891 if (asoc) {
2892 asoc->default_stream = info.snd_sid;
2893 asoc->default_flags = info.snd_flags;
2894 asoc->default_ppid = info.snd_ppid;
2895 asoc->default_context = info.snd_context;
2896 } else {
2897 sp->default_stream = info.snd_sid;
2898 sp->default_flags = info.snd_flags;
2899 sp->default_ppid = info.snd_ppid;
2900 sp->default_context = info.snd_context;
2903 return 0;
2906 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2908 * Requests that the local SCTP stack use the enclosed peer address as
2909 * the association primary. The enclosed address must be one of the
2910 * association peer's addresses.
2912 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2913 unsigned int optlen)
2915 struct sctp_prim prim;
2916 struct sctp_transport *trans;
2918 if (optlen != sizeof(struct sctp_prim))
2919 return -EINVAL;
2921 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2922 return -EFAULT;
2924 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2925 if (!trans)
2926 return -EINVAL;
2928 sctp_assoc_set_primary(trans->asoc, trans);
2930 return 0;
2934 * 7.1.5 SCTP_NODELAY
2936 * Turn on/off any Nagle-like algorithm. This means that packets are
2937 * generally sent as soon as possible and no unnecessary delays are
2938 * introduced, at the cost of more packets in the network. Expects an
2939 * integer boolean flag.
2941 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2942 unsigned int optlen)
2944 int val;
2946 if (optlen < sizeof(int))
2947 return -EINVAL;
2948 if (get_user(val, (int __user *)optval))
2949 return -EFAULT;
2951 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2952 return 0;
2957 * 7.1.1 SCTP_RTOINFO
2959 * The protocol parameters used to initialize and bound retransmission
2960 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2961 * and modify these parameters.
2962 * All parameters are time values, in milliseconds. A value of 0, when
2963 * modifying the parameters, indicates that the current value should not
2964 * be changed.
2967 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2969 struct sctp_rtoinfo rtoinfo;
2970 struct sctp_association *asoc;
2971 unsigned long rto_min, rto_max;
2972 struct sctp_sock *sp = sctp_sk(sk);
2974 if (optlen != sizeof (struct sctp_rtoinfo))
2975 return -EINVAL;
2977 if (copy_from_user(&rtoinfo, optval, optlen))
2978 return -EFAULT;
2980 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2982 /* Set the values to the specific association */
2983 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2984 return -EINVAL;
2986 rto_max = rtoinfo.srto_max;
2987 rto_min = rtoinfo.srto_min;
2989 if (rto_max)
2990 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2991 else
2992 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2994 if (rto_min)
2995 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2996 else
2997 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2999 if (rto_min > rto_max)
3000 return -EINVAL;
3002 if (asoc) {
3003 if (rtoinfo.srto_initial != 0)
3004 asoc->rto_initial =
3005 msecs_to_jiffies(rtoinfo.srto_initial);
3006 asoc->rto_max = rto_max;
3007 asoc->rto_min = rto_min;
3008 } else {
3009 /* If there is no association or the association-id = 0
3010 * set the values to the endpoint.
3012 if (rtoinfo.srto_initial != 0)
3013 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3014 sp->rtoinfo.srto_max = rto_max;
3015 sp->rtoinfo.srto_min = rto_min;
3018 return 0;
3023 * 7.1.2 SCTP_ASSOCINFO
3025 * This option is used to tune the maximum retransmission attempts
3026 * of the association.
3027 * Returns an error if the new association retransmission value is
3028 * greater than the sum of the retransmission value of the peer.
3029 * See [SCTP] for more information.
3032 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3035 struct sctp_assocparams assocparams;
3036 struct sctp_association *asoc;
3038 if (optlen != sizeof(struct sctp_assocparams))
3039 return -EINVAL;
3040 if (copy_from_user(&assocparams, optval, optlen))
3041 return -EFAULT;
3043 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3045 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3046 return -EINVAL;
3048 /* Set the values to the specific association */
3049 if (asoc) {
3050 if (assocparams.sasoc_asocmaxrxt != 0) {
3051 __u32 path_sum = 0;
3052 int paths = 0;
3053 struct sctp_transport *peer_addr;
3055 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3056 transports) {
3057 path_sum += peer_addr->pathmaxrxt;
3058 paths++;
3061 /* Only validate asocmaxrxt if we have more than
3062 * one path/transport. We do this because path
3063 * retransmissions are only counted when we have more
3064 * then one path.
3066 if (paths > 1 &&
3067 assocparams.sasoc_asocmaxrxt > path_sum)
3068 return -EINVAL;
3070 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3073 if (assocparams.sasoc_cookie_life != 0)
3074 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3075 } else {
3076 /* Set the values to the endpoint */
3077 struct sctp_sock *sp = sctp_sk(sk);
3079 if (assocparams.sasoc_asocmaxrxt != 0)
3080 sp->assocparams.sasoc_asocmaxrxt =
3081 assocparams.sasoc_asocmaxrxt;
3082 if (assocparams.sasoc_cookie_life != 0)
3083 sp->assocparams.sasoc_cookie_life =
3084 assocparams.sasoc_cookie_life;
3086 return 0;
3090 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3092 * This socket option is a boolean flag which turns on or off mapped V4
3093 * addresses. If this option is turned on and the socket is type
3094 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3095 * If this option is turned off, then no mapping will be done of V4
3096 * addresses and a user will receive both PF_INET6 and PF_INET type
3097 * addresses on the socket.
3099 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3101 int val;
3102 struct sctp_sock *sp = sctp_sk(sk);
3104 if (optlen < sizeof(int))
3105 return -EINVAL;
3106 if (get_user(val, (int __user *)optval))
3107 return -EFAULT;
3108 if (val)
3109 sp->v4mapped = 1;
3110 else
3111 sp->v4mapped = 0;
3113 return 0;
3117 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3118 * This option will get or set the maximum size to put in any outgoing
3119 * SCTP DATA chunk. If a message is larger than this size it will be
3120 * fragmented by SCTP into the specified size. Note that the underlying
3121 * SCTP implementation may fragment into smaller sized chunks when the
3122 * PMTU of the underlying association is smaller than the value set by
3123 * the user. The default value for this option is '0' which indicates
3124 * the user is NOT limiting fragmentation and only the PMTU will effect
3125 * SCTP's choice of DATA chunk size. Note also that values set larger
3126 * than the maximum size of an IP datagram will effectively let SCTP
3127 * control fragmentation (i.e. the same as setting this option to 0).
3129 * The following structure is used to access and modify this parameter:
3131 * struct sctp_assoc_value {
3132 * sctp_assoc_t assoc_id;
3133 * uint32_t assoc_value;
3134 * };
3136 * assoc_id: This parameter is ignored for one-to-one style sockets.
3137 * For one-to-many style sockets this parameter indicates which
3138 * association the user is performing an action upon. Note that if
3139 * this field's value is zero then the endpoints default value is
3140 * changed (effecting future associations only).
3141 * assoc_value: This parameter specifies the maximum size in bytes.
3143 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3145 struct sctp_sock *sp = sctp_sk(sk);
3146 struct sctp_assoc_value params;
3147 struct sctp_association *asoc;
3148 int val;
3150 if (optlen == sizeof(int)) {
3151 pr_warn_ratelimited(DEPRECATED
3152 "%s (pid %d) "
3153 "Use of int in maxseg socket option.\n"
3154 "Use struct sctp_assoc_value instead\n",
3155 current->comm, task_pid_nr(current));
3156 if (copy_from_user(&val, optval, optlen))
3157 return -EFAULT;
3158 params.assoc_id = 0;
3159 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3160 if (copy_from_user(&params, optval, optlen))
3161 return -EFAULT;
3162 val = params.assoc_value;
3163 } else {
3164 return -EINVAL;
3167 if (val) {
3168 int min_len, max_len;
3170 min_len = SCTP_DEFAULT_MINSEGMENT - sp->pf->af->net_header_len;
3171 min_len -= sizeof(struct sctphdr) +
3172 sizeof(struct sctp_data_chunk);
3174 max_len = SCTP_MAX_CHUNK_LEN - sizeof(struct sctp_data_chunk);
3176 if (val < min_len || val > max_len)
3177 return -EINVAL;
3180 asoc = sctp_id2assoc(sk, params.assoc_id);
3181 if (asoc) {
3182 if (val == 0) {
3183 val = asoc->pathmtu - sp->pf->af->net_header_len;
3184 val -= sizeof(struct sctphdr) +
3185 sizeof(struct sctp_data_chunk);
3187 asoc->user_frag = val;
3188 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3189 } else {
3190 if (params.assoc_id && sctp_style(sk, UDP))
3191 return -EINVAL;
3192 sp->user_frag = val;
3195 return 0;
3200 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3202 * Requests that the peer mark the enclosed address as the association
3203 * primary. The enclosed address must be one of the association's
3204 * locally bound addresses. The following structure is used to make a
3205 * set primary request:
3207 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3208 unsigned int optlen)
3210 struct net *net = sock_net(sk);
3211 struct sctp_sock *sp;
3212 struct sctp_association *asoc = NULL;
3213 struct sctp_setpeerprim prim;
3214 struct sctp_chunk *chunk;
3215 struct sctp_af *af;
3216 int err;
3218 sp = sctp_sk(sk);
3220 if (!net->sctp.addip_enable)
3221 return -EPERM;
3223 if (optlen != sizeof(struct sctp_setpeerprim))
3224 return -EINVAL;
3226 if (copy_from_user(&prim, optval, optlen))
3227 return -EFAULT;
3229 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3230 if (!asoc)
3231 return -EINVAL;
3233 if (!asoc->peer.asconf_capable)
3234 return -EPERM;
3236 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3237 return -EPERM;
3239 if (!sctp_state(asoc, ESTABLISHED))
3240 return -ENOTCONN;
3242 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3243 if (!af)
3244 return -EINVAL;
3246 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3247 return -EADDRNOTAVAIL;
3249 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3250 return -EADDRNOTAVAIL;
3252 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3253 chunk = sctp_make_asconf_set_prim(asoc,
3254 (union sctp_addr *)&prim.sspp_addr);
3255 if (!chunk)
3256 return -ENOMEM;
3258 err = sctp_send_asconf(asoc, chunk);
3260 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3262 return err;
3265 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3266 unsigned int optlen)
3268 struct sctp_setadaptation adaptation;
3270 if (optlen != sizeof(struct sctp_setadaptation))
3271 return -EINVAL;
3272 if (copy_from_user(&adaptation, optval, optlen))
3273 return -EFAULT;
3275 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3277 return 0;
3281 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3283 * The context field in the sctp_sndrcvinfo structure is normally only
3284 * used when a failed message is retrieved holding the value that was
3285 * sent down on the actual send call. This option allows the setting of
3286 * a default context on an association basis that will be received on
3287 * reading messages from the peer. This is especially helpful in the
3288 * one-2-many model for an application to keep some reference to an
3289 * internal state machine that is processing messages on the
3290 * association. Note that the setting of this value only effects
3291 * received messages from the peer and does not effect the value that is
3292 * saved with outbound messages.
3294 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3295 unsigned int optlen)
3297 struct sctp_assoc_value params;
3298 struct sctp_sock *sp;
3299 struct sctp_association *asoc;
3301 if (optlen != sizeof(struct sctp_assoc_value))
3302 return -EINVAL;
3303 if (copy_from_user(&params, optval, optlen))
3304 return -EFAULT;
3306 sp = sctp_sk(sk);
3308 if (params.assoc_id != 0) {
3309 asoc = sctp_id2assoc(sk, params.assoc_id);
3310 if (!asoc)
3311 return -EINVAL;
3312 asoc->default_rcv_context = params.assoc_value;
3313 } else {
3314 sp->default_rcv_context = params.assoc_value;
3317 return 0;
3321 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3323 * This options will at a minimum specify if the implementation is doing
3324 * fragmented interleave. Fragmented interleave, for a one to many
3325 * socket, is when subsequent calls to receive a message may return
3326 * parts of messages from different associations. Some implementations
3327 * may allow you to turn this value on or off. If so, when turned off,
3328 * no fragment interleave will occur (which will cause a head of line
3329 * blocking amongst multiple associations sharing the same one to many
3330 * socket). When this option is turned on, then each receive call may
3331 * come from a different association (thus the user must receive data
3332 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3333 * association each receive belongs to.
3335 * This option takes a boolean value. A non-zero value indicates that
3336 * fragmented interleave is on. A value of zero indicates that
3337 * fragmented interleave is off.
3339 * Note that it is important that an implementation that allows this
3340 * option to be turned on, have it off by default. Otherwise an unaware
3341 * application using the one to many model may become confused and act
3342 * incorrectly.
3344 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3345 char __user *optval,
3346 unsigned int optlen)
3348 int val;
3350 if (optlen != sizeof(int))
3351 return -EINVAL;
3352 if (get_user(val, (int __user *)optval))
3353 return -EFAULT;
3355 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3357 return 0;
3361 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3362 * (SCTP_PARTIAL_DELIVERY_POINT)
3364 * This option will set or get the SCTP partial delivery point. This
3365 * point is the size of a message where the partial delivery API will be
3366 * invoked to help free up rwnd space for the peer. Setting this to a
3367 * lower value will cause partial deliveries to happen more often. The
3368 * calls argument is an integer that sets or gets the partial delivery
3369 * point. Note also that the call will fail if the user attempts to set
3370 * this value larger than the socket receive buffer size.
3372 * Note that any single message having a length smaller than or equal to
3373 * the SCTP partial delivery point will be delivered in one single read
3374 * call as long as the user provided buffer is large enough to hold the
3375 * message.
3377 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3378 char __user *optval,
3379 unsigned int optlen)
3381 u32 val;
3383 if (optlen != sizeof(u32))
3384 return -EINVAL;
3385 if (get_user(val, (int __user *)optval))
3386 return -EFAULT;
3388 /* Note: We double the receive buffer from what the user sets
3389 * it to be, also initial rwnd is based on rcvbuf/2.
3391 if (val > (sk->sk_rcvbuf >> 1))
3392 return -EINVAL;
3394 sctp_sk(sk)->pd_point = val;
3396 return 0; /* is this the right error code? */
3400 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3402 * This option will allow a user to change the maximum burst of packets
3403 * that can be emitted by this association. Note that the default value
3404 * is 4, and some implementations may restrict this setting so that it
3405 * can only be lowered.
3407 * NOTE: This text doesn't seem right. Do this on a socket basis with
3408 * future associations inheriting the socket value.
3410 static int sctp_setsockopt_maxburst(struct sock *sk,
3411 char __user *optval,
3412 unsigned int optlen)
3414 struct sctp_assoc_value params;
3415 struct sctp_sock *sp;
3416 struct sctp_association *asoc;
3417 int val;
3418 int assoc_id = 0;
3420 if (optlen == sizeof(int)) {
3421 pr_warn_ratelimited(DEPRECATED
3422 "%s (pid %d) "
3423 "Use of int in max_burst socket option deprecated.\n"
3424 "Use struct sctp_assoc_value instead\n",
3425 current->comm, task_pid_nr(current));
3426 if (copy_from_user(&val, optval, optlen))
3427 return -EFAULT;
3428 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3429 if (copy_from_user(&params, optval, optlen))
3430 return -EFAULT;
3431 val = params.assoc_value;
3432 assoc_id = params.assoc_id;
3433 } else
3434 return -EINVAL;
3436 sp = sctp_sk(sk);
3438 if (assoc_id != 0) {
3439 asoc = sctp_id2assoc(sk, assoc_id);
3440 if (!asoc)
3441 return -EINVAL;
3442 asoc->max_burst = val;
3443 } else
3444 sp->max_burst = val;
3446 return 0;
3450 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3452 * This set option adds a chunk type that the user is requesting to be
3453 * received only in an authenticated way. Changes to the list of chunks
3454 * will only effect future associations on the socket.
3456 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3457 char __user *optval,
3458 unsigned int optlen)
3460 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3461 struct sctp_authchunk val;
3463 if (!ep->auth_enable)
3464 return -EACCES;
3466 if (optlen != sizeof(struct sctp_authchunk))
3467 return -EINVAL;
3468 if (copy_from_user(&val, optval, optlen))
3469 return -EFAULT;
3471 switch (val.sauth_chunk) {
3472 case SCTP_CID_INIT:
3473 case SCTP_CID_INIT_ACK:
3474 case SCTP_CID_SHUTDOWN_COMPLETE:
3475 case SCTP_CID_AUTH:
3476 return -EINVAL;
3479 /* add this chunk id to the endpoint */
3480 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3484 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3486 * This option gets or sets the list of HMAC algorithms that the local
3487 * endpoint requires the peer to use.
3489 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3490 char __user *optval,
3491 unsigned int optlen)
3493 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3494 struct sctp_hmacalgo *hmacs;
3495 u32 idents;
3496 int err;
3498 if (!ep->auth_enable)
3499 return -EACCES;
3501 if (optlen < sizeof(struct sctp_hmacalgo))
3502 return -EINVAL;
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;
3542 authkey = memdup_user(optval, optlen);
3543 if (IS_ERR(authkey))
3544 return PTR_ERR(authkey);
3546 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3547 ret = -EINVAL;
3548 goto out;
3551 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3552 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3553 ret = -EINVAL;
3554 goto out;
3557 ret = sctp_auth_set_key(ep, asoc, authkey);
3558 out:
3559 kzfree(authkey);
3560 return ret;
3564 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3566 * This option will get or set the active shared key to be used to build
3567 * the association shared key.
3569 static int sctp_setsockopt_active_key(struct sock *sk,
3570 char __user *optval,
3571 unsigned int optlen)
3573 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3574 struct sctp_authkeyid val;
3575 struct sctp_association *asoc;
3577 if (!ep->auth_enable)
3578 return -EACCES;
3580 if (optlen != sizeof(struct sctp_authkeyid))
3581 return -EINVAL;
3582 if (copy_from_user(&val, optval, optlen))
3583 return -EFAULT;
3585 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3586 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3587 return -EINVAL;
3589 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3593 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3595 * This set option will delete a shared secret key from use.
3597 static int sctp_setsockopt_del_key(struct sock *sk,
3598 char __user *optval,
3599 unsigned int optlen)
3601 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3602 struct sctp_authkeyid val;
3603 struct sctp_association *asoc;
3605 if (!ep->auth_enable)
3606 return -EACCES;
3608 if (optlen != sizeof(struct sctp_authkeyid))
3609 return -EINVAL;
3610 if (copy_from_user(&val, optval, optlen))
3611 return -EFAULT;
3613 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3614 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3615 return -EINVAL;
3617 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3622 * 8.1.23 SCTP_AUTO_ASCONF
3624 * This option will enable or disable the use of the automatic generation of
3625 * ASCONF chunks to add and delete addresses to an existing association. Note
3626 * that this option has two caveats namely: a) it only affects sockets that
3627 * are bound to all addresses available to the SCTP stack, and b) the system
3628 * administrator may have an overriding control that turns the ASCONF feature
3629 * off no matter what setting the socket option may have.
3630 * This option expects an integer boolean flag, where a non-zero value turns on
3631 * the option, and a zero value turns off the option.
3632 * Note. In this implementation, socket operation overrides default parameter
3633 * being set by sysctl as well as FreeBSD implementation
3635 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3636 unsigned int optlen)
3638 int val;
3639 struct sctp_sock *sp = sctp_sk(sk);
3641 if (optlen < sizeof(int))
3642 return -EINVAL;
3643 if (get_user(val, (int __user *)optval))
3644 return -EFAULT;
3645 if (!sctp_is_ep_boundall(sk) && val)
3646 return -EINVAL;
3647 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3648 return 0;
3650 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3651 if (val == 0 && sp->do_auto_asconf) {
3652 list_del(&sp->auto_asconf_list);
3653 sp->do_auto_asconf = 0;
3654 } else if (val && !sp->do_auto_asconf) {
3655 list_add_tail(&sp->auto_asconf_list,
3656 &sock_net(sk)->sctp.auto_asconf_splist);
3657 sp->do_auto_asconf = 1;
3659 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3660 return 0;
3664 * SCTP_PEER_ADDR_THLDS
3666 * This option allows us to alter the partially failed threshold for one or all
3667 * transports in an association. See Section 6.1 of:
3668 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3670 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3671 char __user *optval,
3672 unsigned int optlen)
3674 struct sctp_paddrthlds val;
3675 struct sctp_transport *trans;
3676 struct sctp_association *asoc;
3678 if (optlen < sizeof(struct sctp_paddrthlds))
3679 return -EINVAL;
3680 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3681 sizeof(struct sctp_paddrthlds)))
3682 return -EFAULT;
3685 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3686 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3687 if (!asoc)
3688 return -ENOENT;
3689 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3690 transports) {
3691 if (val.spt_pathmaxrxt)
3692 trans->pathmaxrxt = val.spt_pathmaxrxt;
3693 trans->pf_retrans = val.spt_pathpfthld;
3696 if (val.spt_pathmaxrxt)
3697 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3698 asoc->pf_retrans = val.spt_pathpfthld;
3699 } else {
3700 trans = sctp_addr_id2transport(sk, &val.spt_address,
3701 val.spt_assoc_id);
3702 if (!trans)
3703 return -ENOENT;
3705 if (val.spt_pathmaxrxt)
3706 trans->pathmaxrxt = val.spt_pathmaxrxt;
3707 trans->pf_retrans = val.spt_pathpfthld;
3710 return 0;
3713 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3714 char __user *optval,
3715 unsigned int optlen)
3717 int val;
3719 if (optlen < sizeof(int))
3720 return -EINVAL;
3721 if (get_user(val, (int __user *) optval))
3722 return -EFAULT;
3724 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3726 return 0;
3729 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3730 char __user *optval,
3731 unsigned int optlen)
3733 int val;
3735 if (optlen < sizeof(int))
3736 return -EINVAL;
3737 if (get_user(val, (int __user *) optval))
3738 return -EFAULT;
3740 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3742 return 0;
3745 static int sctp_setsockopt_pr_supported(struct sock *sk,
3746 char __user *optval,
3747 unsigned int optlen)
3749 struct sctp_assoc_value params;
3751 if (optlen != sizeof(params))
3752 return -EINVAL;
3754 if (copy_from_user(&params, optval, optlen))
3755 return -EFAULT;
3757 sctp_sk(sk)->ep->prsctp_enable = !!params.assoc_value;
3759 return 0;
3762 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3763 char __user *optval,
3764 unsigned int optlen)
3766 struct sctp_default_prinfo info;
3767 struct sctp_association *asoc;
3768 int retval = -EINVAL;
3770 if (optlen != sizeof(info))
3771 goto out;
3773 if (copy_from_user(&info, optval, sizeof(info))) {
3774 retval = -EFAULT;
3775 goto out;
3778 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3779 goto out;
3781 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3782 info.pr_value = 0;
3784 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3785 if (asoc) {
3786 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3787 asoc->default_timetolive = info.pr_value;
3788 } else if (!info.pr_assoc_id) {
3789 struct sctp_sock *sp = sctp_sk(sk);
3791 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3792 sp->default_timetolive = info.pr_value;
3793 } else {
3794 goto out;
3797 retval = 0;
3799 out:
3800 return retval;
3803 /* API 6.2 setsockopt(), getsockopt()
3805 * Applications use setsockopt() and getsockopt() to set or retrieve
3806 * socket options. Socket options are used to change the default
3807 * behavior of sockets calls. They are described in Section 7.
3809 * The syntax is:
3811 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3812 * int __user *optlen);
3813 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3814 * int optlen);
3816 * sd - the socket descript.
3817 * level - set to IPPROTO_SCTP for all SCTP options.
3818 * optname - the option name.
3819 * optval - the buffer to store the value of the option.
3820 * optlen - the size of the buffer.
3822 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3823 char __user *optval, unsigned int optlen)
3825 int retval = 0;
3827 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3829 /* I can hardly begin to describe how wrong this is. This is
3830 * so broken as to be worse than useless. The API draft
3831 * REALLY is NOT helpful here... I am not convinced that the
3832 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3833 * are at all well-founded.
3835 if (level != SOL_SCTP) {
3836 struct sctp_af *af = sctp_sk(sk)->pf->af;
3837 retval = af->setsockopt(sk, level, optname, optval, optlen);
3838 goto out_nounlock;
3841 lock_sock(sk);
3843 switch (optname) {
3844 case SCTP_SOCKOPT_BINDX_ADD:
3845 /* 'optlen' is the size of the addresses buffer. */
3846 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3847 optlen, SCTP_BINDX_ADD_ADDR);
3848 break;
3850 case SCTP_SOCKOPT_BINDX_REM:
3851 /* 'optlen' is the size of the addresses buffer. */
3852 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3853 optlen, SCTP_BINDX_REM_ADDR);
3854 break;
3856 case SCTP_SOCKOPT_CONNECTX_OLD:
3857 /* 'optlen' is the size of the addresses buffer. */
3858 retval = sctp_setsockopt_connectx_old(sk,
3859 (struct sockaddr __user *)optval,
3860 optlen);
3861 break;
3863 case SCTP_SOCKOPT_CONNECTX:
3864 /* 'optlen' is the size of the addresses buffer. */
3865 retval = sctp_setsockopt_connectx(sk,
3866 (struct sockaddr __user *)optval,
3867 optlen);
3868 break;
3870 case SCTP_DISABLE_FRAGMENTS:
3871 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3872 break;
3874 case SCTP_EVENTS:
3875 retval = sctp_setsockopt_events(sk, optval, optlen);
3876 break;
3878 case SCTP_AUTOCLOSE:
3879 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3880 break;
3882 case SCTP_PEER_ADDR_PARAMS:
3883 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3884 break;
3886 case SCTP_DELAYED_SACK:
3887 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3888 break;
3889 case SCTP_PARTIAL_DELIVERY_POINT:
3890 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3891 break;
3893 case SCTP_INITMSG:
3894 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3895 break;
3896 case SCTP_DEFAULT_SEND_PARAM:
3897 retval = sctp_setsockopt_default_send_param(sk, optval,
3898 optlen);
3899 break;
3900 case SCTP_DEFAULT_SNDINFO:
3901 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
3902 break;
3903 case SCTP_PRIMARY_ADDR:
3904 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3905 break;
3906 case SCTP_SET_PEER_PRIMARY_ADDR:
3907 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3908 break;
3909 case SCTP_NODELAY:
3910 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3911 break;
3912 case SCTP_RTOINFO:
3913 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3914 break;
3915 case SCTP_ASSOCINFO:
3916 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3917 break;
3918 case SCTP_I_WANT_MAPPED_V4_ADDR:
3919 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3920 break;
3921 case SCTP_MAXSEG:
3922 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3923 break;
3924 case SCTP_ADAPTATION_LAYER:
3925 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3926 break;
3927 case SCTP_CONTEXT:
3928 retval = sctp_setsockopt_context(sk, optval, optlen);
3929 break;
3930 case SCTP_FRAGMENT_INTERLEAVE:
3931 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3932 break;
3933 case SCTP_MAX_BURST:
3934 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3935 break;
3936 case SCTP_AUTH_CHUNK:
3937 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3938 break;
3939 case SCTP_HMAC_IDENT:
3940 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3941 break;
3942 case SCTP_AUTH_KEY:
3943 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3944 break;
3945 case SCTP_AUTH_ACTIVE_KEY:
3946 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3947 break;
3948 case SCTP_AUTH_DELETE_KEY:
3949 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3950 break;
3951 case SCTP_AUTO_ASCONF:
3952 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3953 break;
3954 case SCTP_PEER_ADDR_THLDS:
3955 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3956 break;
3957 case SCTP_RECVRCVINFO:
3958 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
3959 break;
3960 case SCTP_RECVNXTINFO:
3961 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
3962 break;
3963 case SCTP_PR_SUPPORTED:
3964 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
3965 break;
3966 case SCTP_DEFAULT_PRINFO:
3967 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
3968 break;
3969 default:
3970 retval = -ENOPROTOOPT;
3971 break;
3974 release_sock(sk);
3976 out_nounlock:
3977 return retval;
3980 /* API 3.1.6 connect() - UDP Style Syntax
3982 * An application may use the connect() call in the UDP model to initiate an
3983 * association without sending data.
3985 * The syntax is:
3987 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3989 * sd: the socket descriptor to have a new association added to.
3991 * nam: the address structure (either struct sockaddr_in or struct
3992 * sockaddr_in6 defined in RFC2553 [7]).
3994 * len: the size of the address.
3996 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3997 int addr_len, int flags)
3999 struct sctp_af *af;
4000 int err = -EINVAL;
4002 lock_sock(sk);
4003 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4004 addr, addr_len);
4006 /* Validate addr_len before calling common connect/connectx routine. */
4007 af = sctp_get_af_specific(addr->sa_family);
4008 if (af && addr_len >= af->sockaddr_len)
4009 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4011 release_sock(sk);
4012 return err;
4015 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4016 int addr_len, int flags)
4018 if (addr_len < sizeof(uaddr->sa_family))
4019 return -EINVAL;
4021 if (uaddr->sa_family == AF_UNSPEC)
4022 return -EOPNOTSUPP;
4024 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4027 /* FIXME: Write comments. */
4028 static int sctp_disconnect(struct sock *sk, int flags)
4030 return -EOPNOTSUPP; /* STUB */
4033 /* 4.1.4 accept() - TCP Style Syntax
4035 * Applications use accept() call to remove an established SCTP
4036 * association from the accept queue of the endpoint. A new socket
4037 * descriptor will be returned from accept() to represent the newly
4038 * formed association.
4040 static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
4042 struct sctp_sock *sp;
4043 struct sctp_endpoint *ep;
4044 struct sock *newsk = NULL;
4045 struct sctp_association *asoc;
4046 long timeo;
4047 int error = 0;
4049 lock_sock(sk);
4051 sp = sctp_sk(sk);
4052 ep = sp->ep;
4054 if (!sctp_style(sk, TCP)) {
4055 error = -EOPNOTSUPP;
4056 goto out;
4059 if (!sctp_sstate(sk, LISTENING)) {
4060 error = -EINVAL;
4061 goto out;
4064 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4066 error = sctp_wait_for_accept(sk, timeo);
4067 if (error)
4068 goto out;
4070 /* We treat the list of associations on the endpoint as the accept
4071 * queue and pick the first association on the list.
4073 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4075 newsk = sp->pf->create_accept_sk(sk, asoc);
4076 if (!newsk) {
4077 error = -ENOMEM;
4078 goto out;
4081 /* Populate the fields of the newsk from the oldsk and migrate the
4082 * asoc to the newsk.
4084 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4086 out:
4087 release_sock(sk);
4088 *err = error;
4089 return newsk;
4092 /* The SCTP ioctl handler. */
4093 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4095 int rc = -ENOTCONN;
4097 lock_sock(sk);
4100 * SEQPACKET-style sockets in LISTENING state are valid, for
4101 * SCTP, so only discard TCP-style sockets in LISTENING state.
4103 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4104 goto out;
4106 switch (cmd) {
4107 case SIOCINQ: {
4108 struct sk_buff *skb;
4109 unsigned int amount = 0;
4111 skb = skb_peek(&sk->sk_receive_queue);
4112 if (skb != NULL) {
4114 * We will only return the amount of this packet since
4115 * that is all that will be read.
4117 amount = skb->len;
4119 rc = put_user(amount, (int __user *)arg);
4120 break;
4122 default:
4123 rc = -ENOIOCTLCMD;
4124 break;
4126 out:
4127 release_sock(sk);
4128 return rc;
4131 /* This is the function which gets called during socket creation to
4132 * initialized the SCTP-specific portion of the sock.
4133 * The sock structure should already be zero-filled memory.
4135 static int sctp_init_sock(struct sock *sk)
4137 struct net *net = sock_net(sk);
4138 struct sctp_sock *sp;
4140 pr_debug("%s: sk:%p\n", __func__, sk);
4142 sp = sctp_sk(sk);
4144 /* Initialize the SCTP per socket area. */
4145 switch (sk->sk_type) {
4146 case SOCK_SEQPACKET:
4147 sp->type = SCTP_SOCKET_UDP;
4148 break;
4149 case SOCK_STREAM:
4150 sp->type = SCTP_SOCKET_TCP;
4151 break;
4152 default:
4153 return -ESOCKTNOSUPPORT;
4156 sk->sk_gso_type = SKB_GSO_SCTP;
4158 /* Initialize default send parameters. These parameters can be
4159 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4161 sp->default_stream = 0;
4162 sp->default_ppid = 0;
4163 sp->default_flags = 0;
4164 sp->default_context = 0;
4165 sp->default_timetolive = 0;
4167 sp->default_rcv_context = 0;
4168 sp->max_burst = net->sctp.max_burst;
4170 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4172 /* Initialize default setup parameters. These parameters
4173 * can be modified with the SCTP_INITMSG socket option or
4174 * overridden by the SCTP_INIT CMSG.
4176 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4177 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4178 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4179 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4181 /* Initialize default RTO related parameters. These parameters can
4182 * be modified for with the SCTP_RTOINFO socket option.
4184 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4185 sp->rtoinfo.srto_max = net->sctp.rto_max;
4186 sp->rtoinfo.srto_min = net->sctp.rto_min;
4188 /* Initialize default association related parameters. These parameters
4189 * can be modified with the SCTP_ASSOCINFO socket option.
4191 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4192 sp->assocparams.sasoc_number_peer_destinations = 0;
4193 sp->assocparams.sasoc_peer_rwnd = 0;
4194 sp->assocparams.sasoc_local_rwnd = 0;
4195 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4197 /* Initialize default event subscriptions. By default, all the
4198 * options are off.
4200 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4202 /* Default Peer Address Parameters. These defaults can
4203 * be modified via SCTP_PEER_ADDR_PARAMS
4205 sp->hbinterval = net->sctp.hb_interval;
4206 sp->pathmaxrxt = net->sctp.max_retrans_path;
4207 sp->pathmtu = 0; /* allow default discovery */
4208 sp->sackdelay = net->sctp.sack_timeout;
4209 sp->sackfreq = 2;
4210 sp->param_flags = SPP_HB_ENABLE |
4211 SPP_PMTUD_ENABLE |
4212 SPP_SACKDELAY_ENABLE;
4214 /* If enabled no SCTP message fragmentation will be performed.
4215 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4217 sp->disable_fragments = 0;
4219 /* Enable Nagle algorithm by default. */
4220 sp->nodelay = 0;
4222 sp->recvrcvinfo = 0;
4223 sp->recvnxtinfo = 0;
4225 /* Enable by default. */
4226 sp->v4mapped = 1;
4228 /* Auto-close idle associations after the configured
4229 * number of seconds. A value of 0 disables this
4230 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4231 * for UDP-style sockets only.
4233 sp->autoclose = 0;
4235 /* User specified fragmentation limit. */
4236 sp->user_frag = 0;
4238 sp->adaptation_ind = 0;
4240 sp->pf = sctp_get_pf_specific(sk->sk_family);
4242 /* Control variables for partial data delivery. */
4243 atomic_set(&sp->pd_mode, 0);
4244 skb_queue_head_init(&sp->pd_lobby);
4245 sp->frag_interleave = 0;
4247 /* Create a per socket endpoint structure. Even if we
4248 * change the data structure relationships, this may still
4249 * be useful for storing pre-connect address information.
4251 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4252 if (!sp->ep)
4253 return -ENOMEM;
4255 sp->hmac = NULL;
4257 sk->sk_destruct = sctp_destruct_sock;
4259 SCTP_DBG_OBJCNT_INC(sock);
4261 local_bh_disable();
4262 sk_sockets_allocated_inc(sk);
4263 sock_prot_inuse_add(net, sk->sk_prot, 1);
4265 /* Nothing can fail after this block, otherwise
4266 * sctp_destroy_sock() will be called without addr_wq_lock held
4268 if (net->sctp.default_auto_asconf) {
4269 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4270 list_add_tail(&sp->auto_asconf_list,
4271 &net->sctp.auto_asconf_splist);
4272 sp->do_auto_asconf = 1;
4273 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4274 } else {
4275 sp->do_auto_asconf = 0;
4278 local_bh_enable();
4280 return 0;
4283 /* Cleanup any SCTP per socket resources. Must be called with
4284 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4286 static void sctp_destroy_sock(struct sock *sk)
4288 struct sctp_sock *sp;
4290 pr_debug("%s: sk:%p\n", __func__, sk);
4292 /* Release our hold on the endpoint. */
4293 sp = sctp_sk(sk);
4294 /* This could happen during socket init, thus we bail out
4295 * early, since the rest of the below is not setup either.
4297 if (sp->ep == NULL)
4298 return;
4300 if (sp->do_auto_asconf) {
4301 sp->do_auto_asconf = 0;
4302 list_del(&sp->auto_asconf_list);
4304 sctp_endpoint_free(sp->ep);
4305 local_bh_disable();
4306 sk_sockets_allocated_dec(sk);
4307 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4308 local_bh_enable();
4311 /* Triggered when there are no references on the socket anymore */
4312 static void sctp_destruct_sock(struct sock *sk)
4314 struct sctp_sock *sp = sctp_sk(sk);
4316 /* Free up the HMAC transform. */
4317 crypto_free_shash(sp->hmac);
4319 inet_sock_destruct(sk);
4322 /* API 4.1.7 shutdown() - TCP Style Syntax
4323 * int shutdown(int socket, int how);
4325 * sd - the socket descriptor of the association to be closed.
4326 * how - Specifies the type of shutdown. The values are
4327 * as follows:
4328 * SHUT_RD
4329 * Disables further receive operations. No SCTP
4330 * protocol action is taken.
4331 * SHUT_WR
4332 * Disables further send operations, and initiates
4333 * the SCTP shutdown sequence.
4334 * SHUT_RDWR
4335 * Disables further send and receive operations
4336 * and initiates the SCTP shutdown sequence.
4338 static void sctp_shutdown(struct sock *sk, int how)
4340 struct net *net = sock_net(sk);
4341 struct sctp_endpoint *ep;
4343 if (!sctp_style(sk, TCP))
4344 return;
4346 ep = sctp_sk(sk)->ep;
4347 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4348 struct sctp_association *asoc;
4350 sk->sk_state = SCTP_SS_CLOSING;
4351 asoc = list_entry(ep->asocs.next,
4352 struct sctp_association, asocs);
4353 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4357 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4358 struct sctp_info *info)
4360 struct sctp_transport *prim;
4361 struct list_head *pos;
4362 int mask;
4364 memset(info, 0, sizeof(*info));
4365 if (!asoc) {
4366 struct sctp_sock *sp = sctp_sk(sk);
4368 info->sctpi_s_autoclose = sp->autoclose;
4369 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4370 info->sctpi_s_pd_point = sp->pd_point;
4371 info->sctpi_s_nodelay = sp->nodelay;
4372 info->sctpi_s_disable_fragments = sp->disable_fragments;
4373 info->sctpi_s_v4mapped = sp->v4mapped;
4374 info->sctpi_s_frag_interleave = sp->frag_interleave;
4375 info->sctpi_s_type = sp->type;
4377 return 0;
4380 info->sctpi_tag = asoc->c.my_vtag;
4381 info->sctpi_state = asoc->state;
4382 info->sctpi_rwnd = asoc->a_rwnd;
4383 info->sctpi_unackdata = asoc->unack_data;
4384 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4385 info->sctpi_instrms = asoc->c.sinit_max_instreams;
4386 info->sctpi_outstrms = asoc->c.sinit_num_ostreams;
4387 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4388 info->sctpi_inqueue++;
4389 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4390 info->sctpi_outqueue++;
4391 info->sctpi_overall_error = asoc->overall_error_count;
4392 info->sctpi_max_burst = asoc->max_burst;
4393 info->sctpi_maxseg = asoc->frag_point;
4394 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4395 info->sctpi_peer_tag = asoc->c.peer_vtag;
4397 mask = asoc->peer.ecn_capable << 1;
4398 mask = (mask | asoc->peer.ipv4_address) << 1;
4399 mask = (mask | asoc->peer.ipv6_address) << 1;
4400 mask = (mask | asoc->peer.hostname_address) << 1;
4401 mask = (mask | asoc->peer.asconf_capable) << 1;
4402 mask = (mask | asoc->peer.prsctp_capable) << 1;
4403 mask = (mask | asoc->peer.auth_capable);
4404 info->sctpi_peer_capable = mask;
4405 mask = asoc->peer.sack_needed << 1;
4406 mask = (mask | asoc->peer.sack_generation) << 1;
4407 mask = (mask | asoc->peer.zero_window_announced);
4408 info->sctpi_peer_sack = mask;
4410 info->sctpi_isacks = asoc->stats.isacks;
4411 info->sctpi_osacks = asoc->stats.osacks;
4412 info->sctpi_opackets = asoc->stats.opackets;
4413 info->sctpi_ipackets = asoc->stats.ipackets;
4414 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4415 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4416 info->sctpi_idupchunks = asoc->stats.idupchunks;
4417 info->sctpi_gapcnt = asoc->stats.gapcnt;
4418 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4419 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4420 info->sctpi_oodchunks = asoc->stats.oodchunks;
4421 info->sctpi_iodchunks = asoc->stats.iodchunks;
4422 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4423 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4425 prim = asoc->peer.primary_path;
4426 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4427 info->sctpi_p_state = prim->state;
4428 info->sctpi_p_cwnd = prim->cwnd;
4429 info->sctpi_p_srtt = prim->srtt;
4430 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4431 info->sctpi_p_hbinterval = prim->hbinterval;
4432 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4433 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4434 info->sctpi_p_ssthresh = prim->ssthresh;
4435 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4436 info->sctpi_p_flight_size = prim->flight_size;
4437 info->sctpi_p_error = prim->error_count;
4439 return 0;
4441 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4443 /* use callback to avoid exporting the core structure */
4444 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4446 int err;
4448 err = rhashtable_walk_init(&sctp_transport_hashtable, iter,
4449 GFP_KERNEL);
4450 if (err)
4451 return err;
4453 err = rhashtable_walk_start(iter);
4454 if (err && err != -EAGAIN) {
4455 rhashtable_walk_stop(iter);
4456 rhashtable_walk_exit(iter);
4457 return err;
4460 return 0;
4463 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4465 rhashtable_walk_stop(iter);
4466 rhashtable_walk_exit(iter);
4469 struct sctp_transport *sctp_transport_get_next(struct net *net,
4470 struct rhashtable_iter *iter)
4472 struct sctp_transport *t;
4474 t = rhashtable_walk_next(iter);
4475 for (; t; t = rhashtable_walk_next(iter)) {
4476 if (IS_ERR(t)) {
4477 if (PTR_ERR(t) == -EAGAIN)
4478 continue;
4479 break;
4482 if (!sctp_transport_hold(t))
4483 continue;
4485 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4486 t->asoc->peer.primary_path == t)
4487 break;
4489 sctp_transport_put(t);
4492 return t;
4495 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4496 struct rhashtable_iter *iter,
4497 int pos)
4499 struct sctp_transport *t;
4501 if (!pos)
4502 return SEQ_START_TOKEN;
4504 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
4505 if (!--pos)
4506 break;
4507 sctp_transport_put(t);
4510 return t;
4513 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4514 void *p) {
4515 int err = 0;
4516 int hash = 0;
4517 struct sctp_ep_common *epb;
4518 struct sctp_hashbucket *head;
4520 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4521 hash++, head++) {
4522 read_lock_bh(&head->lock);
4523 sctp_for_each_hentry(epb, &head->chain) {
4524 err = cb(sctp_ep(epb), p);
4525 if (err)
4526 break;
4528 read_unlock_bh(&head->lock);
4531 return err;
4533 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4535 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4536 struct net *net,
4537 const union sctp_addr *laddr,
4538 const union sctp_addr *paddr, void *p)
4540 struct sctp_transport *transport;
4541 int err = -ENOENT;
4543 rcu_read_lock();
4544 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4545 if (!transport || !sctp_transport_hold(transport)) {
4546 rcu_read_unlock();
4547 goto out;
4549 rcu_read_unlock();
4550 err = cb(transport, p);
4551 sctp_transport_put(transport);
4553 out:
4554 return err;
4556 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4558 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4559 struct net *net, int pos, void *p) {
4560 struct rhashtable_iter hti;
4561 void *obj;
4562 int err;
4564 err = sctp_transport_walk_start(&hti);
4565 if (err)
4566 return err;
4568 obj = sctp_transport_get_idx(net, &hti, pos + 1);
4569 for (; !IS_ERR_OR_NULL(obj); obj = sctp_transport_get_next(net, &hti)) {
4570 struct sctp_transport *transport = obj;
4572 err = cb(transport, p);
4573 sctp_transport_put(transport);
4574 if (err)
4575 break;
4577 sctp_transport_walk_stop(&hti);
4579 return err;
4581 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4583 /* 7.2.1 Association Status (SCTP_STATUS)
4585 * Applications can retrieve current status information about an
4586 * association, including association state, peer receiver window size,
4587 * number of unacked data chunks, and number of data chunks pending
4588 * receipt. This information is read-only.
4590 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4591 char __user *optval,
4592 int __user *optlen)
4594 struct sctp_status status;
4595 struct sctp_association *asoc = NULL;
4596 struct sctp_transport *transport;
4597 sctp_assoc_t associd;
4598 int retval = 0;
4600 if (len < sizeof(status)) {
4601 retval = -EINVAL;
4602 goto out;
4605 len = sizeof(status);
4606 if (copy_from_user(&status, optval, len)) {
4607 retval = -EFAULT;
4608 goto out;
4611 associd = status.sstat_assoc_id;
4612 asoc = sctp_id2assoc(sk, associd);
4613 if (!asoc) {
4614 retval = -EINVAL;
4615 goto out;
4618 transport = asoc->peer.primary_path;
4620 status.sstat_assoc_id = sctp_assoc2id(asoc);
4621 status.sstat_state = sctp_assoc_to_state(asoc);
4622 status.sstat_rwnd = asoc->peer.rwnd;
4623 status.sstat_unackdata = asoc->unack_data;
4625 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4626 status.sstat_instrms = asoc->c.sinit_max_instreams;
4627 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4628 status.sstat_fragmentation_point = asoc->frag_point;
4629 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4630 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4631 transport->af_specific->sockaddr_len);
4632 /* Map ipv4 address into v4-mapped-on-v6 address. */
4633 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4634 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4635 status.sstat_primary.spinfo_state = transport->state;
4636 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4637 status.sstat_primary.spinfo_srtt = transport->srtt;
4638 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4639 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4641 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4642 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4644 if (put_user(len, optlen)) {
4645 retval = -EFAULT;
4646 goto out;
4649 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4650 __func__, len, status.sstat_state, status.sstat_rwnd,
4651 status.sstat_assoc_id);
4653 if (copy_to_user(optval, &status, len)) {
4654 retval = -EFAULT;
4655 goto out;
4658 out:
4659 return retval;
4663 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4665 * Applications can retrieve information about a specific peer address
4666 * of an association, including its reachability state, congestion
4667 * window, and retransmission timer values. This information is
4668 * read-only.
4670 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4671 char __user *optval,
4672 int __user *optlen)
4674 struct sctp_paddrinfo pinfo;
4675 struct sctp_transport *transport;
4676 int retval = 0;
4678 if (len < sizeof(pinfo)) {
4679 retval = -EINVAL;
4680 goto out;
4683 len = sizeof(pinfo);
4684 if (copy_from_user(&pinfo, optval, len)) {
4685 retval = -EFAULT;
4686 goto out;
4689 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4690 pinfo.spinfo_assoc_id);
4691 if (!transport)
4692 return -EINVAL;
4694 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4695 pinfo.spinfo_state = transport->state;
4696 pinfo.spinfo_cwnd = transport->cwnd;
4697 pinfo.spinfo_srtt = transport->srtt;
4698 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4699 pinfo.spinfo_mtu = transport->pathmtu;
4701 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4702 pinfo.spinfo_state = SCTP_ACTIVE;
4704 if (put_user(len, optlen)) {
4705 retval = -EFAULT;
4706 goto out;
4709 if (copy_to_user(optval, &pinfo, len)) {
4710 retval = -EFAULT;
4711 goto out;
4714 out:
4715 return retval;
4718 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4720 * This option is a on/off flag. If enabled no SCTP message
4721 * fragmentation will be performed. Instead if a message being sent
4722 * exceeds the current PMTU size, the message will NOT be sent and
4723 * instead a error will be indicated to the user.
4725 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4726 char __user *optval, int __user *optlen)
4728 int val;
4730 if (len < sizeof(int))
4731 return -EINVAL;
4733 len = sizeof(int);
4734 val = (sctp_sk(sk)->disable_fragments == 1);
4735 if (put_user(len, optlen))
4736 return -EFAULT;
4737 if (copy_to_user(optval, &val, len))
4738 return -EFAULT;
4739 return 0;
4742 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4744 * This socket option is used to specify various notifications and
4745 * ancillary data the user wishes to receive.
4747 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4748 int __user *optlen)
4750 if (len == 0)
4751 return -EINVAL;
4752 if (len > sizeof(struct sctp_event_subscribe))
4753 len = sizeof(struct sctp_event_subscribe);
4754 if (put_user(len, optlen))
4755 return -EFAULT;
4756 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4757 return -EFAULT;
4758 return 0;
4761 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4763 * This socket option is applicable to the UDP-style socket only. When
4764 * set it will cause associations that are idle for more than the
4765 * specified number of seconds to automatically close. An association
4766 * being idle is defined an association that has NOT sent or received
4767 * user data. The special value of '0' indicates that no automatic
4768 * close of any associations should be performed. The option expects an
4769 * integer defining the number of seconds of idle time before an
4770 * association is closed.
4772 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4774 /* Applicable to UDP-style socket only */
4775 if (sctp_style(sk, TCP))
4776 return -EOPNOTSUPP;
4777 if (len < sizeof(int))
4778 return -EINVAL;
4779 len = sizeof(int);
4780 if (put_user(len, optlen))
4781 return -EFAULT;
4782 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
4783 return -EFAULT;
4784 return 0;
4787 /* Helper routine to branch off an association to a new socket. */
4788 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4790 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4791 struct sctp_sock *sp = sctp_sk(sk);
4792 struct socket *sock;
4793 int err = 0;
4795 /* Do not peel off from one netns to another one. */
4796 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
4797 return -EINVAL;
4799 if (!asoc)
4800 return -EINVAL;
4802 /* An association cannot be branched off from an already peeled-off
4803 * socket, nor is this supported for tcp style sockets.
4805 if (!sctp_style(sk, UDP))
4806 return -EINVAL;
4808 /* Create a new socket. */
4809 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4810 if (err < 0)
4811 return err;
4813 sctp_copy_sock(sock->sk, sk, asoc);
4815 /* Make peeled-off sockets more like 1-1 accepted sockets.
4816 * Set the daddr and initialize id to something more random
4818 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4820 /* Populate the fields of the newsk from the oldsk and migrate the
4821 * asoc to the newsk.
4823 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4825 *sockp = sock;
4827 return err;
4829 EXPORT_SYMBOL(sctp_do_peeloff);
4831 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4833 sctp_peeloff_arg_t peeloff;
4834 struct socket *newsock;
4835 struct file *newfile;
4836 int retval = 0;
4838 if (len < sizeof(sctp_peeloff_arg_t))
4839 return -EINVAL;
4840 len = sizeof(sctp_peeloff_arg_t);
4841 if (copy_from_user(&peeloff, optval, len))
4842 return -EFAULT;
4844 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4845 if (retval < 0)
4846 goto out;
4848 /* Map the socket to an unused fd that can be returned to the user. */
4849 retval = get_unused_fd_flags(0);
4850 if (retval < 0) {
4851 sock_release(newsock);
4852 goto out;
4855 newfile = sock_alloc_file(newsock, 0, NULL);
4856 if (IS_ERR(newfile)) {
4857 put_unused_fd(retval);
4858 sock_release(newsock);
4859 return PTR_ERR(newfile);
4862 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4863 retval);
4865 /* Return the fd mapped to the new socket. */
4866 if (put_user(len, optlen)) {
4867 fput(newfile);
4868 put_unused_fd(retval);
4869 return -EFAULT;
4871 peeloff.sd = retval;
4872 if (copy_to_user(optval, &peeloff, len)) {
4873 fput(newfile);
4874 put_unused_fd(retval);
4875 return -EFAULT;
4877 fd_install(retval, newfile);
4878 out:
4879 return retval;
4882 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4884 * Applications can enable or disable heartbeats for any peer address of
4885 * an association, modify an address's heartbeat interval, force a
4886 * heartbeat to be sent immediately, and adjust the address's maximum
4887 * number of retransmissions sent before an address is considered
4888 * unreachable. The following structure is used to access and modify an
4889 * address's parameters:
4891 * struct sctp_paddrparams {
4892 * sctp_assoc_t spp_assoc_id;
4893 * struct sockaddr_storage spp_address;
4894 * uint32_t spp_hbinterval;
4895 * uint16_t spp_pathmaxrxt;
4896 * uint32_t spp_pathmtu;
4897 * uint32_t spp_sackdelay;
4898 * uint32_t spp_flags;
4899 * };
4901 * spp_assoc_id - (one-to-many style socket) This is filled in the
4902 * application, and identifies the association for
4903 * this query.
4904 * spp_address - This specifies which address is of interest.
4905 * spp_hbinterval - This contains the value of the heartbeat interval,
4906 * in milliseconds. If a value of zero
4907 * is present in this field then no changes are to
4908 * be made to this parameter.
4909 * spp_pathmaxrxt - This contains the maximum number of
4910 * retransmissions before this address shall be
4911 * considered unreachable. If a value of zero
4912 * is present in this field then no changes are to
4913 * be made to this parameter.
4914 * spp_pathmtu - When Path MTU discovery is disabled the value
4915 * specified here will be the "fixed" path mtu.
4916 * Note that if the spp_address field is empty
4917 * then all associations on this address will
4918 * have this fixed path mtu set upon them.
4920 * spp_sackdelay - When delayed sack is enabled, this value specifies
4921 * the number of milliseconds that sacks will be delayed
4922 * for. This value will apply to all addresses of an
4923 * association if the spp_address field is empty. Note
4924 * also, that if delayed sack is enabled and this
4925 * value is set to 0, no change is made to the last
4926 * recorded delayed sack timer value.
4928 * spp_flags - These flags are used to control various features
4929 * on an association. The flag field may contain
4930 * zero or more of the following options.
4932 * SPP_HB_ENABLE - Enable heartbeats on the
4933 * specified address. Note that if the address
4934 * field is empty all addresses for the association
4935 * have heartbeats enabled upon them.
4937 * SPP_HB_DISABLE - Disable heartbeats on the
4938 * speicifed address. Note that if the address
4939 * field is empty all addresses for the association
4940 * will have their heartbeats disabled. Note also
4941 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4942 * mutually exclusive, only one of these two should
4943 * be specified. Enabling both fields will have
4944 * undetermined results.
4946 * SPP_HB_DEMAND - Request a user initiated heartbeat
4947 * to be made immediately.
4949 * SPP_PMTUD_ENABLE - This field will enable PMTU
4950 * discovery upon the specified address. Note that
4951 * if the address feild is empty then all addresses
4952 * on the association are effected.
4954 * SPP_PMTUD_DISABLE - This field will disable PMTU
4955 * discovery upon the specified address. Note that
4956 * if the address feild is empty then all addresses
4957 * on the association are effected. Not also that
4958 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4959 * exclusive. Enabling both will have undetermined
4960 * results.
4962 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4963 * on delayed sack. The time specified in spp_sackdelay
4964 * is used to specify the sack delay for this address. Note
4965 * that if spp_address is empty then all addresses will
4966 * enable delayed sack and take on the sack delay
4967 * value specified in spp_sackdelay.
4968 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4969 * off delayed sack. If the spp_address field is blank then
4970 * delayed sack is disabled for the entire association. Note
4971 * also that this field is mutually exclusive to
4972 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4973 * results.
4975 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4976 char __user *optval, int __user *optlen)
4978 struct sctp_paddrparams params;
4979 struct sctp_transport *trans = NULL;
4980 struct sctp_association *asoc = NULL;
4981 struct sctp_sock *sp = sctp_sk(sk);
4983 if (len < sizeof(struct sctp_paddrparams))
4984 return -EINVAL;
4985 len = sizeof(struct sctp_paddrparams);
4986 if (copy_from_user(&params, optval, len))
4987 return -EFAULT;
4989 /* If an address other than INADDR_ANY is specified, and
4990 * no transport is found, then the request is invalid.
4992 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
4993 trans = sctp_addr_id2transport(sk, &params.spp_address,
4994 params.spp_assoc_id);
4995 if (!trans) {
4996 pr_debug("%s: failed no transport\n", __func__);
4997 return -EINVAL;
5001 /* Get association, if assoc_id != 0 and the socket is a one
5002 * to many style socket, and an association was not found, then
5003 * the id was invalid.
5005 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5006 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5007 pr_debug("%s: failed no association\n", __func__);
5008 return -EINVAL;
5011 if (trans) {
5012 /* Fetch transport values. */
5013 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5014 params.spp_pathmtu = trans->pathmtu;
5015 params.spp_pathmaxrxt = trans->pathmaxrxt;
5016 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5018 /*draft-11 doesn't say what to return in spp_flags*/
5019 params.spp_flags = trans->param_flags;
5020 } else if (asoc) {
5021 /* Fetch association values. */
5022 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5023 params.spp_pathmtu = asoc->pathmtu;
5024 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5025 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5027 /*draft-11 doesn't say what to return in spp_flags*/
5028 params.spp_flags = asoc->param_flags;
5029 } else {
5030 /* Fetch socket values. */
5031 params.spp_hbinterval = sp->hbinterval;
5032 params.spp_pathmtu = sp->pathmtu;
5033 params.spp_sackdelay = sp->sackdelay;
5034 params.spp_pathmaxrxt = sp->pathmaxrxt;
5036 /*draft-11 doesn't say what to return in spp_flags*/
5037 params.spp_flags = sp->param_flags;
5040 if (copy_to_user(optval, &params, len))
5041 return -EFAULT;
5043 if (put_user(len, optlen))
5044 return -EFAULT;
5046 return 0;
5050 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5052 * This option will effect the way delayed acks are performed. This
5053 * option allows you to get or set the delayed ack time, in
5054 * milliseconds. It also allows changing the delayed ack frequency.
5055 * Changing the frequency to 1 disables the delayed sack algorithm. If
5056 * the assoc_id is 0, then this sets or gets the endpoints default
5057 * values. If the assoc_id field is non-zero, then the set or get
5058 * effects the specified association for the one to many model (the
5059 * assoc_id field is ignored by the one to one model). Note that if
5060 * sack_delay or sack_freq are 0 when setting this option, then the
5061 * current values will remain unchanged.
5063 * struct sctp_sack_info {
5064 * sctp_assoc_t sack_assoc_id;
5065 * uint32_t sack_delay;
5066 * uint32_t sack_freq;
5067 * };
5069 * sack_assoc_id - This parameter, indicates which association the user
5070 * is performing an action upon. Note that if this field's value is
5071 * zero then the endpoints default value is changed (effecting future
5072 * associations only).
5074 * sack_delay - This parameter contains the number of milliseconds that
5075 * the user is requesting the delayed ACK timer be set to. Note that
5076 * this value is defined in the standard to be between 200 and 500
5077 * milliseconds.
5079 * sack_freq - This parameter contains the number of packets that must
5080 * be received before a sack is sent without waiting for the delay
5081 * timer to expire. The default value for this is 2, setting this
5082 * value to 1 will disable the delayed sack algorithm.
5084 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5085 char __user *optval,
5086 int __user *optlen)
5088 struct sctp_sack_info params;
5089 struct sctp_association *asoc = NULL;
5090 struct sctp_sock *sp = sctp_sk(sk);
5092 if (len >= sizeof(struct sctp_sack_info)) {
5093 len = sizeof(struct sctp_sack_info);
5095 if (copy_from_user(&params, optval, len))
5096 return -EFAULT;
5097 } else if (len == sizeof(struct sctp_assoc_value)) {
5098 pr_warn_ratelimited(DEPRECATED
5099 "%s (pid %d) "
5100 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5101 "Use struct sctp_sack_info instead\n",
5102 current->comm, task_pid_nr(current));
5103 if (copy_from_user(&params, optval, len))
5104 return -EFAULT;
5105 } else
5106 return -EINVAL;
5108 /* Get association, if sack_assoc_id != 0 and the socket is a one
5109 * to many style socket, and an association was not found, then
5110 * the id was invalid.
5112 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5113 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5114 return -EINVAL;
5116 if (asoc) {
5117 /* Fetch association values. */
5118 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5119 params.sack_delay = jiffies_to_msecs(
5120 asoc->sackdelay);
5121 params.sack_freq = asoc->sackfreq;
5123 } else {
5124 params.sack_delay = 0;
5125 params.sack_freq = 1;
5127 } else {
5128 /* Fetch socket values. */
5129 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5130 params.sack_delay = sp->sackdelay;
5131 params.sack_freq = sp->sackfreq;
5132 } else {
5133 params.sack_delay = 0;
5134 params.sack_freq = 1;
5138 if (copy_to_user(optval, &params, len))
5139 return -EFAULT;
5141 if (put_user(len, optlen))
5142 return -EFAULT;
5144 return 0;
5147 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5149 * Applications can specify protocol parameters for the default association
5150 * initialization. The option name argument to setsockopt() and getsockopt()
5151 * is SCTP_INITMSG.
5153 * Setting initialization parameters is effective only on an unconnected
5154 * socket (for UDP-style sockets only future associations are effected
5155 * by the change). With TCP-style sockets, this option is inherited by
5156 * sockets derived from a listener socket.
5158 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5160 if (len < sizeof(struct sctp_initmsg))
5161 return -EINVAL;
5162 len = sizeof(struct sctp_initmsg);
5163 if (put_user(len, optlen))
5164 return -EFAULT;
5165 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5166 return -EFAULT;
5167 return 0;
5171 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5172 char __user *optval, int __user *optlen)
5174 struct sctp_association *asoc;
5175 int cnt = 0;
5176 struct sctp_getaddrs getaddrs;
5177 struct sctp_transport *from;
5178 void __user *to;
5179 union sctp_addr temp;
5180 struct sctp_sock *sp = sctp_sk(sk);
5181 int addrlen;
5182 size_t space_left;
5183 int bytes_copied;
5185 if (len < sizeof(struct sctp_getaddrs))
5186 return -EINVAL;
5188 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5189 return -EFAULT;
5191 /* For UDP-style sockets, id specifies the association to query. */
5192 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5193 if (!asoc)
5194 return -EINVAL;
5196 to = optval + offsetof(struct sctp_getaddrs, addrs);
5197 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5199 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5200 transports) {
5201 memcpy(&temp, &from->ipaddr, sizeof(temp));
5202 addrlen = sctp_get_pf_specific(sk->sk_family)
5203 ->addr_to_user(sp, &temp);
5204 if (space_left < addrlen)
5205 return -ENOMEM;
5206 if (copy_to_user(to, &temp, addrlen))
5207 return -EFAULT;
5208 to += addrlen;
5209 cnt++;
5210 space_left -= addrlen;
5213 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5214 return -EFAULT;
5215 bytes_copied = ((char __user *)to) - optval;
5216 if (put_user(bytes_copied, optlen))
5217 return -EFAULT;
5219 return 0;
5222 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5223 size_t space_left, int *bytes_copied)
5225 struct sctp_sockaddr_entry *addr;
5226 union sctp_addr temp;
5227 int cnt = 0;
5228 int addrlen;
5229 struct net *net = sock_net(sk);
5231 rcu_read_lock();
5232 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5233 if (!addr->valid)
5234 continue;
5236 if ((PF_INET == sk->sk_family) &&
5237 (AF_INET6 == addr->a.sa.sa_family))
5238 continue;
5239 if ((PF_INET6 == sk->sk_family) &&
5240 inet_v6_ipv6only(sk) &&
5241 (AF_INET == addr->a.sa.sa_family))
5242 continue;
5243 memcpy(&temp, &addr->a, sizeof(temp));
5244 if (!temp.v4.sin_port)
5245 temp.v4.sin_port = htons(port);
5247 addrlen = sctp_get_pf_specific(sk->sk_family)
5248 ->addr_to_user(sctp_sk(sk), &temp);
5250 if (space_left < addrlen) {
5251 cnt = -ENOMEM;
5252 break;
5254 memcpy(to, &temp, addrlen);
5256 to += addrlen;
5257 cnt++;
5258 space_left -= addrlen;
5259 *bytes_copied += addrlen;
5261 rcu_read_unlock();
5263 return cnt;
5267 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5268 char __user *optval, int __user *optlen)
5270 struct sctp_bind_addr *bp;
5271 struct sctp_association *asoc;
5272 int cnt = 0;
5273 struct sctp_getaddrs getaddrs;
5274 struct sctp_sockaddr_entry *addr;
5275 void __user *to;
5276 union sctp_addr temp;
5277 struct sctp_sock *sp = sctp_sk(sk);
5278 int addrlen;
5279 int err = 0;
5280 size_t space_left;
5281 int bytes_copied = 0;
5282 void *addrs;
5283 void *buf;
5285 if (len < sizeof(struct sctp_getaddrs))
5286 return -EINVAL;
5288 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5289 return -EFAULT;
5292 * For UDP-style sockets, id specifies the association to query.
5293 * If the id field is set to the value '0' then the locally bound
5294 * addresses are returned without regard to any particular
5295 * association.
5297 if (0 == getaddrs.assoc_id) {
5298 bp = &sctp_sk(sk)->ep->base.bind_addr;
5299 } else {
5300 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5301 if (!asoc)
5302 return -EINVAL;
5303 bp = &asoc->base.bind_addr;
5306 to = optval + offsetof(struct sctp_getaddrs, addrs);
5307 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5309 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5310 if (!addrs)
5311 return -ENOMEM;
5313 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5314 * addresses from the global local address list.
5316 if (sctp_list_single_entry(&bp->address_list)) {
5317 addr = list_entry(bp->address_list.next,
5318 struct sctp_sockaddr_entry, list);
5319 if (sctp_is_any(sk, &addr->a)) {
5320 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5321 space_left, &bytes_copied);
5322 if (cnt < 0) {
5323 err = cnt;
5324 goto out;
5326 goto copy_getaddrs;
5330 buf = addrs;
5331 /* Protection on the bound address list is not needed since
5332 * in the socket option context we hold a socket lock and
5333 * thus the bound address list can't change.
5335 list_for_each_entry(addr, &bp->address_list, list) {
5336 memcpy(&temp, &addr->a, sizeof(temp));
5337 addrlen = sctp_get_pf_specific(sk->sk_family)
5338 ->addr_to_user(sp, &temp);
5339 if (space_left < addrlen) {
5340 err = -ENOMEM; /*fixme: right error?*/
5341 goto out;
5343 memcpy(buf, &temp, addrlen);
5344 buf += addrlen;
5345 bytes_copied += addrlen;
5346 cnt++;
5347 space_left -= addrlen;
5350 copy_getaddrs:
5351 if (copy_to_user(to, addrs, bytes_copied)) {
5352 err = -EFAULT;
5353 goto out;
5355 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5356 err = -EFAULT;
5357 goto out;
5359 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5360 * but we can't change it anymore.
5362 if (put_user(bytes_copied, optlen))
5363 err = -EFAULT;
5364 out:
5365 kfree(addrs);
5366 return err;
5369 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5371 * Requests that the local SCTP stack use the enclosed peer address as
5372 * the association primary. The enclosed address must be one of the
5373 * association peer's addresses.
5375 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5376 char __user *optval, int __user *optlen)
5378 struct sctp_prim prim;
5379 struct sctp_association *asoc;
5380 struct sctp_sock *sp = sctp_sk(sk);
5382 if (len < sizeof(struct sctp_prim))
5383 return -EINVAL;
5385 len = sizeof(struct sctp_prim);
5387 if (copy_from_user(&prim, optval, len))
5388 return -EFAULT;
5390 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5391 if (!asoc)
5392 return -EINVAL;
5394 if (!asoc->peer.primary_path)
5395 return -ENOTCONN;
5397 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5398 asoc->peer.primary_path->af_specific->sockaddr_len);
5400 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5401 (union sctp_addr *)&prim.ssp_addr);
5403 if (put_user(len, optlen))
5404 return -EFAULT;
5405 if (copy_to_user(optval, &prim, len))
5406 return -EFAULT;
5408 return 0;
5412 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5414 * Requests that the local endpoint set the specified Adaptation Layer
5415 * Indication parameter for all future INIT and INIT-ACK exchanges.
5417 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5418 char __user *optval, int __user *optlen)
5420 struct sctp_setadaptation adaptation;
5422 if (len < sizeof(struct sctp_setadaptation))
5423 return -EINVAL;
5425 len = sizeof(struct sctp_setadaptation);
5427 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5429 if (put_user(len, optlen))
5430 return -EFAULT;
5431 if (copy_to_user(optval, &adaptation, len))
5432 return -EFAULT;
5434 return 0;
5439 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5441 * Applications that wish to use the sendto() system call may wish to
5442 * specify a default set of parameters that would normally be supplied
5443 * through the inclusion of ancillary data. This socket option allows
5444 * such an application to set the default sctp_sndrcvinfo structure.
5447 * The application that wishes to use this socket option simply passes
5448 * in to this call the sctp_sndrcvinfo structure defined in Section
5449 * 5.2.2) The input parameters accepted by this call include
5450 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5451 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5452 * to this call if the caller is using the UDP model.
5454 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5456 static int sctp_getsockopt_default_send_param(struct sock *sk,
5457 int len, char __user *optval,
5458 int __user *optlen)
5460 struct sctp_sock *sp = sctp_sk(sk);
5461 struct sctp_association *asoc;
5462 struct sctp_sndrcvinfo info;
5464 if (len < sizeof(info))
5465 return -EINVAL;
5467 len = sizeof(info);
5469 if (copy_from_user(&info, optval, len))
5470 return -EFAULT;
5472 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5473 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5474 return -EINVAL;
5475 if (asoc) {
5476 info.sinfo_stream = asoc->default_stream;
5477 info.sinfo_flags = asoc->default_flags;
5478 info.sinfo_ppid = asoc->default_ppid;
5479 info.sinfo_context = asoc->default_context;
5480 info.sinfo_timetolive = asoc->default_timetolive;
5481 } else {
5482 info.sinfo_stream = sp->default_stream;
5483 info.sinfo_flags = sp->default_flags;
5484 info.sinfo_ppid = sp->default_ppid;
5485 info.sinfo_context = sp->default_context;
5486 info.sinfo_timetolive = sp->default_timetolive;
5489 if (put_user(len, optlen))
5490 return -EFAULT;
5491 if (copy_to_user(optval, &info, len))
5492 return -EFAULT;
5494 return 0;
5497 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5498 * (SCTP_DEFAULT_SNDINFO)
5500 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5501 char __user *optval,
5502 int __user *optlen)
5504 struct sctp_sock *sp = sctp_sk(sk);
5505 struct sctp_association *asoc;
5506 struct sctp_sndinfo info;
5508 if (len < sizeof(info))
5509 return -EINVAL;
5511 len = sizeof(info);
5513 if (copy_from_user(&info, optval, len))
5514 return -EFAULT;
5516 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5517 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5518 return -EINVAL;
5519 if (asoc) {
5520 info.snd_sid = asoc->default_stream;
5521 info.snd_flags = asoc->default_flags;
5522 info.snd_ppid = asoc->default_ppid;
5523 info.snd_context = asoc->default_context;
5524 } else {
5525 info.snd_sid = sp->default_stream;
5526 info.snd_flags = sp->default_flags;
5527 info.snd_ppid = sp->default_ppid;
5528 info.snd_context = sp->default_context;
5531 if (put_user(len, optlen))
5532 return -EFAULT;
5533 if (copy_to_user(optval, &info, len))
5534 return -EFAULT;
5536 return 0;
5541 * 7.1.5 SCTP_NODELAY
5543 * Turn on/off any Nagle-like algorithm. This means that packets are
5544 * generally sent as soon as possible and no unnecessary delays are
5545 * introduced, at the cost of more packets in the network. Expects an
5546 * integer boolean flag.
5549 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5550 char __user *optval, int __user *optlen)
5552 int val;
5554 if (len < sizeof(int))
5555 return -EINVAL;
5557 len = sizeof(int);
5558 val = (sctp_sk(sk)->nodelay == 1);
5559 if (put_user(len, optlen))
5560 return -EFAULT;
5561 if (copy_to_user(optval, &val, len))
5562 return -EFAULT;
5563 return 0;
5568 * 7.1.1 SCTP_RTOINFO
5570 * The protocol parameters used to initialize and bound retransmission
5571 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5572 * and modify these parameters.
5573 * All parameters are time values, in milliseconds. A value of 0, when
5574 * modifying the parameters, indicates that the current value should not
5575 * be changed.
5578 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5579 char __user *optval,
5580 int __user *optlen) {
5581 struct sctp_rtoinfo rtoinfo;
5582 struct sctp_association *asoc;
5584 if (len < sizeof (struct sctp_rtoinfo))
5585 return -EINVAL;
5587 len = sizeof(struct sctp_rtoinfo);
5589 if (copy_from_user(&rtoinfo, optval, len))
5590 return -EFAULT;
5592 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5594 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5595 return -EINVAL;
5597 /* Values corresponding to the specific association. */
5598 if (asoc) {
5599 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5600 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5601 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5602 } else {
5603 /* Values corresponding to the endpoint. */
5604 struct sctp_sock *sp = sctp_sk(sk);
5606 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5607 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5608 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5611 if (put_user(len, optlen))
5612 return -EFAULT;
5614 if (copy_to_user(optval, &rtoinfo, len))
5615 return -EFAULT;
5617 return 0;
5622 * 7.1.2 SCTP_ASSOCINFO
5624 * This option is used to tune the maximum retransmission attempts
5625 * of the association.
5626 * Returns an error if the new association retransmission value is
5627 * greater than the sum of the retransmission value of the peer.
5628 * See [SCTP] for more information.
5631 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5632 char __user *optval,
5633 int __user *optlen)
5636 struct sctp_assocparams assocparams;
5637 struct sctp_association *asoc;
5638 struct list_head *pos;
5639 int cnt = 0;
5641 if (len < sizeof (struct sctp_assocparams))
5642 return -EINVAL;
5644 len = sizeof(struct sctp_assocparams);
5646 if (copy_from_user(&assocparams, optval, len))
5647 return -EFAULT;
5649 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5651 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5652 return -EINVAL;
5654 /* Values correspoinding to the specific association */
5655 if (asoc) {
5656 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5657 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5658 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5659 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5661 list_for_each(pos, &asoc->peer.transport_addr_list) {
5662 cnt++;
5665 assocparams.sasoc_number_peer_destinations = cnt;
5666 } else {
5667 /* Values corresponding to the endpoint */
5668 struct sctp_sock *sp = sctp_sk(sk);
5670 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5671 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5672 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5673 assocparams.sasoc_cookie_life =
5674 sp->assocparams.sasoc_cookie_life;
5675 assocparams.sasoc_number_peer_destinations =
5676 sp->assocparams.
5677 sasoc_number_peer_destinations;
5680 if (put_user(len, optlen))
5681 return -EFAULT;
5683 if (copy_to_user(optval, &assocparams, len))
5684 return -EFAULT;
5686 return 0;
5690 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5692 * This socket option is a boolean flag which turns on or off mapped V4
5693 * addresses. If this option is turned on and the socket is type
5694 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5695 * If this option is turned off, then no mapping will be done of V4
5696 * addresses and a user will receive both PF_INET6 and PF_INET type
5697 * addresses on the socket.
5699 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5700 char __user *optval, int __user *optlen)
5702 int val;
5703 struct sctp_sock *sp = sctp_sk(sk);
5705 if (len < sizeof(int))
5706 return -EINVAL;
5708 len = sizeof(int);
5709 val = sp->v4mapped;
5710 if (put_user(len, optlen))
5711 return -EFAULT;
5712 if (copy_to_user(optval, &val, len))
5713 return -EFAULT;
5715 return 0;
5719 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5720 * (chapter and verse is quoted at sctp_setsockopt_context())
5722 static int sctp_getsockopt_context(struct sock *sk, int len,
5723 char __user *optval, int __user *optlen)
5725 struct sctp_assoc_value params;
5726 struct sctp_sock *sp;
5727 struct sctp_association *asoc;
5729 if (len < sizeof(struct sctp_assoc_value))
5730 return -EINVAL;
5732 len = sizeof(struct sctp_assoc_value);
5734 if (copy_from_user(&params, optval, len))
5735 return -EFAULT;
5737 sp = sctp_sk(sk);
5739 if (params.assoc_id != 0) {
5740 asoc = sctp_id2assoc(sk, params.assoc_id);
5741 if (!asoc)
5742 return -EINVAL;
5743 params.assoc_value = asoc->default_rcv_context;
5744 } else {
5745 params.assoc_value = sp->default_rcv_context;
5748 if (put_user(len, optlen))
5749 return -EFAULT;
5750 if (copy_to_user(optval, &params, len))
5751 return -EFAULT;
5753 return 0;
5757 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5758 * This option will get or set the maximum size to put in any outgoing
5759 * SCTP DATA chunk. If a message is larger than this size it will be
5760 * fragmented by SCTP into the specified size. Note that the underlying
5761 * SCTP implementation may fragment into smaller sized chunks when the
5762 * PMTU of the underlying association is smaller than the value set by
5763 * the user. The default value for this option is '0' which indicates
5764 * the user is NOT limiting fragmentation and only the PMTU will effect
5765 * SCTP's choice of DATA chunk size. Note also that values set larger
5766 * than the maximum size of an IP datagram will effectively let SCTP
5767 * control fragmentation (i.e. the same as setting this option to 0).
5769 * The following structure is used to access and modify this parameter:
5771 * struct sctp_assoc_value {
5772 * sctp_assoc_t assoc_id;
5773 * uint32_t assoc_value;
5774 * };
5776 * assoc_id: This parameter is ignored for one-to-one style sockets.
5777 * For one-to-many style sockets this parameter indicates which
5778 * association the user is performing an action upon. Note that if
5779 * this field's value is zero then the endpoints default value is
5780 * changed (effecting future associations only).
5781 * assoc_value: This parameter specifies the maximum size in bytes.
5783 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5784 char __user *optval, int __user *optlen)
5786 struct sctp_assoc_value params;
5787 struct sctp_association *asoc;
5789 if (len == sizeof(int)) {
5790 pr_warn_ratelimited(DEPRECATED
5791 "%s (pid %d) "
5792 "Use of int in maxseg socket option.\n"
5793 "Use struct sctp_assoc_value instead\n",
5794 current->comm, task_pid_nr(current));
5795 params.assoc_id = 0;
5796 } else if (len >= sizeof(struct sctp_assoc_value)) {
5797 len = sizeof(struct sctp_assoc_value);
5798 if (copy_from_user(&params, optval, len))
5799 return -EFAULT;
5800 } else
5801 return -EINVAL;
5803 asoc = sctp_id2assoc(sk, params.assoc_id);
5804 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5805 return -EINVAL;
5807 if (asoc)
5808 params.assoc_value = asoc->frag_point;
5809 else
5810 params.assoc_value = sctp_sk(sk)->user_frag;
5812 if (put_user(len, optlen))
5813 return -EFAULT;
5814 if (len == sizeof(int)) {
5815 if (copy_to_user(optval, &params.assoc_value, len))
5816 return -EFAULT;
5817 } else {
5818 if (copy_to_user(optval, &params, len))
5819 return -EFAULT;
5822 return 0;
5826 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5827 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5829 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5830 char __user *optval, int __user *optlen)
5832 int val;
5834 if (len < sizeof(int))
5835 return -EINVAL;
5837 len = sizeof(int);
5839 val = sctp_sk(sk)->frag_interleave;
5840 if (put_user(len, optlen))
5841 return -EFAULT;
5842 if (copy_to_user(optval, &val, len))
5843 return -EFAULT;
5845 return 0;
5849 * 7.1.25. Set or Get the sctp partial delivery point
5850 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5852 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5853 char __user *optval,
5854 int __user *optlen)
5856 u32 val;
5858 if (len < sizeof(u32))
5859 return -EINVAL;
5861 len = sizeof(u32);
5863 val = sctp_sk(sk)->pd_point;
5864 if (put_user(len, optlen))
5865 return -EFAULT;
5866 if (copy_to_user(optval, &val, len))
5867 return -EFAULT;
5869 return 0;
5873 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5874 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5876 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5877 char __user *optval,
5878 int __user *optlen)
5880 struct sctp_assoc_value params;
5881 struct sctp_sock *sp;
5882 struct sctp_association *asoc;
5884 if (len == sizeof(int)) {
5885 pr_warn_ratelimited(DEPRECATED
5886 "%s (pid %d) "
5887 "Use of int in max_burst socket option.\n"
5888 "Use struct sctp_assoc_value instead\n",
5889 current->comm, task_pid_nr(current));
5890 params.assoc_id = 0;
5891 } else if (len >= sizeof(struct sctp_assoc_value)) {
5892 len = sizeof(struct sctp_assoc_value);
5893 if (copy_from_user(&params, optval, len))
5894 return -EFAULT;
5895 } else
5896 return -EINVAL;
5898 sp = sctp_sk(sk);
5900 if (params.assoc_id != 0) {
5901 asoc = sctp_id2assoc(sk, params.assoc_id);
5902 if (!asoc)
5903 return -EINVAL;
5904 params.assoc_value = asoc->max_burst;
5905 } else
5906 params.assoc_value = sp->max_burst;
5908 if (len == sizeof(int)) {
5909 if (copy_to_user(optval, &params.assoc_value, len))
5910 return -EFAULT;
5911 } else {
5912 if (copy_to_user(optval, &params, len))
5913 return -EFAULT;
5916 return 0;
5920 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5921 char __user *optval, int __user *optlen)
5923 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5924 struct sctp_hmacalgo __user *p = (void __user *)optval;
5925 struct sctp_hmac_algo_param *hmacs;
5926 __u16 data_len = 0;
5927 u32 num_idents;
5928 int i;
5930 if (!ep->auth_enable)
5931 return -EACCES;
5933 hmacs = ep->auth_hmacs_list;
5934 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5936 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5937 return -EINVAL;
5939 len = sizeof(struct sctp_hmacalgo) + data_len;
5940 num_idents = data_len / sizeof(u16);
5942 if (put_user(len, optlen))
5943 return -EFAULT;
5944 if (put_user(num_idents, &p->shmac_num_idents))
5945 return -EFAULT;
5946 for (i = 0; i < num_idents; i++) {
5947 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
5949 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
5950 return -EFAULT;
5952 return 0;
5955 static int sctp_getsockopt_active_key(struct sock *sk, int len,
5956 char __user *optval, int __user *optlen)
5958 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5959 struct sctp_authkeyid val;
5960 struct sctp_association *asoc;
5962 if (!ep->auth_enable)
5963 return -EACCES;
5965 if (len < sizeof(struct sctp_authkeyid))
5966 return -EINVAL;
5968 len = sizeof(struct sctp_authkeyid);
5969 if (copy_from_user(&val, optval, len))
5970 return -EFAULT;
5972 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5973 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5974 return -EINVAL;
5976 if (asoc)
5977 val.scact_keynumber = asoc->active_key_id;
5978 else
5979 val.scact_keynumber = ep->active_key_id;
5981 if (put_user(len, optlen))
5982 return -EFAULT;
5983 if (copy_to_user(optval, &val, len))
5984 return -EFAULT;
5986 return 0;
5989 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5990 char __user *optval, int __user *optlen)
5992 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5993 struct sctp_authchunks __user *p = (void __user *)optval;
5994 struct sctp_authchunks val;
5995 struct sctp_association *asoc;
5996 struct sctp_chunks_param *ch;
5997 u32 num_chunks = 0;
5998 char __user *to;
6000 if (!ep->auth_enable)
6001 return -EACCES;
6003 if (len < sizeof(struct sctp_authchunks))
6004 return -EINVAL;
6006 if (copy_from_user(&val, optval, sizeof(val)))
6007 return -EFAULT;
6009 to = p->gauth_chunks;
6010 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6011 if (!asoc)
6012 return -EINVAL;
6014 ch = asoc->peer.peer_chunks;
6015 if (!ch)
6016 goto num;
6018 /* See if the user provided enough room for all the data */
6019 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6020 if (len < num_chunks)
6021 return -EINVAL;
6023 if (copy_to_user(to, ch->chunks, num_chunks))
6024 return -EFAULT;
6025 num:
6026 len = sizeof(struct sctp_authchunks) + num_chunks;
6027 if (put_user(len, optlen))
6028 return -EFAULT;
6029 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6030 return -EFAULT;
6031 return 0;
6034 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6035 char __user *optval, int __user *optlen)
6037 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6038 struct sctp_authchunks __user *p = (void __user *)optval;
6039 struct sctp_authchunks val;
6040 struct sctp_association *asoc;
6041 struct sctp_chunks_param *ch;
6042 u32 num_chunks = 0;
6043 char __user *to;
6045 if (!ep->auth_enable)
6046 return -EACCES;
6048 if (len < sizeof(struct sctp_authchunks))
6049 return -EINVAL;
6051 if (copy_from_user(&val, optval, sizeof(val)))
6052 return -EFAULT;
6054 to = p->gauth_chunks;
6055 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6056 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6057 return -EINVAL;
6059 if (asoc)
6060 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6061 else
6062 ch = ep->auth_chunk_list;
6064 if (!ch)
6065 goto num;
6067 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
6068 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6069 return -EINVAL;
6071 if (copy_to_user(to, ch->chunks, num_chunks))
6072 return -EFAULT;
6073 num:
6074 len = sizeof(struct sctp_authchunks) + num_chunks;
6075 if (put_user(len, optlen))
6076 return -EFAULT;
6077 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6078 return -EFAULT;
6080 return 0;
6084 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6085 * This option gets the current number of associations that are attached
6086 * to a one-to-many style socket. The option value is an uint32_t.
6088 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6089 char __user *optval, int __user *optlen)
6091 struct sctp_sock *sp = sctp_sk(sk);
6092 struct sctp_association *asoc;
6093 u32 val = 0;
6095 if (sctp_style(sk, TCP))
6096 return -EOPNOTSUPP;
6098 if (len < sizeof(u32))
6099 return -EINVAL;
6101 len = sizeof(u32);
6103 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6104 val++;
6107 if (put_user(len, optlen))
6108 return -EFAULT;
6109 if (copy_to_user(optval, &val, len))
6110 return -EFAULT;
6112 return 0;
6116 * 8.1.23 SCTP_AUTO_ASCONF
6117 * See the corresponding setsockopt entry as description
6119 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6120 char __user *optval, int __user *optlen)
6122 int val = 0;
6124 if (len < sizeof(int))
6125 return -EINVAL;
6127 len = sizeof(int);
6128 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6129 val = 1;
6130 if (put_user(len, optlen))
6131 return -EFAULT;
6132 if (copy_to_user(optval, &val, len))
6133 return -EFAULT;
6134 return 0;
6138 * 8.2.6. Get the Current Identifiers of Associations
6139 * (SCTP_GET_ASSOC_ID_LIST)
6141 * This option gets the current list of SCTP association identifiers of
6142 * the SCTP associations handled by a one-to-many style socket.
6144 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6145 char __user *optval, int __user *optlen)
6147 struct sctp_sock *sp = sctp_sk(sk);
6148 struct sctp_association *asoc;
6149 struct sctp_assoc_ids *ids;
6150 u32 num = 0;
6152 if (sctp_style(sk, TCP))
6153 return -EOPNOTSUPP;
6155 if (len < sizeof(struct sctp_assoc_ids))
6156 return -EINVAL;
6158 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6159 num++;
6162 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6163 return -EINVAL;
6165 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6167 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6168 if (unlikely(!ids))
6169 return -ENOMEM;
6171 ids->gaids_number_of_ids = num;
6172 num = 0;
6173 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6174 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6177 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6178 kfree(ids);
6179 return -EFAULT;
6182 kfree(ids);
6183 return 0;
6187 * SCTP_PEER_ADDR_THLDS
6189 * This option allows us to fetch the partially failed threshold for one or all
6190 * transports in an association. See Section 6.1 of:
6191 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6193 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6194 char __user *optval,
6195 int len,
6196 int __user *optlen)
6198 struct sctp_paddrthlds val;
6199 struct sctp_transport *trans;
6200 struct sctp_association *asoc;
6202 if (len < sizeof(struct sctp_paddrthlds))
6203 return -EINVAL;
6204 len = sizeof(struct sctp_paddrthlds);
6205 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6206 return -EFAULT;
6208 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6209 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6210 if (!asoc)
6211 return -ENOENT;
6213 val.spt_pathpfthld = asoc->pf_retrans;
6214 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6215 } else {
6216 trans = sctp_addr_id2transport(sk, &val.spt_address,
6217 val.spt_assoc_id);
6218 if (!trans)
6219 return -ENOENT;
6221 val.spt_pathmaxrxt = trans->pathmaxrxt;
6222 val.spt_pathpfthld = trans->pf_retrans;
6225 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6226 return -EFAULT;
6228 return 0;
6232 * SCTP_GET_ASSOC_STATS
6234 * This option retrieves local per endpoint statistics. It is modeled
6235 * after OpenSolaris' implementation
6237 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6238 char __user *optval,
6239 int __user *optlen)
6241 struct sctp_assoc_stats sas;
6242 struct sctp_association *asoc = NULL;
6244 /* User must provide at least the assoc id */
6245 if (len < sizeof(sctp_assoc_t))
6246 return -EINVAL;
6248 /* Allow the struct to grow and fill in as much as possible */
6249 len = min_t(size_t, len, sizeof(sas));
6251 if (copy_from_user(&sas, optval, len))
6252 return -EFAULT;
6254 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6255 if (!asoc)
6256 return -EINVAL;
6258 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6259 sas.sas_gapcnt = asoc->stats.gapcnt;
6260 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6261 sas.sas_osacks = asoc->stats.osacks;
6262 sas.sas_isacks = asoc->stats.isacks;
6263 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6264 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6265 sas.sas_oodchunks = asoc->stats.oodchunks;
6266 sas.sas_iodchunks = asoc->stats.iodchunks;
6267 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6268 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6269 sas.sas_idupchunks = asoc->stats.idupchunks;
6270 sas.sas_opackets = asoc->stats.opackets;
6271 sas.sas_ipackets = asoc->stats.ipackets;
6273 /* New high max rto observed, will return 0 if not a single
6274 * RTO update took place. obs_rto_ipaddr will be bogus
6275 * in such a case
6277 sas.sas_maxrto = asoc->stats.max_obs_rto;
6278 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6279 sizeof(struct sockaddr_storage));
6281 /* Mark beginning of a new observation period */
6282 asoc->stats.max_obs_rto = asoc->rto_min;
6284 if (put_user(len, optlen))
6285 return -EFAULT;
6287 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6289 if (copy_to_user(optval, &sas, len))
6290 return -EFAULT;
6292 return 0;
6295 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6296 char __user *optval,
6297 int __user *optlen)
6299 int val = 0;
6301 if (len < sizeof(int))
6302 return -EINVAL;
6304 len = sizeof(int);
6305 if (sctp_sk(sk)->recvrcvinfo)
6306 val = 1;
6307 if (put_user(len, optlen))
6308 return -EFAULT;
6309 if (copy_to_user(optval, &val, len))
6310 return -EFAULT;
6312 return 0;
6315 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6316 char __user *optval,
6317 int __user *optlen)
6319 int val = 0;
6321 if (len < sizeof(int))
6322 return -EINVAL;
6324 len = sizeof(int);
6325 if (sctp_sk(sk)->recvnxtinfo)
6326 val = 1;
6327 if (put_user(len, optlen))
6328 return -EFAULT;
6329 if (copy_to_user(optval, &val, len))
6330 return -EFAULT;
6332 return 0;
6335 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6336 char __user *optval,
6337 int __user *optlen)
6339 struct sctp_assoc_value params;
6340 struct sctp_association *asoc;
6341 int retval = -EFAULT;
6343 if (len < sizeof(params)) {
6344 retval = -EINVAL;
6345 goto out;
6348 len = sizeof(params);
6349 if (copy_from_user(&params, optval, len))
6350 goto out;
6352 asoc = sctp_id2assoc(sk, params.assoc_id);
6353 if (asoc) {
6354 params.assoc_value = asoc->prsctp_enable;
6355 } else if (!params.assoc_id) {
6356 struct sctp_sock *sp = sctp_sk(sk);
6358 params.assoc_value = sp->ep->prsctp_enable;
6359 } else {
6360 retval = -EINVAL;
6361 goto out;
6364 if (put_user(len, optlen))
6365 goto out;
6367 if (copy_to_user(optval, &params, len))
6368 goto out;
6370 retval = 0;
6372 out:
6373 return retval;
6376 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6377 char __user *optval,
6378 int __user *optlen)
6380 struct sctp_default_prinfo info;
6381 struct sctp_association *asoc;
6382 int retval = -EFAULT;
6384 if (len < sizeof(info)) {
6385 retval = -EINVAL;
6386 goto out;
6389 len = sizeof(info);
6390 if (copy_from_user(&info, optval, len))
6391 goto out;
6393 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6394 if (asoc) {
6395 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6396 info.pr_value = asoc->default_timetolive;
6397 } else if (!info.pr_assoc_id) {
6398 struct sctp_sock *sp = sctp_sk(sk);
6400 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6401 info.pr_value = sp->default_timetolive;
6402 } else {
6403 retval = -EINVAL;
6404 goto out;
6407 if (put_user(len, optlen))
6408 goto out;
6410 if (copy_to_user(optval, &info, len))
6411 goto out;
6413 retval = 0;
6415 out:
6416 return retval;
6419 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6420 char __user *optval,
6421 int __user *optlen)
6423 struct sctp_prstatus params;
6424 struct sctp_association *asoc;
6425 int policy;
6426 int retval = -EINVAL;
6428 if (len < sizeof(params))
6429 goto out;
6431 len = sizeof(params);
6432 if (copy_from_user(&params, optval, len)) {
6433 retval = -EFAULT;
6434 goto out;
6437 policy = params.sprstat_policy;
6438 if (policy & ~SCTP_PR_SCTP_MASK)
6439 goto out;
6441 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6442 if (!asoc)
6443 goto out;
6445 if (policy == SCTP_PR_SCTP_NONE) {
6446 params.sprstat_abandoned_unsent = 0;
6447 params.sprstat_abandoned_sent = 0;
6448 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6449 params.sprstat_abandoned_unsent +=
6450 asoc->abandoned_unsent[policy];
6451 params.sprstat_abandoned_sent +=
6452 asoc->abandoned_sent[policy];
6454 } else {
6455 params.sprstat_abandoned_unsent =
6456 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6457 params.sprstat_abandoned_sent =
6458 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6461 if (put_user(len, optlen)) {
6462 retval = -EFAULT;
6463 goto out;
6466 if (copy_to_user(optval, &params, len)) {
6467 retval = -EFAULT;
6468 goto out;
6471 retval = 0;
6473 out:
6474 return retval;
6477 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6478 char __user *optval, int __user *optlen)
6480 int retval = 0;
6481 int len;
6483 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6485 /* I can hardly begin to describe how wrong this is. This is
6486 * so broken as to be worse than useless. The API draft
6487 * REALLY is NOT helpful here... I am not convinced that the
6488 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6489 * are at all well-founded.
6491 if (level != SOL_SCTP) {
6492 struct sctp_af *af = sctp_sk(sk)->pf->af;
6494 retval = af->getsockopt(sk, level, optname, optval, optlen);
6495 return retval;
6498 if (get_user(len, optlen))
6499 return -EFAULT;
6501 if (len < 0)
6502 return -EINVAL;
6504 lock_sock(sk);
6506 switch (optname) {
6507 case SCTP_STATUS:
6508 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6509 break;
6510 case SCTP_DISABLE_FRAGMENTS:
6511 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6512 optlen);
6513 break;
6514 case SCTP_EVENTS:
6515 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6516 break;
6517 case SCTP_AUTOCLOSE:
6518 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6519 break;
6520 case SCTP_SOCKOPT_PEELOFF:
6521 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6522 break;
6523 case SCTP_PEER_ADDR_PARAMS:
6524 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6525 optlen);
6526 break;
6527 case SCTP_DELAYED_SACK:
6528 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6529 optlen);
6530 break;
6531 case SCTP_INITMSG:
6532 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6533 break;
6534 case SCTP_GET_PEER_ADDRS:
6535 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6536 optlen);
6537 break;
6538 case SCTP_GET_LOCAL_ADDRS:
6539 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6540 optlen);
6541 break;
6542 case SCTP_SOCKOPT_CONNECTX3:
6543 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6544 break;
6545 case SCTP_DEFAULT_SEND_PARAM:
6546 retval = sctp_getsockopt_default_send_param(sk, len,
6547 optval, optlen);
6548 break;
6549 case SCTP_DEFAULT_SNDINFO:
6550 retval = sctp_getsockopt_default_sndinfo(sk, len,
6551 optval, optlen);
6552 break;
6553 case SCTP_PRIMARY_ADDR:
6554 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6555 break;
6556 case SCTP_NODELAY:
6557 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6558 break;
6559 case SCTP_RTOINFO:
6560 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6561 break;
6562 case SCTP_ASSOCINFO:
6563 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6564 break;
6565 case SCTP_I_WANT_MAPPED_V4_ADDR:
6566 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6567 break;
6568 case SCTP_MAXSEG:
6569 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6570 break;
6571 case SCTP_GET_PEER_ADDR_INFO:
6572 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6573 optlen);
6574 break;
6575 case SCTP_ADAPTATION_LAYER:
6576 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6577 optlen);
6578 break;
6579 case SCTP_CONTEXT:
6580 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6581 break;
6582 case SCTP_FRAGMENT_INTERLEAVE:
6583 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6584 optlen);
6585 break;
6586 case SCTP_PARTIAL_DELIVERY_POINT:
6587 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6588 optlen);
6589 break;
6590 case SCTP_MAX_BURST:
6591 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6592 break;
6593 case SCTP_AUTH_KEY:
6594 case SCTP_AUTH_CHUNK:
6595 case SCTP_AUTH_DELETE_KEY:
6596 retval = -EOPNOTSUPP;
6597 break;
6598 case SCTP_HMAC_IDENT:
6599 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6600 break;
6601 case SCTP_AUTH_ACTIVE_KEY:
6602 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6603 break;
6604 case SCTP_PEER_AUTH_CHUNKS:
6605 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6606 optlen);
6607 break;
6608 case SCTP_LOCAL_AUTH_CHUNKS:
6609 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6610 optlen);
6611 break;
6612 case SCTP_GET_ASSOC_NUMBER:
6613 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6614 break;
6615 case SCTP_GET_ASSOC_ID_LIST:
6616 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6617 break;
6618 case SCTP_AUTO_ASCONF:
6619 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6620 break;
6621 case SCTP_PEER_ADDR_THLDS:
6622 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6623 break;
6624 case SCTP_GET_ASSOC_STATS:
6625 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6626 break;
6627 case SCTP_RECVRCVINFO:
6628 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6629 break;
6630 case SCTP_RECVNXTINFO:
6631 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6632 break;
6633 case SCTP_PR_SUPPORTED:
6634 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
6635 break;
6636 case SCTP_DEFAULT_PRINFO:
6637 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
6638 optlen);
6639 break;
6640 case SCTP_PR_ASSOC_STATUS:
6641 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
6642 optlen);
6643 break;
6644 default:
6645 retval = -ENOPROTOOPT;
6646 break;
6649 release_sock(sk);
6650 return retval;
6653 static int sctp_hash(struct sock *sk)
6655 /* STUB */
6656 return 0;
6659 static void sctp_unhash(struct sock *sk)
6661 /* STUB */
6664 /* Check if port is acceptable. Possibly find first available port.
6666 * The port hash table (contained in the 'global' SCTP protocol storage
6667 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
6668 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
6669 * list (the list number is the port number hashed out, so as you
6670 * would expect from a hash function, all the ports in a given list have
6671 * such a number that hashes out to the same list number; you were
6672 * expecting that, right?); so each list has a set of ports, with a
6673 * link to the socket (struct sock) that uses it, the port number and
6674 * a fastreuse flag (FIXME: NPI ipg).
6676 static struct sctp_bind_bucket *sctp_bucket_create(
6677 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
6679 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
6681 struct sctp_bind_hashbucket *head; /* hash list */
6682 struct sctp_bind_bucket *pp;
6683 unsigned short snum;
6684 int ret;
6686 snum = ntohs(addr->v4.sin_port);
6688 pr_debug("%s: begins, snum:%d\n", __func__, snum);
6690 local_bh_disable();
6692 if (snum == 0) {
6693 /* Search for an available port. */
6694 int low, high, remaining, index;
6695 unsigned int rover;
6696 struct net *net = sock_net(sk);
6698 inet_get_local_port_range(net, &low, &high);
6699 remaining = (high - low) + 1;
6700 rover = prandom_u32() % remaining + low;
6702 do {
6703 rover++;
6704 if ((rover < low) || (rover > high))
6705 rover = low;
6706 if (inet_is_local_reserved_port(net, rover))
6707 continue;
6708 index = sctp_phashfn(sock_net(sk), rover);
6709 head = &sctp_port_hashtable[index];
6710 spin_lock(&head->lock);
6711 sctp_for_each_hentry(pp, &head->chain)
6712 if ((pp->port == rover) &&
6713 net_eq(sock_net(sk), pp->net))
6714 goto next;
6715 break;
6716 next:
6717 spin_unlock(&head->lock);
6718 } while (--remaining > 0);
6720 /* Exhausted local port range during search? */
6721 ret = 1;
6722 if (remaining <= 0)
6723 goto fail;
6725 /* OK, here is the one we will use. HEAD (the port
6726 * hash table list entry) is non-NULL and we hold it's
6727 * mutex.
6729 snum = rover;
6730 } else {
6731 /* We are given an specific port number; we verify
6732 * that it is not being used. If it is used, we will
6733 * exahust the search in the hash list corresponding
6734 * to the port number (snum) - we detect that with the
6735 * port iterator, pp being NULL.
6737 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
6738 spin_lock(&head->lock);
6739 sctp_for_each_hentry(pp, &head->chain) {
6740 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
6741 goto pp_found;
6744 pp = NULL;
6745 goto pp_not_found;
6746 pp_found:
6747 if (!hlist_empty(&pp->owner)) {
6748 /* We had a port hash table hit - there is an
6749 * available port (pp != NULL) and it is being
6750 * used by other socket (pp->owner not empty); that other
6751 * socket is going to be sk2.
6753 int reuse = sk->sk_reuse;
6754 struct sock *sk2;
6756 pr_debug("%s: found a possible match\n", __func__);
6758 if (pp->fastreuse && sk->sk_reuse &&
6759 sk->sk_state != SCTP_SS_LISTENING)
6760 goto success;
6762 /* Run through the list of sockets bound to the port
6763 * (pp->port) [via the pointers bind_next and
6764 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6765 * we get the endpoint they describe and run through
6766 * the endpoint's list of IP (v4 or v6) addresses,
6767 * comparing each of the addresses with the address of
6768 * the socket sk. If we find a match, then that means
6769 * that this port/socket (sk) combination are already
6770 * in an endpoint.
6772 sk_for_each_bound(sk2, &pp->owner) {
6773 struct sctp_endpoint *ep2;
6774 ep2 = sctp_sk(sk2)->ep;
6776 if (sk == sk2 ||
6777 (reuse && sk2->sk_reuse &&
6778 sk2->sk_state != SCTP_SS_LISTENING))
6779 continue;
6781 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6782 sctp_sk(sk2), sctp_sk(sk))) {
6783 ret = (long)sk2;
6784 goto fail_unlock;
6788 pr_debug("%s: found a match\n", __func__);
6790 pp_not_found:
6791 /* If there was a hash table miss, create a new port. */
6792 ret = 1;
6793 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6794 goto fail_unlock;
6796 /* In either case (hit or miss), make sure fastreuse is 1 only
6797 * if sk->sk_reuse is too (that is, if the caller requested
6798 * SO_REUSEADDR on this socket -sk-).
6800 if (hlist_empty(&pp->owner)) {
6801 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6802 pp->fastreuse = 1;
6803 else
6804 pp->fastreuse = 0;
6805 } else if (pp->fastreuse &&
6806 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6807 pp->fastreuse = 0;
6809 /* We are set, so fill up all the data in the hash table
6810 * entry, tie the socket list information with the rest of the
6811 * sockets FIXME: Blurry, NPI (ipg).
6813 success:
6814 if (!sctp_sk(sk)->bind_hash) {
6815 inet_sk(sk)->inet_num = snum;
6816 sk_add_bind_node(sk, &pp->owner);
6817 sctp_sk(sk)->bind_hash = pp;
6819 ret = 0;
6821 fail_unlock:
6822 spin_unlock(&head->lock);
6824 fail:
6825 local_bh_enable();
6826 return ret;
6829 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6830 * port is requested.
6832 static int sctp_get_port(struct sock *sk, unsigned short snum)
6834 union sctp_addr addr;
6835 struct sctp_af *af = sctp_sk(sk)->pf->af;
6837 /* Set up a dummy address struct from the sk. */
6838 af->from_sk(&addr, sk);
6839 addr.v4.sin_port = htons(snum);
6841 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6842 return !!sctp_get_port_local(sk, &addr);
6846 * Move a socket to LISTENING state.
6848 static int sctp_listen_start(struct sock *sk, int backlog)
6850 struct sctp_sock *sp = sctp_sk(sk);
6851 struct sctp_endpoint *ep = sp->ep;
6852 struct crypto_shash *tfm = NULL;
6853 char alg[32];
6855 /* Allocate HMAC for generating cookie. */
6856 if (!sp->hmac && sp->sctp_hmac_alg) {
6857 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6858 tfm = crypto_alloc_shash(alg, 0, 0);
6859 if (IS_ERR(tfm)) {
6860 net_info_ratelimited("failed to load transform for %s: %ld\n",
6861 sp->sctp_hmac_alg, PTR_ERR(tfm));
6862 return -ENOSYS;
6864 sctp_sk(sk)->hmac = tfm;
6868 * If a bind() or sctp_bindx() is not called prior to a listen()
6869 * call that allows new associations to be accepted, the system
6870 * picks an ephemeral port and will choose an address set equivalent
6871 * to binding with a wildcard address.
6873 * This is not currently spelled out in the SCTP sockets
6874 * extensions draft, but follows the practice as seen in TCP
6875 * sockets.
6878 sk->sk_state = SCTP_SS_LISTENING;
6879 if (!ep->base.bind_addr.port) {
6880 if (sctp_autobind(sk))
6881 return -EAGAIN;
6882 } else {
6883 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6884 sk->sk_state = SCTP_SS_CLOSED;
6885 return -EADDRINUSE;
6889 sk->sk_max_ack_backlog = backlog;
6890 sctp_hash_endpoint(ep);
6891 return 0;
6895 * 4.1.3 / 5.1.3 listen()
6897 * By default, new associations are not accepted for UDP style sockets.
6898 * An application uses listen() to mark a socket as being able to
6899 * accept new associations.
6901 * On TCP style sockets, applications use listen() to ready the SCTP
6902 * endpoint for accepting inbound associations.
6904 * On both types of endpoints a backlog of '0' disables listening.
6906 * Move a socket to LISTENING state.
6908 int sctp_inet_listen(struct socket *sock, int backlog)
6910 struct sock *sk = sock->sk;
6911 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6912 int err = -EINVAL;
6914 if (unlikely(backlog < 0))
6915 return err;
6917 lock_sock(sk);
6919 /* Peeled-off sockets are not allowed to listen(). */
6920 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6921 goto out;
6923 if (sock->state != SS_UNCONNECTED)
6924 goto out;
6926 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
6927 goto out;
6929 /* If backlog is zero, disable listening. */
6930 if (!backlog) {
6931 if (sctp_sstate(sk, CLOSED))
6932 goto out;
6934 err = 0;
6935 sctp_unhash_endpoint(ep);
6936 sk->sk_state = SCTP_SS_CLOSED;
6937 if (sk->sk_reuse)
6938 sctp_sk(sk)->bind_hash->fastreuse = 1;
6939 goto out;
6942 /* If we are already listening, just update the backlog */
6943 if (sctp_sstate(sk, LISTENING))
6944 sk->sk_max_ack_backlog = backlog;
6945 else {
6946 err = sctp_listen_start(sk, backlog);
6947 if (err)
6948 goto out;
6951 err = 0;
6952 out:
6953 release_sock(sk);
6954 return err;
6958 * This function is done by modeling the current datagram_poll() and the
6959 * tcp_poll(). Note that, based on these implementations, we don't
6960 * lock the socket in this function, even though it seems that,
6961 * ideally, locking or some other mechanisms can be used to ensure
6962 * the integrity of the counters (sndbuf and wmem_alloc) used
6963 * in this place. We assume that we don't need locks either until proven
6964 * otherwise.
6966 * Another thing to note is that we include the Async I/O support
6967 * here, again, by modeling the current TCP/UDP code. We don't have
6968 * a good way to test with it yet.
6970 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6972 struct sock *sk = sock->sk;
6973 struct sctp_sock *sp = sctp_sk(sk);
6974 unsigned int mask;
6976 poll_wait(file, sk_sleep(sk), wait);
6978 sock_rps_record_flow(sk);
6980 /* A TCP-style listening socket becomes readable when the accept queue
6981 * is not empty.
6983 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6984 return (!list_empty(&sp->ep->asocs)) ?
6985 (POLLIN | POLLRDNORM) : 0;
6987 mask = 0;
6989 /* Is there any exceptional events? */
6990 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6991 mask |= POLLERR |
6992 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6993 if (sk->sk_shutdown & RCV_SHUTDOWN)
6994 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6995 if (sk->sk_shutdown == SHUTDOWN_MASK)
6996 mask |= POLLHUP;
6998 /* Is it readable? Reconsider this code with TCP-style support. */
6999 if (!skb_queue_empty(&sk->sk_receive_queue))
7000 mask |= POLLIN | POLLRDNORM;
7002 /* The association is either gone or not ready. */
7003 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7004 return mask;
7006 /* Is it writable? */
7007 if (sctp_writeable(sk)) {
7008 mask |= POLLOUT | POLLWRNORM;
7009 } else {
7010 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7012 * Since the socket is not locked, the buffer
7013 * might be made available after the writeable check and
7014 * before the bit is set. This could cause a lost I/O
7015 * signal. tcp_poll() has a race breaker for this race
7016 * condition. Based on their implementation, we put
7017 * in the following code to cover it as well.
7019 if (sctp_writeable(sk))
7020 mask |= POLLOUT | POLLWRNORM;
7022 return mask;
7025 /********************************************************************
7026 * 2nd Level Abstractions
7027 ********************************************************************/
7029 static struct sctp_bind_bucket *sctp_bucket_create(
7030 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7032 struct sctp_bind_bucket *pp;
7034 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7035 if (pp) {
7036 SCTP_DBG_OBJCNT_INC(bind_bucket);
7037 pp->port = snum;
7038 pp->fastreuse = 0;
7039 INIT_HLIST_HEAD(&pp->owner);
7040 pp->net = net;
7041 hlist_add_head(&pp->node, &head->chain);
7043 return pp;
7046 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7047 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7049 if (pp && hlist_empty(&pp->owner)) {
7050 __hlist_del(&pp->node);
7051 kmem_cache_free(sctp_bucket_cachep, pp);
7052 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7056 /* Release this socket's reference to a local port. */
7057 static inline void __sctp_put_port(struct sock *sk)
7059 struct sctp_bind_hashbucket *head =
7060 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7061 inet_sk(sk)->inet_num)];
7062 struct sctp_bind_bucket *pp;
7064 spin_lock(&head->lock);
7065 pp = sctp_sk(sk)->bind_hash;
7066 __sk_del_bind_node(sk);
7067 sctp_sk(sk)->bind_hash = NULL;
7068 inet_sk(sk)->inet_num = 0;
7069 sctp_bucket_destroy(pp);
7070 spin_unlock(&head->lock);
7073 void sctp_put_port(struct sock *sk)
7075 local_bh_disable();
7076 __sctp_put_port(sk);
7077 local_bh_enable();
7081 * The system picks an ephemeral port and choose an address set equivalent
7082 * to binding with a wildcard address.
7083 * One of those addresses will be the primary address for the association.
7084 * This automatically enables the multihoming capability of SCTP.
7086 static int sctp_autobind(struct sock *sk)
7088 union sctp_addr autoaddr;
7089 struct sctp_af *af;
7090 __be16 port;
7092 /* Initialize a local sockaddr structure to INADDR_ANY. */
7093 af = sctp_sk(sk)->pf->af;
7095 port = htons(inet_sk(sk)->inet_num);
7096 af->inaddr_any(&autoaddr, port);
7098 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7101 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7103 * From RFC 2292
7104 * 4.2 The cmsghdr Structure *
7106 * When ancillary data is sent or received, any number of ancillary data
7107 * objects can be specified by the msg_control and msg_controllen members of
7108 * the msghdr structure, because each object is preceded by
7109 * a cmsghdr structure defining the object's length (the cmsg_len member).
7110 * Historically Berkeley-derived implementations have passed only one object
7111 * at a time, but this API allows multiple objects to be
7112 * passed in a single call to sendmsg() or recvmsg(). The following example
7113 * shows two ancillary data objects in a control buffer.
7115 * |<--------------------------- msg_controllen -------------------------->|
7116 * | |
7118 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7120 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7121 * | | |
7123 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7125 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7126 * | | | | |
7128 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7129 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7131 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7133 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7137 * msg_control
7138 * points here
7140 static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
7142 struct cmsghdr *cmsg;
7143 struct msghdr *my_msg = (struct msghdr *)msg;
7145 for_each_cmsghdr(cmsg, my_msg) {
7146 if (!CMSG_OK(my_msg, cmsg))
7147 return -EINVAL;
7149 /* Should we parse this header or ignore? */
7150 if (cmsg->cmsg_level != IPPROTO_SCTP)
7151 continue;
7153 /* Strictly check lengths following example in SCM code. */
7154 switch (cmsg->cmsg_type) {
7155 case SCTP_INIT:
7156 /* SCTP Socket API Extension
7157 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7159 * This cmsghdr structure provides information for
7160 * initializing new SCTP associations with sendmsg().
7161 * The SCTP_INITMSG socket option uses this same data
7162 * structure. This structure is not used for
7163 * recvmsg().
7165 * cmsg_level cmsg_type cmsg_data[]
7166 * ------------ ------------ ----------------------
7167 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7169 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7170 return -EINVAL;
7172 cmsgs->init = CMSG_DATA(cmsg);
7173 break;
7175 case SCTP_SNDRCV:
7176 /* SCTP Socket API Extension
7177 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7179 * This cmsghdr structure specifies SCTP options for
7180 * sendmsg() and describes SCTP header information
7181 * about a received message through recvmsg().
7183 * cmsg_level cmsg_type cmsg_data[]
7184 * ------------ ------------ ----------------------
7185 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7187 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7188 return -EINVAL;
7190 cmsgs->srinfo = CMSG_DATA(cmsg);
7192 if (cmsgs->srinfo->sinfo_flags &
7193 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7194 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7195 SCTP_ABORT | SCTP_EOF))
7196 return -EINVAL;
7197 break;
7199 case SCTP_SNDINFO:
7200 /* SCTP Socket API Extension
7201 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7203 * This cmsghdr structure specifies SCTP options for
7204 * sendmsg(). This structure and SCTP_RCVINFO replaces
7205 * SCTP_SNDRCV which has been deprecated.
7207 * cmsg_level cmsg_type cmsg_data[]
7208 * ------------ ------------ ---------------------
7209 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7211 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7212 return -EINVAL;
7214 cmsgs->sinfo = CMSG_DATA(cmsg);
7216 if (cmsgs->sinfo->snd_flags &
7217 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7218 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7219 SCTP_ABORT | SCTP_EOF))
7220 return -EINVAL;
7221 break;
7222 default:
7223 return -EINVAL;
7227 return 0;
7231 * Wait for a packet..
7232 * Note: This function is the same function as in core/datagram.c
7233 * with a few modifications to make lksctp work.
7235 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7237 int error;
7238 DEFINE_WAIT(wait);
7240 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7242 /* Socket errors? */
7243 error = sock_error(sk);
7244 if (error)
7245 goto out;
7247 if (!skb_queue_empty(&sk->sk_receive_queue))
7248 goto ready;
7250 /* Socket shut down? */
7251 if (sk->sk_shutdown & RCV_SHUTDOWN)
7252 goto out;
7254 /* Sequenced packets can come disconnected. If so we report the
7255 * problem.
7257 error = -ENOTCONN;
7259 /* Is there a good reason to think that we may receive some data? */
7260 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7261 goto out;
7263 /* Handle signals. */
7264 if (signal_pending(current))
7265 goto interrupted;
7267 /* Let another process have a go. Since we are going to sleep
7268 * anyway. Note: This may cause odd behaviors if the message
7269 * does not fit in the user's buffer, but this seems to be the
7270 * only way to honor MSG_DONTWAIT realistically.
7272 release_sock(sk);
7273 *timeo_p = schedule_timeout(*timeo_p);
7274 lock_sock(sk);
7276 ready:
7277 finish_wait(sk_sleep(sk), &wait);
7278 return 0;
7280 interrupted:
7281 error = sock_intr_errno(*timeo_p);
7283 out:
7284 finish_wait(sk_sleep(sk), &wait);
7285 *err = error;
7286 return error;
7289 /* Receive a datagram.
7290 * Note: This is pretty much the same routine as in core/datagram.c
7291 * with a few changes to make lksctp work.
7293 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7294 int noblock, int *err)
7296 int error;
7297 struct sk_buff *skb;
7298 long timeo;
7300 timeo = sock_rcvtimeo(sk, noblock);
7302 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7303 MAX_SCHEDULE_TIMEOUT);
7305 do {
7306 /* Again only user level code calls this function,
7307 * so nothing interrupt level
7308 * will suddenly eat the receive_queue.
7310 * Look at current nfs client by the way...
7311 * However, this function was correct in any case. 8)
7313 if (flags & MSG_PEEK) {
7314 skb = skb_peek(&sk->sk_receive_queue);
7315 if (skb)
7316 atomic_inc(&skb->users);
7317 } else {
7318 skb = __skb_dequeue(&sk->sk_receive_queue);
7321 if (skb)
7322 return skb;
7324 /* Caller is allowed not to check sk->sk_err before calling. */
7325 error = sock_error(sk);
7326 if (error)
7327 goto no_packet;
7329 if (sk->sk_shutdown & RCV_SHUTDOWN)
7330 break;
7332 if (sk_can_busy_loop(sk) &&
7333 sk_busy_loop(sk, noblock))
7334 continue;
7336 /* User doesn't want to wait. */
7337 error = -EAGAIN;
7338 if (!timeo)
7339 goto no_packet;
7340 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7342 return NULL;
7344 no_packet:
7345 *err = error;
7346 return NULL;
7349 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7350 static void __sctp_write_space(struct sctp_association *asoc)
7352 struct sock *sk = asoc->base.sk;
7354 if (sctp_wspace(asoc) <= 0)
7355 return;
7357 if (waitqueue_active(&asoc->wait))
7358 wake_up_interruptible(&asoc->wait);
7360 if (sctp_writeable(sk)) {
7361 struct socket_wq *wq;
7363 rcu_read_lock();
7364 wq = rcu_dereference(sk->sk_wq);
7365 if (wq) {
7366 if (waitqueue_active(&wq->wait))
7367 wake_up_interruptible(&wq->wait);
7369 /* Note that we try to include the Async I/O support
7370 * here by modeling from the current TCP/UDP code.
7371 * We have not tested with it yet.
7373 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7374 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7376 rcu_read_unlock();
7380 static void sctp_wake_up_waiters(struct sock *sk,
7381 struct sctp_association *asoc)
7383 struct sctp_association *tmp = asoc;
7385 /* We do accounting for the sndbuf space per association,
7386 * so we only need to wake our own association.
7388 if (asoc->ep->sndbuf_policy)
7389 return __sctp_write_space(asoc);
7391 /* If association goes down and is just flushing its
7392 * outq, then just normally notify others.
7394 if (asoc->base.dead)
7395 return sctp_write_space(sk);
7397 /* Accounting for the sndbuf space is per socket, so we
7398 * need to wake up others, try to be fair and in case of
7399 * other associations, let them have a go first instead
7400 * of just doing a sctp_write_space() call.
7402 * Note that we reach sctp_wake_up_waiters() only when
7403 * associations free up queued chunks, thus we are under
7404 * lock and the list of associations on a socket is
7405 * guaranteed not to change.
7407 for (tmp = list_next_entry(tmp, asocs); 1;
7408 tmp = list_next_entry(tmp, asocs)) {
7409 /* Manually skip the head element. */
7410 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7411 continue;
7412 /* Wake up association. */
7413 __sctp_write_space(tmp);
7414 /* We've reached the end. */
7415 if (tmp == asoc)
7416 break;
7420 /* Do accounting for the sndbuf space.
7421 * Decrement the used sndbuf space of the corresponding association by the
7422 * data size which was just transmitted(freed).
7424 static void sctp_wfree(struct sk_buff *skb)
7426 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7427 struct sctp_association *asoc = chunk->asoc;
7428 struct sock *sk = asoc->base.sk;
7430 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7431 sizeof(struct sk_buff) +
7432 sizeof(struct sctp_chunk);
7434 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
7437 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7439 sk->sk_wmem_queued -= skb->truesize;
7440 sk_mem_uncharge(sk, skb->truesize);
7442 sock_wfree(skb);
7443 sctp_wake_up_waiters(sk, asoc);
7445 sctp_association_put(asoc);
7448 /* Do accounting for the receive space on the socket.
7449 * Accounting for the association is done in ulpevent.c
7450 * We set this as a destructor for the cloned data skbs so that
7451 * accounting is done at the correct time.
7453 void sctp_sock_rfree(struct sk_buff *skb)
7455 struct sock *sk = skb->sk;
7456 struct sctp_ulpevent *event = sctp_skb2event(skb);
7458 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7461 * Mimic the behavior of sock_rfree
7463 sk_mem_uncharge(sk, event->rmem_len);
7467 /* Helper function to wait for space in the sndbuf. */
7468 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7469 size_t msg_len)
7471 struct sock *sk = asoc->base.sk;
7472 long current_timeo = *timeo_p;
7473 DEFINE_WAIT(wait);
7474 int err = 0;
7476 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7477 *timeo_p, msg_len);
7479 /* Increment the association's refcnt. */
7480 sctp_association_hold(asoc);
7482 /* Wait on the association specific sndbuf space. */
7483 for (;;) {
7484 prepare_to_wait_exclusive(&asoc->wait, &wait,
7485 TASK_INTERRUPTIBLE);
7486 if (asoc->base.dead)
7487 goto do_dead;
7488 if (!*timeo_p)
7489 goto do_nonblock;
7490 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
7491 goto do_error;
7492 if (signal_pending(current))
7493 goto do_interrupted;
7494 if (msg_len <= sctp_wspace(asoc))
7495 break;
7497 /* Let another process have a go. Since we are going
7498 * to sleep anyway.
7500 release_sock(sk);
7501 current_timeo = schedule_timeout(current_timeo);
7502 lock_sock(sk);
7503 if (sk != asoc->base.sk)
7504 goto do_error;
7506 *timeo_p = current_timeo;
7509 out:
7510 finish_wait(&asoc->wait, &wait);
7512 /* Release the association's refcnt. */
7513 sctp_association_put(asoc);
7515 return err;
7517 do_dead:
7518 err = -ESRCH;
7519 goto out;
7521 do_error:
7522 err = -EPIPE;
7523 goto out;
7525 do_interrupted:
7526 err = sock_intr_errno(*timeo_p);
7527 goto out;
7529 do_nonblock:
7530 err = -EAGAIN;
7531 goto out;
7534 void sctp_data_ready(struct sock *sk)
7536 struct socket_wq *wq;
7538 rcu_read_lock();
7539 wq = rcu_dereference(sk->sk_wq);
7540 if (skwq_has_sleeper(wq))
7541 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7542 POLLRDNORM | POLLRDBAND);
7543 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7544 rcu_read_unlock();
7547 /* If socket sndbuf has changed, wake up all per association waiters. */
7548 void sctp_write_space(struct sock *sk)
7550 struct sctp_association *asoc;
7552 /* Wake up the tasks in each wait queue. */
7553 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7554 __sctp_write_space(asoc);
7558 /* Is there any sndbuf space available on the socket?
7560 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7561 * associations on the same socket. For a UDP-style socket with
7562 * multiple associations, it is possible for it to be "unwriteable"
7563 * prematurely. I assume that this is acceptable because
7564 * a premature "unwriteable" is better than an accidental "writeable" which
7565 * would cause an unwanted block under certain circumstances. For the 1-1
7566 * UDP-style sockets or TCP-style sockets, this code should work.
7567 * - Daisy
7569 static int sctp_writeable(struct sock *sk)
7571 int amt = 0;
7573 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7574 if (amt < 0)
7575 amt = 0;
7576 return amt;
7579 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7580 * returns immediately with EINPROGRESS.
7582 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7584 struct sock *sk = asoc->base.sk;
7585 int err = 0;
7586 long current_timeo = *timeo_p;
7587 DEFINE_WAIT(wait);
7589 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7591 /* Increment the association's refcnt. */
7592 sctp_association_hold(asoc);
7594 for (;;) {
7595 prepare_to_wait_exclusive(&asoc->wait, &wait,
7596 TASK_INTERRUPTIBLE);
7597 if (!*timeo_p)
7598 goto do_nonblock;
7599 if (sk->sk_shutdown & RCV_SHUTDOWN)
7600 break;
7601 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7602 asoc->base.dead)
7603 goto do_error;
7604 if (signal_pending(current))
7605 goto do_interrupted;
7607 if (sctp_state(asoc, ESTABLISHED))
7608 break;
7610 /* Let another process have a go. Since we are going
7611 * to sleep anyway.
7613 release_sock(sk);
7614 current_timeo = schedule_timeout(current_timeo);
7615 lock_sock(sk);
7617 *timeo_p = current_timeo;
7620 out:
7621 finish_wait(&asoc->wait, &wait);
7623 /* Release the association's refcnt. */
7624 sctp_association_put(asoc);
7626 return err;
7628 do_error:
7629 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
7630 err = -ETIMEDOUT;
7631 else
7632 err = -ECONNREFUSED;
7633 goto out;
7635 do_interrupted:
7636 err = sock_intr_errno(*timeo_p);
7637 goto out;
7639 do_nonblock:
7640 err = -EINPROGRESS;
7641 goto out;
7644 static int sctp_wait_for_accept(struct sock *sk, long timeo)
7646 struct sctp_endpoint *ep;
7647 int err = 0;
7648 DEFINE_WAIT(wait);
7650 ep = sctp_sk(sk)->ep;
7653 for (;;) {
7654 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
7655 TASK_INTERRUPTIBLE);
7657 if (list_empty(&ep->asocs)) {
7658 release_sock(sk);
7659 timeo = schedule_timeout(timeo);
7660 lock_sock(sk);
7663 err = -EINVAL;
7664 if (!sctp_sstate(sk, LISTENING))
7665 break;
7667 err = 0;
7668 if (!list_empty(&ep->asocs))
7669 break;
7671 err = sock_intr_errno(timeo);
7672 if (signal_pending(current))
7673 break;
7675 err = -EAGAIN;
7676 if (!timeo)
7677 break;
7680 finish_wait(sk_sleep(sk), &wait);
7682 return err;
7685 static void sctp_wait_for_close(struct sock *sk, long timeout)
7687 DEFINE_WAIT(wait);
7689 do {
7690 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7691 if (list_empty(&sctp_sk(sk)->ep->asocs))
7692 break;
7693 release_sock(sk);
7694 timeout = schedule_timeout(timeout);
7695 lock_sock(sk);
7696 } while (!signal_pending(current) && timeout);
7698 finish_wait(sk_sleep(sk), &wait);
7701 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
7703 struct sk_buff *frag;
7705 if (!skb->data_len)
7706 goto done;
7708 /* Don't forget the fragments. */
7709 skb_walk_frags(skb, frag)
7710 sctp_skb_set_owner_r_frag(frag, sk);
7712 done:
7713 sctp_skb_set_owner_r(skb, sk);
7716 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
7717 struct sctp_association *asoc)
7719 struct inet_sock *inet = inet_sk(sk);
7720 struct inet_sock *newinet;
7722 newsk->sk_type = sk->sk_type;
7723 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
7724 newsk->sk_flags = sk->sk_flags;
7725 newsk->sk_tsflags = sk->sk_tsflags;
7726 newsk->sk_no_check_tx = sk->sk_no_check_tx;
7727 newsk->sk_no_check_rx = sk->sk_no_check_rx;
7728 newsk->sk_reuse = sk->sk_reuse;
7730 newsk->sk_shutdown = sk->sk_shutdown;
7731 newsk->sk_destruct = sctp_destruct_sock;
7732 newsk->sk_family = sk->sk_family;
7733 newsk->sk_protocol = IPPROTO_SCTP;
7734 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
7735 newsk->sk_sndbuf = sk->sk_sndbuf;
7736 newsk->sk_rcvbuf = sk->sk_rcvbuf;
7737 newsk->sk_lingertime = sk->sk_lingertime;
7738 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
7739 newsk->sk_sndtimeo = sk->sk_sndtimeo;
7740 newsk->sk_rxhash = sk->sk_rxhash;
7742 newinet = inet_sk(newsk);
7744 /* Initialize sk's sport, dport, rcv_saddr and daddr for
7745 * getsockname() and getpeername()
7747 newinet->inet_sport = inet->inet_sport;
7748 newinet->inet_saddr = inet->inet_saddr;
7749 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
7750 newinet->inet_dport = htons(asoc->peer.port);
7751 newinet->pmtudisc = inet->pmtudisc;
7752 newinet->inet_id = prandom_u32();
7754 newinet->uc_ttl = inet->uc_ttl;
7755 newinet->mc_loop = 1;
7756 newinet->mc_ttl = 1;
7757 newinet->mc_index = 0;
7758 newinet->mc_list = NULL;
7760 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
7761 net_enable_timestamp();
7763 security_sk_clone(sk, newsk);
7766 static inline void sctp_copy_descendant(struct sock *sk_to,
7767 const struct sock *sk_from)
7769 int ancestor_size = sizeof(struct inet_sock) +
7770 sizeof(struct sctp_sock) -
7771 offsetof(struct sctp_sock, auto_asconf_list);
7773 if (sk_from->sk_family == PF_INET6)
7774 ancestor_size += sizeof(struct ipv6_pinfo);
7776 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
7779 /* Populate the fields of the newsk from the oldsk and migrate the assoc
7780 * and its messages to the newsk.
7782 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
7783 struct sctp_association *assoc,
7784 sctp_socket_type_t type)
7786 struct sctp_sock *oldsp = sctp_sk(oldsk);
7787 struct sctp_sock *newsp = sctp_sk(newsk);
7788 struct sctp_bind_bucket *pp; /* hash list port iterator */
7789 struct sctp_endpoint *newep = newsp->ep;
7790 struct sk_buff *skb, *tmp;
7791 struct sctp_ulpevent *event;
7792 struct sctp_bind_hashbucket *head;
7794 /* Migrate socket buffer sizes and all the socket level options to the
7795 * new socket.
7797 newsk->sk_sndbuf = oldsk->sk_sndbuf;
7798 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
7799 /* Brute force copy old sctp opt. */
7800 sctp_copy_descendant(newsk, oldsk);
7802 /* Restore the ep value that was overwritten with the above structure
7803 * copy.
7805 newsp->ep = newep;
7806 newsp->hmac = NULL;
7808 /* Hook this new socket in to the bind_hash list. */
7809 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7810 inet_sk(oldsk)->inet_num)];
7811 spin_lock_bh(&head->lock);
7812 pp = sctp_sk(oldsk)->bind_hash;
7813 sk_add_bind_node(newsk, &pp->owner);
7814 sctp_sk(newsk)->bind_hash = pp;
7815 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7816 spin_unlock_bh(&head->lock);
7818 /* Copy the bind_addr list from the original endpoint to the new
7819 * endpoint so that we can handle restarts properly
7821 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7822 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7824 /* Move any messages in the old socket's receive queue that are for the
7825 * peeled off association to the new socket's receive queue.
7827 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7828 event = sctp_skb2event(skb);
7829 if (event->asoc == assoc) {
7830 __skb_unlink(skb, &oldsk->sk_receive_queue);
7831 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7832 sctp_skb_set_owner_r_frag(skb, newsk);
7836 /* Clean up any messages pending delivery due to partial
7837 * delivery. Three cases:
7838 * 1) No partial deliver; no work.
7839 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7840 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7842 skb_queue_head_init(&newsp->pd_lobby);
7843 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7845 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7846 struct sk_buff_head *queue;
7848 /* Decide which queue to move pd_lobby skbs to. */
7849 if (assoc->ulpq.pd_mode) {
7850 queue = &newsp->pd_lobby;
7851 } else
7852 queue = &newsk->sk_receive_queue;
7854 /* Walk through the pd_lobby, looking for skbs that
7855 * need moved to the new socket.
7857 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7858 event = sctp_skb2event(skb);
7859 if (event->asoc == assoc) {
7860 __skb_unlink(skb, &oldsp->pd_lobby);
7861 __skb_queue_tail(queue, skb);
7862 sctp_skb_set_owner_r_frag(skb, newsk);
7866 /* Clear up any skbs waiting for the partial
7867 * delivery to finish.
7869 if (assoc->ulpq.pd_mode)
7870 sctp_clear_pd(oldsk, NULL);
7874 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7875 sctp_skb_set_owner_r_frag(skb, newsk);
7877 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7878 sctp_skb_set_owner_r_frag(skb, newsk);
7880 /* Set the type of socket to indicate that it is peeled off from the
7881 * original UDP-style socket or created with the accept() call on a
7882 * TCP-style socket..
7884 newsp->type = type;
7886 /* Mark the new socket "in-use" by the user so that any packets
7887 * that may arrive on the association after we've moved it are
7888 * queued to the backlog. This prevents a potential race between
7889 * backlog processing on the old socket and new-packet processing
7890 * on the new socket.
7892 * The caller has just allocated newsk so we can guarantee that other
7893 * paths won't try to lock it and then oldsk.
7895 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7896 sctp_for_each_tx_datachunk(assoc, true, sctp_clear_owner_w);
7897 sctp_assoc_migrate(assoc, newsk);
7898 sctp_for_each_tx_datachunk(assoc, false, sctp_set_owner_w);
7900 /* If the association on the newsk is already closed before accept()
7901 * is called, set RCV_SHUTDOWN flag.
7903 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
7904 newsk->sk_state = SCTP_SS_CLOSED;
7905 newsk->sk_shutdown |= RCV_SHUTDOWN;
7906 } else {
7907 newsk->sk_state = SCTP_SS_ESTABLISHED;
7910 release_sock(newsk);
7914 /* This proto struct describes the ULP interface for SCTP. */
7915 struct proto sctp_prot = {
7916 .name = "SCTP",
7917 .owner = THIS_MODULE,
7918 .close = sctp_close,
7919 .disconnect = sctp_disconnect,
7920 .accept = sctp_accept,
7921 .ioctl = sctp_ioctl,
7922 .init = sctp_init_sock,
7923 .destroy = sctp_destroy_sock,
7924 .shutdown = sctp_shutdown,
7925 .setsockopt = sctp_setsockopt,
7926 .getsockopt = sctp_getsockopt,
7927 .sendmsg = sctp_sendmsg,
7928 .recvmsg = sctp_recvmsg,
7929 .bind = sctp_bind,
7930 .backlog_rcv = sctp_backlog_rcv,
7931 .hash = sctp_hash,
7932 .unhash = sctp_unhash,
7933 .no_autobind = true,
7934 .obj_size = sizeof(struct sctp_sock),
7935 .sysctl_mem = sysctl_sctp_mem,
7936 .sysctl_rmem = sysctl_sctp_rmem,
7937 .sysctl_wmem = sysctl_sctp_wmem,
7938 .memory_pressure = &sctp_memory_pressure,
7939 .enter_memory_pressure = sctp_enter_memory_pressure,
7940 .memory_allocated = &sctp_memory_allocated,
7941 .sockets_allocated = &sctp_sockets_allocated,
7944 #if IS_ENABLED(CONFIG_IPV6)
7946 #include <net/transp_v6.h>
7947 static void sctp_v6_destroy_sock(struct sock *sk)
7949 sctp_destroy_sock(sk);
7950 inet6_destroy_sock(sk);
7953 struct proto sctpv6_prot = {
7954 .name = "SCTPv6",
7955 .owner = THIS_MODULE,
7956 .close = sctp_close,
7957 .disconnect = sctp_disconnect,
7958 .accept = sctp_accept,
7959 .ioctl = sctp_ioctl,
7960 .init = sctp_init_sock,
7961 .destroy = sctp_v6_destroy_sock,
7962 .shutdown = sctp_shutdown,
7963 .setsockopt = sctp_setsockopt,
7964 .getsockopt = sctp_getsockopt,
7965 .sendmsg = sctp_sendmsg,
7966 .recvmsg = sctp_recvmsg,
7967 .bind = sctp_bind,
7968 .backlog_rcv = sctp_backlog_rcv,
7969 .hash = sctp_hash,
7970 .unhash = sctp_unhash,
7971 .no_autobind = true,
7972 .obj_size = sizeof(struct sctp6_sock),
7973 .sysctl_mem = sysctl_sctp_mem,
7974 .sysctl_rmem = sysctl_sctp_rmem,
7975 .sysctl_wmem = sysctl_sctp_wmem,
7976 .memory_pressure = &sctp_memory_pressure,
7977 .enter_memory_pressure = sctp_enter_memory_pressure,
7978 .memory_allocated = &sctp_memory_allocated,
7979 .sockets_allocated = &sctp_sockets_allocated,
7981 #endif /* IS_ENABLED(CONFIG_IPV6) */