x86/speculation/mds: Fix documentation typo
[linux/fpc-iii.git] / net / sctp / socket.c
blob6a25323705459a974fe7505f11ad6615e1c2d069
1 /* SCTP kernel implementation
2 * (C) Copyright IBM Corp. 2001, 2004
3 * Copyright (c) 1999-2000 Cisco, Inc.
4 * Copyright (c) 1999-2001 Motorola, Inc.
5 * Copyright (c) 2001-2003 Intel Corp.
6 * Copyright (c) 2001-2002 Nokia, Inc.
7 * Copyright (c) 2001 La Monte H.P. Yarroll
9 * This file is part of the SCTP kernel implementation
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
14 * Note that the descriptions from the specification are USER level
15 * functions--this file is the functions which populate the struct proto
16 * for SCTP which is the BOTTOM of the sockets interface.
18 * This SCTP implementation is free software;
19 * you can redistribute it and/or modify it under the terms of
20 * the GNU General Public License as published by
21 * the Free Software Foundation; either version 2, or (at your option)
22 * any later version.
24 * This SCTP implementation is distributed in the hope that it
25 * will be useful, but WITHOUT ANY WARRANTY; without even the implied
26 * ************************
27 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
28 * See the GNU General Public License for more details.
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, see
32 * <http://www.gnu.org/licenses/>.
34 * Please send any bug reports or fixes you make to the
35 * email address(es):
36 * lksctp developers <linux-sctp@vger.kernel.org>
38 * Written or modified by:
39 * La Monte H.P. Yarroll <piggy@acm.org>
40 * Narasimha Budihal <narsi@refcode.org>
41 * Karl Knutson <karl@athena.chicago.il.us>
42 * Jon Grimm <jgrimm@us.ibm.com>
43 * Xingang Guo <xingang.guo@intel.com>
44 * Daisy Chang <daisyc@us.ibm.com>
45 * Sridhar Samudrala <samudrala@us.ibm.com>
46 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
47 * Ardelle Fan <ardelle.fan@intel.com>
48 * Ryan Layer <rmlayer@us.ibm.com>
49 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
50 * Kevin Gao <kevin.gao@intel.com>
53 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
55 #include <crypto/hash.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/wait.h>
59 #include <linux/time.h>
60 #include <linux/sched/signal.h>
61 #include <linux/ip.h>
62 #include <linux/capability.h>
63 #include <linux/fcntl.h>
64 #include <linux/poll.h>
65 #include <linux/init.h>
66 #include <linux/slab.h>
67 #include <linux/file.h>
68 #include <linux/compat.h>
70 #include <net/ip.h>
71 #include <net/icmp.h>
72 #include <net/route.h>
73 #include <net/ipv6.h>
74 #include <net/inet_common.h>
75 #include <net/busy_poll.h>
77 #include <linux/socket.h> /* for sa_family_t */
78 #include <linux/export.h>
79 #include <net/sock.h>
80 #include <net/sctp/sctp.h>
81 #include <net/sctp/sm.h>
83 /* Forward declarations for internal helper functions. */
84 static int sctp_writeable(struct sock *sk);
85 static void sctp_wfree(struct sk_buff *skb);
86 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
87 size_t msg_len);
88 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
89 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
90 static int sctp_wait_for_accept(struct sock *sk, long timeo);
91 static void sctp_wait_for_close(struct sock *sk, long timeo);
92 static void sctp_destruct_sock(struct sock *sk);
93 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
94 union sctp_addr *addr, int len);
95 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
96 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
97 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
98 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf(struct sctp_association *asoc,
100 struct sctp_chunk *chunk);
101 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
102 static int sctp_autobind(struct sock *sk);
103 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
104 struct sctp_association *assoc,
105 enum sctp_socket_type type);
107 static unsigned long sctp_memory_pressure;
108 static atomic_long_t sctp_memory_allocated;
109 struct percpu_counter sctp_sockets_allocated;
111 static void sctp_enter_memory_pressure(struct sock *sk)
113 sctp_memory_pressure = 1;
117 /* Get the sndbuf space available at the time on the association. */
118 static inline int sctp_wspace(struct sctp_association *asoc)
120 int amt;
122 if (asoc->ep->sndbuf_policy)
123 amt = asoc->sndbuf_used;
124 else
125 amt = sk_wmem_alloc_get(asoc->base.sk);
127 if (amt >= asoc->base.sk->sk_sndbuf) {
128 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
129 amt = 0;
130 else {
131 amt = sk_stream_wspace(asoc->base.sk);
132 if (amt < 0)
133 amt = 0;
135 } else {
136 amt = asoc->base.sk->sk_sndbuf - amt;
138 return amt;
141 /* Increment the used sndbuf space count of the corresponding association by
142 * the size of the outgoing data chunk.
143 * Also, set the skb destructor for sndbuf accounting later.
145 * Since it is always 1-1 between chunk and skb, and also a new skb is always
146 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
147 * destructor in the data chunk skb for the purpose of the sndbuf space
148 * tracking.
150 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
152 struct sctp_association *asoc = chunk->asoc;
153 struct sock *sk = asoc->base.sk;
155 /* The sndbuf space is tracked per association. */
156 sctp_association_hold(asoc);
158 skb_set_owner_w(chunk->skb, sk);
160 chunk->skb->destructor = sctp_wfree;
161 /* Save the chunk pointer in skb for sctp_wfree to use later. */
162 skb_shinfo(chunk->skb)->destructor_arg = chunk;
164 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
165 sizeof(struct sk_buff) +
166 sizeof(struct sctp_chunk);
168 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
169 sk->sk_wmem_queued += chunk->skb->truesize;
170 sk_mem_charge(sk, chunk->skb->truesize);
173 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
175 skb_orphan(chunk->skb);
178 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
179 void (*cb)(struct sctp_chunk *))
182 struct sctp_outq *q = &asoc->outqueue;
183 struct sctp_transport *t;
184 struct sctp_chunk *chunk;
186 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
187 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
188 cb(chunk);
190 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
191 cb(chunk);
193 list_for_each_entry(chunk, &q->sacked, transmitted_list)
194 cb(chunk);
196 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
197 cb(chunk);
199 list_for_each_entry(chunk, &q->out_chunk_list, list)
200 cb(chunk);
203 /* Verify that this is a valid address. */
204 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
205 int len)
207 struct sctp_af *af;
209 /* Verify basic sockaddr. */
210 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
211 if (!af)
212 return -EINVAL;
214 /* Is this a valid SCTP address? */
215 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
216 return -EINVAL;
218 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
219 return -EINVAL;
221 return 0;
224 /* Look up the association by its id. If this is not a UDP-style
225 * socket, the ID field is always ignored.
227 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
229 struct sctp_association *asoc = NULL;
231 /* If this is not a UDP-style socket, assoc id should be ignored. */
232 if (!sctp_style(sk, UDP)) {
233 /* Return NULL if the socket state is not ESTABLISHED. It
234 * could be a TCP-style listening socket or a socket which
235 * hasn't yet called connect() to establish an association.
237 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
238 return NULL;
240 /* Get the first and the only association from the list. */
241 if (!list_empty(&sctp_sk(sk)->ep->asocs))
242 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
243 struct sctp_association, asocs);
244 return asoc;
247 /* Otherwise this is a UDP-style socket. */
248 if (!id || (id == (sctp_assoc_t)-1))
249 return NULL;
251 spin_lock_bh(&sctp_assocs_id_lock);
252 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
253 if (asoc && (asoc->base.sk != sk || asoc->base.dead))
254 asoc = NULL;
255 spin_unlock_bh(&sctp_assocs_id_lock);
257 return asoc;
260 /* Look up the transport from an address and an assoc id. If both address and
261 * id are specified, the associations matching the address and the id should be
262 * the same.
264 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
265 struct sockaddr_storage *addr,
266 sctp_assoc_t id)
268 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
269 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
270 union sctp_addr *laddr = (union sctp_addr *)addr;
271 struct sctp_transport *transport;
273 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
274 return NULL;
276 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
277 laddr,
278 &transport);
280 if (!addr_asoc)
281 return NULL;
283 id_asoc = sctp_id2assoc(sk, id);
284 if (id_asoc && (id_asoc != addr_asoc))
285 return NULL;
287 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
288 (union sctp_addr *)addr);
290 return transport;
293 /* API 3.1.2 bind() - UDP Style Syntax
294 * The syntax of bind() is,
296 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
298 * sd - the socket descriptor returned by socket().
299 * addr - the address structure (struct sockaddr_in or struct
300 * sockaddr_in6 [RFC 2553]),
301 * addr_len - the size of the address structure.
303 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
305 int retval = 0;
307 lock_sock(sk);
309 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
310 addr, addr_len);
312 /* Disallow binding twice. */
313 if (!sctp_sk(sk)->ep->base.bind_addr.port)
314 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
315 addr_len);
316 else
317 retval = -EINVAL;
319 release_sock(sk);
321 return retval;
324 static long sctp_get_port_local(struct sock *, union sctp_addr *);
326 /* Verify this is a valid sockaddr. */
327 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
328 union sctp_addr *addr, int len)
330 struct sctp_af *af;
332 /* Check minimum size. */
333 if (len < sizeof (struct sockaddr))
334 return NULL;
336 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
337 return NULL;
339 if (addr->sa.sa_family == AF_INET6) {
340 if (len < SIN6_LEN_RFC2133)
341 return NULL;
342 /* V4 mapped address are really of AF_INET family */
343 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
344 !opt->pf->af_supported(AF_INET, opt))
345 return NULL;
348 /* If we get this far, af is valid. */
349 af = sctp_get_af_specific(addr->sa.sa_family);
351 if (len < af->sockaddr_len)
352 return NULL;
354 return af;
357 /* Bind a local address either to an endpoint or to an association. */
358 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
360 struct net *net = sock_net(sk);
361 struct sctp_sock *sp = sctp_sk(sk);
362 struct sctp_endpoint *ep = sp->ep;
363 struct sctp_bind_addr *bp = &ep->base.bind_addr;
364 struct sctp_af *af;
365 unsigned short snum;
366 int ret = 0;
368 /* Common sockaddr verification. */
369 af = sctp_sockaddr_af(sp, addr, len);
370 if (!af) {
371 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
372 __func__, sk, addr, len);
373 return -EINVAL;
376 snum = ntohs(addr->v4.sin_port);
378 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
379 __func__, sk, &addr->sa, bp->port, snum, len);
381 /* PF specific bind() address verification. */
382 if (!sp->pf->bind_verify(sp, addr))
383 return -EADDRNOTAVAIL;
385 /* We must either be unbound, or bind to the same port.
386 * It's OK to allow 0 ports if we are already bound.
387 * We'll just inhert an already bound port in this case
389 if (bp->port) {
390 if (!snum)
391 snum = bp->port;
392 else if (snum != bp->port) {
393 pr_debug("%s: new port %d doesn't match existing port "
394 "%d\n", __func__, snum, bp->port);
395 return -EINVAL;
399 if (snum && snum < inet_prot_sock(net) &&
400 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
401 return -EACCES;
403 /* See if the address matches any of the addresses we may have
404 * already bound before checking against other endpoints.
406 if (sctp_bind_addr_match(bp, addr, sp))
407 return -EINVAL;
409 /* Make sure we are allowed to bind here.
410 * The function sctp_get_port_local() does duplicate address
411 * detection.
413 addr->v4.sin_port = htons(snum);
414 if ((ret = sctp_get_port_local(sk, addr))) {
415 return -EADDRINUSE;
418 /* Refresh ephemeral port. */
419 if (!bp->port)
420 bp->port = inet_sk(sk)->inet_num;
422 /* Add the address to the bind address list.
423 * Use GFP_ATOMIC since BHs will be disabled.
425 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
426 SCTP_ADDR_SRC, GFP_ATOMIC);
428 /* Copy back into socket for getsockname() use. */
429 if (!ret) {
430 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
431 sp->pf->to_sk_saddr(addr, sk);
434 return ret;
437 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
439 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
440 * at any one time. If a sender, after sending an ASCONF chunk, decides
441 * it needs to transfer another ASCONF Chunk, it MUST wait until the
442 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
443 * subsequent ASCONF. Note this restriction binds each side, so at any
444 * time two ASCONF may be in-transit on any given association (one sent
445 * from each endpoint).
447 static int sctp_send_asconf(struct sctp_association *asoc,
448 struct sctp_chunk *chunk)
450 struct net *net = sock_net(asoc->base.sk);
451 int retval = 0;
453 /* If there is an outstanding ASCONF chunk, queue it for later
454 * transmission.
456 if (asoc->addip_last_asconf) {
457 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
458 goto out;
461 /* Hold the chunk until an ASCONF_ACK is received. */
462 sctp_chunk_hold(chunk);
463 retval = sctp_primitive_ASCONF(net, asoc, chunk);
464 if (retval)
465 sctp_chunk_free(chunk);
466 else
467 asoc->addip_last_asconf = chunk;
469 out:
470 return retval;
473 /* Add a list of addresses as bind addresses to local endpoint or
474 * association.
476 * Basically run through each address specified in the addrs/addrcnt
477 * array/length pair, determine if it is IPv6 or IPv4 and call
478 * sctp_do_bind() on it.
480 * If any of them fails, then the operation will be reversed and the
481 * ones that were added will be removed.
483 * Only sctp_setsockopt_bindx() is supposed to call this function.
485 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
487 int cnt;
488 int retval = 0;
489 void *addr_buf;
490 struct sockaddr *sa_addr;
491 struct sctp_af *af;
493 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
494 addrs, addrcnt);
496 addr_buf = addrs;
497 for (cnt = 0; cnt < addrcnt; cnt++) {
498 /* The list may contain either IPv4 or IPv6 address;
499 * determine the address length for walking thru the list.
501 sa_addr = addr_buf;
502 af = sctp_get_af_specific(sa_addr->sa_family);
503 if (!af) {
504 retval = -EINVAL;
505 goto err_bindx_add;
508 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
509 af->sockaddr_len);
511 addr_buf += af->sockaddr_len;
513 err_bindx_add:
514 if (retval < 0) {
515 /* Failed. Cleanup the ones that have been added */
516 if (cnt > 0)
517 sctp_bindx_rem(sk, addrs, cnt);
518 return retval;
522 return retval;
525 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
526 * associations that are part of the endpoint indicating that a list of local
527 * addresses are added to the endpoint.
529 * If any of the addresses is already in the bind address list of the
530 * association, we do not send the chunk for that association. But it will not
531 * affect other associations.
533 * Only sctp_setsockopt_bindx() is supposed to call this function.
535 static int sctp_send_asconf_add_ip(struct sock *sk,
536 struct sockaddr *addrs,
537 int addrcnt)
539 struct net *net = sock_net(sk);
540 struct sctp_sock *sp;
541 struct sctp_endpoint *ep;
542 struct sctp_association *asoc;
543 struct sctp_bind_addr *bp;
544 struct sctp_chunk *chunk;
545 struct sctp_sockaddr_entry *laddr;
546 union sctp_addr *addr;
547 union sctp_addr saveaddr;
548 void *addr_buf;
549 struct sctp_af *af;
550 struct list_head *p;
551 int i;
552 int retval = 0;
554 if (!net->sctp.addip_enable)
555 return retval;
557 sp = sctp_sk(sk);
558 ep = sp->ep;
560 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
561 __func__, sk, addrs, addrcnt);
563 list_for_each_entry(asoc, &ep->asocs, asocs) {
564 if (!asoc->peer.asconf_capable)
565 continue;
567 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
568 continue;
570 if (!sctp_state(asoc, ESTABLISHED))
571 continue;
573 /* Check if any address in the packed array of addresses is
574 * in the bind address list of the association. If so,
575 * do not send the asconf chunk to its peer, but continue with
576 * other associations.
578 addr_buf = addrs;
579 for (i = 0; i < addrcnt; i++) {
580 addr = addr_buf;
581 af = sctp_get_af_specific(addr->v4.sin_family);
582 if (!af) {
583 retval = -EINVAL;
584 goto out;
587 if (sctp_assoc_lookup_laddr(asoc, addr))
588 break;
590 addr_buf += af->sockaddr_len;
592 if (i < addrcnt)
593 continue;
595 /* Use the first valid address in bind addr list of
596 * association as Address Parameter of ASCONF CHUNK.
598 bp = &asoc->base.bind_addr;
599 p = bp->address_list.next;
600 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
601 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
602 addrcnt, SCTP_PARAM_ADD_IP);
603 if (!chunk) {
604 retval = -ENOMEM;
605 goto out;
608 /* Add the new addresses to the bind address list with
609 * use_as_src set to 0.
611 addr_buf = addrs;
612 for (i = 0; i < addrcnt; i++) {
613 addr = addr_buf;
614 af = sctp_get_af_specific(addr->v4.sin_family);
615 memcpy(&saveaddr, addr, af->sockaddr_len);
616 retval = sctp_add_bind_addr(bp, &saveaddr,
617 sizeof(saveaddr),
618 SCTP_ADDR_NEW, GFP_ATOMIC);
619 addr_buf += af->sockaddr_len;
621 if (asoc->src_out_of_asoc_ok) {
622 struct sctp_transport *trans;
624 list_for_each_entry(trans,
625 &asoc->peer.transport_addr_list, transports) {
626 /* Clear the source and route cache */
627 sctp_transport_dst_release(trans);
628 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
629 2*asoc->pathmtu, 4380));
630 trans->ssthresh = asoc->peer.i.a_rwnd;
631 trans->rto = asoc->rto_initial;
632 sctp_max_rto(asoc, trans);
633 trans->rtt = trans->srtt = trans->rttvar = 0;
634 sctp_transport_route(trans, NULL,
635 sctp_sk(asoc->base.sk));
638 retval = sctp_send_asconf(asoc, chunk);
641 out:
642 return retval;
645 /* Remove a list of addresses from bind addresses list. Do not remove the
646 * last address.
648 * Basically run through each address specified in the addrs/addrcnt
649 * array/length pair, determine if it is IPv6 or IPv4 and call
650 * sctp_del_bind() on it.
652 * If any of them fails, then the operation will be reversed and the
653 * ones that were removed will be added back.
655 * At least one address has to be left; if only one address is
656 * available, the operation will return -EBUSY.
658 * Only sctp_setsockopt_bindx() is supposed to call this function.
660 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
662 struct sctp_sock *sp = sctp_sk(sk);
663 struct sctp_endpoint *ep = sp->ep;
664 int cnt;
665 struct sctp_bind_addr *bp = &ep->base.bind_addr;
666 int retval = 0;
667 void *addr_buf;
668 union sctp_addr *sa_addr;
669 struct sctp_af *af;
671 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
672 __func__, sk, addrs, addrcnt);
674 addr_buf = addrs;
675 for (cnt = 0; cnt < addrcnt; cnt++) {
676 /* If the bind address list is empty or if there is only one
677 * bind address, there is nothing more to be removed (we need
678 * at least one address here).
680 if (list_empty(&bp->address_list) ||
681 (sctp_list_single_entry(&bp->address_list))) {
682 retval = -EBUSY;
683 goto err_bindx_rem;
686 sa_addr = addr_buf;
687 af = sctp_get_af_specific(sa_addr->sa.sa_family);
688 if (!af) {
689 retval = -EINVAL;
690 goto err_bindx_rem;
693 if (!af->addr_valid(sa_addr, sp, NULL)) {
694 retval = -EADDRNOTAVAIL;
695 goto err_bindx_rem;
698 if (sa_addr->v4.sin_port &&
699 sa_addr->v4.sin_port != htons(bp->port)) {
700 retval = -EINVAL;
701 goto err_bindx_rem;
704 if (!sa_addr->v4.sin_port)
705 sa_addr->v4.sin_port = htons(bp->port);
707 /* FIXME - There is probably a need to check if sk->sk_saddr and
708 * sk->sk_rcv_addr are currently set to one of the addresses to
709 * be removed. This is something which needs to be looked into
710 * when we are fixing the outstanding issues with multi-homing
711 * socket routing and failover schemes. Refer to comments in
712 * sctp_do_bind(). -daisy
714 retval = sctp_del_bind_addr(bp, sa_addr);
716 addr_buf += af->sockaddr_len;
717 err_bindx_rem:
718 if (retval < 0) {
719 /* Failed. Add the ones that has been removed back */
720 if (cnt > 0)
721 sctp_bindx_add(sk, addrs, cnt);
722 return retval;
726 return retval;
729 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
730 * the associations that are part of the endpoint indicating that a list of
731 * local addresses are removed from the endpoint.
733 * If any of the addresses is already in the bind address list of the
734 * association, we do not send the chunk for that association. But it will not
735 * affect other associations.
737 * Only sctp_setsockopt_bindx() is supposed to call this function.
739 static int sctp_send_asconf_del_ip(struct sock *sk,
740 struct sockaddr *addrs,
741 int addrcnt)
743 struct net *net = sock_net(sk);
744 struct sctp_sock *sp;
745 struct sctp_endpoint *ep;
746 struct sctp_association *asoc;
747 struct sctp_transport *transport;
748 struct sctp_bind_addr *bp;
749 struct sctp_chunk *chunk;
750 union sctp_addr *laddr;
751 void *addr_buf;
752 struct sctp_af *af;
753 struct sctp_sockaddr_entry *saddr;
754 int i;
755 int retval = 0;
756 int stored = 0;
758 chunk = NULL;
759 if (!net->sctp.addip_enable)
760 return retval;
762 sp = sctp_sk(sk);
763 ep = sp->ep;
765 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
766 __func__, sk, addrs, addrcnt);
768 list_for_each_entry(asoc, &ep->asocs, asocs) {
770 if (!asoc->peer.asconf_capable)
771 continue;
773 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
774 continue;
776 if (!sctp_state(asoc, ESTABLISHED))
777 continue;
779 /* Check if any address in the packed array of addresses is
780 * not present in the bind address list of the association.
781 * If so, do not send the asconf chunk to its peer, but
782 * continue with other associations.
784 addr_buf = addrs;
785 for (i = 0; i < addrcnt; i++) {
786 laddr = addr_buf;
787 af = sctp_get_af_specific(laddr->v4.sin_family);
788 if (!af) {
789 retval = -EINVAL;
790 goto out;
793 if (!sctp_assoc_lookup_laddr(asoc, laddr))
794 break;
796 addr_buf += af->sockaddr_len;
798 if (i < addrcnt)
799 continue;
801 /* Find one address in the association's bind address list
802 * that is not in the packed array of addresses. This is to
803 * make sure that we do not delete all the addresses in the
804 * association.
806 bp = &asoc->base.bind_addr;
807 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
808 addrcnt, sp);
809 if ((laddr == NULL) && (addrcnt == 1)) {
810 if (asoc->asconf_addr_del_pending)
811 continue;
812 asoc->asconf_addr_del_pending =
813 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
814 if (asoc->asconf_addr_del_pending == NULL) {
815 retval = -ENOMEM;
816 goto out;
818 asoc->asconf_addr_del_pending->sa.sa_family =
819 addrs->sa_family;
820 asoc->asconf_addr_del_pending->v4.sin_port =
821 htons(bp->port);
822 if (addrs->sa_family == AF_INET) {
823 struct sockaddr_in *sin;
825 sin = (struct sockaddr_in *)addrs;
826 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
827 } else if (addrs->sa_family == AF_INET6) {
828 struct sockaddr_in6 *sin6;
830 sin6 = (struct sockaddr_in6 *)addrs;
831 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
834 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
835 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
836 asoc->asconf_addr_del_pending);
838 asoc->src_out_of_asoc_ok = 1;
839 stored = 1;
840 goto skip_mkasconf;
843 if (laddr == NULL)
844 return -EINVAL;
846 /* We do not need RCU protection throughout this loop
847 * because this is done under a socket lock from the
848 * setsockopt call.
850 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
851 SCTP_PARAM_DEL_IP);
852 if (!chunk) {
853 retval = -ENOMEM;
854 goto out;
857 skip_mkasconf:
858 /* Reset use_as_src flag for the addresses in the bind address
859 * list that are to be deleted.
861 addr_buf = addrs;
862 for (i = 0; i < addrcnt; i++) {
863 laddr = addr_buf;
864 af = sctp_get_af_specific(laddr->v4.sin_family);
865 list_for_each_entry(saddr, &bp->address_list, list) {
866 if (sctp_cmp_addr_exact(&saddr->a, laddr))
867 saddr->state = SCTP_ADDR_DEL;
869 addr_buf += af->sockaddr_len;
872 /* Update the route and saddr entries for all the transports
873 * as some of the addresses in the bind address list are
874 * about to be deleted and cannot be used as source addresses.
876 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
877 transports) {
878 sctp_transport_dst_release(transport);
879 sctp_transport_route(transport, NULL,
880 sctp_sk(asoc->base.sk));
883 if (stored)
884 /* We don't need to transmit ASCONF */
885 continue;
886 retval = sctp_send_asconf(asoc, chunk);
888 out:
889 return retval;
892 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
893 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
895 struct sock *sk = sctp_opt2sk(sp);
896 union sctp_addr *addr;
897 struct sctp_af *af;
899 /* It is safe to write port space in caller. */
900 addr = &addrw->a;
901 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
902 af = sctp_get_af_specific(addr->sa.sa_family);
903 if (!af)
904 return -EINVAL;
905 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
906 return -EINVAL;
908 if (addrw->state == SCTP_ADDR_NEW)
909 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
910 else
911 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
914 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
916 * API 8.1
917 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
918 * int flags);
920 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
921 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
922 * or IPv6 addresses.
924 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
925 * Section 3.1.2 for this usage.
927 * addrs is a pointer to an array of one or more socket addresses. Each
928 * address is contained in its appropriate structure (i.e. struct
929 * sockaddr_in or struct sockaddr_in6) the family of the address type
930 * must be used to distinguish the address length (note that this
931 * representation is termed a "packed array" of addresses). The caller
932 * specifies the number of addresses in the array with addrcnt.
934 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
935 * -1, and sets errno to the appropriate error code.
937 * For SCTP, the port given in each socket address must be the same, or
938 * sctp_bindx() will fail, setting errno to EINVAL.
940 * The flags parameter is formed from the bitwise OR of zero or more of
941 * the following currently defined flags:
943 * SCTP_BINDX_ADD_ADDR
945 * SCTP_BINDX_REM_ADDR
947 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
948 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
949 * addresses from the association. The two flags are mutually exclusive;
950 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
951 * not remove all addresses from an association; sctp_bindx() will
952 * reject such an attempt with EINVAL.
954 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
955 * additional addresses with an endpoint after calling bind(). Or use
956 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
957 * socket is associated with so that no new association accepted will be
958 * associated with those addresses. If the endpoint supports dynamic
959 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
960 * endpoint to send the appropriate message to the peer to change the
961 * peers address lists.
963 * Adding and removing addresses from a connected association is
964 * optional functionality. Implementations that do not support this
965 * functionality should return EOPNOTSUPP.
967 * Basically do nothing but copying the addresses from user to kernel
968 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
969 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
970 * from userspace.
972 * We don't use copy_from_user() for optimization: we first do the
973 * sanity checks (buffer size -fast- and access check-healthy
974 * pointer); if all of those succeed, then we can alloc the memory
975 * (expensive operation) needed to copy the data to kernel. Then we do
976 * the copying without checking the user space area
977 * (__copy_from_user()).
979 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
980 * it.
982 * sk The sk of the socket
983 * addrs The pointer to the addresses in user land
984 * addrssize Size of the addrs buffer
985 * op Operation to perform (add or remove, see the flags of
986 * sctp_bindx)
988 * Returns 0 if ok, <0 errno code on error.
990 static int sctp_setsockopt_bindx(struct sock *sk,
991 struct sockaddr __user *addrs,
992 int addrs_size, int op)
994 struct sockaddr *kaddrs;
995 int err;
996 int addrcnt = 0;
997 int walk_size = 0;
998 struct sockaddr *sa_addr;
999 void *addr_buf;
1000 struct sctp_af *af;
1002 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1003 __func__, sk, addrs, addrs_size, op);
1005 if (unlikely(addrs_size <= 0))
1006 return -EINVAL;
1008 /* Check the user passed a healthy pointer. */
1009 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1010 return -EFAULT;
1012 /* Alloc space for the address array in kernel memory. */
1013 kaddrs = kmalloc(addrs_size, GFP_USER | __GFP_NOWARN);
1014 if (unlikely(!kaddrs))
1015 return -ENOMEM;
1017 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1018 kfree(kaddrs);
1019 return -EFAULT;
1022 /* Walk through the addrs buffer and count the number of addresses. */
1023 addr_buf = kaddrs;
1024 while (walk_size < addrs_size) {
1025 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1026 kfree(kaddrs);
1027 return -EINVAL;
1030 sa_addr = addr_buf;
1031 af = sctp_get_af_specific(sa_addr->sa_family);
1033 /* If the address family is not supported or if this address
1034 * causes the address buffer to overflow return EINVAL.
1036 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1037 kfree(kaddrs);
1038 return -EINVAL;
1040 addrcnt++;
1041 addr_buf += af->sockaddr_len;
1042 walk_size += af->sockaddr_len;
1045 /* Do the work. */
1046 switch (op) {
1047 case SCTP_BINDX_ADD_ADDR:
1048 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1049 if (err)
1050 goto out;
1051 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1052 break;
1054 case SCTP_BINDX_REM_ADDR:
1055 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1056 if (err)
1057 goto out;
1058 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1059 break;
1061 default:
1062 err = -EINVAL;
1063 break;
1066 out:
1067 kfree(kaddrs);
1069 return err;
1072 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1074 * Common routine for handling connect() and sctp_connectx().
1075 * Connect will come in with just a single address.
1077 static int __sctp_connect(struct sock *sk,
1078 struct sockaddr *kaddrs,
1079 int addrs_size,
1080 sctp_assoc_t *assoc_id)
1082 struct net *net = sock_net(sk);
1083 struct sctp_sock *sp;
1084 struct sctp_endpoint *ep;
1085 struct sctp_association *asoc = NULL;
1086 struct sctp_association *asoc2;
1087 struct sctp_transport *transport;
1088 union sctp_addr to;
1089 enum sctp_scope scope;
1090 long timeo;
1091 int err = 0;
1092 int addrcnt = 0;
1093 int walk_size = 0;
1094 union sctp_addr *sa_addr = NULL;
1095 void *addr_buf;
1096 unsigned short port;
1097 unsigned int f_flags = 0;
1099 sp = sctp_sk(sk);
1100 ep = sp->ep;
1102 /* connect() cannot be done on a socket that is already in ESTABLISHED
1103 * state - UDP-style peeled off socket or a TCP-style socket that
1104 * is already connected.
1105 * It cannot be done even on a TCP-style listening socket.
1107 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1108 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1109 err = -EISCONN;
1110 goto out_free;
1113 /* Walk through the addrs buffer and count the number of addresses. */
1114 addr_buf = kaddrs;
1115 while (walk_size < addrs_size) {
1116 struct sctp_af *af;
1118 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1119 err = -EINVAL;
1120 goto out_free;
1123 sa_addr = addr_buf;
1124 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1126 /* If the address family is not supported or if this address
1127 * causes the address buffer to overflow return EINVAL.
1129 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1130 err = -EINVAL;
1131 goto out_free;
1134 port = ntohs(sa_addr->v4.sin_port);
1136 /* Save current address so we can work with it */
1137 memcpy(&to, sa_addr, af->sockaddr_len);
1139 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1140 if (err)
1141 goto out_free;
1143 /* Make sure the destination port is correctly set
1144 * in all addresses.
1146 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1147 err = -EINVAL;
1148 goto out_free;
1151 /* Check if there already is a matching association on the
1152 * endpoint (other than the one created here).
1154 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1155 if (asoc2 && asoc2 != asoc) {
1156 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1157 err = -EISCONN;
1158 else
1159 err = -EALREADY;
1160 goto out_free;
1163 /* If we could not find a matching association on the endpoint,
1164 * make sure that there is no peeled-off association matching
1165 * the peer address even on another socket.
1167 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1168 err = -EADDRNOTAVAIL;
1169 goto out_free;
1172 if (!asoc) {
1173 /* If a bind() or sctp_bindx() is not called prior to
1174 * an sctp_connectx() call, the system picks an
1175 * ephemeral port and will choose an address set
1176 * equivalent to binding with a wildcard address.
1178 if (!ep->base.bind_addr.port) {
1179 if (sctp_autobind(sk)) {
1180 err = -EAGAIN;
1181 goto out_free;
1183 } else {
1185 * If an unprivileged user inherits a 1-many
1186 * style socket with open associations on a
1187 * privileged port, it MAY be permitted to
1188 * accept new associations, but it SHOULD NOT
1189 * be permitted to open new associations.
1191 if (ep->base.bind_addr.port <
1192 inet_prot_sock(net) &&
1193 !ns_capable(net->user_ns,
1194 CAP_NET_BIND_SERVICE)) {
1195 err = -EACCES;
1196 goto out_free;
1200 scope = sctp_scope(&to);
1201 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1202 if (!asoc) {
1203 err = -ENOMEM;
1204 goto out_free;
1207 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1208 GFP_KERNEL);
1209 if (err < 0) {
1210 goto out_free;
1215 /* Prime the peer's transport structures. */
1216 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1217 SCTP_UNKNOWN);
1218 if (!transport) {
1219 err = -ENOMEM;
1220 goto out_free;
1223 addrcnt++;
1224 addr_buf += af->sockaddr_len;
1225 walk_size += af->sockaddr_len;
1228 /* In case the user of sctp_connectx() wants an association
1229 * id back, assign one now.
1231 if (assoc_id) {
1232 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1233 if (err < 0)
1234 goto out_free;
1237 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1238 if (err < 0) {
1239 goto out_free;
1242 /* Initialize sk's dport and daddr for getpeername() */
1243 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1244 sp->pf->to_sk_daddr(sa_addr, sk);
1245 sk->sk_err = 0;
1247 /* in-kernel sockets don't generally have a file allocated to them
1248 * if all they do is call sock_create_kern().
1250 if (sk->sk_socket->file)
1251 f_flags = sk->sk_socket->file->f_flags;
1253 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1255 if (assoc_id)
1256 *assoc_id = asoc->assoc_id;
1257 err = sctp_wait_for_connect(asoc, &timeo);
1258 /* Note: the asoc may be freed after the return of
1259 * sctp_wait_for_connect.
1262 /* Don't free association on exit. */
1263 asoc = NULL;
1265 out_free:
1266 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1267 __func__, asoc, kaddrs, err);
1269 if (asoc) {
1270 /* sctp_primitive_ASSOCIATE may have added this association
1271 * To the hash table, try to unhash it, just in case, its a noop
1272 * if it wasn't hashed so we're safe
1274 sctp_association_free(asoc);
1276 return err;
1279 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1281 * API 8.9
1282 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1283 * sctp_assoc_t *asoc);
1285 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1286 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1287 * or IPv6 addresses.
1289 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1290 * Section 3.1.2 for this usage.
1292 * addrs is a pointer to an array of one or more socket addresses. Each
1293 * address is contained in its appropriate structure (i.e. struct
1294 * sockaddr_in or struct sockaddr_in6) the family of the address type
1295 * must be used to distengish the address length (note that this
1296 * representation is termed a "packed array" of addresses). The caller
1297 * specifies the number of addresses in the array with addrcnt.
1299 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1300 * the association id of the new association. On failure, sctp_connectx()
1301 * returns -1, and sets errno to the appropriate error code. The assoc_id
1302 * is not touched by the kernel.
1304 * For SCTP, the port given in each socket address must be the same, or
1305 * sctp_connectx() will fail, setting errno to EINVAL.
1307 * An application can use sctp_connectx to initiate an association with
1308 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1309 * allows a caller to specify multiple addresses at which a peer can be
1310 * reached. The way the SCTP stack uses the list of addresses to set up
1311 * the association is implementation dependent. This function only
1312 * specifies that the stack will try to make use of all the addresses in
1313 * the list when needed.
1315 * Note that the list of addresses passed in is only used for setting up
1316 * the association. It does not necessarily equal the set of addresses
1317 * the peer uses for the resulting association. If the caller wants to
1318 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1319 * retrieve them after the association has been set up.
1321 * Basically do nothing but copying the addresses from user to kernel
1322 * land and invoking either sctp_connectx(). This is used for tunneling
1323 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1325 * We don't use copy_from_user() for optimization: we first do the
1326 * sanity checks (buffer size -fast- and access check-healthy
1327 * pointer); if all of those succeed, then we can alloc the memory
1328 * (expensive operation) needed to copy the data to kernel. Then we do
1329 * the copying without checking the user space area
1330 * (__copy_from_user()).
1332 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1333 * it.
1335 * sk The sk of the socket
1336 * addrs The pointer to the addresses in user land
1337 * addrssize Size of the addrs buffer
1339 * Returns >=0 if ok, <0 errno code on error.
1341 static int __sctp_setsockopt_connectx(struct sock *sk,
1342 struct sockaddr __user *addrs,
1343 int addrs_size,
1344 sctp_assoc_t *assoc_id)
1346 struct sockaddr *kaddrs;
1347 gfp_t gfp = GFP_KERNEL;
1348 int err = 0;
1350 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1351 __func__, sk, addrs, addrs_size);
1353 if (unlikely(addrs_size <= 0))
1354 return -EINVAL;
1356 /* Check the user passed a healthy pointer. */
1357 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1358 return -EFAULT;
1360 /* Alloc space for the address array in kernel memory. */
1361 if (sk->sk_socket->file)
1362 gfp = GFP_USER | __GFP_NOWARN;
1363 kaddrs = kmalloc(addrs_size, gfp);
1364 if (unlikely(!kaddrs))
1365 return -ENOMEM;
1367 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1368 err = -EFAULT;
1369 } else {
1370 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1373 kfree(kaddrs);
1375 return err;
1379 * This is an older interface. It's kept for backward compatibility
1380 * to the option that doesn't provide association id.
1382 static int sctp_setsockopt_connectx_old(struct sock *sk,
1383 struct sockaddr __user *addrs,
1384 int addrs_size)
1386 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1390 * New interface for the API. The since the API is done with a socket
1391 * option, to make it simple we feed back the association id is as a return
1392 * indication to the call. Error is always negative and association id is
1393 * always positive.
1395 static int sctp_setsockopt_connectx(struct sock *sk,
1396 struct sockaddr __user *addrs,
1397 int addrs_size)
1399 sctp_assoc_t assoc_id = 0;
1400 int err = 0;
1402 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1404 if (err)
1405 return err;
1406 else
1407 return assoc_id;
1411 * New (hopefully final) interface for the API.
1412 * We use the sctp_getaddrs_old structure so that use-space library
1413 * can avoid any unnecessary allocations. The only different part
1414 * is that we store the actual length of the address buffer into the
1415 * addrs_num structure member. That way we can re-use the existing
1416 * code.
1418 #ifdef CONFIG_COMPAT
1419 struct compat_sctp_getaddrs_old {
1420 sctp_assoc_t assoc_id;
1421 s32 addr_num;
1422 compat_uptr_t addrs; /* struct sockaddr * */
1424 #endif
1426 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1427 char __user *optval,
1428 int __user *optlen)
1430 struct sctp_getaddrs_old param;
1431 sctp_assoc_t assoc_id = 0;
1432 int err = 0;
1434 #ifdef CONFIG_COMPAT
1435 if (in_compat_syscall()) {
1436 struct compat_sctp_getaddrs_old param32;
1438 if (len < sizeof(param32))
1439 return -EINVAL;
1440 if (copy_from_user(&param32, optval, sizeof(param32)))
1441 return -EFAULT;
1443 param.assoc_id = param32.assoc_id;
1444 param.addr_num = param32.addr_num;
1445 param.addrs = compat_ptr(param32.addrs);
1446 } else
1447 #endif
1449 if (len < sizeof(param))
1450 return -EINVAL;
1451 if (copy_from_user(&param, optval, sizeof(param)))
1452 return -EFAULT;
1455 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1456 param.addrs, param.addr_num,
1457 &assoc_id);
1458 if (err == 0 || err == -EINPROGRESS) {
1459 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1460 return -EFAULT;
1461 if (put_user(sizeof(assoc_id), optlen))
1462 return -EFAULT;
1465 return err;
1468 /* API 3.1.4 close() - UDP Style Syntax
1469 * Applications use close() to perform graceful shutdown (as described in
1470 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1471 * by a UDP-style socket.
1473 * The syntax is
1475 * ret = close(int sd);
1477 * sd - the socket descriptor of the associations to be closed.
1479 * To gracefully shutdown a specific association represented by the
1480 * UDP-style socket, an application should use the sendmsg() call,
1481 * passing no user data, but including the appropriate flag in the
1482 * ancillary data (see Section xxxx).
1484 * If sd in the close() call is a branched-off socket representing only
1485 * one association, the shutdown is performed on that association only.
1487 * 4.1.6 close() - TCP Style Syntax
1489 * Applications use close() to gracefully close down an association.
1491 * The syntax is:
1493 * int close(int sd);
1495 * sd - the socket descriptor of the association to be closed.
1497 * After an application calls close() on a socket descriptor, no further
1498 * socket operations will succeed on that descriptor.
1500 * API 7.1.4 SO_LINGER
1502 * An application using the TCP-style socket can use this option to
1503 * perform the SCTP ABORT primitive. The linger option structure is:
1505 * struct linger {
1506 * int l_onoff; // option on/off
1507 * int l_linger; // linger time
1508 * };
1510 * To enable the option, set l_onoff to 1. If the l_linger value is set
1511 * to 0, calling close() is the same as the ABORT primitive. If the
1512 * value is set to a negative value, the setsockopt() call will return
1513 * an error. If the value is set to a positive value linger_time, the
1514 * close() can be blocked for at most linger_time ms. If the graceful
1515 * shutdown phase does not finish during this period, close() will
1516 * return but the graceful shutdown phase continues in the system.
1518 static void sctp_close(struct sock *sk, long timeout)
1520 struct net *net = sock_net(sk);
1521 struct sctp_endpoint *ep;
1522 struct sctp_association *asoc;
1523 struct list_head *pos, *temp;
1524 unsigned int data_was_unread;
1526 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1528 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1529 sk->sk_shutdown = SHUTDOWN_MASK;
1530 sk->sk_state = SCTP_SS_CLOSING;
1532 ep = sctp_sk(sk)->ep;
1534 /* Clean up any skbs sitting on the receive queue. */
1535 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1536 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1538 /* Walk all associations on an endpoint. */
1539 list_for_each_safe(pos, temp, &ep->asocs) {
1540 asoc = list_entry(pos, struct sctp_association, asocs);
1542 if (sctp_style(sk, TCP)) {
1543 /* A closed association can still be in the list if
1544 * it belongs to a TCP-style listening socket that is
1545 * not yet accepted. If so, free it. If not, send an
1546 * ABORT or SHUTDOWN based on the linger options.
1548 if (sctp_state(asoc, CLOSED)) {
1549 sctp_association_free(asoc);
1550 continue;
1554 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1555 !skb_queue_empty(&asoc->ulpq.reasm) ||
1556 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1557 struct sctp_chunk *chunk;
1559 chunk = sctp_make_abort_user(asoc, NULL, 0);
1560 sctp_primitive_ABORT(net, asoc, chunk);
1561 } else
1562 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1565 /* On a TCP-style socket, block for at most linger_time if set. */
1566 if (sctp_style(sk, TCP) && timeout)
1567 sctp_wait_for_close(sk, timeout);
1569 /* This will run the backlog queue. */
1570 release_sock(sk);
1572 /* Supposedly, no process has access to the socket, but
1573 * the net layers still may.
1574 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1575 * held and that should be grabbed before socket lock.
1577 spin_lock_bh(&net->sctp.addr_wq_lock);
1578 bh_lock_sock_nested(sk);
1580 /* Hold the sock, since sk_common_release() will put sock_put()
1581 * and we have just a little more cleanup.
1583 sock_hold(sk);
1584 sk_common_release(sk);
1586 bh_unlock_sock(sk);
1587 spin_unlock_bh(&net->sctp.addr_wq_lock);
1589 sock_put(sk);
1591 SCTP_DBG_OBJCNT_DEC(sock);
1594 /* Handle EPIPE error. */
1595 static int sctp_error(struct sock *sk, int flags, int err)
1597 if (err == -EPIPE)
1598 err = sock_error(sk) ? : -EPIPE;
1599 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1600 send_sig(SIGPIPE, current, 0);
1601 return err;
1604 /* API 3.1.3 sendmsg() - UDP Style Syntax
1606 * An application uses sendmsg() and recvmsg() calls to transmit data to
1607 * and receive data from its peer.
1609 * ssize_t sendmsg(int socket, const struct msghdr *message,
1610 * int flags);
1612 * socket - the socket descriptor of the endpoint.
1613 * message - pointer to the msghdr structure which contains a single
1614 * user message and possibly some ancillary data.
1616 * See Section 5 for complete description of the data
1617 * structures.
1619 * flags - flags sent or received with the user message, see Section
1620 * 5 for complete description of the flags.
1622 * Note: This function could use a rewrite especially when explicit
1623 * connect support comes in.
1625 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1627 static int sctp_msghdr_parse(const struct msghdr *msg,
1628 struct sctp_cmsgs *cmsgs);
1630 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
1632 struct net *net = sock_net(sk);
1633 struct sctp_sock *sp;
1634 struct sctp_endpoint *ep;
1635 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1636 struct sctp_transport *transport, *chunk_tp;
1637 struct sctp_chunk *chunk;
1638 union sctp_addr to;
1639 struct sockaddr *msg_name = NULL;
1640 struct sctp_sndrcvinfo default_sinfo;
1641 struct sctp_sndrcvinfo *sinfo;
1642 struct sctp_initmsg *sinit;
1643 sctp_assoc_t associd = 0;
1644 struct sctp_cmsgs cmsgs = { NULL };
1645 enum sctp_scope scope;
1646 bool fill_sinfo_ttl = false, wait_connect = false;
1647 struct sctp_datamsg *datamsg;
1648 int msg_flags = msg->msg_flags;
1649 __u16 sinfo_flags = 0;
1650 long timeo;
1651 int err;
1653 err = 0;
1654 sp = sctp_sk(sk);
1655 ep = sp->ep;
1657 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1658 msg, msg_len, ep);
1660 /* We cannot send a message over a TCP-style listening socket. */
1661 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1662 err = -EPIPE;
1663 goto out_nounlock;
1666 /* Parse out the SCTP CMSGs. */
1667 err = sctp_msghdr_parse(msg, &cmsgs);
1668 if (err) {
1669 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1670 goto out_nounlock;
1673 /* Fetch the destination address for this packet. This
1674 * address only selects the association--it is not necessarily
1675 * the address we will send to.
1676 * For a peeled-off socket, msg_name is ignored.
1678 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1679 int msg_namelen = msg->msg_namelen;
1681 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1682 msg_namelen);
1683 if (err)
1684 return err;
1686 if (msg_namelen > sizeof(to))
1687 msg_namelen = sizeof(to);
1688 memcpy(&to, msg->msg_name, msg_namelen);
1689 msg_name = msg->msg_name;
1692 sinit = cmsgs.init;
1693 if (cmsgs.sinfo != NULL) {
1694 memset(&default_sinfo, 0, sizeof(default_sinfo));
1695 default_sinfo.sinfo_stream = cmsgs.sinfo->snd_sid;
1696 default_sinfo.sinfo_flags = cmsgs.sinfo->snd_flags;
1697 default_sinfo.sinfo_ppid = cmsgs.sinfo->snd_ppid;
1698 default_sinfo.sinfo_context = cmsgs.sinfo->snd_context;
1699 default_sinfo.sinfo_assoc_id = cmsgs.sinfo->snd_assoc_id;
1701 sinfo = &default_sinfo;
1702 fill_sinfo_ttl = true;
1703 } else {
1704 sinfo = cmsgs.srinfo;
1706 /* Did the user specify SNDINFO/SNDRCVINFO? */
1707 if (sinfo) {
1708 sinfo_flags = sinfo->sinfo_flags;
1709 associd = sinfo->sinfo_assoc_id;
1712 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1713 msg_len, sinfo_flags);
1715 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1716 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1717 err = -EINVAL;
1718 goto out_nounlock;
1721 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1722 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1723 * If SCTP_ABORT is set, the message length could be non zero with
1724 * the msg_iov set to the user abort reason.
1726 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1727 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1728 err = -EINVAL;
1729 goto out_nounlock;
1732 /* If SCTP_ADDR_OVER is set, there must be an address
1733 * specified in msg_name.
1735 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1736 err = -EINVAL;
1737 goto out_nounlock;
1740 transport = NULL;
1742 pr_debug("%s: about to look up association\n", __func__);
1744 lock_sock(sk);
1746 /* If a msg_name has been specified, assume this is to be used. */
1747 if (msg_name) {
1748 /* Look for a matching association on the endpoint. */
1749 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1751 /* If we could not find a matching association on the
1752 * endpoint, make sure that it is not a TCP-style
1753 * socket that already has an association or there is
1754 * no peeled-off association on another socket.
1756 if (!asoc &&
1757 ((sctp_style(sk, TCP) &&
1758 (sctp_sstate(sk, ESTABLISHED) ||
1759 sctp_sstate(sk, CLOSING))) ||
1760 sctp_endpoint_is_peeled_off(ep, &to))) {
1761 err = -EADDRNOTAVAIL;
1762 goto out_unlock;
1764 } else {
1765 asoc = sctp_id2assoc(sk, associd);
1766 if (!asoc) {
1767 err = -EPIPE;
1768 goto out_unlock;
1772 if (asoc) {
1773 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1775 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1776 * socket that has an association in CLOSED state. This can
1777 * happen when an accepted socket has an association that is
1778 * already CLOSED.
1780 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1781 err = -EPIPE;
1782 goto out_unlock;
1785 if (sinfo_flags & SCTP_EOF) {
1786 pr_debug("%s: shutting down association:%p\n",
1787 __func__, asoc);
1789 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1790 err = 0;
1791 goto out_unlock;
1793 if (sinfo_flags & SCTP_ABORT) {
1795 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1796 if (!chunk) {
1797 err = -ENOMEM;
1798 goto out_unlock;
1801 pr_debug("%s: aborting association:%p\n",
1802 __func__, asoc);
1804 sctp_primitive_ABORT(net, asoc, chunk);
1805 err = 0;
1806 goto out_unlock;
1810 /* Do we need to create the association? */
1811 if (!asoc) {
1812 pr_debug("%s: there is no association yet\n", __func__);
1814 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1815 err = -EINVAL;
1816 goto out_unlock;
1819 /* Check for invalid stream against the stream counts,
1820 * either the default or the user specified stream counts.
1822 if (sinfo) {
1823 if (!sinit || !sinit->sinit_num_ostreams) {
1824 /* Check against the defaults. */
1825 if (sinfo->sinfo_stream >=
1826 sp->initmsg.sinit_num_ostreams) {
1827 err = -EINVAL;
1828 goto out_unlock;
1830 } else {
1831 /* Check against the requested. */
1832 if (sinfo->sinfo_stream >=
1833 sinit->sinit_num_ostreams) {
1834 err = -EINVAL;
1835 goto out_unlock;
1841 * API 3.1.2 bind() - UDP Style Syntax
1842 * If a bind() or sctp_bindx() is not called prior to a
1843 * sendmsg() call that initiates a new association, the
1844 * system picks an ephemeral port and will choose an address
1845 * set equivalent to binding with a wildcard address.
1847 if (!ep->base.bind_addr.port) {
1848 if (sctp_autobind(sk)) {
1849 err = -EAGAIN;
1850 goto out_unlock;
1852 } else {
1854 * If an unprivileged user inherits a one-to-many
1855 * style socket with open associations on a privileged
1856 * port, it MAY be permitted to accept new associations,
1857 * but it SHOULD NOT be permitted to open new
1858 * associations.
1860 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1861 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1862 err = -EACCES;
1863 goto out_unlock;
1867 scope = sctp_scope(&to);
1868 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1869 if (!new_asoc) {
1870 err = -ENOMEM;
1871 goto out_unlock;
1873 asoc = new_asoc;
1874 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1875 if (err < 0) {
1876 err = -ENOMEM;
1877 goto out_free;
1880 /* If the SCTP_INIT ancillary data is specified, set all
1881 * the association init values accordingly.
1883 if (sinit) {
1884 if (sinit->sinit_num_ostreams) {
1885 __u16 outcnt = sinit->sinit_num_ostreams;
1887 asoc->c.sinit_num_ostreams = outcnt;
1888 /* outcnt has been changed, so re-init stream */
1889 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1890 GFP_KERNEL);
1891 if (err)
1892 goto out_free;
1894 if (sinit->sinit_max_instreams) {
1895 asoc->c.sinit_max_instreams =
1896 sinit->sinit_max_instreams;
1898 if (sinit->sinit_max_attempts) {
1899 asoc->max_init_attempts
1900 = sinit->sinit_max_attempts;
1902 if (sinit->sinit_max_init_timeo) {
1903 asoc->max_init_timeo =
1904 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1908 /* Prime the peer's transport structures. */
1909 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1910 if (!transport) {
1911 err = -ENOMEM;
1912 goto out_free;
1916 /* ASSERT: we have a valid association at this point. */
1917 pr_debug("%s: we have a valid association\n", __func__);
1919 if (!sinfo) {
1920 /* If the user didn't specify SNDINFO/SNDRCVINFO, make up
1921 * one with some defaults.
1923 memset(&default_sinfo, 0, sizeof(default_sinfo));
1924 default_sinfo.sinfo_stream = asoc->default_stream;
1925 default_sinfo.sinfo_flags = asoc->default_flags;
1926 default_sinfo.sinfo_ppid = asoc->default_ppid;
1927 default_sinfo.sinfo_context = asoc->default_context;
1928 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1929 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1931 sinfo = &default_sinfo;
1932 } else if (fill_sinfo_ttl) {
1933 /* In case SNDINFO was specified, we still need to fill
1934 * it with a default ttl from the assoc here.
1936 sinfo->sinfo_timetolive = asoc->default_timetolive;
1939 /* API 7.1.7, the sndbuf size per association bounds the
1940 * maximum size of data that can be sent in a single send call.
1942 if (msg_len > sk->sk_sndbuf) {
1943 err = -EMSGSIZE;
1944 goto out_free;
1947 if (asoc->pmtu_pending)
1948 sctp_assoc_pending_pmtu(asoc);
1950 /* If fragmentation is disabled and the message length exceeds the
1951 * association fragmentation point, return EMSGSIZE. The I-D
1952 * does not specify what this error is, but this looks like
1953 * a great fit.
1955 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1956 err = -EMSGSIZE;
1957 goto out_free;
1960 /* Check for invalid stream. */
1961 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1962 err = -EINVAL;
1963 goto out_free;
1966 if (sctp_wspace(asoc) < msg_len)
1967 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1969 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1970 if (!sctp_wspace(asoc)) {
1971 /* sk can be changed by peel off when waiting for buf. */
1972 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1973 if (err) {
1974 if (err == -ESRCH) {
1975 /* asoc is already dead. */
1976 new_asoc = NULL;
1977 err = -EPIPE;
1979 goto out_free;
1983 /* If an address is passed with the sendto/sendmsg call, it is used
1984 * to override the primary destination address in the TCP model, or
1985 * when SCTP_ADDR_OVER flag is set in the UDP model.
1987 if ((sctp_style(sk, TCP) && msg_name) ||
1988 (sinfo_flags & SCTP_ADDR_OVER)) {
1989 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1990 if (!chunk_tp) {
1991 err = -EINVAL;
1992 goto out_free;
1994 } else
1995 chunk_tp = NULL;
1997 /* Auto-connect, if we aren't connected already. */
1998 if (sctp_state(asoc, CLOSED)) {
1999 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
2000 if (err < 0)
2001 goto out_free;
2003 wait_connect = true;
2004 pr_debug("%s: we associated primitively\n", __func__);
2007 /* Break the message into multiple chunks of maximum size. */
2008 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
2009 if (IS_ERR(datamsg)) {
2010 err = PTR_ERR(datamsg);
2011 goto out_free;
2013 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
2015 /* Now send the (possibly) fragmented message. */
2016 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
2017 sctp_chunk_hold(chunk);
2019 /* Do accounting for the write space. */
2020 sctp_set_owner_w(chunk);
2022 chunk->transport = chunk_tp;
2025 /* Send it to the lower layers. Note: all chunks
2026 * must either fail or succeed. The lower layer
2027 * works that way today. Keep it that way or this
2028 * breaks.
2030 err = sctp_primitive_SEND(net, asoc, datamsg);
2031 /* Did the lower layer accept the chunk? */
2032 if (err) {
2033 sctp_datamsg_free(datamsg);
2034 goto out_free;
2037 pr_debug("%s: we sent primitively\n", __func__);
2039 sctp_datamsg_put(datamsg);
2040 err = msg_len;
2042 if (unlikely(wait_connect)) {
2043 timeo = sock_sndtimeo(sk, msg_flags & MSG_DONTWAIT);
2044 sctp_wait_for_connect(asoc, &timeo);
2047 /* If we are already past ASSOCIATE, the lower
2048 * layers are responsible for association cleanup.
2050 goto out_unlock;
2052 out_free:
2053 if (new_asoc)
2054 sctp_association_free(asoc);
2055 out_unlock:
2056 release_sock(sk);
2058 out_nounlock:
2059 return sctp_error(sk, msg_flags, err);
2061 #if 0
2062 do_sock_err:
2063 if (msg_len)
2064 err = msg_len;
2065 else
2066 err = sock_error(sk);
2067 goto out;
2069 do_interrupted:
2070 if (msg_len)
2071 err = msg_len;
2072 goto out;
2073 #endif /* 0 */
2076 /* This is an extended version of skb_pull() that removes the data from the
2077 * start of a skb even when data is spread across the list of skb's in the
2078 * frag_list. len specifies the total amount of data that needs to be removed.
2079 * when 'len' bytes could be removed from the skb, it returns 0.
2080 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2081 * could not be removed.
2083 static int sctp_skb_pull(struct sk_buff *skb, int len)
2085 struct sk_buff *list;
2086 int skb_len = skb_headlen(skb);
2087 int rlen;
2089 if (len <= skb_len) {
2090 __skb_pull(skb, len);
2091 return 0;
2093 len -= skb_len;
2094 __skb_pull(skb, skb_len);
2096 skb_walk_frags(skb, list) {
2097 rlen = sctp_skb_pull(list, len);
2098 skb->len -= (len-rlen);
2099 skb->data_len -= (len-rlen);
2101 if (!rlen)
2102 return 0;
2104 len = rlen;
2107 return len;
2110 /* API 3.1.3 recvmsg() - UDP Style Syntax
2112 * ssize_t recvmsg(int socket, struct msghdr *message,
2113 * int flags);
2115 * socket - the socket descriptor of the endpoint.
2116 * message - pointer to the msghdr structure which contains a single
2117 * user message and possibly some ancillary data.
2119 * See Section 5 for complete description of the data
2120 * structures.
2122 * flags - flags sent or received with the user message, see Section
2123 * 5 for complete description of the flags.
2125 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2126 int noblock, int flags, int *addr_len)
2128 struct sctp_ulpevent *event = NULL;
2129 struct sctp_sock *sp = sctp_sk(sk);
2130 struct sk_buff *skb, *head_skb;
2131 int copied;
2132 int err = 0;
2133 int skb_len;
2135 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2136 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2137 addr_len);
2139 lock_sock(sk);
2141 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2142 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2143 err = -ENOTCONN;
2144 goto out;
2147 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2148 if (!skb)
2149 goto out;
2151 /* Get the total length of the skb including any skb's in the
2152 * frag_list.
2154 skb_len = skb->len;
2156 copied = skb_len;
2157 if (copied > len)
2158 copied = len;
2160 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2162 event = sctp_skb2event(skb);
2164 if (err)
2165 goto out_free;
2167 if (event->chunk && event->chunk->head_skb)
2168 head_skb = event->chunk->head_skb;
2169 else
2170 head_skb = skb;
2171 sock_recv_ts_and_drops(msg, sk, head_skb);
2172 if (sctp_ulpevent_is_notification(event)) {
2173 msg->msg_flags |= MSG_NOTIFICATION;
2174 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2175 } else {
2176 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2179 /* Check if we allow SCTP_NXTINFO. */
2180 if (sp->recvnxtinfo)
2181 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2182 /* Check if we allow SCTP_RCVINFO. */
2183 if (sp->recvrcvinfo)
2184 sctp_ulpevent_read_rcvinfo(event, msg);
2185 /* Check if we allow SCTP_SNDRCVINFO. */
2186 if (sp->subscribe.sctp_data_io_event)
2187 sctp_ulpevent_read_sndrcvinfo(event, msg);
2189 err = copied;
2191 /* If skb's length exceeds the user's buffer, update the skb and
2192 * push it back to the receive_queue so that the next call to
2193 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2195 if (skb_len > copied) {
2196 msg->msg_flags &= ~MSG_EOR;
2197 if (flags & MSG_PEEK)
2198 goto out_free;
2199 sctp_skb_pull(skb, copied);
2200 skb_queue_head(&sk->sk_receive_queue, skb);
2202 /* When only partial message is copied to the user, increase
2203 * rwnd by that amount. If all the data in the skb is read,
2204 * rwnd is updated when the event is freed.
2206 if (!sctp_ulpevent_is_notification(event))
2207 sctp_assoc_rwnd_increase(event->asoc, copied);
2208 goto out;
2209 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2210 (event->msg_flags & MSG_EOR))
2211 msg->msg_flags |= MSG_EOR;
2212 else
2213 msg->msg_flags &= ~MSG_EOR;
2215 out_free:
2216 if (flags & MSG_PEEK) {
2217 /* Release the skb reference acquired after peeking the skb in
2218 * sctp_skb_recv_datagram().
2220 kfree_skb(skb);
2221 } else {
2222 /* Free the event which includes releasing the reference to
2223 * the owner of the skb, freeing the skb and updating the
2224 * rwnd.
2226 sctp_ulpevent_free(event);
2228 out:
2229 release_sock(sk);
2230 return err;
2233 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2235 * This option is a on/off flag. If enabled no SCTP message
2236 * fragmentation will be performed. Instead if a message being sent
2237 * exceeds the current PMTU size, the message will NOT be sent and
2238 * instead a error will be indicated to the user.
2240 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2241 char __user *optval,
2242 unsigned int optlen)
2244 int val;
2246 if (optlen < sizeof(int))
2247 return -EINVAL;
2249 if (get_user(val, (int __user *)optval))
2250 return -EFAULT;
2252 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2254 return 0;
2257 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2258 unsigned int optlen)
2260 struct sctp_association *asoc;
2261 struct sctp_ulpevent *event;
2263 if (optlen > sizeof(struct sctp_event_subscribe))
2264 return -EINVAL;
2265 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2266 return -EFAULT;
2268 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2269 * if there is no data to be sent or retransmit, the stack will
2270 * immediately send up this notification.
2272 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2273 &sctp_sk(sk)->subscribe)) {
2274 asoc = sctp_id2assoc(sk, 0);
2276 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2277 event = sctp_ulpevent_make_sender_dry_event(asoc,
2278 GFP_ATOMIC);
2279 if (!event)
2280 return -ENOMEM;
2282 sctp_ulpq_tail_event(&asoc->ulpq, event);
2286 return 0;
2289 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2291 * This socket option is applicable to the UDP-style socket only. When
2292 * set it will cause associations that are idle for more than the
2293 * specified number of seconds to automatically close. An association
2294 * being idle is defined an association that has NOT sent or received
2295 * user data. The special value of '0' indicates that no automatic
2296 * close of any associations should be performed. The option expects an
2297 * integer defining the number of seconds of idle time before an
2298 * association is closed.
2300 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2301 unsigned int optlen)
2303 struct sctp_sock *sp = sctp_sk(sk);
2304 struct net *net = sock_net(sk);
2306 /* Applicable to UDP-style socket only */
2307 if (sctp_style(sk, TCP))
2308 return -EOPNOTSUPP;
2309 if (optlen != sizeof(int))
2310 return -EINVAL;
2311 if (copy_from_user(&sp->autoclose, optval, optlen))
2312 return -EFAULT;
2314 if (sp->autoclose > net->sctp.max_autoclose)
2315 sp->autoclose = net->sctp.max_autoclose;
2317 return 0;
2320 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2322 * Applications can enable or disable heartbeats for any peer address of
2323 * an association, modify an address's heartbeat interval, force a
2324 * heartbeat to be sent immediately, and adjust the address's maximum
2325 * number of retransmissions sent before an address is considered
2326 * unreachable. The following structure is used to access and modify an
2327 * address's parameters:
2329 * struct sctp_paddrparams {
2330 * sctp_assoc_t spp_assoc_id;
2331 * struct sockaddr_storage spp_address;
2332 * uint32_t spp_hbinterval;
2333 * uint16_t spp_pathmaxrxt;
2334 * uint32_t spp_pathmtu;
2335 * uint32_t spp_sackdelay;
2336 * uint32_t spp_flags;
2337 * };
2339 * spp_assoc_id - (one-to-many style socket) This is filled in the
2340 * application, and identifies the association for
2341 * this query.
2342 * spp_address - This specifies which address is of interest.
2343 * spp_hbinterval - This contains the value of the heartbeat interval,
2344 * in milliseconds. If a value of zero
2345 * is present in this field then no changes are to
2346 * be made to this parameter.
2347 * spp_pathmaxrxt - This contains the maximum number of
2348 * retransmissions before this address shall be
2349 * considered unreachable. If a value of zero
2350 * is present in this field then no changes are to
2351 * be made to this parameter.
2352 * spp_pathmtu - When Path MTU discovery is disabled the value
2353 * specified here will be the "fixed" path mtu.
2354 * Note that if the spp_address field is empty
2355 * then all associations on this address will
2356 * have this fixed path mtu set upon them.
2358 * spp_sackdelay - When delayed sack is enabled, this value specifies
2359 * the number of milliseconds that sacks will be delayed
2360 * for. This value will apply to all addresses of an
2361 * association if the spp_address field is empty. Note
2362 * also, that if delayed sack is enabled and this
2363 * value is set to 0, no change is made to the last
2364 * recorded delayed sack timer value.
2366 * spp_flags - These flags are used to control various features
2367 * on an association. The flag field may contain
2368 * zero or more of the following options.
2370 * SPP_HB_ENABLE - Enable heartbeats on the
2371 * specified address. Note that if the address
2372 * field is empty all addresses for the association
2373 * have heartbeats enabled upon them.
2375 * SPP_HB_DISABLE - Disable heartbeats on the
2376 * speicifed address. Note that if the address
2377 * field is empty all addresses for the association
2378 * will have their heartbeats disabled. Note also
2379 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2380 * mutually exclusive, only one of these two should
2381 * be specified. Enabling both fields will have
2382 * undetermined results.
2384 * SPP_HB_DEMAND - Request a user initiated heartbeat
2385 * to be made immediately.
2387 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2388 * heartbeat delayis to be set to the value of 0
2389 * milliseconds.
2391 * SPP_PMTUD_ENABLE - This field will enable PMTU
2392 * discovery upon the specified address. Note that
2393 * if the address feild is empty then all addresses
2394 * on the association are effected.
2396 * SPP_PMTUD_DISABLE - This field will disable PMTU
2397 * discovery upon the specified address. Note that
2398 * if the address feild is empty then all addresses
2399 * on the association are effected. Not also that
2400 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2401 * exclusive. Enabling both will have undetermined
2402 * results.
2404 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2405 * on delayed sack. The time specified in spp_sackdelay
2406 * is used to specify the sack delay for this address. Note
2407 * that if spp_address is empty then all addresses will
2408 * enable delayed sack and take on the sack delay
2409 * value specified in spp_sackdelay.
2410 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2411 * off delayed sack. If the spp_address field is blank then
2412 * delayed sack is disabled for the entire association. Note
2413 * also that this field is mutually exclusive to
2414 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2415 * results.
2417 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2418 struct sctp_transport *trans,
2419 struct sctp_association *asoc,
2420 struct sctp_sock *sp,
2421 int hb_change,
2422 int pmtud_change,
2423 int sackdelay_change)
2425 int error;
2427 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2428 struct net *net = sock_net(trans->asoc->base.sk);
2430 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2431 if (error)
2432 return error;
2435 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2436 * this field is ignored. Note also that a value of zero indicates
2437 * the current setting should be left unchanged.
2439 if (params->spp_flags & SPP_HB_ENABLE) {
2441 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2442 * set. This lets us use 0 value when this flag
2443 * is set.
2445 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2446 params->spp_hbinterval = 0;
2448 if (params->spp_hbinterval ||
2449 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2450 if (trans) {
2451 trans->hbinterval =
2452 msecs_to_jiffies(params->spp_hbinterval);
2453 } else if (asoc) {
2454 asoc->hbinterval =
2455 msecs_to_jiffies(params->spp_hbinterval);
2456 } else {
2457 sp->hbinterval = params->spp_hbinterval;
2462 if (hb_change) {
2463 if (trans) {
2464 trans->param_flags =
2465 (trans->param_flags & ~SPP_HB) | hb_change;
2466 } else if (asoc) {
2467 asoc->param_flags =
2468 (asoc->param_flags & ~SPP_HB) | hb_change;
2469 } else {
2470 sp->param_flags =
2471 (sp->param_flags & ~SPP_HB) | hb_change;
2475 /* When Path MTU discovery is disabled the value specified here will
2476 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2477 * include the flag SPP_PMTUD_DISABLE for this field to have any
2478 * effect).
2480 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2481 if (trans) {
2482 trans->pathmtu = params->spp_pathmtu;
2483 sctp_assoc_sync_pmtu(asoc);
2484 } else if (asoc) {
2485 asoc->pathmtu = params->spp_pathmtu;
2486 } else {
2487 sp->pathmtu = params->spp_pathmtu;
2491 if (pmtud_change) {
2492 if (trans) {
2493 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2494 (params->spp_flags & SPP_PMTUD_ENABLE);
2495 trans->param_flags =
2496 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2497 if (update) {
2498 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2499 sctp_assoc_sync_pmtu(asoc);
2501 } else if (asoc) {
2502 asoc->param_flags =
2503 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2504 } else {
2505 sp->param_flags =
2506 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2510 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2511 * value of this field is ignored. Note also that a value of zero
2512 * indicates the current setting should be left unchanged.
2514 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2515 if (trans) {
2516 trans->sackdelay =
2517 msecs_to_jiffies(params->spp_sackdelay);
2518 } else if (asoc) {
2519 asoc->sackdelay =
2520 msecs_to_jiffies(params->spp_sackdelay);
2521 } else {
2522 sp->sackdelay = params->spp_sackdelay;
2526 if (sackdelay_change) {
2527 if (trans) {
2528 trans->param_flags =
2529 (trans->param_flags & ~SPP_SACKDELAY) |
2530 sackdelay_change;
2531 } else if (asoc) {
2532 asoc->param_flags =
2533 (asoc->param_flags & ~SPP_SACKDELAY) |
2534 sackdelay_change;
2535 } else {
2536 sp->param_flags =
2537 (sp->param_flags & ~SPP_SACKDELAY) |
2538 sackdelay_change;
2542 /* Note that a value of zero indicates the current setting should be
2543 left unchanged.
2545 if (params->spp_pathmaxrxt) {
2546 if (trans) {
2547 trans->pathmaxrxt = params->spp_pathmaxrxt;
2548 } else if (asoc) {
2549 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2550 } else {
2551 sp->pathmaxrxt = params->spp_pathmaxrxt;
2555 return 0;
2558 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2559 char __user *optval,
2560 unsigned int optlen)
2562 struct sctp_paddrparams params;
2563 struct sctp_transport *trans = NULL;
2564 struct sctp_association *asoc = NULL;
2565 struct sctp_sock *sp = sctp_sk(sk);
2566 int error;
2567 int hb_change, pmtud_change, sackdelay_change;
2569 if (optlen != sizeof(struct sctp_paddrparams))
2570 return -EINVAL;
2572 if (copy_from_user(&params, optval, optlen))
2573 return -EFAULT;
2575 /* Validate flags and value parameters. */
2576 hb_change = params.spp_flags & SPP_HB;
2577 pmtud_change = params.spp_flags & SPP_PMTUD;
2578 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2580 if (hb_change == SPP_HB ||
2581 pmtud_change == SPP_PMTUD ||
2582 sackdelay_change == SPP_SACKDELAY ||
2583 params.spp_sackdelay > 500 ||
2584 (params.spp_pathmtu &&
2585 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2586 return -EINVAL;
2588 /* If an address other than INADDR_ANY is specified, and
2589 * no transport is found, then the request is invalid.
2591 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2592 trans = sctp_addr_id2transport(sk, &params.spp_address,
2593 params.spp_assoc_id);
2594 if (!trans)
2595 return -EINVAL;
2598 /* Get association, if assoc_id != 0 and the socket is a one
2599 * to many style socket, and an association was not found, then
2600 * the id was invalid.
2602 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2603 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2604 return -EINVAL;
2606 /* Heartbeat demand can only be sent on a transport or
2607 * association, but not a socket.
2609 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2610 return -EINVAL;
2612 /* Process parameters. */
2613 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2614 hb_change, pmtud_change,
2615 sackdelay_change);
2617 if (error)
2618 return error;
2620 /* If changes are for association, also apply parameters to each
2621 * transport.
2623 if (!trans && asoc) {
2624 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2625 transports) {
2626 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2627 hb_change, pmtud_change,
2628 sackdelay_change);
2632 return 0;
2635 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2637 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2640 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2642 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2646 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2648 * This option will effect the way delayed acks are performed. This
2649 * option allows you to get or set the delayed ack time, in
2650 * milliseconds. It also allows changing the delayed ack frequency.
2651 * Changing the frequency to 1 disables the delayed sack algorithm. If
2652 * the assoc_id is 0, then this sets or gets the endpoints default
2653 * values. If the assoc_id field is non-zero, then the set or get
2654 * effects the specified association for the one to many model (the
2655 * assoc_id field is ignored by the one to one model). Note that if
2656 * sack_delay or sack_freq are 0 when setting this option, then the
2657 * current values will remain unchanged.
2659 * struct sctp_sack_info {
2660 * sctp_assoc_t sack_assoc_id;
2661 * uint32_t sack_delay;
2662 * uint32_t sack_freq;
2663 * };
2665 * sack_assoc_id - This parameter, indicates which association the user
2666 * is performing an action upon. Note that if this field's value is
2667 * zero then the endpoints default value is changed (effecting future
2668 * associations only).
2670 * sack_delay - This parameter contains the number of milliseconds that
2671 * the user is requesting the delayed ACK timer be set to. Note that
2672 * this value is defined in the standard to be between 200 and 500
2673 * milliseconds.
2675 * sack_freq - This parameter contains the number of packets that must
2676 * be received before a sack is sent without waiting for the delay
2677 * timer to expire. The default value for this is 2, setting this
2678 * value to 1 will disable the delayed sack algorithm.
2681 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2682 char __user *optval, unsigned int optlen)
2684 struct sctp_sack_info params;
2685 struct sctp_transport *trans = NULL;
2686 struct sctp_association *asoc = NULL;
2687 struct sctp_sock *sp = sctp_sk(sk);
2689 if (optlen == sizeof(struct sctp_sack_info)) {
2690 if (copy_from_user(&params, optval, optlen))
2691 return -EFAULT;
2693 if (params.sack_delay == 0 && params.sack_freq == 0)
2694 return 0;
2695 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2696 pr_warn_ratelimited(DEPRECATED
2697 "%s (pid %d) "
2698 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2699 "Use struct sctp_sack_info instead\n",
2700 current->comm, task_pid_nr(current));
2701 if (copy_from_user(&params, optval, optlen))
2702 return -EFAULT;
2704 if (params.sack_delay == 0)
2705 params.sack_freq = 1;
2706 else
2707 params.sack_freq = 0;
2708 } else
2709 return -EINVAL;
2711 /* Validate value parameter. */
2712 if (params.sack_delay > 500)
2713 return -EINVAL;
2715 /* Get association, if sack_assoc_id != 0 and the socket is a one
2716 * to many style socket, and an association was not found, then
2717 * the id was invalid.
2719 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2720 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2721 return -EINVAL;
2723 if (params.sack_delay) {
2724 if (asoc) {
2725 asoc->sackdelay =
2726 msecs_to_jiffies(params.sack_delay);
2727 asoc->param_flags =
2728 sctp_spp_sackdelay_enable(asoc->param_flags);
2729 } else {
2730 sp->sackdelay = params.sack_delay;
2731 sp->param_flags =
2732 sctp_spp_sackdelay_enable(sp->param_flags);
2736 if (params.sack_freq == 1) {
2737 if (asoc) {
2738 asoc->param_flags =
2739 sctp_spp_sackdelay_disable(asoc->param_flags);
2740 } else {
2741 sp->param_flags =
2742 sctp_spp_sackdelay_disable(sp->param_flags);
2744 } else if (params.sack_freq > 1) {
2745 if (asoc) {
2746 asoc->sackfreq = params.sack_freq;
2747 asoc->param_flags =
2748 sctp_spp_sackdelay_enable(asoc->param_flags);
2749 } else {
2750 sp->sackfreq = params.sack_freq;
2751 sp->param_flags =
2752 sctp_spp_sackdelay_enable(sp->param_flags);
2756 /* If change is for association, also apply to each transport. */
2757 if (asoc) {
2758 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2759 transports) {
2760 if (params.sack_delay) {
2761 trans->sackdelay =
2762 msecs_to_jiffies(params.sack_delay);
2763 trans->param_flags =
2764 sctp_spp_sackdelay_enable(trans->param_flags);
2766 if (params.sack_freq == 1) {
2767 trans->param_flags =
2768 sctp_spp_sackdelay_disable(trans->param_flags);
2769 } else if (params.sack_freq > 1) {
2770 trans->sackfreq = params.sack_freq;
2771 trans->param_flags =
2772 sctp_spp_sackdelay_enable(trans->param_flags);
2777 return 0;
2780 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2782 * Applications can specify protocol parameters for the default association
2783 * initialization. The option name argument to setsockopt() and getsockopt()
2784 * is SCTP_INITMSG.
2786 * Setting initialization parameters is effective only on an unconnected
2787 * socket (for UDP-style sockets only future associations are effected
2788 * by the change). With TCP-style sockets, this option is inherited by
2789 * sockets derived from a listener socket.
2791 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2793 struct sctp_initmsg sinit;
2794 struct sctp_sock *sp = sctp_sk(sk);
2796 if (optlen != sizeof(struct sctp_initmsg))
2797 return -EINVAL;
2798 if (copy_from_user(&sinit, optval, optlen))
2799 return -EFAULT;
2801 if (sinit.sinit_num_ostreams)
2802 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2803 if (sinit.sinit_max_instreams)
2804 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2805 if (sinit.sinit_max_attempts)
2806 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2807 if (sinit.sinit_max_init_timeo)
2808 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2810 return 0;
2814 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2816 * Applications that wish to use the sendto() system call may wish to
2817 * specify a default set of parameters that would normally be supplied
2818 * through the inclusion of ancillary data. This socket option allows
2819 * such an application to set the default sctp_sndrcvinfo structure.
2820 * The application that wishes to use this socket option simply passes
2821 * in to this call the sctp_sndrcvinfo structure defined in Section
2822 * 5.2.2) The input parameters accepted by this call include
2823 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2824 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2825 * to this call if the caller is using the UDP model.
2827 static int sctp_setsockopt_default_send_param(struct sock *sk,
2828 char __user *optval,
2829 unsigned int optlen)
2831 struct sctp_sock *sp = sctp_sk(sk);
2832 struct sctp_association *asoc;
2833 struct sctp_sndrcvinfo info;
2835 if (optlen != sizeof(info))
2836 return -EINVAL;
2837 if (copy_from_user(&info, optval, optlen))
2838 return -EFAULT;
2839 if (info.sinfo_flags &
2840 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2841 SCTP_ABORT | SCTP_EOF))
2842 return -EINVAL;
2844 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2845 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2846 return -EINVAL;
2847 if (asoc) {
2848 asoc->default_stream = info.sinfo_stream;
2849 asoc->default_flags = info.sinfo_flags;
2850 asoc->default_ppid = info.sinfo_ppid;
2851 asoc->default_context = info.sinfo_context;
2852 asoc->default_timetolive = info.sinfo_timetolive;
2853 } else {
2854 sp->default_stream = info.sinfo_stream;
2855 sp->default_flags = info.sinfo_flags;
2856 sp->default_ppid = info.sinfo_ppid;
2857 sp->default_context = info.sinfo_context;
2858 sp->default_timetolive = info.sinfo_timetolive;
2861 return 0;
2864 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
2865 * (SCTP_DEFAULT_SNDINFO)
2867 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
2868 char __user *optval,
2869 unsigned int optlen)
2871 struct sctp_sock *sp = sctp_sk(sk);
2872 struct sctp_association *asoc;
2873 struct sctp_sndinfo info;
2875 if (optlen != sizeof(info))
2876 return -EINVAL;
2877 if (copy_from_user(&info, optval, optlen))
2878 return -EFAULT;
2879 if (info.snd_flags &
2880 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2881 SCTP_ABORT | SCTP_EOF))
2882 return -EINVAL;
2884 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
2885 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
2886 return -EINVAL;
2887 if (asoc) {
2888 asoc->default_stream = info.snd_sid;
2889 asoc->default_flags = info.snd_flags;
2890 asoc->default_ppid = info.snd_ppid;
2891 asoc->default_context = info.snd_context;
2892 } else {
2893 sp->default_stream = info.snd_sid;
2894 sp->default_flags = info.snd_flags;
2895 sp->default_ppid = info.snd_ppid;
2896 sp->default_context = info.snd_context;
2899 return 0;
2902 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2904 * Requests that the local SCTP stack use the enclosed peer address as
2905 * the association primary. The enclosed address must be one of the
2906 * association peer's addresses.
2908 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2909 unsigned int optlen)
2911 struct sctp_prim prim;
2912 struct sctp_transport *trans;
2914 if (optlen != sizeof(struct sctp_prim))
2915 return -EINVAL;
2917 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2918 return -EFAULT;
2920 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2921 if (!trans)
2922 return -EINVAL;
2924 sctp_assoc_set_primary(trans->asoc, trans);
2926 return 0;
2930 * 7.1.5 SCTP_NODELAY
2932 * Turn on/off any Nagle-like algorithm. This means that packets are
2933 * generally sent as soon as possible and no unnecessary delays are
2934 * introduced, at the cost of more packets in the network. Expects an
2935 * integer boolean flag.
2937 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2938 unsigned int optlen)
2940 int val;
2942 if (optlen < sizeof(int))
2943 return -EINVAL;
2944 if (get_user(val, (int __user *)optval))
2945 return -EFAULT;
2947 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2948 return 0;
2953 * 7.1.1 SCTP_RTOINFO
2955 * The protocol parameters used to initialize and bound retransmission
2956 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2957 * and modify these parameters.
2958 * All parameters are time values, in milliseconds. A value of 0, when
2959 * modifying the parameters, indicates that the current value should not
2960 * be changed.
2963 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2965 struct sctp_rtoinfo rtoinfo;
2966 struct sctp_association *asoc;
2967 unsigned long rto_min, rto_max;
2968 struct sctp_sock *sp = sctp_sk(sk);
2970 if (optlen != sizeof (struct sctp_rtoinfo))
2971 return -EINVAL;
2973 if (copy_from_user(&rtoinfo, optval, optlen))
2974 return -EFAULT;
2976 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2978 /* Set the values to the specific association */
2979 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2980 return -EINVAL;
2982 rto_max = rtoinfo.srto_max;
2983 rto_min = rtoinfo.srto_min;
2985 if (rto_max)
2986 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2987 else
2988 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2990 if (rto_min)
2991 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2992 else
2993 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2995 if (rto_min > rto_max)
2996 return -EINVAL;
2998 if (asoc) {
2999 if (rtoinfo.srto_initial != 0)
3000 asoc->rto_initial =
3001 msecs_to_jiffies(rtoinfo.srto_initial);
3002 asoc->rto_max = rto_max;
3003 asoc->rto_min = rto_min;
3004 } else {
3005 /* If there is no association or the association-id = 0
3006 * set the values to the endpoint.
3008 if (rtoinfo.srto_initial != 0)
3009 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3010 sp->rtoinfo.srto_max = rto_max;
3011 sp->rtoinfo.srto_min = rto_min;
3014 return 0;
3019 * 7.1.2 SCTP_ASSOCINFO
3021 * This option is used to tune the maximum retransmission attempts
3022 * of the association.
3023 * Returns an error if the new association retransmission value is
3024 * greater than the sum of the retransmission value of the peer.
3025 * See [SCTP] for more information.
3028 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3031 struct sctp_assocparams assocparams;
3032 struct sctp_association *asoc;
3034 if (optlen != sizeof(struct sctp_assocparams))
3035 return -EINVAL;
3036 if (copy_from_user(&assocparams, optval, optlen))
3037 return -EFAULT;
3039 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3041 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3042 return -EINVAL;
3044 /* Set the values to the specific association */
3045 if (asoc) {
3046 if (assocparams.sasoc_asocmaxrxt != 0) {
3047 __u32 path_sum = 0;
3048 int paths = 0;
3049 struct sctp_transport *peer_addr;
3051 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3052 transports) {
3053 path_sum += peer_addr->pathmaxrxt;
3054 paths++;
3057 /* Only validate asocmaxrxt if we have more than
3058 * one path/transport. We do this because path
3059 * retransmissions are only counted when we have more
3060 * then one path.
3062 if (paths > 1 &&
3063 assocparams.sasoc_asocmaxrxt > path_sum)
3064 return -EINVAL;
3066 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3069 if (assocparams.sasoc_cookie_life != 0)
3070 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3071 } else {
3072 /* Set the values to the endpoint */
3073 struct sctp_sock *sp = sctp_sk(sk);
3075 if (assocparams.sasoc_asocmaxrxt != 0)
3076 sp->assocparams.sasoc_asocmaxrxt =
3077 assocparams.sasoc_asocmaxrxt;
3078 if (assocparams.sasoc_cookie_life != 0)
3079 sp->assocparams.sasoc_cookie_life =
3080 assocparams.sasoc_cookie_life;
3082 return 0;
3086 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3088 * This socket option is a boolean flag which turns on or off mapped V4
3089 * addresses. If this option is turned on and the socket is type
3090 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3091 * If this option is turned off, then no mapping will be done of V4
3092 * addresses and a user will receive both PF_INET6 and PF_INET type
3093 * addresses on the socket.
3095 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3097 int val;
3098 struct sctp_sock *sp = sctp_sk(sk);
3100 if (optlen < sizeof(int))
3101 return -EINVAL;
3102 if (get_user(val, (int __user *)optval))
3103 return -EFAULT;
3104 if (val)
3105 sp->v4mapped = 1;
3106 else
3107 sp->v4mapped = 0;
3109 return 0;
3113 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3114 * This option will get or set the maximum size to put in any outgoing
3115 * SCTP DATA chunk. If a message is larger than this size it will be
3116 * fragmented by SCTP into the specified size. Note that the underlying
3117 * SCTP implementation may fragment into smaller sized chunks when the
3118 * PMTU of the underlying association is smaller than the value set by
3119 * the user. The default value for this option is '0' which indicates
3120 * the user is NOT limiting fragmentation and only the PMTU will effect
3121 * SCTP's choice of DATA chunk size. Note also that values set larger
3122 * than the maximum size of an IP datagram will effectively let SCTP
3123 * control fragmentation (i.e. the same as setting this option to 0).
3125 * The following structure is used to access and modify this parameter:
3127 * struct sctp_assoc_value {
3128 * sctp_assoc_t assoc_id;
3129 * uint32_t assoc_value;
3130 * };
3132 * assoc_id: This parameter is ignored for one-to-one style sockets.
3133 * For one-to-many style sockets this parameter indicates which
3134 * association the user is performing an action upon. Note that if
3135 * this field's value is zero then the endpoints default value is
3136 * changed (effecting future associations only).
3137 * assoc_value: This parameter specifies the maximum size in bytes.
3139 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3141 struct sctp_sock *sp = sctp_sk(sk);
3142 struct sctp_assoc_value params;
3143 struct sctp_association *asoc;
3144 int val;
3146 if (optlen == sizeof(int)) {
3147 pr_warn_ratelimited(DEPRECATED
3148 "%s (pid %d) "
3149 "Use of int in maxseg socket option.\n"
3150 "Use struct sctp_assoc_value instead\n",
3151 current->comm, task_pid_nr(current));
3152 if (copy_from_user(&val, optval, optlen))
3153 return -EFAULT;
3154 params.assoc_id = 0;
3155 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3156 if (copy_from_user(&params, optval, optlen))
3157 return -EFAULT;
3158 val = params.assoc_value;
3159 } else {
3160 return -EINVAL;
3163 if (val) {
3164 int min_len, max_len;
3166 min_len = SCTP_DEFAULT_MINSEGMENT - sp->pf->af->net_header_len;
3167 min_len -= sizeof(struct sctphdr) +
3168 sizeof(struct sctp_data_chunk);
3170 max_len = SCTP_MAX_CHUNK_LEN - sizeof(struct sctp_data_chunk);
3172 if (val < min_len || val > max_len)
3173 return -EINVAL;
3176 asoc = sctp_id2assoc(sk, params.assoc_id);
3177 if (asoc) {
3178 if (val == 0) {
3179 val = asoc->pathmtu - sp->pf->af->net_header_len;
3180 val -= sizeof(struct sctphdr) +
3181 sizeof(struct sctp_data_chunk);
3183 asoc->user_frag = val;
3184 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3185 } else {
3186 if (params.assoc_id && sctp_style(sk, UDP))
3187 return -EINVAL;
3188 sp->user_frag = val;
3191 return 0;
3196 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3198 * Requests that the peer mark the enclosed address as the association
3199 * primary. The enclosed address must be one of the association's
3200 * locally bound addresses. The following structure is used to make a
3201 * set primary request:
3203 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3204 unsigned int optlen)
3206 struct net *net = sock_net(sk);
3207 struct sctp_sock *sp;
3208 struct sctp_association *asoc = NULL;
3209 struct sctp_setpeerprim prim;
3210 struct sctp_chunk *chunk;
3211 struct sctp_af *af;
3212 int err;
3214 sp = sctp_sk(sk);
3216 if (!net->sctp.addip_enable)
3217 return -EPERM;
3219 if (optlen != sizeof(struct sctp_setpeerprim))
3220 return -EINVAL;
3222 if (copy_from_user(&prim, optval, optlen))
3223 return -EFAULT;
3225 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3226 if (!asoc)
3227 return -EINVAL;
3229 if (!asoc->peer.asconf_capable)
3230 return -EPERM;
3232 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3233 return -EPERM;
3235 if (!sctp_state(asoc, ESTABLISHED))
3236 return -ENOTCONN;
3238 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3239 if (!af)
3240 return -EINVAL;
3242 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3243 return -EADDRNOTAVAIL;
3245 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3246 return -EADDRNOTAVAIL;
3248 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3249 chunk = sctp_make_asconf_set_prim(asoc,
3250 (union sctp_addr *)&prim.sspp_addr);
3251 if (!chunk)
3252 return -ENOMEM;
3254 err = sctp_send_asconf(asoc, chunk);
3256 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3258 return err;
3261 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3262 unsigned int optlen)
3264 struct sctp_setadaptation adaptation;
3266 if (optlen != sizeof(struct sctp_setadaptation))
3267 return -EINVAL;
3268 if (copy_from_user(&adaptation, optval, optlen))
3269 return -EFAULT;
3271 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3273 return 0;
3277 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3279 * The context field in the sctp_sndrcvinfo structure is normally only
3280 * used when a failed message is retrieved holding the value that was
3281 * sent down on the actual send call. This option allows the setting of
3282 * a default context on an association basis that will be received on
3283 * reading messages from the peer. This is especially helpful in the
3284 * one-2-many model for an application to keep some reference to an
3285 * internal state machine that is processing messages on the
3286 * association. Note that the setting of this value only effects
3287 * received messages from the peer and does not effect the value that is
3288 * saved with outbound messages.
3290 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3291 unsigned int optlen)
3293 struct sctp_assoc_value params;
3294 struct sctp_sock *sp;
3295 struct sctp_association *asoc;
3297 if (optlen != sizeof(struct sctp_assoc_value))
3298 return -EINVAL;
3299 if (copy_from_user(&params, optval, optlen))
3300 return -EFAULT;
3302 sp = sctp_sk(sk);
3304 if (params.assoc_id != 0) {
3305 asoc = sctp_id2assoc(sk, params.assoc_id);
3306 if (!asoc)
3307 return -EINVAL;
3308 asoc->default_rcv_context = params.assoc_value;
3309 } else {
3310 sp->default_rcv_context = params.assoc_value;
3313 return 0;
3317 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3319 * This options will at a minimum specify if the implementation is doing
3320 * fragmented interleave. Fragmented interleave, for a one to many
3321 * socket, is when subsequent calls to receive a message may return
3322 * parts of messages from different associations. Some implementations
3323 * may allow you to turn this value on or off. If so, when turned off,
3324 * no fragment interleave will occur (which will cause a head of line
3325 * blocking amongst multiple associations sharing the same one to many
3326 * socket). When this option is turned on, then each receive call may
3327 * come from a different association (thus the user must receive data
3328 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3329 * association each receive belongs to.
3331 * This option takes a boolean value. A non-zero value indicates that
3332 * fragmented interleave is on. A value of zero indicates that
3333 * fragmented interleave is off.
3335 * Note that it is important that an implementation that allows this
3336 * option to be turned on, have it off by default. Otherwise an unaware
3337 * application using the one to many model may become confused and act
3338 * incorrectly.
3340 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3341 char __user *optval,
3342 unsigned int optlen)
3344 int val;
3346 if (optlen != sizeof(int))
3347 return -EINVAL;
3348 if (get_user(val, (int __user *)optval))
3349 return -EFAULT;
3351 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3353 return 0;
3357 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3358 * (SCTP_PARTIAL_DELIVERY_POINT)
3360 * This option will set or get the SCTP partial delivery point. This
3361 * point is the size of a message where the partial delivery API will be
3362 * invoked to help free up rwnd space for the peer. Setting this to a
3363 * lower value will cause partial deliveries to happen more often. The
3364 * calls argument is an integer that sets or gets the partial delivery
3365 * point. Note also that the call will fail if the user attempts to set
3366 * this value larger than the socket receive buffer size.
3368 * Note that any single message having a length smaller than or equal to
3369 * the SCTP partial delivery point will be delivered in one single read
3370 * call as long as the user provided buffer is large enough to hold the
3371 * message.
3373 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3374 char __user *optval,
3375 unsigned int optlen)
3377 u32 val;
3379 if (optlen != sizeof(u32))
3380 return -EINVAL;
3381 if (get_user(val, (int __user *)optval))
3382 return -EFAULT;
3384 /* Note: We double the receive buffer from what the user sets
3385 * it to be, also initial rwnd is based on rcvbuf/2.
3387 if (val > (sk->sk_rcvbuf >> 1))
3388 return -EINVAL;
3390 sctp_sk(sk)->pd_point = val;
3392 return 0; /* is this the right error code? */
3396 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3398 * This option will allow a user to change the maximum burst of packets
3399 * that can be emitted by this association. Note that the default value
3400 * is 4, and some implementations may restrict this setting so that it
3401 * can only be lowered.
3403 * NOTE: This text doesn't seem right. Do this on a socket basis with
3404 * future associations inheriting the socket value.
3406 static int sctp_setsockopt_maxburst(struct sock *sk,
3407 char __user *optval,
3408 unsigned int optlen)
3410 struct sctp_assoc_value params;
3411 struct sctp_sock *sp;
3412 struct sctp_association *asoc;
3413 int val;
3414 int assoc_id = 0;
3416 if (optlen == sizeof(int)) {
3417 pr_warn_ratelimited(DEPRECATED
3418 "%s (pid %d) "
3419 "Use of int in max_burst socket option deprecated.\n"
3420 "Use struct sctp_assoc_value instead\n",
3421 current->comm, task_pid_nr(current));
3422 if (copy_from_user(&val, optval, optlen))
3423 return -EFAULT;
3424 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3425 if (copy_from_user(&params, optval, optlen))
3426 return -EFAULT;
3427 val = params.assoc_value;
3428 assoc_id = params.assoc_id;
3429 } else
3430 return -EINVAL;
3432 sp = sctp_sk(sk);
3434 if (assoc_id != 0) {
3435 asoc = sctp_id2assoc(sk, assoc_id);
3436 if (!asoc)
3437 return -EINVAL;
3438 asoc->max_burst = val;
3439 } else
3440 sp->max_burst = val;
3442 return 0;
3446 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3448 * This set option adds a chunk type that the user is requesting to be
3449 * received only in an authenticated way. Changes to the list of chunks
3450 * will only effect future associations on the socket.
3452 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3453 char __user *optval,
3454 unsigned int optlen)
3456 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3457 struct sctp_authchunk val;
3459 if (!ep->auth_enable)
3460 return -EACCES;
3462 if (optlen != sizeof(struct sctp_authchunk))
3463 return -EINVAL;
3464 if (copy_from_user(&val, optval, optlen))
3465 return -EFAULT;
3467 switch (val.sauth_chunk) {
3468 case SCTP_CID_INIT:
3469 case SCTP_CID_INIT_ACK:
3470 case SCTP_CID_SHUTDOWN_COMPLETE:
3471 case SCTP_CID_AUTH:
3472 return -EINVAL;
3475 /* add this chunk id to the endpoint */
3476 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3480 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3482 * This option gets or sets the list of HMAC algorithms that the local
3483 * endpoint requires the peer to use.
3485 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3486 char __user *optval,
3487 unsigned int optlen)
3489 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3490 struct sctp_hmacalgo *hmacs;
3491 u32 idents;
3492 int err;
3494 if (!ep->auth_enable)
3495 return -EACCES;
3497 if (optlen < sizeof(struct sctp_hmacalgo))
3498 return -EINVAL;
3499 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3500 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3502 hmacs = memdup_user(optval, optlen);
3503 if (IS_ERR(hmacs))
3504 return PTR_ERR(hmacs);
3506 idents = hmacs->shmac_num_idents;
3507 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3508 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3509 err = -EINVAL;
3510 goto out;
3513 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3514 out:
3515 kfree(hmacs);
3516 return err;
3520 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3522 * This option will set a shared secret key which is used to build an
3523 * association shared key.
3525 static int sctp_setsockopt_auth_key(struct sock *sk,
3526 char __user *optval,
3527 unsigned int optlen)
3529 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3530 struct sctp_authkey *authkey;
3531 struct sctp_association *asoc;
3532 int ret;
3534 if (!ep->auth_enable)
3535 return -EACCES;
3537 if (optlen <= sizeof(struct sctp_authkey))
3538 return -EINVAL;
3539 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3540 * this.
3542 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3543 sizeof(struct sctp_authkey));
3545 authkey = memdup_user(optval, optlen);
3546 if (IS_ERR(authkey))
3547 return PTR_ERR(authkey);
3549 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3550 ret = -EINVAL;
3551 goto out;
3554 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3555 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3556 ret = -EINVAL;
3557 goto out;
3560 ret = sctp_auth_set_key(ep, asoc, authkey);
3561 out:
3562 kzfree(authkey);
3563 return ret;
3567 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3569 * This option will get or set the active shared key to be used to build
3570 * the association shared key.
3572 static int sctp_setsockopt_active_key(struct sock *sk,
3573 char __user *optval,
3574 unsigned int optlen)
3576 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3577 struct sctp_authkeyid val;
3578 struct sctp_association *asoc;
3580 if (!ep->auth_enable)
3581 return -EACCES;
3583 if (optlen != sizeof(struct sctp_authkeyid))
3584 return -EINVAL;
3585 if (copy_from_user(&val, optval, optlen))
3586 return -EFAULT;
3588 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3589 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3590 return -EINVAL;
3592 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3596 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3598 * This set option will delete a shared secret key from use.
3600 static int sctp_setsockopt_del_key(struct sock *sk,
3601 char __user *optval,
3602 unsigned int optlen)
3604 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3605 struct sctp_authkeyid val;
3606 struct sctp_association *asoc;
3608 if (!ep->auth_enable)
3609 return -EACCES;
3611 if (optlen != sizeof(struct sctp_authkeyid))
3612 return -EINVAL;
3613 if (copy_from_user(&val, optval, optlen))
3614 return -EFAULT;
3616 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3617 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3618 return -EINVAL;
3620 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3625 * 8.1.23 SCTP_AUTO_ASCONF
3627 * This option will enable or disable the use of the automatic generation of
3628 * ASCONF chunks to add and delete addresses to an existing association. Note
3629 * that this option has two caveats namely: a) it only affects sockets that
3630 * are bound to all addresses available to the SCTP stack, and b) the system
3631 * administrator may have an overriding control that turns the ASCONF feature
3632 * off no matter what setting the socket option may have.
3633 * This option expects an integer boolean flag, where a non-zero value turns on
3634 * the option, and a zero value turns off the option.
3635 * Note. In this implementation, socket operation overrides default parameter
3636 * being set by sysctl as well as FreeBSD implementation
3638 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3639 unsigned int optlen)
3641 int val;
3642 struct sctp_sock *sp = sctp_sk(sk);
3644 if (optlen < sizeof(int))
3645 return -EINVAL;
3646 if (get_user(val, (int __user *)optval))
3647 return -EFAULT;
3648 if (!sctp_is_ep_boundall(sk) && val)
3649 return -EINVAL;
3650 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3651 return 0;
3653 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3654 if (val == 0 && sp->do_auto_asconf) {
3655 list_del(&sp->auto_asconf_list);
3656 sp->do_auto_asconf = 0;
3657 } else if (val && !sp->do_auto_asconf) {
3658 list_add_tail(&sp->auto_asconf_list,
3659 &sock_net(sk)->sctp.auto_asconf_splist);
3660 sp->do_auto_asconf = 1;
3662 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3663 return 0;
3667 * SCTP_PEER_ADDR_THLDS
3669 * This option allows us to alter the partially failed threshold for one or all
3670 * transports in an association. See Section 6.1 of:
3671 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3673 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3674 char __user *optval,
3675 unsigned int optlen)
3677 struct sctp_paddrthlds val;
3678 struct sctp_transport *trans;
3679 struct sctp_association *asoc;
3681 if (optlen < sizeof(struct sctp_paddrthlds))
3682 return -EINVAL;
3683 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3684 sizeof(struct sctp_paddrthlds)))
3685 return -EFAULT;
3688 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3689 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3690 if (!asoc)
3691 return -ENOENT;
3692 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3693 transports) {
3694 if (val.spt_pathmaxrxt)
3695 trans->pathmaxrxt = val.spt_pathmaxrxt;
3696 trans->pf_retrans = val.spt_pathpfthld;
3699 if (val.spt_pathmaxrxt)
3700 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3701 asoc->pf_retrans = val.spt_pathpfthld;
3702 } else {
3703 trans = sctp_addr_id2transport(sk, &val.spt_address,
3704 val.spt_assoc_id);
3705 if (!trans)
3706 return -ENOENT;
3708 if (val.spt_pathmaxrxt)
3709 trans->pathmaxrxt = val.spt_pathmaxrxt;
3710 trans->pf_retrans = val.spt_pathpfthld;
3713 return 0;
3716 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3717 char __user *optval,
3718 unsigned int optlen)
3720 int val;
3722 if (optlen < sizeof(int))
3723 return -EINVAL;
3724 if (get_user(val, (int __user *) optval))
3725 return -EFAULT;
3727 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3729 return 0;
3732 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3733 char __user *optval,
3734 unsigned int optlen)
3736 int val;
3738 if (optlen < sizeof(int))
3739 return -EINVAL;
3740 if (get_user(val, (int __user *) optval))
3741 return -EFAULT;
3743 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3745 return 0;
3748 static int sctp_setsockopt_pr_supported(struct sock *sk,
3749 char __user *optval,
3750 unsigned int optlen)
3752 struct sctp_assoc_value params;
3754 if (optlen != sizeof(params))
3755 return -EINVAL;
3757 if (copy_from_user(&params, optval, optlen))
3758 return -EFAULT;
3760 sctp_sk(sk)->ep->prsctp_enable = !!params.assoc_value;
3762 return 0;
3765 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3766 char __user *optval,
3767 unsigned int optlen)
3769 struct sctp_default_prinfo info;
3770 struct sctp_association *asoc;
3771 int retval = -EINVAL;
3773 if (optlen != sizeof(info))
3774 goto out;
3776 if (copy_from_user(&info, optval, sizeof(info))) {
3777 retval = -EFAULT;
3778 goto out;
3781 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3782 goto out;
3784 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3785 info.pr_value = 0;
3787 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3788 if (asoc) {
3789 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3790 asoc->default_timetolive = info.pr_value;
3791 } else if (!info.pr_assoc_id) {
3792 struct sctp_sock *sp = sctp_sk(sk);
3794 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3795 sp->default_timetolive = info.pr_value;
3796 } else {
3797 goto out;
3800 retval = 0;
3802 out:
3803 return retval;
3806 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3807 char __user *optval,
3808 unsigned int optlen)
3810 struct sctp_assoc_value params;
3811 struct sctp_association *asoc;
3812 int retval = -EINVAL;
3814 if (optlen != sizeof(params))
3815 goto out;
3817 if (copy_from_user(&params, optval, optlen)) {
3818 retval = -EFAULT;
3819 goto out;
3822 asoc = sctp_id2assoc(sk, params.assoc_id);
3823 if (asoc) {
3824 asoc->reconf_enable = !!params.assoc_value;
3825 } else if (!params.assoc_id) {
3826 struct sctp_sock *sp = sctp_sk(sk);
3828 sp->ep->reconf_enable = !!params.assoc_value;
3829 } else {
3830 goto out;
3833 retval = 0;
3835 out:
3836 return retval;
3839 static int sctp_setsockopt_enable_strreset(struct sock *sk,
3840 char __user *optval,
3841 unsigned int optlen)
3843 struct sctp_assoc_value params;
3844 struct sctp_association *asoc;
3845 int retval = -EINVAL;
3847 if (optlen != sizeof(params))
3848 goto out;
3850 if (copy_from_user(&params, optval, optlen)) {
3851 retval = -EFAULT;
3852 goto out;
3855 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
3856 goto out;
3858 asoc = sctp_id2assoc(sk, params.assoc_id);
3859 if (asoc) {
3860 asoc->strreset_enable = params.assoc_value;
3861 } else if (!params.assoc_id) {
3862 struct sctp_sock *sp = sctp_sk(sk);
3864 sp->ep->strreset_enable = params.assoc_value;
3865 } else {
3866 goto out;
3869 retval = 0;
3871 out:
3872 return retval;
3875 static int sctp_setsockopt_reset_streams(struct sock *sk,
3876 char __user *optval,
3877 unsigned int optlen)
3879 struct sctp_reset_streams *params;
3880 struct sctp_association *asoc;
3881 int retval = -EINVAL;
3883 if (optlen < sizeof(*params))
3884 return -EINVAL;
3885 /* srs_number_streams is u16, so optlen can't be bigger than this. */
3886 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3887 sizeof(__u16) * sizeof(*params));
3889 params = memdup_user(optval, optlen);
3890 if (IS_ERR(params))
3891 return PTR_ERR(params);
3893 if (params->srs_number_streams * sizeof(__u16) >
3894 optlen - sizeof(*params))
3895 goto out;
3897 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
3898 if (!asoc)
3899 goto out;
3901 retval = sctp_send_reset_streams(asoc, params);
3903 out:
3904 kfree(params);
3905 return retval;
3908 static int sctp_setsockopt_reset_assoc(struct sock *sk,
3909 char __user *optval,
3910 unsigned int optlen)
3912 struct sctp_association *asoc;
3913 sctp_assoc_t associd;
3914 int retval = -EINVAL;
3916 if (optlen != sizeof(associd))
3917 goto out;
3919 if (copy_from_user(&associd, optval, optlen)) {
3920 retval = -EFAULT;
3921 goto out;
3924 asoc = sctp_id2assoc(sk, associd);
3925 if (!asoc)
3926 goto out;
3928 retval = sctp_send_reset_assoc(asoc);
3930 out:
3931 return retval;
3934 static int sctp_setsockopt_add_streams(struct sock *sk,
3935 char __user *optval,
3936 unsigned int optlen)
3938 struct sctp_association *asoc;
3939 struct sctp_add_streams params;
3940 int retval = -EINVAL;
3942 if (optlen != sizeof(params))
3943 goto out;
3945 if (copy_from_user(&params, optval, optlen)) {
3946 retval = -EFAULT;
3947 goto out;
3950 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
3951 if (!asoc)
3952 goto out;
3954 retval = sctp_send_add_streams(asoc, &params);
3956 out:
3957 return retval;
3960 /* API 6.2 setsockopt(), getsockopt()
3962 * Applications use setsockopt() and getsockopt() to set or retrieve
3963 * socket options. Socket options are used to change the default
3964 * behavior of sockets calls. They are described in Section 7.
3966 * The syntax is:
3968 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3969 * int __user *optlen);
3970 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3971 * int optlen);
3973 * sd - the socket descript.
3974 * level - set to IPPROTO_SCTP for all SCTP options.
3975 * optname - the option name.
3976 * optval - the buffer to store the value of the option.
3977 * optlen - the size of the buffer.
3979 static int sctp_setsockopt(struct sock *sk, int level, int optname,
3980 char __user *optval, unsigned int optlen)
3982 int retval = 0;
3984 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3986 /* I can hardly begin to describe how wrong this is. This is
3987 * so broken as to be worse than useless. The API draft
3988 * REALLY is NOT helpful here... I am not convinced that the
3989 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3990 * are at all well-founded.
3992 if (level != SOL_SCTP) {
3993 struct sctp_af *af = sctp_sk(sk)->pf->af;
3994 retval = af->setsockopt(sk, level, optname, optval, optlen);
3995 goto out_nounlock;
3998 lock_sock(sk);
4000 switch (optname) {
4001 case SCTP_SOCKOPT_BINDX_ADD:
4002 /* 'optlen' is the size of the addresses buffer. */
4003 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4004 optlen, SCTP_BINDX_ADD_ADDR);
4005 break;
4007 case SCTP_SOCKOPT_BINDX_REM:
4008 /* 'optlen' is the size of the addresses buffer. */
4009 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4010 optlen, SCTP_BINDX_REM_ADDR);
4011 break;
4013 case SCTP_SOCKOPT_CONNECTX_OLD:
4014 /* 'optlen' is the size of the addresses buffer. */
4015 retval = sctp_setsockopt_connectx_old(sk,
4016 (struct sockaddr __user *)optval,
4017 optlen);
4018 break;
4020 case SCTP_SOCKOPT_CONNECTX:
4021 /* 'optlen' is the size of the addresses buffer. */
4022 retval = sctp_setsockopt_connectx(sk,
4023 (struct sockaddr __user *)optval,
4024 optlen);
4025 break;
4027 case SCTP_DISABLE_FRAGMENTS:
4028 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4029 break;
4031 case SCTP_EVENTS:
4032 retval = sctp_setsockopt_events(sk, optval, optlen);
4033 break;
4035 case SCTP_AUTOCLOSE:
4036 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4037 break;
4039 case SCTP_PEER_ADDR_PARAMS:
4040 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4041 break;
4043 case SCTP_DELAYED_SACK:
4044 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4045 break;
4046 case SCTP_PARTIAL_DELIVERY_POINT:
4047 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4048 break;
4050 case SCTP_INITMSG:
4051 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4052 break;
4053 case SCTP_DEFAULT_SEND_PARAM:
4054 retval = sctp_setsockopt_default_send_param(sk, optval,
4055 optlen);
4056 break;
4057 case SCTP_DEFAULT_SNDINFO:
4058 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4059 break;
4060 case SCTP_PRIMARY_ADDR:
4061 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4062 break;
4063 case SCTP_SET_PEER_PRIMARY_ADDR:
4064 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4065 break;
4066 case SCTP_NODELAY:
4067 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4068 break;
4069 case SCTP_RTOINFO:
4070 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4071 break;
4072 case SCTP_ASSOCINFO:
4073 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4074 break;
4075 case SCTP_I_WANT_MAPPED_V4_ADDR:
4076 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4077 break;
4078 case SCTP_MAXSEG:
4079 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4080 break;
4081 case SCTP_ADAPTATION_LAYER:
4082 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4083 break;
4084 case SCTP_CONTEXT:
4085 retval = sctp_setsockopt_context(sk, optval, optlen);
4086 break;
4087 case SCTP_FRAGMENT_INTERLEAVE:
4088 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4089 break;
4090 case SCTP_MAX_BURST:
4091 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4092 break;
4093 case SCTP_AUTH_CHUNK:
4094 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4095 break;
4096 case SCTP_HMAC_IDENT:
4097 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4098 break;
4099 case SCTP_AUTH_KEY:
4100 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4101 break;
4102 case SCTP_AUTH_ACTIVE_KEY:
4103 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4104 break;
4105 case SCTP_AUTH_DELETE_KEY:
4106 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4107 break;
4108 case SCTP_AUTO_ASCONF:
4109 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4110 break;
4111 case SCTP_PEER_ADDR_THLDS:
4112 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4113 break;
4114 case SCTP_RECVRCVINFO:
4115 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4116 break;
4117 case SCTP_RECVNXTINFO:
4118 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4119 break;
4120 case SCTP_PR_SUPPORTED:
4121 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4122 break;
4123 case SCTP_DEFAULT_PRINFO:
4124 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4125 break;
4126 case SCTP_RECONFIG_SUPPORTED:
4127 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4128 break;
4129 case SCTP_ENABLE_STREAM_RESET:
4130 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4131 break;
4132 case SCTP_RESET_STREAMS:
4133 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4134 break;
4135 case SCTP_RESET_ASSOC:
4136 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4137 break;
4138 case SCTP_ADD_STREAMS:
4139 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4140 break;
4141 default:
4142 retval = -ENOPROTOOPT;
4143 break;
4146 release_sock(sk);
4148 out_nounlock:
4149 return retval;
4152 /* API 3.1.6 connect() - UDP Style Syntax
4154 * An application may use the connect() call in the UDP model to initiate an
4155 * association without sending data.
4157 * The syntax is:
4159 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4161 * sd: the socket descriptor to have a new association added to.
4163 * nam: the address structure (either struct sockaddr_in or struct
4164 * sockaddr_in6 defined in RFC2553 [7]).
4166 * len: the size of the address.
4168 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4169 int addr_len)
4171 int err = 0;
4172 struct sctp_af *af;
4174 lock_sock(sk);
4176 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4177 addr, addr_len);
4179 /* Validate addr_len before calling common connect/connectx routine. */
4180 af = sctp_get_af_specific(addr->sa_family);
4181 if (!af || addr_len < af->sockaddr_len) {
4182 err = -EINVAL;
4183 } else {
4184 /* Pass correct addr len to common routine (so it knows there
4185 * is only one address being passed.
4187 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
4190 release_sock(sk);
4191 return err;
4194 /* FIXME: Write comments. */
4195 static int sctp_disconnect(struct sock *sk, int flags)
4197 return -EOPNOTSUPP; /* STUB */
4200 /* 4.1.4 accept() - TCP Style Syntax
4202 * Applications use accept() call to remove an established SCTP
4203 * association from the accept queue of the endpoint. A new socket
4204 * descriptor will be returned from accept() to represent the newly
4205 * formed association.
4207 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4209 struct sctp_sock *sp;
4210 struct sctp_endpoint *ep;
4211 struct sock *newsk = NULL;
4212 struct sctp_association *asoc;
4213 long timeo;
4214 int error = 0;
4216 lock_sock(sk);
4218 sp = sctp_sk(sk);
4219 ep = sp->ep;
4221 if (!sctp_style(sk, TCP)) {
4222 error = -EOPNOTSUPP;
4223 goto out;
4226 if (!sctp_sstate(sk, LISTENING)) {
4227 error = -EINVAL;
4228 goto out;
4231 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4233 error = sctp_wait_for_accept(sk, timeo);
4234 if (error)
4235 goto out;
4237 /* We treat the list of associations on the endpoint as the accept
4238 * queue and pick the first association on the list.
4240 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4242 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4243 if (!newsk) {
4244 error = -ENOMEM;
4245 goto out;
4248 /* Populate the fields of the newsk from the oldsk and migrate the
4249 * asoc to the newsk.
4251 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4253 out:
4254 release_sock(sk);
4255 *err = error;
4256 return newsk;
4259 /* The SCTP ioctl handler. */
4260 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4262 int rc = -ENOTCONN;
4264 lock_sock(sk);
4267 * SEQPACKET-style sockets in LISTENING state are valid, for
4268 * SCTP, so only discard TCP-style sockets in LISTENING state.
4270 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4271 goto out;
4273 switch (cmd) {
4274 case SIOCINQ: {
4275 struct sk_buff *skb;
4276 unsigned int amount = 0;
4278 skb = skb_peek(&sk->sk_receive_queue);
4279 if (skb != NULL) {
4281 * We will only return the amount of this packet since
4282 * that is all that will be read.
4284 amount = skb->len;
4286 rc = put_user(amount, (int __user *)arg);
4287 break;
4289 default:
4290 rc = -ENOIOCTLCMD;
4291 break;
4293 out:
4294 release_sock(sk);
4295 return rc;
4298 /* This is the function which gets called during socket creation to
4299 * initialized the SCTP-specific portion of the sock.
4300 * The sock structure should already be zero-filled memory.
4302 static int sctp_init_sock(struct sock *sk)
4304 struct net *net = sock_net(sk);
4305 struct sctp_sock *sp;
4307 pr_debug("%s: sk:%p\n", __func__, sk);
4309 sp = sctp_sk(sk);
4311 /* Initialize the SCTP per socket area. */
4312 switch (sk->sk_type) {
4313 case SOCK_SEQPACKET:
4314 sp->type = SCTP_SOCKET_UDP;
4315 break;
4316 case SOCK_STREAM:
4317 sp->type = SCTP_SOCKET_TCP;
4318 break;
4319 default:
4320 return -ESOCKTNOSUPPORT;
4323 sk->sk_gso_type = SKB_GSO_SCTP;
4325 /* Initialize default send parameters. These parameters can be
4326 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4328 sp->default_stream = 0;
4329 sp->default_ppid = 0;
4330 sp->default_flags = 0;
4331 sp->default_context = 0;
4332 sp->default_timetolive = 0;
4334 sp->default_rcv_context = 0;
4335 sp->max_burst = net->sctp.max_burst;
4337 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4339 /* Initialize default setup parameters. These parameters
4340 * can be modified with the SCTP_INITMSG socket option or
4341 * overridden by the SCTP_INIT CMSG.
4343 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4344 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4345 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4346 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4348 /* Initialize default RTO related parameters. These parameters can
4349 * be modified for with the SCTP_RTOINFO socket option.
4351 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4352 sp->rtoinfo.srto_max = net->sctp.rto_max;
4353 sp->rtoinfo.srto_min = net->sctp.rto_min;
4355 /* Initialize default association related parameters. These parameters
4356 * can be modified with the SCTP_ASSOCINFO socket option.
4358 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4359 sp->assocparams.sasoc_number_peer_destinations = 0;
4360 sp->assocparams.sasoc_peer_rwnd = 0;
4361 sp->assocparams.sasoc_local_rwnd = 0;
4362 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4364 /* Initialize default event subscriptions. By default, all the
4365 * options are off.
4367 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4369 /* Default Peer Address Parameters. These defaults can
4370 * be modified via SCTP_PEER_ADDR_PARAMS
4372 sp->hbinterval = net->sctp.hb_interval;
4373 sp->pathmaxrxt = net->sctp.max_retrans_path;
4374 sp->pathmtu = 0; /* allow default discovery */
4375 sp->sackdelay = net->sctp.sack_timeout;
4376 sp->sackfreq = 2;
4377 sp->param_flags = SPP_HB_ENABLE |
4378 SPP_PMTUD_ENABLE |
4379 SPP_SACKDELAY_ENABLE;
4381 /* If enabled no SCTP message fragmentation will be performed.
4382 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4384 sp->disable_fragments = 0;
4386 /* Enable Nagle algorithm by default. */
4387 sp->nodelay = 0;
4389 sp->recvrcvinfo = 0;
4390 sp->recvnxtinfo = 0;
4392 /* Enable by default. */
4393 sp->v4mapped = 1;
4395 /* Auto-close idle associations after the configured
4396 * number of seconds. A value of 0 disables this
4397 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4398 * for UDP-style sockets only.
4400 sp->autoclose = 0;
4402 /* User specified fragmentation limit. */
4403 sp->user_frag = 0;
4405 sp->adaptation_ind = 0;
4407 sp->pf = sctp_get_pf_specific(sk->sk_family);
4409 /* Control variables for partial data delivery. */
4410 atomic_set(&sp->pd_mode, 0);
4411 skb_queue_head_init(&sp->pd_lobby);
4412 sp->frag_interleave = 0;
4414 /* Create a per socket endpoint structure. Even if we
4415 * change the data structure relationships, this may still
4416 * be useful for storing pre-connect address information.
4418 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4419 if (!sp->ep)
4420 return -ENOMEM;
4422 sp->hmac = NULL;
4424 sk->sk_destruct = sctp_destruct_sock;
4426 SCTP_DBG_OBJCNT_INC(sock);
4428 local_bh_disable();
4429 sk_sockets_allocated_inc(sk);
4430 sock_prot_inuse_add(net, sk->sk_prot, 1);
4432 /* Nothing can fail after this block, otherwise
4433 * sctp_destroy_sock() will be called without addr_wq_lock held
4435 if (net->sctp.default_auto_asconf) {
4436 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4437 list_add_tail(&sp->auto_asconf_list,
4438 &net->sctp.auto_asconf_splist);
4439 sp->do_auto_asconf = 1;
4440 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4441 } else {
4442 sp->do_auto_asconf = 0;
4445 local_bh_enable();
4447 return 0;
4450 /* Cleanup any SCTP per socket resources. Must be called with
4451 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4453 static void sctp_destroy_sock(struct sock *sk)
4455 struct sctp_sock *sp;
4457 pr_debug("%s: sk:%p\n", __func__, sk);
4459 /* Release our hold on the endpoint. */
4460 sp = sctp_sk(sk);
4461 /* This could happen during socket init, thus we bail out
4462 * early, since the rest of the below is not setup either.
4464 if (sp->ep == NULL)
4465 return;
4467 if (sp->do_auto_asconf) {
4468 sp->do_auto_asconf = 0;
4469 list_del(&sp->auto_asconf_list);
4471 sctp_endpoint_free(sp->ep);
4472 local_bh_disable();
4473 sk_sockets_allocated_dec(sk);
4474 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4475 local_bh_enable();
4478 /* Triggered when there are no references on the socket anymore */
4479 static void sctp_destruct_sock(struct sock *sk)
4481 struct sctp_sock *sp = sctp_sk(sk);
4483 /* Free up the HMAC transform. */
4484 crypto_free_shash(sp->hmac);
4486 inet_sock_destruct(sk);
4489 /* API 4.1.7 shutdown() - TCP Style Syntax
4490 * int shutdown(int socket, int how);
4492 * sd - the socket descriptor of the association to be closed.
4493 * how - Specifies the type of shutdown. The values are
4494 * as follows:
4495 * SHUT_RD
4496 * Disables further receive operations. No SCTP
4497 * protocol action is taken.
4498 * SHUT_WR
4499 * Disables further send operations, and initiates
4500 * the SCTP shutdown sequence.
4501 * SHUT_RDWR
4502 * Disables further send and receive operations
4503 * and initiates the SCTP shutdown sequence.
4505 static void sctp_shutdown(struct sock *sk, int how)
4507 struct net *net = sock_net(sk);
4508 struct sctp_endpoint *ep;
4510 if (!sctp_style(sk, TCP))
4511 return;
4513 ep = sctp_sk(sk)->ep;
4514 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4515 struct sctp_association *asoc;
4517 sk->sk_state = SCTP_SS_CLOSING;
4518 asoc = list_entry(ep->asocs.next,
4519 struct sctp_association, asocs);
4520 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4524 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4525 struct sctp_info *info)
4527 struct sctp_transport *prim;
4528 struct list_head *pos;
4529 int mask;
4531 memset(info, 0, sizeof(*info));
4532 if (!asoc) {
4533 struct sctp_sock *sp = sctp_sk(sk);
4535 info->sctpi_s_autoclose = sp->autoclose;
4536 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4537 info->sctpi_s_pd_point = sp->pd_point;
4538 info->sctpi_s_nodelay = sp->nodelay;
4539 info->sctpi_s_disable_fragments = sp->disable_fragments;
4540 info->sctpi_s_v4mapped = sp->v4mapped;
4541 info->sctpi_s_frag_interleave = sp->frag_interleave;
4542 info->sctpi_s_type = sp->type;
4544 return 0;
4547 info->sctpi_tag = asoc->c.my_vtag;
4548 info->sctpi_state = asoc->state;
4549 info->sctpi_rwnd = asoc->a_rwnd;
4550 info->sctpi_unackdata = asoc->unack_data;
4551 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4552 info->sctpi_instrms = asoc->stream.incnt;
4553 info->sctpi_outstrms = asoc->stream.outcnt;
4554 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4555 info->sctpi_inqueue++;
4556 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4557 info->sctpi_outqueue++;
4558 info->sctpi_overall_error = asoc->overall_error_count;
4559 info->sctpi_max_burst = asoc->max_burst;
4560 info->sctpi_maxseg = asoc->frag_point;
4561 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4562 info->sctpi_peer_tag = asoc->c.peer_vtag;
4564 mask = asoc->peer.ecn_capable << 1;
4565 mask = (mask | asoc->peer.ipv4_address) << 1;
4566 mask = (mask | asoc->peer.ipv6_address) << 1;
4567 mask = (mask | asoc->peer.hostname_address) << 1;
4568 mask = (mask | asoc->peer.asconf_capable) << 1;
4569 mask = (mask | asoc->peer.prsctp_capable) << 1;
4570 mask = (mask | asoc->peer.auth_capable);
4571 info->sctpi_peer_capable = mask;
4572 mask = asoc->peer.sack_needed << 1;
4573 mask = (mask | asoc->peer.sack_generation) << 1;
4574 mask = (mask | asoc->peer.zero_window_announced);
4575 info->sctpi_peer_sack = mask;
4577 info->sctpi_isacks = asoc->stats.isacks;
4578 info->sctpi_osacks = asoc->stats.osacks;
4579 info->sctpi_opackets = asoc->stats.opackets;
4580 info->sctpi_ipackets = asoc->stats.ipackets;
4581 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4582 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4583 info->sctpi_idupchunks = asoc->stats.idupchunks;
4584 info->sctpi_gapcnt = asoc->stats.gapcnt;
4585 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4586 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4587 info->sctpi_oodchunks = asoc->stats.oodchunks;
4588 info->sctpi_iodchunks = asoc->stats.iodchunks;
4589 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4590 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4592 prim = asoc->peer.primary_path;
4593 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4594 info->sctpi_p_state = prim->state;
4595 info->sctpi_p_cwnd = prim->cwnd;
4596 info->sctpi_p_srtt = prim->srtt;
4597 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4598 info->sctpi_p_hbinterval = prim->hbinterval;
4599 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4600 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4601 info->sctpi_p_ssthresh = prim->ssthresh;
4602 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4603 info->sctpi_p_flight_size = prim->flight_size;
4604 info->sctpi_p_error = prim->error_count;
4606 return 0;
4608 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4610 /* use callback to avoid exporting the core structure */
4611 int sctp_transport_walk_start(struct rhashtable_iter *iter)
4613 int err;
4615 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4617 err = rhashtable_walk_start(iter);
4618 if (err && err != -EAGAIN) {
4619 rhashtable_walk_stop(iter);
4620 rhashtable_walk_exit(iter);
4621 return err;
4624 return 0;
4627 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4629 rhashtable_walk_stop(iter);
4630 rhashtable_walk_exit(iter);
4633 struct sctp_transport *sctp_transport_get_next(struct net *net,
4634 struct rhashtable_iter *iter)
4636 struct sctp_transport *t;
4638 t = rhashtable_walk_next(iter);
4639 for (; t; t = rhashtable_walk_next(iter)) {
4640 if (IS_ERR(t)) {
4641 if (PTR_ERR(t) == -EAGAIN)
4642 continue;
4643 break;
4646 if (!sctp_transport_hold(t))
4647 continue;
4649 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4650 t->asoc->peer.primary_path == t)
4651 break;
4653 sctp_transport_put(t);
4656 return t;
4659 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4660 struct rhashtable_iter *iter,
4661 int pos)
4663 struct sctp_transport *t;
4665 if (!pos)
4666 return SEQ_START_TOKEN;
4668 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
4669 if (!--pos)
4670 break;
4671 sctp_transport_put(t);
4674 return t;
4677 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
4678 void *p) {
4679 int err = 0;
4680 int hash = 0;
4681 struct sctp_ep_common *epb;
4682 struct sctp_hashbucket *head;
4684 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
4685 hash++, head++) {
4686 read_lock_bh(&head->lock);
4687 sctp_for_each_hentry(epb, &head->chain) {
4688 err = cb(sctp_ep(epb), p);
4689 if (err)
4690 break;
4692 read_unlock_bh(&head->lock);
4695 return err;
4697 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
4699 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
4700 struct net *net,
4701 const union sctp_addr *laddr,
4702 const union sctp_addr *paddr, void *p)
4704 struct sctp_transport *transport;
4705 int err;
4707 rcu_read_lock();
4708 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
4709 rcu_read_unlock();
4710 if (!transport)
4711 return -ENOENT;
4713 err = cb(transport, p);
4714 sctp_transport_put(transport);
4716 return err;
4718 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
4720 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
4721 int (*cb_done)(struct sctp_transport *, void *),
4722 struct net *net, int *pos, void *p) {
4723 struct rhashtable_iter hti;
4724 struct sctp_transport *tsp;
4725 int ret;
4727 again:
4728 ret = sctp_transport_walk_start(&hti);
4729 if (ret)
4730 return ret;
4732 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
4733 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
4734 ret = cb(tsp, p);
4735 if (ret)
4736 break;
4737 (*pos)++;
4738 sctp_transport_put(tsp);
4740 sctp_transport_walk_stop(&hti);
4742 if (ret) {
4743 if (cb_done && !cb_done(tsp, p)) {
4744 (*pos)++;
4745 sctp_transport_put(tsp);
4746 goto again;
4748 sctp_transport_put(tsp);
4751 return ret;
4753 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
4755 /* 7.2.1 Association Status (SCTP_STATUS)
4757 * Applications can retrieve current status information about an
4758 * association, including association state, peer receiver window size,
4759 * number of unacked data chunks, and number of data chunks pending
4760 * receipt. This information is read-only.
4762 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4763 char __user *optval,
4764 int __user *optlen)
4766 struct sctp_status status;
4767 struct sctp_association *asoc = NULL;
4768 struct sctp_transport *transport;
4769 sctp_assoc_t associd;
4770 int retval = 0;
4772 if (len < sizeof(status)) {
4773 retval = -EINVAL;
4774 goto out;
4777 len = sizeof(status);
4778 if (copy_from_user(&status, optval, len)) {
4779 retval = -EFAULT;
4780 goto out;
4783 associd = status.sstat_assoc_id;
4784 asoc = sctp_id2assoc(sk, associd);
4785 if (!asoc) {
4786 retval = -EINVAL;
4787 goto out;
4790 transport = asoc->peer.primary_path;
4792 status.sstat_assoc_id = sctp_assoc2id(asoc);
4793 status.sstat_state = sctp_assoc_to_state(asoc);
4794 status.sstat_rwnd = asoc->peer.rwnd;
4795 status.sstat_unackdata = asoc->unack_data;
4797 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4798 status.sstat_instrms = asoc->stream.incnt;
4799 status.sstat_outstrms = asoc->stream.outcnt;
4800 status.sstat_fragmentation_point = asoc->frag_point;
4801 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4802 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4803 transport->af_specific->sockaddr_len);
4804 /* Map ipv4 address into v4-mapped-on-v6 address. */
4805 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
4806 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4807 status.sstat_primary.spinfo_state = transport->state;
4808 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4809 status.sstat_primary.spinfo_srtt = transport->srtt;
4810 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4811 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4813 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4814 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4816 if (put_user(len, optlen)) {
4817 retval = -EFAULT;
4818 goto out;
4821 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4822 __func__, len, status.sstat_state, status.sstat_rwnd,
4823 status.sstat_assoc_id);
4825 if (copy_to_user(optval, &status, len)) {
4826 retval = -EFAULT;
4827 goto out;
4830 out:
4831 return retval;
4835 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4837 * Applications can retrieve information about a specific peer address
4838 * of an association, including its reachability state, congestion
4839 * window, and retransmission timer values. This information is
4840 * read-only.
4842 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4843 char __user *optval,
4844 int __user *optlen)
4846 struct sctp_paddrinfo pinfo;
4847 struct sctp_transport *transport;
4848 int retval = 0;
4850 if (len < sizeof(pinfo)) {
4851 retval = -EINVAL;
4852 goto out;
4855 len = sizeof(pinfo);
4856 if (copy_from_user(&pinfo, optval, len)) {
4857 retval = -EFAULT;
4858 goto out;
4861 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4862 pinfo.spinfo_assoc_id);
4863 if (!transport)
4864 return -EINVAL;
4866 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4867 pinfo.spinfo_state = transport->state;
4868 pinfo.spinfo_cwnd = transport->cwnd;
4869 pinfo.spinfo_srtt = transport->srtt;
4870 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4871 pinfo.spinfo_mtu = transport->pathmtu;
4873 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4874 pinfo.spinfo_state = SCTP_ACTIVE;
4876 if (put_user(len, optlen)) {
4877 retval = -EFAULT;
4878 goto out;
4881 if (copy_to_user(optval, &pinfo, len)) {
4882 retval = -EFAULT;
4883 goto out;
4886 out:
4887 return retval;
4890 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4892 * This option is a on/off flag. If enabled no SCTP message
4893 * fragmentation will be performed. Instead if a message being sent
4894 * exceeds the current PMTU size, the message will NOT be sent and
4895 * instead a error will be indicated to the user.
4897 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4898 char __user *optval, int __user *optlen)
4900 int val;
4902 if (len < sizeof(int))
4903 return -EINVAL;
4905 len = sizeof(int);
4906 val = (sctp_sk(sk)->disable_fragments == 1);
4907 if (put_user(len, optlen))
4908 return -EFAULT;
4909 if (copy_to_user(optval, &val, len))
4910 return -EFAULT;
4911 return 0;
4914 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4916 * This socket option is used to specify various notifications and
4917 * ancillary data the user wishes to receive.
4919 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4920 int __user *optlen)
4922 if (len == 0)
4923 return -EINVAL;
4924 if (len > sizeof(struct sctp_event_subscribe))
4925 len = sizeof(struct sctp_event_subscribe);
4926 if (put_user(len, optlen))
4927 return -EFAULT;
4928 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4929 return -EFAULT;
4930 return 0;
4933 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4935 * This socket option is applicable to the UDP-style socket only. When
4936 * set it will cause associations that are idle for more than the
4937 * specified number of seconds to automatically close. An association
4938 * being idle is defined an association that has NOT sent or received
4939 * user data. The special value of '0' indicates that no automatic
4940 * close of any associations should be performed. The option expects an
4941 * integer defining the number of seconds of idle time before an
4942 * association is closed.
4944 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4946 /* Applicable to UDP-style socket only */
4947 if (sctp_style(sk, TCP))
4948 return -EOPNOTSUPP;
4949 if (len < sizeof(int))
4950 return -EINVAL;
4951 len = sizeof(int);
4952 if (put_user(len, optlen))
4953 return -EFAULT;
4954 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, len))
4955 return -EFAULT;
4956 return 0;
4959 /* Helper routine to branch off an association to a new socket. */
4960 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4962 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4963 struct sctp_sock *sp = sctp_sk(sk);
4964 struct socket *sock;
4965 int err = 0;
4967 /* Do not peel off from one netns to another one. */
4968 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
4969 return -EINVAL;
4971 if (!asoc)
4972 return -EINVAL;
4974 /* An association cannot be branched off from an already peeled-off
4975 * socket, nor is this supported for tcp style sockets.
4977 if (!sctp_style(sk, UDP))
4978 return -EINVAL;
4980 /* Create a new socket. */
4981 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4982 if (err < 0)
4983 return err;
4985 sctp_copy_sock(sock->sk, sk, asoc);
4987 /* Make peeled-off sockets more like 1-1 accepted sockets.
4988 * Set the daddr and initialize id to something more random
4990 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
4992 /* Populate the fields of the newsk from the oldsk and migrate the
4993 * asoc to the newsk.
4995 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4997 *sockp = sock;
4999 return err;
5001 EXPORT_SYMBOL(sctp_do_peeloff);
5003 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5004 struct file **newfile, unsigned flags)
5006 struct socket *newsock;
5007 int retval;
5009 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5010 if (retval < 0)
5011 goto out;
5013 /* Map the socket to an unused fd that can be returned to the user. */
5014 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5015 if (retval < 0) {
5016 sock_release(newsock);
5017 goto out;
5020 *newfile = sock_alloc_file(newsock, 0, NULL);
5021 if (IS_ERR(*newfile)) {
5022 put_unused_fd(retval);
5023 sock_release(newsock);
5024 retval = PTR_ERR(*newfile);
5025 *newfile = NULL;
5026 return retval;
5029 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5030 retval);
5032 peeloff->sd = retval;
5034 if (flags & SOCK_NONBLOCK)
5035 (*newfile)->f_flags |= O_NONBLOCK;
5036 out:
5037 return retval;
5040 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5042 sctp_peeloff_arg_t peeloff;
5043 struct file *newfile = NULL;
5044 int retval = 0;
5046 if (len < sizeof(sctp_peeloff_arg_t))
5047 return -EINVAL;
5048 len = sizeof(sctp_peeloff_arg_t);
5049 if (copy_from_user(&peeloff, optval, len))
5050 return -EFAULT;
5052 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5053 if (retval < 0)
5054 goto out;
5056 /* Return the fd mapped to the new socket. */
5057 if (put_user(len, optlen)) {
5058 fput(newfile);
5059 put_unused_fd(retval);
5060 return -EFAULT;
5063 if (copy_to_user(optval, &peeloff, len)) {
5064 fput(newfile);
5065 put_unused_fd(retval);
5066 return -EFAULT;
5068 fd_install(retval, newfile);
5069 out:
5070 return retval;
5073 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5074 char __user *optval, int __user *optlen)
5076 sctp_peeloff_flags_arg_t peeloff;
5077 struct file *newfile = NULL;
5078 int retval = 0;
5080 if (len < sizeof(sctp_peeloff_flags_arg_t))
5081 return -EINVAL;
5082 len = sizeof(sctp_peeloff_flags_arg_t);
5083 if (copy_from_user(&peeloff, optval, len))
5084 return -EFAULT;
5086 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5087 &newfile, peeloff.flags);
5088 if (retval < 0)
5089 goto out;
5091 /* Return the fd mapped to the new socket. */
5092 if (put_user(len, optlen)) {
5093 fput(newfile);
5094 put_unused_fd(retval);
5095 return -EFAULT;
5098 if (copy_to_user(optval, &peeloff, len)) {
5099 fput(newfile);
5100 put_unused_fd(retval);
5101 return -EFAULT;
5103 fd_install(retval, newfile);
5104 out:
5105 return retval;
5108 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5110 * Applications can enable or disable heartbeats for any peer address of
5111 * an association, modify an address's heartbeat interval, force a
5112 * heartbeat to be sent immediately, and adjust the address's maximum
5113 * number of retransmissions sent before an address is considered
5114 * unreachable. The following structure is used to access and modify an
5115 * address's parameters:
5117 * struct sctp_paddrparams {
5118 * sctp_assoc_t spp_assoc_id;
5119 * struct sockaddr_storage spp_address;
5120 * uint32_t spp_hbinterval;
5121 * uint16_t spp_pathmaxrxt;
5122 * uint32_t spp_pathmtu;
5123 * uint32_t spp_sackdelay;
5124 * uint32_t spp_flags;
5125 * };
5127 * spp_assoc_id - (one-to-many style socket) This is filled in the
5128 * application, and identifies the association for
5129 * this query.
5130 * spp_address - This specifies which address is of interest.
5131 * spp_hbinterval - This contains the value of the heartbeat interval,
5132 * in milliseconds. If a value of zero
5133 * is present in this field then no changes are to
5134 * be made to this parameter.
5135 * spp_pathmaxrxt - This contains the maximum number of
5136 * retransmissions before this address shall be
5137 * considered unreachable. If a value of zero
5138 * is present in this field then no changes are to
5139 * be made to this parameter.
5140 * spp_pathmtu - When Path MTU discovery is disabled the value
5141 * specified here will be the "fixed" path mtu.
5142 * Note that if the spp_address field is empty
5143 * then all associations on this address will
5144 * have this fixed path mtu set upon them.
5146 * spp_sackdelay - When delayed sack is enabled, this value specifies
5147 * the number of milliseconds that sacks will be delayed
5148 * for. This value will apply to all addresses of an
5149 * association if the spp_address field is empty. Note
5150 * also, that if delayed sack is enabled and this
5151 * value is set to 0, no change is made to the last
5152 * recorded delayed sack timer value.
5154 * spp_flags - These flags are used to control various features
5155 * on an association. The flag field may contain
5156 * zero or more of the following options.
5158 * SPP_HB_ENABLE - Enable heartbeats on the
5159 * specified address. Note that if the address
5160 * field is empty all addresses for the association
5161 * have heartbeats enabled upon them.
5163 * SPP_HB_DISABLE - Disable heartbeats on the
5164 * speicifed address. Note that if the address
5165 * field is empty all addresses for the association
5166 * will have their heartbeats disabled. Note also
5167 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5168 * mutually exclusive, only one of these two should
5169 * be specified. Enabling both fields will have
5170 * undetermined results.
5172 * SPP_HB_DEMAND - Request a user initiated heartbeat
5173 * to be made immediately.
5175 * SPP_PMTUD_ENABLE - This field will enable PMTU
5176 * discovery upon the specified address. Note that
5177 * if the address feild is empty then all addresses
5178 * on the association are effected.
5180 * SPP_PMTUD_DISABLE - This field will disable PMTU
5181 * discovery upon the specified address. Note that
5182 * if the address feild is empty then all addresses
5183 * on the association are effected. Not also that
5184 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5185 * exclusive. Enabling both will have undetermined
5186 * results.
5188 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5189 * on delayed sack. The time specified in spp_sackdelay
5190 * is used to specify the sack delay for this address. Note
5191 * that if spp_address is empty then all addresses will
5192 * enable delayed sack and take on the sack delay
5193 * value specified in spp_sackdelay.
5194 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5195 * off delayed sack. If the spp_address field is blank then
5196 * delayed sack is disabled for the entire association. Note
5197 * also that this field is mutually exclusive to
5198 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5199 * results.
5201 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5202 char __user *optval, int __user *optlen)
5204 struct sctp_paddrparams params;
5205 struct sctp_transport *trans = NULL;
5206 struct sctp_association *asoc = NULL;
5207 struct sctp_sock *sp = sctp_sk(sk);
5209 if (len < sizeof(struct sctp_paddrparams))
5210 return -EINVAL;
5211 len = sizeof(struct sctp_paddrparams);
5212 if (copy_from_user(&params, optval, len))
5213 return -EFAULT;
5215 /* If an address other than INADDR_ANY is specified, and
5216 * no transport is found, then the request is invalid.
5218 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
5219 trans = sctp_addr_id2transport(sk, &params.spp_address,
5220 params.spp_assoc_id);
5221 if (!trans) {
5222 pr_debug("%s: failed no transport\n", __func__);
5223 return -EINVAL;
5227 /* Get association, if assoc_id != 0 and the socket is a one
5228 * to many style socket, and an association was not found, then
5229 * the id was invalid.
5231 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5232 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5233 pr_debug("%s: failed no association\n", __func__);
5234 return -EINVAL;
5237 if (trans) {
5238 /* Fetch transport values. */
5239 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5240 params.spp_pathmtu = trans->pathmtu;
5241 params.spp_pathmaxrxt = trans->pathmaxrxt;
5242 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5244 /*draft-11 doesn't say what to return in spp_flags*/
5245 params.spp_flags = trans->param_flags;
5246 } else if (asoc) {
5247 /* Fetch association values. */
5248 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5249 params.spp_pathmtu = asoc->pathmtu;
5250 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5251 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5253 /*draft-11 doesn't say what to return in spp_flags*/
5254 params.spp_flags = asoc->param_flags;
5255 } else {
5256 /* Fetch socket values. */
5257 params.spp_hbinterval = sp->hbinterval;
5258 params.spp_pathmtu = sp->pathmtu;
5259 params.spp_sackdelay = sp->sackdelay;
5260 params.spp_pathmaxrxt = sp->pathmaxrxt;
5262 /*draft-11 doesn't say what to return in spp_flags*/
5263 params.spp_flags = sp->param_flags;
5266 if (copy_to_user(optval, &params, len))
5267 return -EFAULT;
5269 if (put_user(len, optlen))
5270 return -EFAULT;
5272 return 0;
5276 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5278 * This option will effect the way delayed acks are performed. This
5279 * option allows you to get or set the delayed ack time, in
5280 * milliseconds. It also allows changing the delayed ack frequency.
5281 * Changing the frequency to 1 disables the delayed sack algorithm. If
5282 * the assoc_id is 0, then this sets or gets the endpoints default
5283 * values. If the assoc_id field is non-zero, then the set or get
5284 * effects the specified association for the one to many model (the
5285 * assoc_id field is ignored by the one to one model). Note that if
5286 * sack_delay or sack_freq are 0 when setting this option, then the
5287 * current values will remain unchanged.
5289 * struct sctp_sack_info {
5290 * sctp_assoc_t sack_assoc_id;
5291 * uint32_t sack_delay;
5292 * uint32_t sack_freq;
5293 * };
5295 * sack_assoc_id - This parameter, indicates which association the user
5296 * is performing an action upon. Note that if this field's value is
5297 * zero then the endpoints default value is changed (effecting future
5298 * associations only).
5300 * sack_delay - This parameter contains the number of milliseconds that
5301 * the user is requesting the delayed ACK timer be set to. Note that
5302 * this value is defined in the standard to be between 200 and 500
5303 * milliseconds.
5305 * sack_freq - This parameter contains the number of packets that must
5306 * be received before a sack is sent without waiting for the delay
5307 * timer to expire. The default value for this is 2, setting this
5308 * value to 1 will disable the delayed sack algorithm.
5310 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5311 char __user *optval,
5312 int __user *optlen)
5314 struct sctp_sack_info params;
5315 struct sctp_association *asoc = NULL;
5316 struct sctp_sock *sp = sctp_sk(sk);
5318 if (len >= sizeof(struct sctp_sack_info)) {
5319 len = sizeof(struct sctp_sack_info);
5321 if (copy_from_user(&params, optval, len))
5322 return -EFAULT;
5323 } else if (len == sizeof(struct sctp_assoc_value)) {
5324 pr_warn_ratelimited(DEPRECATED
5325 "%s (pid %d) "
5326 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5327 "Use struct sctp_sack_info instead\n",
5328 current->comm, task_pid_nr(current));
5329 if (copy_from_user(&params, optval, len))
5330 return -EFAULT;
5331 } else
5332 return -EINVAL;
5334 /* Get association, if sack_assoc_id != 0 and the socket is a one
5335 * to many style socket, and an association was not found, then
5336 * the id was invalid.
5338 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5339 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5340 return -EINVAL;
5342 if (asoc) {
5343 /* Fetch association values. */
5344 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5345 params.sack_delay = jiffies_to_msecs(
5346 asoc->sackdelay);
5347 params.sack_freq = asoc->sackfreq;
5349 } else {
5350 params.sack_delay = 0;
5351 params.sack_freq = 1;
5353 } else {
5354 /* Fetch socket values. */
5355 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5356 params.sack_delay = sp->sackdelay;
5357 params.sack_freq = sp->sackfreq;
5358 } else {
5359 params.sack_delay = 0;
5360 params.sack_freq = 1;
5364 if (copy_to_user(optval, &params, len))
5365 return -EFAULT;
5367 if (put_user(len, optlen))
5368 return -EFAULT;
5370 return 0;
5373 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5375 * Applications can specify protocol parameters for the default association
5376 * initialization. The option name argument to setsockopt() and getsockopt()
5377 * is SCTP_INITMSG.
5379 * Setting initialization parameters is effective only on an unconnected
5380 * socket (for UDP-style sockets only future associations are effected
5381 * by the change). With TCP-style sockets, this option is inherited by
5382 * sockets derived from a listener socket.
5384 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5386 if (len < sizeof(struct sctp_initmsg))
5387 return -EINVAL;
5388 len = sizeof(struct sctp_initmsg);
5389 if (put_user(len, optlen))
5390 return -EFAULT;
5391 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5392 return -EFAULT;
5393 return 0;
5397 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5398 char __user *optval, int __user *optlen)
5400 struct sctp_association *asoc;
5401 int cnt = 0;
5402 struct sctp_getaddrs getaddrs;
5403 struct sctp_transport *from;
5404 void __user *to;
5405 union sctp_addr temp;
5406 struct sctp_sock *sp = sctp_sk(sk);
5407 int addrlen;
5408 size_t space_left;
5409 int bytes_copied;
5411 if (len < sizeof(struct sctp_getaddrs))
5412 return -EINVAL;
5414 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5415 return -EFAULT;
5417 /* For UDP-style sockets, id specifies the association to query. */
5418 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5419 if (!asoc)
5420 return -EINVAL;
5422 to = optval + offsetof(struct sctp_getaddrs, addrs);
5423 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5425 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5426 transports) {
5427 memcpy(&temp, &from->ipaddr, sizeof(temp));
5428 addrlen = sctp_get_pf_specific(sk->sk_family)
5429 ->addr_to_user(sp, &temp);
5430 if (space_left < addrlen)
5431 return -ENOMEM;
5432 if (copy_to_user(to, &temp, addrlen))
5433 return -EFAULT;
5434 to += addrlen;
5435 cnt++;
5436 space_left -= addrlen;
5439 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5440 return -EFAULT;
5441 bytes_copied = ((char __user *)to) - optval;
5442 if (put_user(bytes_copied, optlen))
5443 return -EFAULT;
5445 return 0;
5448 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5449 size_t space_left, int *bytes_copied)
5451 struct sctp_sockaddr_entry *addr;
5452 union sctp_addr temp;
5453 int cnt = 0;
5454 int addrlen;
5455 struct net *net = sock_net(sk);
5457 rcu_read_lock();
5458 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5459 if (!addr->valid)
5460 continue;
5462 if ((PF_INET == sk->sk_family) &&
5463 (AF_INET6 == addr->a.sa.sa_family))
5464 continue;
5465 if ((PF_INET6 == sk->sk_family) &&
5466 inet_v6_ipv6only(sk) &&
5467 (AF_INET == addr->a.sa.sa_family))
5468 continue;
5469 memcpy(&temp, &addr->a, sizeof(temp));
5470 if (!temp.v4.sin_port)
5471 temp.v4.sin_port = htons(port);
5473 addrlen = sctp_get_pf_specific(sk->sk_family)
5474 ->addr_to_user(sctp_sk(sk), &temp);
5476 if (space_left < addrlen) {
5477 cnt = -ENOMEM;
5478 break;
5480 memcpy(to, &temp, addrlen);
5482 to += addrlen;
5483 cnt++;
5484 space_left -= addrlen;
5485 *bytes_copied += addrlen;
5487 rcu_read_unlock();
5489 return cnt;
5493 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5494 char __user *optval, int __user *optlen)
5496 struct sctp_bind_addr *bp;
5497 struct sctp_association *asoc;
5498 int cnt = 0;
5499 struct sctp_getaddrs getaddrs;
5500 struct sctp_sockaddr_entry *addr;
5501 void __user *to;
5502 union sctp_addr temp;
5503 struct sctp_sock *sp = sctp_sk(sk);
5504 int addrlen;
5505 int err = 0;
5506 size_t space_left;
5507 int bytes_copied = 0;
5508 void *addrs;
5509 void *buf;
5511 if (len < sizeof(struct sctp_getaddrs))
5512 return -EINVAL;
5514 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5515 return -EFAULT;
5518 * For UDP-style sockets, id specifies the association to query.
5519 * If the id field is set to the value '0' then the locally bound
5520 * addresses are returned without regard to any particular
5521 * association.
5523 if (0 == getaddrs.assoc_id) {
5524 bp = &sctp_sk(sk)->ep->base.bind_addr;
5525 } else {
5526 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5527 if (!asoc)
5528 return -EINVAL;
5529 bp = &asoc->base.bind_addr;
5532 to = optval + offsetof(struct sctp_getaddrs, addrs);
5533 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5535 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5536 if (!addrs)
5537 return -ENOMEM;
5539 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5540 * addresses from the global local address list.
5542 if (sctp_list_single_entry(&bp->address_list)) {
5543 addr = list_entry(bp->address_list.next,
5544 struct sctp_sockaddr_entry, list);
5545 if (sctp_is_any(sk, &addr->a)) {
5546 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5547 space_left, &bytes_copied);
5548 if (cnt < 0) {
5549 err = cnt;
5550 goto out;
5552 goto copy_getaddrs;
5556 buf = addrs;
5557 /* Protection on the bound address list is not needed since
5558 * in the socket option context we hold a socket lock and
5559 * thus the bound address list can't change.
5561 list_for_each_entry(addr, &bp->address_list, list) {
5562 memcpy(&temp, &addr->a, sizeof(temp));
5563 addrlen = sctp_get_pf_specific(sk->sk_family)
5564 ->addr_to_user(sp, &temp);
5565 if (space_left < addrlen) {
5566 err = -ENOMEM; /*fixme: right error?*/
5567 goto out;
5569 memcpy(buf, &temp, addrlen);
5570 buf += addrlen;
5571 bytes_copied += addrlen;
5572 cnt++;
5573 space_left -= addrlen;
5576 copy_getaddrs:
5577 if (copy_to_user(to, addrs, bytes_copied)) {
5578 err = -EFAULT;
5579 goto out;
5581 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5582 err = -EFAULT;
5583 goto out;
5585 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5586 * but we can't change it anymore.
5588 if (put_user(bytes_copied, optlen))
5589 err = -EFAULT;
5590 out:
5591 kfree(addrs);
5592 return err;
5595 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
5597 * Requests that the local SCTP stack use the enclosed peer address as
5598 * the association primary. The enclosed address must be one of the
5599 * association peer's addresses.
5601 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
5602 char __user *optval, int __user *optlen)
5604 struct sctp_prim prim;
5605 struct sctp_association *asoc;
5606 struct sctp_sock *sp = sctp_sk(sk);
5608 if (len < sizeof(struct sctp_prim))
5609 return -EINVAL;
5611 len = sizeof(struct sctp_prim);
5613 if (copy_from_user(&prim, optval, len))
5614 return -EFAULT;
5616 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
5617 if (!asoc)
5618 return -EINVAL;
5620 if (!asoc->peer.primary_path)
5621 return -ENOTCONN;
5623 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
5624 asoc->peer.primary_path->af_specific->sockaddr_len);
5626 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
5627 (union sctp_addr *)&prim.ssp_addr);
5629 if (put_user(len, optlen))
5630 return -EFAULT;
5631 if (copy_to_user(optval, &prim, len))
5632 return -EFAULT;
5634 return 0;
5638 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
5640 * Requests that the local endpoint set the specified Adaptation Layer
5641 * Indication parameter for all future INIT and INIT-ACK exchanges.
5643 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
5644 char __user *optval, int __user *optlen)
5646 struct sctp_setadaptation adaptation;
5648 if (len < sizeof(struct sctp_setadaptation))
5649 return -EINVAL;
5651 len = sizeof(struct sctp_setadaptation);
5653 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
5655 if (put_user(len, optlen))
5656 return -EFAULT;
5657 if (copy_to_user(optval, &adaptation, len))
5658 return -EFAULT;
5660 return 0;
5665 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
5667 * Applications that wish to use the sendto() system call may wish to
5668 * specify a default set of parameters that would normally be supplied
5669 * through the inclusion of ancillary data. This socket option allows
5670 * such an application to set the default sctp_sndrcvinfo structure.
5673 * The application that wishes to use this socket option simply passes
5674 * in to this call the sctp_sndrcvinfo structure defined in Section
5675 * 5.2.2) The input parameters accepted by this call include
5676 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
5677 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
5678 * to this call if the caller is using the UDP model.
5680 * For getsockopt, it get the default sctp_sndrcvinfo structure.
5682 static int sctp_getsockopt_default_send_param(struct sock *sk,
5683 int len, char __user *optval,
5684 int __user *optlen)
5686 struct sctp_sock *sp = sctp_sk(sk);
5687 struct sctp_association *asoc;
5688 struct sctp_sndrcvinfo info;
5690 if (len < sizeof(info))
5691 return -EINVAL;
5693 len = sizeof(info);
5695 if (copy_from_user(&info, optval, len))
5696 return -EFAULT;
5698 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
5699 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
5700 return -EINVAL;
5701 if (asoc) {
5702 info.sinfo_stream = asoc->default_stream;
5703 info.sinfo_flags = asoc->default_flags;
5704 info.sinfo_ppid = asoc->default_ppid;
5705 info.sinfo_context = asoc->default_context;
5706 info.sinfo_timetolive = asoc->default_timetolive;
5707 } else {
5708 info.sinfo_stream = sp->default_stream;
5709 info.sinfo_flags = sp->default_flags;
5710 info.sinfo_ppid = sp->default_ppid;
5711 info.sinfo_context = sp->default_context;
5712 info.sinfo_timetolive = sp->default_timetolive;
5715 if (put_user(len, optlen))
5716 return -EFAULT;
5717 if (copy_to_user(optval, &info, len))
5718 return -EFAULT;
5720 return 0;
5723 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
5724 * (SCTP_DEFAULT_SNDINFO)
5726 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
5727 char __user *optval,
5728 int __user *optlen)
5730 struct sctp_sock *sp = sctp_sk(sk);
5731 struct sctp_association *asoc;
5732 struct sctp_sndinfo info;
5734 if (len < sizeof(info))
5735 return -EINVAL;
5737 len = sizeof(info);
5739 if (copy_from_user(&info, optval, len))
5740 return -EFAULT;
5742 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
5743 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
5744 return -EINVAL;
5745 if (asoc) {
5746 info.snd_sid = asoc->default_stream;
5747 info.snd_flags = asoc->default_flags;
5748 info.snd_ppid = asoc->default_ppid;
5749 info.snd_context = asoc->default_context;
5750 } else {
5751 info.snd_sid = sp->default_stream;
5752 info.snd_flags = sp->default_flags;
5753 info.snd_ppid = sp->default_ppid;
5754 info.snd_context = sp->default_context;
5757 if (put_user(len, optlen))
5758 return -EFAULT;
5759 if (copy_to_user(optval, &info, len))
5760 return -EFAULT;
5762 return 0;
5767 * 7.1.5 SCTP_NODELAY
5769 * Turn on/off any Nagle-like algorithm. This means that packets are
5770 * generally sent as soon as possible and no unnecessary delays are
5771 * introduced, at the cost of more packets in the network. Expects an
5772 * integer boolean flag.
5775 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5776 char __user *optval, int __user *optlen)
5778 int val;
5780 if (len < sizeof(int))
5781 return -EINVAL;
5783 len = sizeof(int);
5784 val = (sctp_sk(sk)->nodelay == 1);
5785 if (put_user(len, optlen))
5786 return -EFAULT;
5787 if (copy_to_user(optval, &val, len))
5788 return -EFAULT;
5789 return 0;
5794 * 7.1.1 SCTP_RTOINFO
5796 * The protocol parameters used to initialize and bound retransmission
5797 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5798 * and modify these parameters.
5799 * All parameters are time values, in milliseconds. A value of 0, when
5800 * modifying the parameters, indicates that the current value should not
5801 * be changed.
5804 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5805 char __user *optval,
5806 int __user *optlen) {
5807 struct sctp_rtoinfo rtoinfo;
5808 struct sctp_association *asoc;
5810 if (len < sizeof (struct sctp_rtoinfo))
5811 return -EINVAL;
5813 len = sizeof(struct sctp_rtoinfo);
5815 if (copy_from_user(&rtoinfo, optval, len))
5816 return -EFAULT;
5818 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5820 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5821 return -EINVAL;
5823 /* Values corresponding to the specific association. */
5824 if (asoc) {
5825 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5826 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5827 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5828 } else {
5829 /* Values corresponding to the endpoint. */
5830 struct sctp_sock *sp = sctp_sk(sk);
5832 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5833 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5834 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5837 if (put_user(len, optlen))
5838 return -EFAULT;
5840 if (copy_to_user(optval, &rtoinfo, len))
5841 return -EFAULT;
5843 return 0;
5848 * 7.1.2 SCTP_ASSOCINFO
5850 * This option is used to tune the maximum retransmission attempts
5851 * of the association.
5852 * Returns an error if the new association retransmission value is
5853 * greater than the sum of the retransmission value of the peer.
5854 * See [SCTP] for more information.
5857 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5858 char __user *optval,
5859 int __user *optlen)
5862 struct sctp_assocparams assocparams;
5863 struct sctp_association *asoc;
5864 struct list_head *pos;
5865 int cnt = 0;
5867 if (len < sizeof (struct sctp_assocparams))
5868 return -EINVAL;
5870 len = sizeof(struct sctp_assocparams);
5872 if (copy_from_user(&assocparams, optval, len))
5873 return -EFAULT;
5875 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5877 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5878 return -EINVAL;
5880 /* Values correspoinding to the specific association */
5881 if (asoc) {
5882 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5883 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5884 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5885 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5887 list_for_each(pos, &asoc->peer.transport_addr_list) {
5888 cnt++;
5891 assocparams.sasoc_number_peer_destinations = cnt;
5892 } else {
5893 /* Values corresponding to the endpoint */
5894 struct sctp_sock *sp = sctp_sk(sk);
5896 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5897 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5898 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5899 assocparams.sasoc_cookie_life =
5900 sp->assocparams.sasoc_cookie_life;
5901 assocparams.sasoc_number_peer_destinations =
5902 sp->assocparams.
5903 sasoc_number_peer_destinations;
5906 if (put_user(len, optlen))
5907 return -EFAULT;
5909 if (copy_to_user(optval, &assocparams, len))
5910 return -EFAULT;
5912 return 0;
5916 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5918 * This socket option is a boolean flag which turns on or off mapped V4
5919 * addresses. If this option is turned on and the socket is type
5920 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5921 * If this option is turned off, then no mapping will be done of V4
5922 * addresses and a user will receive both PF_INET6 and PF_INET type
5923 * addresses on the socket.
5925 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5926 char __user *optval, int __user *optlen)
5928 int val;
5929 struct sctp_sock *sp = sctp_sk(sk);
5931 if (len < sizeof(int))
5932 return -EINVAL;
5934 len = sizeof(int);
5935 val = sp->v4mapped;
5936 if (put_user(len, optlen))
5937 return -EFAULT;
5938 if (copy_to_user(optval, &val, len))
5939 return -EFAULT;
5941 return 0;
5945 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5946 * (chapter and verse is quoted at sctp_setsockopt_context())
5948 static int sctp_getsockopt_context(struct sock *sk, int len,
5949 char __user *optval, int __user *optlen)
5951 struct sctp_assoc_value params;
5952 struct sctp_sock *sp;
5953 struct sctp_association *asoc;
5955 if (len < sizeof(struct sctp_assoc_value))
5956 return -EINVAL;
5958 len = sizeof(struct sctp_assoc_value);
5960 if (copy_from_user(&params, optval, len))
5961 return -EFAULT;
5963 sp = sctp_sk(sk);
5965 if (params.assoc_id != 0) {
5966 asoc = sctp_id2assoc(sk, params.assoc_id);
5967 if (!asoc)
5968 return -EINVAL;
5969 params.assoc_value = asoc->default_rcv_context;
5970 } else {
5971 params.assoc_value = sp->default_rcv_context;
5974 if (put_user(len, optlen))
5975 return -EFAULT;
5976 if (copy_to_user(optval, &params, len))
5977 return -EFAULT;
5979 return 0;
5983 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5984 * This option will get or set the maximum size to put in any outgoing
5985 * SCTP DATA chunk. If a message is larger than this size it will be
5986 * fragmented by SCTP into the specified size. Note that the underlying
5987 * SCTP implementation may fragment into smaller sized chunks when the
5988 * PMTU of the underlying association is smaller than the value set by
5989 * the user. The default value for this option is '0' which indicates
5990 * the user is NOT limiting fragmentation and only the PMTU will effect
5991 * SCTP's choice of DATA chunk size. Note also that values set larger
5992 * than the maximum size of an IP datagram will effectively let SCTP
5993 * control fragmentation (i.e. the same as setting this option to 0).
5995 * The following structure is used to access and modify this parameter:
5997 * struct sctp_assoc_value {
5998 * sctp_assoc_t assoc_id;
5999 * uint32_t assoc_value;
6000 * };
6002 * assoc_id: This parameter is ignored for one-to-one style sockets.
6003 * For one-to-many style sockets this parameter indicates which
6004 * association the user is performing an action upon. Note that if
6005 * this field's value is zero then the endpoints default value is
6006 * changed (effecting future associations only).
6007 * assoc_value: This parameter specifies the maximum size in bytes.
6009 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6010 char __user *optval, int __user *optlen)
6012 struct sctp_assoc_value params;
6013 struct sctp_association *asoc;
6015 if (len == sizeof(int)) {
6016 pr_warn_ratelimited(DEPRECATED
6017 "%s (pid %d) "
6018 "Use of int in maxseg socket option.\n"
6019 "Use struct sctp_assoc_value instead\n",
6020 current->comm, task_pid_nr(current));
6021 params.assoc_id = 0;
6022 } else if (len >= sizeof(struct sctp_assoc_value)) {
6023 len = sizeof(struct sctp_assoc_value);
6024 if (copy_from_user(&params, optval, len))
6025 return -EFAULT;
6026 } else
6027 return -EINVAL;
6029 asoc = sctp_id2assoc(sk, params.assoc_id);
6030 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6031 return -EINVAL;
6033 if (asoc)
6034 params.assoc_value = asoc->frag_point;
6035 else
6036 params.assoc_value = sctp_sk(sk)->user_frag;
6038 if (put_user(len, optlen))
6039 return -EFAULT;
6040 if (len == sizeof(int)) {
6041 if (copy_to_user(optval, &params.assoc_value, len))
6042 return -EFAULT;
6043 } else {
6044 if (copy_to_user(optval, &params, len))
6045 return -EFAULT;
6048 return 0;
6052 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6053 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6055 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6056 char __user *optval, int __user *optlen)
6058 int val;
6060 if (len < sizeof(int))
6061 return -EINVAL;
6063 len = sizeof(int);
6065 val = sctp_sk(sk)->frag_interleave;
6066 if (put_user(len, optlen))
6067 return -EFAULT;
6068 if (copy_to_user(optval, &val, len))
6069 return -EFAULT;
6071 return 0;
6075 * 7.1.25. Set or Get the sctp partial delivery point
6076 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6078 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6079 char __user *optval,
6080 int __user *optlen)
6082 u32 val;
6084 if (len < sizeof(u32))
6085 return -EINVAL;
6087 len = sizeof(u32);
6089 val = sctp_sk(sk)->pd_point;
6090 if (put_user(len, optlen))
6091 return -EFAULT;
6092 if (copy_to_user(optval, &val, len))
6093 return -EFAULT;
6095 return 0;
6099 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6100 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6102 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6103 char __user *optval,
6104 int __user *optlen)
6106 struct sctp_assoc_value params;
6107 struct sctp_sock *sp;
6108 struct sctp_association *asoc;
6110 if (len == sizeof(int)) {
6111 pr_warn_ratelimited(DEPRECATED
6112 "%s (pid %d) "
6113 "Use of int in max_burst socket option.\n"
6114 "Use struct sctp_assoc_value instead\n",
6115 current->comm, task_pid_nr(current));
6116 params.assoc_id = 0;
6117 } else if (len >= sizeof(struct sctp_assoc_value)) {
6118 len = sizeof(struct sctp_assoc_value);
6119 if (copy_from_user(&params, optval, len))
6120 return -EFAULT;
6121 } else
6122 return -EINVAL;
6124 sp = sctp_sk(sk);
6126 if (params.assoc_id != 0) {
6127 asoc = sctp_id2assoc(sk, params.assoc_id);
6128 if (!asoc)
6129 return -EINVAL;
6130 params.assoc_value = asoc->max_burst;
6131 } else
6132 params.assoc_value = sp->max_burst;
6134 if (len == sizeof(int)) {
6135 if (copy_to_user(optval, &params.assoc_value, len))
6136 return -EFAULT;
6137 } else {
6138 if (copy_to_user(optval, &params, len))
6139 return -EFAULT;
6142 return 0;
6146 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6147 char __user *optval, int __user *optlen)
6149 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6150 struct sctp_hmacalgo __user *p = (void __user *)optval;
6151 struct sctp_hmac_algo_param *hmacs;
6152 __u16 data_len = 0;
6153 u32 num_idents;
6154 int i;
6156 if (!ep->auth_enable)
6157 return -EACCES;
6159 hmacs = ep->auth_hmacs_list;
6160 data_len = ntohs(hmacs->param_hdr.length) -
6161 sizeof(struct sctp_paramhdr);
6163 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6164 return -EINVAL;
6166 len = sizeof(struct sctp_hmacalgo) + data_len;
6167 num_idents = data_len / sizeof(u16);
6169 if (put_user(len, optlen))
6170 return -EFAULT;
6171 if (put_user(num_idents, &p->shmac_num_idents))
6172 return -EFAULT;
6173 for (i = 0; i < num_idents; i++) {
6174 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6176 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6177 return -EFAULT;
6179 return 0;
6182 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6183 char __user *optval, int __user *optlen)
6185 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6186 struct sctp_authkeyid val;
6187 struct sctp_association *asoc;
6189 if (!ep->auth_enable)
6190 return -EACCES;
6192 if (len < sizeof(struct sctp_authkeyid))
6193 return -EINVAL;
6195 len = sizeof(struct sctp_authkeyid);
6196 if (copy_from_user(&val, optval, len))
6197 return -EFAULT;
6199 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6200 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6201 return -EINVAL;
6203 if (asoc)
6204 val.scact_keynumber = asoc->active_key_id;
6205 else
6206 val.scact_keynumber = ep->active_key_id;
6208 if (put_user(len, optlen))
6209 return -EFAULT;
6210 if (copy_to_user(optval, &val, len))
6211 return -EFAULT;
6213 return 0;
6216 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6217 char __user *optval, int __user *optlen)
6219 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6220 struct sctp_authchunks __user *p = (void __user *)optval;
6221 struct sctp_authchunks val;
6222 struct sctp_association *asoc;
6223 struct sctp_chunks_param *ch;
6224 u32 num_chunks = 0;
6225 char __user *to;
6227 if (!ep->auth_enable)
6228 return -EACCES;
6230 if (len < sizeof(struct sctp_authchunks))
6231 return -EINVAL;
6233 if (copy_from_user(&val, optval, sizeof(val)))
6234 return -EFAULT;
6236 to = p->gauth_chunks;
6237 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6238 if (!asoc)
6239 return -EINVAL;
6241 ch = asoc->peer.peer_chunks;
6242 if (!ch)
6243 goto num;
6245 /* See if the user provided enough room for all the data */
6246 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6247 if (len < num_chunks)
6248 return -EINVAL;
6250 if (copy_to_user(to, ch->chunks, num_chunks))
6251 return -EFAULT;
6252 num:
6253 len = sizeof(struct sctp_authchunks) + num_chunks;
6254 if (put_user(len, optlen))
6255 return -EFAULT;
6256 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6257 return -EFAULT;
6258 return 0;
6261 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6262 char __user *optval, int __user *optlen)
6264 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6265 struct sctp_authchunks __user *p = (void __user *)optval;
6266 struct sctp_authchunks val;
6267 struct sctp_association *asoc;
6268 struct sctp_chunks_param *ch;
6269 u32 num_chunks = 0;
6270 char __user *to;
6272 if (!ep->auth_enable)
6273 return -EACCES;
6275 if (len < sizeof(struct sctp_authchunks))
6276 return -EINVAL;
6278 if (copy_from_user(&val, optval, sizeof(val)))
6279 return -EFAULT;
6281 to = p->gauth_chunks;
6282 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6283 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6284 return -EINVAL;
6286 if (asoc)
6287 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6288 else
6289 ch = ep->auth_chunk_list;
6291 if (!ch)
6292 goto num;
6294 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6295 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6296 return -EINVAL;
6298 if (copy_to_user(to, ch->chunks, num_chunks))
6299 return -EFAULT;
6300 num:
6301 len = sizeof(struct sctp_authchunks) + num_chunks;
6302 if (put_user(len, optlen))
6303 return -EFAULT;
6304 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6305 return -EFAULT;
6307 return 0;
6311 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6312 * This option gets the current number of associations that are attached
6313 * to a one-to-many style socket. The option value is an uint32_t.
6315 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6316 char __user *optval, int __user *optlen)
6318 struct sctp_sock *sp = sctp_sk(sk);
6319 struct sctp_association *asoc;
6320 u32 val = 0;
6322 if (sctp_style(sk, TCP))
6323 return -EOPNOTSUPP;
6325 if (len < sizeof(u32))
6326 return -EINVAL;
6328 len = sizeof(u32);
6330 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6331 val++;
6334 if (put_user(len, optlen))
6335 return -EFAULT;
6336 if (copy_to_user(optval, &val, len))
6337 return -EFAULT;
6339 return 0;
6343 * 8.1.23 SCTP_AUTO_ASCONF
6344 * See the corresponding setsockopt entry as description
6346 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6347 char __user *optval, int __user *optlen)
6349 int val = 0;
6351 if (len < sizeof(int))
6352 return -EINVAL;
6354 len = sizeof(int);
6355 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6356 val = 1;
6357 if (put_user(len, optlen))
6358 return -EFAULT;
6359 if (copy_to_user(optval, &val, len))
6360 return -EFAULT;
6361 return 0;
6365 * 8.2.6. Get the Current Identifiers of Associations
6366 * (SCTP_GET_ASSOC_ID_LIST)
6368 * This option gets the current list of SCTP association identifiers of
6369 * the SCTP associations handled by a one-to-many style socket.
6371 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6372 char __user *optval, int __user *optlen)
6374 struct sctp_sock *sp = sctp_sk(sk);
6375 struct sctp_association *asoc;
6376 struct sctp_assoc_ids *ids;
6377 u32 num = 0;
6379 if (sctp_style(sk, TCP))
6380 return -EOPNOTSUPP;
6382 if (len < sizeof(struct sctp_assoc_ids))
6383 return -EINVAL;
6385 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6386 num++;
6389 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6390 return -EINVAL;
6392 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6394 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6395 if (unlikely(!ids))
6396 return -ENOMEM;
6398 ids->gaids_number_of_ids = num;
6399 num = 0;
6400 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6401 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6404 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6405 kfree(ids);
6406 return -EFAULT;
6409 kfree(ids);
6410 return 0;
6414 * SCTP_PEER_ADDR_THLDS
6416 * This option allows us to fetch the partially failed threshold for one or all
6417 * transports in an association. See Section 6.1 of:
6418 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6420 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6421 char __user *optval,
6422 int len,
6423 int __user *optlen)
6425 struct sctp_paddrthlds val;
6426 struct sctp_transport *trans;
6427 struct sctp_association *asoc;
6429 if (len < sizeof(struct sctp_paddrthlds))
6430 return -EINVAL;
6431 len = sizeof(struct sctp_paddrthlds);
6432 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6433 return -EFAULT;
6435 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6436 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6437 if (!asoc)
6438 return -ENOENT;
6440 val.spt_pathpfthld = asoc->pf_retrans;
6441 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6442 } else {
6443 trans = sctp_addr_id2transport(sk, &val.spt_address,
6444 val.spt_assoc_id);
6445 if (!trans)
6446 return -ENOENT;
6448 val.spt_pathmaxrxt = trans->pathmaxrxt;
6449 val.spt_pathpfthld = trans->pf_retrans;
6452 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6453 return -EFAULT;
6455 return 0;
6459 * SCTP_GET_ASSOC_STATS
6461 * This option retrieves local per endpoint statistics. It is modeled
6462 * after OpenSolaris' implementation
6464 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6465 char __user *optval,
6466 int __user *optlen)
6468 struct sctp_assoc_stats sas;
6469 struct sctp_association *asoc = NULL;
6471 /* User must provide at least the assoc id */
6472 if (len < sizeof(sctp_assoc_t))
6473 return -EINVAL;
6475 /* Allow the struct to grow and fill in as much as possible */
6476 len = min_t(size_t, len, sizeof(sas));
6478 if (copy_from_user(&sas, optval, len))
6479 return -EFAULT;
6481 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6482 if (!asoc)
6483 return -EINVAL;
6485 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6486 sas.sas_gapcnt = asoc->stats.gapcnt;
6487 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6488 sas.sas_osacks = asoc->stats.osacks;
6489 sas.sas_isacks = asoc->stats.isacks;
6490 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6491 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6492 sas.sas_oodchunks = asoc->stats.oodchunks;
6493 sas.sas_iodchunks = asoc->stats.iodchunks;
6494 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6495 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6496 sas.sas_idupchunks = asoc->stats.idupchunks;
6497 sas.sas_opackets = asoc->stats.opackets;
6498 sas.sas_ipackets = asoc->stats.ipackets;
6500 /* New high max rto observed, will return 0 if not a single
6501 * RTO update took place. obs_rto_ipaddr will be bogus
6502 * in such a case
6504 sas.sas_maxrto = asoc->stats.max_obs_rto;
6505 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6506 sizeof(struct sockaddr_storage));
6508 /* Mark beginning of a new observation period */
6509 asoc->stats.max_obs_rto = asoc->rto_min;
6511 if (put_user(len, optlen))
6512 return -EFAULT;
6514 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6516 if (copy_to_user(optval, &sas, len))
6517 return -EFAULT;
6519 return 0;
6522 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6523 char __user *optval,
6524 int __user *optlen)
6526 int val = 0;
6528 if (len < sizeof(int))
6529 return -EINVAL;
6531 len = sizeof(int);
6532 if (sctp_sk(sk)->recvrcvinfo)
6533 val = 1;
6534 if (put_user(len, optlen))
6535 return -EFAULT;
6536 if (copy_to_user(optval, &val, len))
6537 return -EFAULT;
6539 return 0;
6542 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6543 char __user *optval,
6544 int __user *optlen)
6546 int val = 0;
6548 if (len < sizeof(int))
6549 return -EINVAL;
6551 len = sizeof(int);
6552 if (sctp_sk(sk)->recvnxtinfo)
6553 val = 1;
6554 if (put_user(len, optlen))
6555 return -EFAULT;
6556 if (copy_to_user(optval, &val, len))
6557 return -EFAULT;
6559 return 0;
6562 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6563 char __user *optval,
6564 int __user *optlen)
6566 struct sctp_assoc_value params;
6567 struct sctp_association *asoc;
6568 int retval = -EFAULT;
6570 if (len < sizeof(params)) {
6571 retval = -EINVAL;
6572 goto out;
6575 len = sizeof(params);
6576 if (copy_from_user(&params, optval, len))
6577 goto out;
6579 asoc = sctp_id2assoc(sk, params.assoc_id);
6580 if (asoc) {
6581 params.assoc_value = asoc->prsctp_enable;
6582 } else if (!params.assoc_id) {
6583 struct sctp_sock *sp = sctp_sk(sk);
6585 params.assoc_value = sp->ep->prsctp_enable;
6586 } else {
6587 retval = -EINVAL;
6588 goto out;
6591 if (put_user(len, optlen))
6592 goto out;
6594 if (copy_to_user(optval, &params, len))
6595 goto out;
6597 retval = 0;
6599 out:
6600 return retval;
6603 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
6604 char __user *optval,
6605 int __user *optlen)
6607 struct sctp_default_prinfo info;
6608 struct sctp_association *asoc;
6609 int retval = -EFAULT;
6611 if (len < sizeof(info)) {
6612 retval = -EINVAL;
6613 goto out;
6616 len = sizeof(info);
6617 if (copy_from_user(&info, optval, len))
6618 goto out;
6620 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
6621 if (asoc) {
6622 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
6623 info.pr_value = asoc->default_timetolive;
6624 } else if (!info.pr_assoc_id) {
6625 struct sctp_sock *sp = sctp_sk(sk);
6627 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
6628 info.pr_value = sp->default_timetolive;
6629 } else {
6630 retval = -EINVAL;
6631 goto out;
6634 if (put_user(len, optlen))
6635 goto out;
6637 if (copy_to_user(optval, &info, len))
6638 goto out;
6640 retval = 0;
6642 out:
6643 return retval;
6646 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
6647 char __user *optval,
6648 int __user *optlen)
6650 struct sctp_prstatus params;
6651 struct sctp_association *asoc;
6652 int policy;
6653 int retval = -EINVAL;
6655 if (len < sizeof(params))
6656 goto out;
6658 len = sizeof(params);
6659 if (copy_from_user(&params, optval, len)) {
6660 retval = -EFAULT;
6661 goto out;
6664 policy = params.sprstat_policy;
6665 if (policy & ~SCTP_PR_SCTP_MASK)
6666 goto out;
6668 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6669 if (!asoc)
6670 goto out;
6672 if (policy == SCTP_PR_SCTP_NONE) {
6673 params.sprstat_abandoned_unsent = 0;
6674 params.sprstat_abandoned_sent = 0;
6675 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6676 params.sprstat_abandoned_unsent +=
6677 asoc->abandoned_unsent[policy];
6678 params.sprstat_abandoned_sent +=
6679 asoc->abandoned_sent[policy];
6681 } else {
6682 params.sprstat_abandoned_unsent =
6683 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6684 params.sprstat_abandoned_sent =
6685 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
6688 if (put_user(len, optlen)) {
6689 retval = -EFAULT;
6690 goto out;
6693 if (copy_to_user(optval, &params, len)) {
6694 retval = -EFAULT;
6695 goto out;
6698 retval = 0;
6700 out:
6701 return retval;
6704 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
6705 char __user *optval,
6706 int __user *optlen)
6708 struct sctp_stream_out *streamout;
6709 struct sctp_association *asoc;
6710 struct sctp_prstatus params;
6711 int retval = -EINVAL;
6712 int policy;
6714 if (len < sizeof(params))
6715 goto out;
6717 len = sizeof(params);
6718 if (copy_from_user(&params, optval, len)) {
6719 retval = -EFAULT;
6720 goto out;
6723 policy = params.sprstat_policy;
6724 if (policy & ~SCTP_PR_SCTP_MASK)
6725 goto out;
6727 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
6728 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
6729 goto out;
6731 streamout = &asoc->stream.out[params.sprstat_sid];
6732 if (policy == SCTP_PR_SCTP_NONE) {
6733 params.sprstat_abandoned_unsent = 0;
6734 params.sprstat_abandoned_sent = 0;
6735 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
6736 params.sprstat_abandoned_unsent +=
6737 streamout->abandoned_unsent[policy];
6738 params.sprstat_abandoned_sent +=
6739 streamout->abandoned_sent[policy];
6741 } else {
6742 params.sprstat_abandoned_unsent =
6743 streamout->abandoned_unsent[__SCTP_PR_INDEX(policy)];
6744 params.sprstat_abandoned_sent =
6745 streamout->abandoned_sent[__SCTP_PR_INDEX(policy)];
6748 if (put_user(len, optlen) || copy_to_user(optval, &params, len)) {
6749 retval = -EFAULT;
6750 goto out;
6753 retval = 0;
6755 out:
6756 return retval;
6759 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
6760 char __user *optval,
6761 int __user *optlen)
6763 struct sctp_assoc_value params;
6764 struct sctp_association *asoc;
6765 int retval = -EFAULT;
6767 if (len < sizeof(params)) {
6768 retval = -EINVAL;
6769 goto out;
6772 len = sizeof(params);
6773 if (copy_from_user(&params, optval, len))
6774 goto out;
6776 asoc = sctp_id2assoc(sk, params.assoc_id);
6777 if (asoc) {
6778 params.assoc_value = asoc->reconf_enable;
6779 } else if (!params.assoc_id) {
6780 struct sctp_sock *sp = sctp_sk(sk);
6782 params.assoc_value = sp->ep->reconf_enable;
6783 } else {
6784 retval = -EINVAL;
6785 goto out;
6788 if (put_user(len, optlen))
6789 goto out;
6791 if (copy_to_user(optval, &params, len))
6792 goto out;
6794 retval = 0;
6796 out:
6797 return retval;
6800 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
6801 char __user *optval,
6802 int __user *optlen)
6804 struct sctp_assoc_value params;
6805 struct sctp_association *asoc;
6806 int retval = -EFAULT;
6808 if (len < sizeof(params)) {
6809 retval = -EINVAL;
6810 goto out;
6813 len = sizeof(params);
6814 if (copy_from_user(&params, optval, len))
6815 goto out;
6817 asoc = sctp_id2assoc(sk, params.assoc_id);
6818 if (asoc) {
6819 params.assoc_value = asoc->strreset_enable;
6820 } else if (!params.assoc_id) {
6821 struct sctp_sock *sp = sctp_sk(sk);
6823 params.assoc_value = sp->ep->strreset_enable;
6824 } else {
6825 retval = -EINVAL;
6826 goto out;
6829 if (put_user(len, optlen))
6830 goto out;
6832 if (copy_to_user(optval, &params, len))
6833 goto out;
6835 retval = 0;
6837 out:
6838 return retval;
6841 static int sctp_getsockopt(struct sock *sk, int level, int optname,
6842 char __user *optval, int __user *optlen)
6844 int retval = 0;
6845 int len;
6847 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
6849 /* I can hardly begin to describe how wrong this is. This is
6850 * so broken as to be worse than useless. The API draft
6851 * REALLY is NOT helpful here... I am not convinced that the
6852 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
6853 * are at all well-founded.
6855 if (level != SOL_SCTP) {
6856 struct sctp_af *af = sctp_sk(sk)->pf->af;
6858 retval = af->getsockopt(sk, level, optname, optval, optlen);
6859 return retval;
6862 if (get_user(len, optlen))
6863 return -EFAULT;
6865 if (len < 0)
6866 return -EINVAL;
6868 lock_sock(sk);
6870 switch (optname) {
6871 case SCTP_STATUS:
6872 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
6873 break;
6874 case SCTP_DISABLE_FRAGMENTS:
6875 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
6876 optlen);
6877 break;
6878 case SCTP_EVENTS:
6879 retval = sctp_getsockopt_events(sk, len, optval, optlen);
6880 break;
6881 case SCTP_AUTOCLOSE:
6882 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
6883 break;
6884 case SCTP_SOCKOPT_PEELOFF:
6885 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
6886 break;
6887 case SCTP_SOCKOPT_PEELOFF_FLAGS:
6888 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
6889 break;
6890 case SCTP_PEER_ADDR_PARAMS:
6891 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
6892 optlen);
6893 break;
6894 case SCTP_DELAYED_SACK:
6895 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
6896 optlen);
6897 break;
6898 case SCTP_INITMSG:
6899 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
6900 break;
6901 case SCTP_GET_PEER_ADDRS:
6902 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
6903 optlen);
6904 break;
6905 case SCTP_GET_LOCAL_ADDRS:
6906 retval = sctp_getsockopt_local_addrs(sk, len, optval,
6907 optlen);
6908 break;
6909 case SCTP_SOCKOPT_CONNECTX3:
6910 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
6911 break;
6912 case SCTP_DEFAULT_SEND_PARAM:
6913 retval = sctp_getsockopt_default_send_param(sk, len,
6914 optval, optlen);
6915 break;
6916 case SCTP_DEFAULT_SNDINFO:
6917 retval = sctp_getsockopt_default_sndinfo(sk, len,
6918 optval, optlen);
6919 break;
6920 case SCTP_PRIMARY_ADDR:
6921 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
6922 break;
6923 case SCTP_NODELAY:
6924 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
6925 break;
6926 case SCTP_RTOINFO:
6927 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
6928 break;
6929 case SCTP_ASSOCINFO:
6930 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
6931 break;
6932 case SCTP_I_WANT_MAPPED_V4_ADDR:
6933 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
6934 break;
6935 case SCTP_MAXSEG:
6936 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
6937 break;
6938 case SCTP_GET_PEER_ADDR_INFO:
6939 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
6940 optlen);
6941 break;
6942 case SCTP_ADAPTATION_LAYER:
6943 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
6944 optlen);
6945 break;
6946 case SCTP_CONTEXT:
6947 retval = sctp_getsockopt_context(sk, len, optval, optlen);
6948 break;
6949 case SCTP_FRAGMENT_INTERLEAVE:
6950 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
6951 optlen);
6952 break;
6953 case SCTP_PARTIAL_DELIVERY_POINT:
6954 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
6955 optlen);
6956 break;
6957 case SCTP_MAX_BURST:
6958 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
6959 break;
6960 case SCTP_AUTH_KEY:
6961 case SCTP_AUTH_CHUNK:
6962 case SCTP_AUTH_DELETE_KEY:
6963 retval = -EOPNOTSUPP;
6964 break;
6965 case SCTP_HMAC_IDENT:
6966 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
6967 break;
6968 case SCTP_AUTH_ACTIVE_KEY:
6969 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
6970 break;
6971 case SCTP_PEER_AUTH_CHUNKS:
6972 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
6973 optlen);
6974 break;
6975 case SCTP_LOCAL_AUTH_CHUNKS:
6976 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
6977 optlen);
6978 break;
6979 case SCTP_GET_ASSOC_NUMBER:
6980 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
6981 break;
6982 case SCTP_GET_ASSOC_ID_LIST:
6983 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
6984 break;
6985 case SCTP_AUTO_ASCONF:
6986 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
6987 break;
6988 case SCTP_PEER_ADDR_THLDS:
6989 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
6990 break;
6991 case SCTP_GET_ASSOC_STATS:
6992 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
6993 break;
6994 case SCTP_RECVRCVINFO:
6995 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
6996 break;
6997 case SCTP_RECVNXTINFO:
6998 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
6999 break;
7000 case SCTP_PR_SUPPORTED:
7001 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7002 break;
7003 case SCTP_DEFAULT_PRINFO:
7004 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7005 optlen);
7006 break;
7007 case SCTP_PR_ASSOC_STATUS:
7008 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7009 optlen);
7010 break;
7011 case SCTP_PR_STREAM_STATUS:
7012 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7013 optlen);
7014 break;
7015 case SCTP_RECONFIG_SUPPORTED:
7016 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7017 optlen);
7018 break;
7019 case SCTP_ENABLE_STREAM_RESET:
7020 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7021 optlen);
7022 break;
7023 default:
7024 retval = -ENOPROTOOPT;
7025 break;
7028 release_sock(sk);
7029 return retval;
7032 static int sctp_hash(struct sock *sk)
7034 /* STUB */
7035 return 0;
7038 static void sctp_unhash(struct sock *sk)
7040 /* STUB */
7043 /* Check if port is acceptable. Possibly find first available port.
7045 * The port hash table (contained in the 'global' SCTP protocol storage
7046 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7047 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7048 * list (the list number is the port number hashed out, so as you
7049 * would expect from a hash function, all the ports in a given list have
7050 * such a number that hashes out to the same list number; you were
7051 * expecting that, right?); so each list has a set of ports, with a
7052 * link to the socket (struct sock) that uses it, the port number and
7053 * a fastreuse flag (FIXME: NPI ipg).
7055 static struct sctp_bind_bucket *sctp_bucket_create(
7056 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7058 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7060 struct sctp_bind_hashbucket *head; /* hash list */
7061 struct sctp_bind_bucket *pp;
7062 unsigned short snum;
7063 int ret;
7065 snum = ntohs(addr->v4.sin_port);
7067 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7069 local_bh_disable();
7071 if (snum == 0) {
7072 /* Search for an available port. */
7073 int low, high, remaining, index;
7074 unsigned int rover;
7075 struct net *net = sock_net(sk);
7077 inet_get_local_port_range(net, &low, &high);
7078 remaining = (high - low) + 1;
7079 rover = prandom_u32() % remaining + low;
7081 do {
7082 rover++;
7083 if ((rover < low) || (rover > high))
7084 rover = low;
7085 if (inet_is_local_reserved_port(net, rover))
7086 continue;
7087 index = sctp_phashfn(sock_net(sk), rover);
7088 head = &sctp_port_hashtable[index];
7089 spin_lock(&head->lock);
7090 sctp_for_each_hentry(pp, &head->chain)
7091 if ((pp->port == rover) &&
7092 net_eq(sock_net(sk), pp->net))
7093 goto next;
7094 break;
7095 next:
7096 spin_unlock(&head->lock);
7097 } while (--remaining > 0);
7099 /* Exhausted local port range during search? */
7100 ret = 1;
7101 if (remaining <= 0)
7102 goto fail;
7104 /* OK, here is the one we will use. HEAD (the port
7105 * hash table list entry) is non-NULL and we hold it's
7106 * mutex.
7108 snum = rover;
7109 } else {
7110 /* We are given an specific port number; we verify
7111 * that it is not being used. If it is used, we will
7112 * exahust the search in the hash list corresponding
7113 * to the port number (snum) - we detect that with the
7114 * port iterator, pp being NULL.
7116 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7117 spin_lock(&head->lock);
7118 sctp_for_each_hentry(pp, &head->chain) {
7119 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7120 goto pp_found;
7123 pp = NULL;
7124 goto pp_not_found;
7125 pp_found:
7126 if (!hlist_empty(&pp->owner)) {
7127 /* We had a port hash table hit - there is an
7128 * available port (pp != NULL) and it is being
7129 * used by other socket (pp->owner not empty); that other
7130 * socket is going to be sk2.
7132 int reuse = sk->sk_reuse;
7133 struct sock *sk2;
7135 pr_debug("%s: found a possible match\n", __func__);
7137 if (pp->fastreuse && sk->sk_reuse &&
7138 sk->sk_state != SCTP_SS_LISTENING)
7139 goto success;
7141 /* Run through the list of sockets bound to the port
7142 * (pp->port) [via the pointers bind_next and
7143 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7144 * we get the endpoint they describe and run through
7145 * the endpoint's list of IP (v4 or v6) addresses,
7146 * comparing each of the addresses with the address of
7147 * the socket sk. If we find a match, then that means
7148 * that this port/socket (sk) combination are already
7149 * in an endpoint.
7151 sk_for_each_bound(sk2, &pp->owner) {
7152 struct sctp_endpoint *ep2;
7153 ep2 = sctp_sk(sk2)->ep;
7155 if (sk == sk2 ||
7156 (reuse && sk2->sk_reuse &&
7157 sk2->sk_state != SCTP_SS_LISTENING))
7158 continue;
7160 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7161 sctp_sk(sk2), sctp_sk(sk))) {
7162 ret = (long)sk2;
7163 goto fail_unlock;
7167 pr_debug("%s: found a match\n", __func__);
7169 pp_not_found:
7170 /* If there was a hash table miss, create a new port. */
7171 ret = 1;
7172 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7173 goto fail_unlock;
7175 /* In either case (hit or miss), make sure fastreuse is 1 only
7176 * if sk->sk_reuse is too (that is, if the caller requested
7177 * SO_REUSEADDR on this socket -sk-).
7179 if (hlist_empty(&pp->owner)) {
7180 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
7181 pp->fastreuse = 1;
7182 else
7183 pp->fastreuse = 0;
7184 } else if (pp->fastreuse &&
7185 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
7186 pp->fastreuse = 0;
7188 /* We are set, so fill up all the data in the hash table
7189 * entry, tie the socket list information with the rest of the
7190 * sockets FIXME: Blurry, NPI (ipg).
7192 success:
7193 if (!sctp_sk(sk)->bind_hash) {
7194 inet_sk(sk)->inet_num = snum;
7195 sk_add_bind_node(sk, &pp->owner);
7196 sctp_sk(sk)->bind_hash = pp;
7198 ret = 0;
7200 fail_unlock:
7201 spin_unlock(&head->lock);
7203 fail:
7204 local_bh_enable();
7205 return ret;
7208 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7209 * port is requested.
7211 static int sctp_get_port(struct sock *sk, unsigned short snum)
7213 union sctp_addr addr;
7214 struct sctp_af *af = sctp_sk(sk)->pf->af;
7216 /* Set up a dummy address struct from the sk. */
7217 af->from_sk(&addr, sk);
7218 addr.v4.sin_port = htons(snum);
7220 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7221 return !!sctp_get_port_local(sk, &addr);
7225 * Move a socket to LISTENING state.
7227 static int sctp_listen_start(struct sock *sk, int backlog)
7229 struct sctp_sock *sp = sctp_sk(sk);
7230 struct sctp_endpoint *ep = sp->ep;
7231 struct crypto_shash *tfm = NULL;
7232 char alg[32];
7234 /* Allocate HMAC for generating cookie. */
7235 if (!sp->hmac && sp->sctp_hmac_alg) {
7236 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7237 tfm = crypto_alloc_shash(alg, 0, 0);
7238 if (IS_ERR(tfm)) {
7239 net_info_ratelimited("failed to load transform for %s: %ld\n",
7240 sp->sctp_hmac_alg, PTR_ERR(tfm));
7241 return -ENOSYS;
7243 sctp_sk(sk)->hmac = tfm;
7247 * If a bind() or sctp_bindx() is not called prior to a listen()
7248 * call that allows new associations to be accepted, the system
7249 * picks an ephemeral port and will choose an address set equivalent
7250 * to binding with a wildcard address.
7252 * This is not currently spelled out in the SCTP sockets
7253 * extensions draft, but follows the practice as seen in TCP
7254 * sockets.
7257 sk->sk_state = SCTP_SS_LISTENING;
7258 if (!ep->base.bind_addr.port) {
7259 if (sctp_autobind(sk))
7260 return -EAGAIN;
7261 } else {
7262 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7263 sk->sk_state = SCTP_SS_CLOSED;
7264 return -EADDRINUSE;
7268 sk->sk_max_ack_backlog = backlog;
7269 sctp_hash_endpoint(ep);
7270 return 0;
7274 * 4.1.3 / 5.1.3 listen()
7276 * By default, new associations are not accepted for UDP style sockets.
7277 * An application uses listen() to mark a socket as being able to
7278 * accept new associations.
7280 * On TCP style sockets, applications use listen() to ready the SCTP
7281 * endpoint for accepting inbound associations.
7283 * On both types of endpoints a backlog of '0' disables listening.
7285 * Move a socket to LISTENING state.
7287 int sctp_inet_listen(struct socket *sock, int backlog)
7289 struct sock *sk = sock->sk;
7290 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7291 int err = -EINVAL;
7293 if (unlikely(backlog < 0))
7294 return err;
7296 lock_sock(sk);
7298 /* Peeled-off sockets are not allowed to listen(). */
7299 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7300 goto out;
7302 if (sock->state != SS_UNCONNECTED)
7303 goto out;
7305 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7306 goto out;
7308 /* If backlog is zero, disable listening. */
7309 if (!backlog) {
7310 if (sctp_sstate(sk, CLOSED))
7311 goto out;
7313 err = 0;
7314 sctp_unhash_endpoint(ep);
7315 sk->sk_state = SCTP_SS_CLOSED;
7316 if (sk->sk_reuse)
7317 sctp_sk(sk)->bind_hash->fastreuse = 1;
7318 goto out;
7321 /* If we are already listening, just update the backlog */
7322 if (sctp_sstate(sk, LISTENING))
7323 sk->sk_max_ack_backlog = backlog;
7324 else {
7325 err = sctp_listen_start(sk, backlog);
7326 if (err)
7327 goto out;
7330 err = 0;
7331 out:
7332 release_sock(sk);
7333 return err;
7337 * This function is done by modeling the current datagram_poll() and the
7338 * tcp_poll(). Note that, based on these implementations, we don't
7339 * lock the socket in this function, even though it seems that,
7340 * ideally, locking or some other mechanisms can be used to ensure
7341 * the integrity of the counters (sndbuf and wmem_alloc) used
7342 * in this place. We assume that we don't need locks either until proven
7343 * otherwise.
7345 * Another thing to note is that we include the Async I/O support
7346 * here, again, by modeling the current TCP/UDP code. We don't have
7347 * a good way to test with it yet.
7349 unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7351 struct sock *sk = sock->sk;
7352 struct sctp_sock *sp = sctp_sk(sk);
7353 unsigned int mask;
7355 poll_wait(file, sk_sleep(sk), wait);
7357 sock_rps_record_flow(sk);
7359 /* A TCP-style listening socket becomes readable when the accept queue
7360 * is not empty.
7362 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7363 return (!list_empty(&sp->ep->asocs)) ?
7364 (POLLIN | POLLRDNORM) : 0;
7366 mask = 0;
7368 /* Is there any exceptional events? */
7369 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7370 mask |= POLLERR |
7371 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7372 if (sk->sk_shutdown & RCV_SHUTDOWN)
7373 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
7374 if (sk->sk_shutdown == SHUTDOWN_MASK)
7375 mask |= POLLHUP;
7377 /* Is it readable? Reconsider this code with TCP-style support. */
7378 if (!skb_queue_empty(&sk->sk_receive_queue))
7379 mask |= POLLIN | POLLRDNORM;
7381 /* The association is either gone or not ready. */
7382 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7383 return mask;
7385 /* Is it writable? */
7386 if (sctp_writeable(sk)) {
7387 mask |= POLLOUT | POLLWRNORM;
7388 } else {
7389 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7391 * Since the socket is not locked, the buffer
7392 * might be made available after the writeable check and
7393 * before the bit is set. This could cause a lost I/O
7394 * signal. tcp_poll() has a race breaker for this race
7395 * condition. Based on their implementation, we put
7396 * in the following code to cover it as well.
7398 if (sctp_writeable(sk))
7399 mask |= POLLOUT | POLLWRNORM;
7401 return mask;
7404 /********************************************************************
7405 * 2nd Level Abstractions
7406 ********************************************************************/
7408 static struct sctp_bind_bucket *sctp_bucket_create(
7409 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7411 struct sctp_bind_bucket *pp;
7413 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7414 if (pp) {
7415 SCTP_DBG_OBJCNT_INC(bind_bucket);
7416 pp->port = snum;
7417 pp->fastreuse = 0;
7418 INIT_HLIST_HEAD(&pp->owner);
7419 pp->net = net;
7420 hlist_add_head(&pp->node, &head->chain);
7422 return pp;
7425 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7426 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7428 if (pp && hlist_empty(&pp->owner)) {
7429 __hlist_del(&pp->node);
7430 kmem_cache_free(sctp_bucket_cachep, pp);
7431 SCTP_DBG_OBJCNT_DEC(bind_bucket);
7435 /* Release this socket's reference to a local port. */
7436 static inline void __sctp_put_port(struct sock *sk)
7438 struct sctp_bind_hashbucket *head =
7439 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
7440 inet_sk(sk)->inet_num)];
7441 struct sctp_bind_bucket *pp;
7443 spin_lock(&head->lock);
7444 pp = sctp_sk(sk)->bind_hash;
7445 __sk_del_bind_node(sk);
7446 sctp_sk(sk)->bind_hash = NULL;
7447 inet_sk(sk)->inet_num = 0;
7448 sctp_bucket_destroy(pp);
7449 spin_unlock(&head->lock);
7452 void sctp_put_port(struct sock *sk)
7454 local_bh_disable();
7455 __sctp_put_port(sk);
7456 local_bh_enable();
7460 * The system picks an ephemeral port and choose an address set equivalent
7461 * to binding with a wildcard address.
7462 * One of those addresses will be the primary address for the association.
7463 * This automatically enables the multihoming capability of SCTP.
7465 static int sctp_autobind(struct sock *sk)
7467 union sctp_addr autoaddr;
7468 struct sctp_af *af;
7469 __be16 port;
7471 /* Initialize a local sockaddr structure to INADDR_ANY. */
7472 af = sctp_sk(sk)->pf->af;
7474 port = htons(inet_sk(sk)->inet_num);
7475 af->inaddr_any(&autoaddr, port);
7477 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
7480 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
7482 * From RFC 2292
7483 * 4.2 The cmsghdr Structure *
7485 * When ancillary data is sent or received, any number of ancillary data
7486 * objects can be specified by the msg_control and msg_controllen members of
7487 * the msghdr structure, because each object is preceded by
7488 * a cmsghdr structure defining the object's length (the cmsg_len member).
7489 * Historically Berkeley-derived implementations have passed only one object
7490 * at a time, but this API allows multiple objects to be
7491 * passed in a single call to sendmsg() or recvmsg(). The following example
7492 * shows two ancillary data objects in a control buffer.
7494 * |<--------------------------- msg_controllen -------------------------->|
7495 * | |
7497 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
7499 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
7500 * | | |
7502 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
7504 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
7505 * | | | | |
7507 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7508 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
7510 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
7512 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
7516 * msg_control
7517 * points here
7519 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
7521 struct msghdr *my_msg = (struct msghdr *)msg;
7522 struct cmsghdr *cmsg;
7524 for_each_cmsghdr(cmsg, my_msg) {
7525 if (!CMSG_OK(my_msg, cmsg))
7526 return -EINVAL;
7528 /* Should we parse this header or ignore? */
7529 if (cmsg->cmsg_level != IPPROTO_SCTP)
7530 continue;
7532 /* Strictly check lengths following example in SCM code. */
7533 switch (cmsg->cmsg_type) {
7534 case SCTP_INIT:
7535 /* SCTP Socket API Extension
7536 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
7538 * This cmsghdr structure provides information for
7539 * initializing new SCTP associations with sendmsg().
7540 * The SCTP_INITMSG socket option uses this same data
7541 * structure. This structure is not used for
7542 * recvmsg().
7544 * cmsg_level cmsg_type cmsg_data[]
7545 * ------------ ------------ ----------------------
7546 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
7548 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
7549 return -EINVAL;
7551 cmsgs->init = CMSG_DATA(cmsg);
7552 break;
7554 case SCTP_SNDRCV:
7555 /* SCTP Socket API Extension
7556 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
7558 * This cmsghdr structure specifies SCTP options for
7559 * sendmsg() and describes SCTP header information
7560 * about a received message through recvmsg().
7562 * cmsg_level cmsg_type cmsg_data[]
7563 * ------------ ------------ ----------------------
7564 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
7566 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
7567 return -EINVAL;
7569 cmsgs->srinfo = CMSG_DATA(cmsg);
7571 if (cmsgs->srinfo->sinfo_flags &
7572 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7573 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7574 SCTP_ABORT | SCTP_EOF))
7575 return -EINVAL;
7576 break;
7578 case SCTP_SNDINFO:
7579 /* SCTP Socket API Extension
7580 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
7582 * This cmsghdr structure specifies SCTP options for
7583 * sendmsg(). This structure and SCTP_RCVINFO replaces
7584 * SCTP_SNDRCV which has been deprecated.
7586 * cmsg_level cmsg_type cmsg_data[]
7587 * ------------ ------------ ---------------------
7588 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
7590 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
7591 return -EINVAL;
7593 cmsgs->sinfo = CMSG_DATA(cmsg);
7595 if (cmsgs->sinfo->snd_flags &
7596 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
7597 SCTP_SACK_IMMEDIATELY | SCTP_PR_SCTP_MASK |
7598 SCTP_ABORT | SCTP_EOF))
7599 return -EINVAL;
7600 break;
7601 default:
7602 return -EINVAL;
7606 return 0;
7610 * Wait for a packet..
7611 * Note: This function is the same function as in core/datagram.c
7612 * with a few modifications to make lksctp work.
7614 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
7616 int error;
7617 DEFINE_WAIT(wait);
7619 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
7621 /* Socket errors? */
7622 error = sock_error(sk);
7623 if (error)
7624 goto out;
7626 if (!skb_queue_empty(&sk->sk_receive_queue))
7627 goto ready;
7629 /* Socket shut down? */
7630 if (sk->sk_shutdown & RCV_SHUTDOWN)
7631 goto out;
7633 /* Sequenced packets can come disconnected. If so we report the
7634 * problem.
7636 error = -ENOTCONN;
7638 /* Is there a good reason to think that we may receive some data? */
7639 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
7640 goto out;
7642 /* Handle signals. */
7643 if (signal_pending(current))
7644 goto interrupted;
7646 /* Let another process have a go. Since we are going to sleep
7647 * anyway. Note: This may cause odd behaviors if the message
7648 * does not fit in the user's buffer, but this seems to be the
7649 * only way to honor MSG_DONTWAIT realistically.
7651 release_sock(sk);
7652 *timeo_p = schedule_timeout(*timeo_p);
7653 lock_sock(sk);
7655 ready:
7656 finish_wait(sk_sleep(sk), &wait);
7657 return 0;
7659 interrupted:
7660 error = sock_intr_errno(*timeo_p);
7662 out:
7663 finish_wait(sk_sleep(sk), &wait);
7664 *err = error;
7665 return error;
7668 /* Receive a datagram.
7669 * Note: This is pretty much the same routine as in core/datagram.c
7670 * with a few changes to make lksctp work.
7672 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
7673 int noblock, int *err)
7675 int error;
7676 struct sk_buff *skb;
7677 long timeo;
7679 timeo = sock_rcvtimeo(sk, noblock);
7681 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
7682 MAX_SCHEDULE_TIMEOUT);
7684 do {
7685 /* Again only user level code calls this function,
7686 * so nothing interrupt level
7687 * will suddenly eat the receive_queue.
7689 * Look at current nfs client by the way...
7690 * However, this function was correct in any case. 8)
7692 if (flags & MSG_PEEK) {
7693 skb = skb_peek(&sk->sk_receive_queue);
7694 if (skb)
7695 refcount_inc(&skb->users);
7696 } else {
7697 skb = __skb_dequeue(&sk->sk_receive_queue);
7700 if (skb)
7701 return skb;
7703 /* Caller is allowed not to check sk->sk_err before calling. */
7704 error = sock_error(sk);
7705 if (error)
7706 goto no_packet;
7708 if (sk->sk_shutdown & RCV_SHUTDOWN)
7709 break;
7711 if (sk_can_busy_loop(sk)) {
7712 sk_busy_loop(sk, noblock);
7714 if (!skb_queue_empty(&sk->sk_receive_queue))
7715 continue;
7718 /* User doesn't want to wait. */
7719 error = -EAGAIN;
7720 if (!timeo)
7721 goto no_packet;
7722 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
7724 return NULL;
7726 no_packet:
7727 *err = error;
7728 return NULL;
7731 /* If sndbuf has changed, wake up per association sndbuf waiters. */
7732 static void __sctp_write_space(struct sctp_association *asoc)
7734 struct sock *sk = asoc->base.sk;
7736 if (sctp_wspace(asoc) <= 0)
7737 return;
7739 if (waitqueue_active(&asoc->wait))
7740 wake_up_interruptible(&asoc->wait);
7742 if (sctp_writeable(sk)) {
7743 struct socket_wq *wq;
7745 rcu_read_lock();
7746 wq = rcu_dereference(sk->sk_wq);
7747 if (wq) {
7748 if (waitqueue_active(&wq->wait))
7749 wake_up_interruptible(&wq->wait);
7751 /* Note that we try to include the Async I/O support
7752 * here by modeling from the current TCP/UDP code.
7753 * We have not tested with it yet.
7755 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
7756 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
7758 rcu_read_unlock();
7762 static void sctp_wake_up_waiters(struct sock *sk,
7763 struct sctp_association *asoc)
7765 struct sctp_association *tmp = asoc;
7767 /* We do accounting for the sndbuf space per association,
7768 * so we only need to wake our own association.
7770 if (asoc->ep->sndbuf_policy)
7771 return __sctp_write_space(asoc);
7773 /* If association goes down and is just flushing its
7774 * outq, then just normally notify others.
7776 if (asoc->base.dead)
7777 return sctp_write_space(sk);
7779 /* Accounting for the sndbuf space is per socket, so we
7780 * need to wake up others, try to be fair and in case of
7781 * other associations, let them have a go first instead
7782 * of just doing a sctp_write_space() call.
7784 * Note that we reach sctp_wake_up_waiters() only when
7785 * associations free up queued chunks, thus we are under
7786 * lock and the list of associations on a socket is
7787 * guaranteed not to change.
7789 for (tmp = list_next_entry(tmp, asocs); 1;
7790 tmp = list_next_entry(tmp, asocs)) {
7791 /* Manually skip the head element. */
7792 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
7793 continue;
7794 /* Wake up association. */
7795 __sctp_write_space(tmp);
7796 /* We've reached the end. */
7797 if (tmp == asoc)
7798 break;
7802 /* Do accounting for the sndbuf space.
7803 * Decrement the used sndbuf space of the corresponding association by the
7804 * data size which was just transmitted(freed).
7806 static void sctp_wfree(struct sk_buff *skb)
7808 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
7809 struct sctp_association *asoc = chunk->asoc;
7810 struct sock *sk = asoc->base.sk;
7812 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
7813 sizeof(struct sk_buff) +
7814 sizeof(struct sctp_chunk);
7816 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
7819 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
7821 sk->sk_wmem_queued -= skb->truesize;
7822 sk_mem_uncharge(sk, skb->truesize);
7824 sock_wfree(skb);
7825 sctp_wake_up_waiters(sk, asoc);
7827 sctp_association_put(asoc);
7830 /* Do accounting for the receive space on the socket.
7831 * Accounting for the association is done in ulpevent.c
7832 * We set this as a destructor for the cloned data skbs so that
7833 * accounting is done at the correct time.
7835 void sctp_sock_rfree(struct sk_buff *skb)
7837 struct sock *sk = skb->sk;
7838 struct sctp_ulpevent *event = sctp_skb2event(skb);
7840 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
7843 * Mimic the behavior of sock_rfree
7845 sk_mem_uncharge(sk, event->rmem_len);
7849 /* Helper function to wait for space in the sndbuf. */
7850 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
7851 size_t msg_len)
7853 struct sock *sk = asoc->base.sk;
7854 long current_timeo = *timeo_p;
7855 DEFINE_WAIT(wait);
7856 int err = 0;
7858 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
7859 *timeo_p, msg_len);
7861 /* Increment the association's refcnt. */
7862 sctp_association_hold(asoc);
7864 /* Wait on the association specific sndbuf space. */
7865 for (;;) {
7866 prepare_to_wait_exclusive(&asoc->wait, &wait,
7867 TASK_INTERRUPTIBLE);
7868 if (asoc->base.dead)
7869 goto do_dead;
7870 if (!*timeo_p)
7871 goto do_nonblock;
7872 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
7873 goto do_error;
7874 if (signal_pending(current))
7875 goto do_interrupted;
7876 if (msg_len <= sctp_wspace(asoc))
7877 break;
7879 /* Let another process have a go. Since we are going
7880 * to sleep anyway.
7882 release_sock(sk);
7883 current_timeo = schedule_timeout(current_timeo);
7884 lock_sock(sk);
7885 if (sk != asoc->base.sk)
7886 goto do_error;
7888 *timeo_p = current_timeo;
7891 out:
7892 finish_wait(&asoc->wait, &wait);
7894 /* Release the association's refcnt. */
7895 sctp_association_put(asoc);
7897 return err;
7899 do_dead:
7900 err = -ESRCH;
7901 goto out;
7903 do_error:
7904 err = -EPIPE;
7905 goto out;
7907 do_interrupted:
7908 err = sock_intr_errno(*timeo_p);
7909 goto out;
7911 do_nonblock:
7912 err = -EAGAIN;
7913 goto out;
7916 void sctp_data_ready(struct sock *sk)
7918 struct socket_wq *wq;
7920 rcu_read_lock();
7921 wq = rcu_dereference(sk->sk_wq);
7922 if (skwq_has_sleeper(wq))
7923 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
7924 POLLRDNORM | POLLRDBAND);
7925 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
7926 rcu_read_unlock();
7929 /* If socket sndbuf has changed, wake up all per association waiters. */
7930 void sctp_write_space(struct sock *sk)
7932 struct sctp_association *asoc;
7934 /* Wake up the tasks in each wait queue. */
7935 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
7936 __sctp_write_space(asoc);
7940 /* Is there any sndbuf space available on the socket?
7942 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
7943 * associations on the same socket. For a UDP-style socket with
7944 * multiple associations, it is possible for it to be "unwriteable"
7945 * prematurely. I assume that this is acceptable because
7946 * a premature "unwriteable" is better than an accidental "writeable" which
7947 * would cause an unwanted block under certain circumstances. For the 1-1
7948 * UDP-style sockets or TCP-style sockets, this code should work.
7949 * - Daisy
7951 static int sctp_writeable(struct sock *sk)
7953 int amt = 0;
7955 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
7956 if (amt < 0)
7957 amt = 0;
7958 return amt;
7961 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
7962 * returns immediately with EINPROGRESS.
7964 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
7966 struct sock *sk = asoc->base.sk;
7967 int err = 0;
7968 long current_timeo = *timeo_p;
7969 DEFINE_WAIT(wait);
7971 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
7973 /* Increment the association's refcnt. */
7974 sctp_association_hold(asoc);
7976 for (;;) {
7977 prepare_to_wait_exclusive(&asoc->wait, &wait,
7978 TASK_INTERRUPTIBLE);
7979 if (!*timeo_p)
7980 goto do_nonblock;
7981 if (sk->sk_shutdown & RCV_SHUTDOWN)
7982 break;
7983 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
7984 asoc->base.dead)
7985 goto do_error;
7986 if (signal_pending(current))
7987 goto do_interrupted;
7989 if (sctp_state(asoc, ESTABLISHED))
7990 break;
7992 /* Let another process have a go. Since we are going
7993 * to sleep anyway.
7995 release_sock(sk);
7996 current_timeo = schedule_timeout(current_timeo);
7997 lock_sock(sk);
7999 *timeo_p = current_timeo;
8002 out:
8003 finish_wait(&asoc->wait, &wait);
8005 /* Release the association's refcnt. */
8006 sctp_association_put(asoc);
8008 return err;
8010 do_error:
8011 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8012 err = -ETIMEDOUT;
8013 else
8014 err = -ECONNREFUSED;
8015 goto out;
8017 do_interrupted:
8018 err = sock_intr_errno(*timeo_p);
8019 goto out;
8021 do_nonblock:
8022 err = -EINPROGRESS;
8023 goto out;
8026 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8028 struct sctp_endpoint *ep;
8029 int err = 0;
8030 DEFINE_WAIT(wait);
8032 ep = sctp_sk(sk)->ep;
8035 for (;;) {
8036 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8037 TASK_INTERRUPTIBLE);
8039 if (list_empty(&ep->asocs)) {
8040 release_sock(sk);
8041 timeo = schedule_timeout(timeo);
8042 lock_sock(sk);
8045 err = -EINVAL;
8046 if (!sctp_sstate(sk, LISTENING))
8047 break;
8049 err = 0;
8050 if (!list_empty(&ep->asocs))
8051 break;
8053 err = sock_intr_errno(timeo);
8054 if (signal_pending(current))
8055 break;
8057 err = -EAGAIN;
8058 if (!timeo)
8059 break;
8062 finish_wait(sk_sleep(sk), &wait);
8064 return err;
8067 static void sctp_wait_for_close(struct sock *sk, long timeout)
8069 DEFINE_WAIT(wait);
8071 do {
8072 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8073 if (list_empty(&sctp_sk(sk)->ep->asocs))
8074 break;
8075 release_sock(sk);
8076 timeout = schedule_timeout(timeout);
8077 lock_sock(sk);
8078 } while (!signal_pending(current) && timeout);
8080 finish_wait(sk_sleep(sk), &wait);
8083 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8085 struct sk_buff *frag;
8087 if (!skb->data_len)
8088 goto done;
8090 /* Don't forget the fragments. */
8091 skb_walk_frags(skb, frag)
8092 sctp_skb_set_owner_r_frag(frag, sk);
8094 done:
8095 sctp_skb_set_owner_r(skb, sk);
8098 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8099 struct sctp_association *asoc)
8101 struct inet_sock *inet = inet_sk(sk);
8102 struct inet_sock *newinet;
8104 newsk->sk_type = sk->sk_type;
8105 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8106 newsk->sk_flags = sk->sk_flags;
8107 newsk->sk_tsflags = sk->sk_tsflags;
8108 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8109 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8110 newsk->sk_reuse = sk->sk_reuse;
8112 newsk->sk_shutdown = sk->sk_shutdown;
8113 newsk->sk_destruct = sctp_destruct_sock;
8114 newsk->sk_family = sk->sk_family;
8115 newsk->sk_protocol = IPPROTO_SCTP;
8116 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8117 newsk->sk_sndbuf = sk->sk_sndbuf;
8118 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8119 newsk->sk_lingertime = sk->sk_lingertime;
8120 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8121 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8122 newsk->sk_rxhash = sk->sk_rxhash;
8124 newinet = inet_sk(newsk);
8126 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8127 * getsockname() and getpeername()
8129 newinet->inet_sport = inet->inet_sport;
8130 newinet->inet_saddr = inet->inet_saddr;
8131 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8132 newinet->inet_dport = htons(asoc->peer.port);
8133 newinet->pmtudisc = inet->pmtudisc;
8134 newinet->inet_id = asoc->next_tsn ^ jiffies;
8136 newinet->uc_ttl = inet->uc_ttl;
8137 newinet->mc_loop = 1;
8138 newinet->mc_ttl = 1;
8139 newinet->mc_index = 0;
8140 newinet->mc_list = NULL;
8142 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8143 net_enable_timestamp();
8145 security_sk_clone(sk, newsk);
8148 static inline void sctp_copy_descendant(struct sock *sk_to,
8149 const struct sock *sk_from)
8151 int ancestor_size = sizeof(struct inet_sock) +
8152 sizeof(struct sctp_sock) -
8153 offsetof(struct sctp_sock, auto_asconf_list);
8155 if (sk_from->sk_family == PF_INET6)
8156 ancestor_size += sizeof(struct ipv6_pinfo);
8158 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8161 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8162 * and its messages to the newsk.
8164 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8165 struct sctp_association *assoc,
8166 enum sctp_socket_type type)
8168 struct sctp_sock *oldsp = sctp_sk(oldsk);
8169 struct sctp_sock *newsp = sctp_sk(newsk);
8170 struct sctp_bind_bucket *pp; /* hash list port iterator */
8171 struct sctp_endpoint *newep = newsp->ep;
8172 struct sk_buff *skb, *tmp;
8173 struct sctp_ulpevent *event;
8174 struct sctp_bind_hashbucket *head;
8176 /* Migrate socket buffer sizes and all the socket level options to the
8177 * new socket.
8179 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8180 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8181 /* Brute force copy old sctp opt. */
8182 sctp_copy_descendant(newsk, oldsk);
8184 /* Restore the ep value that was overwritten with the above structure
8185 * copy.
8187 newsp->ep = newep;
8188 newsp->hmac = NULL;
8190 /* Hook this new socket in to the bind_hash list. */
8191 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8192 inet_sk(oldsk)->inet_num)];
8193 spin_lock_bh(&head->lock);
8194 pp = sctp_sk(oldsk)->bind_hash;
8195 sk_add_bind_node(newsk, &pp->owner);
8196 sctp_sk(newsk)->bind_hash = pp;
8197 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8198 spin_unlock_bh(&head->lock);
8200 /* Copy the bind_addr list from the original endpoint to the new
8201 * endpoint so that we can handle restarts properly
8203 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8204 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8206 /* Move any messages in the old socket's receive queue that are for the
8207 * peeled off association to the new socket's receive queue.
8209 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8210 event = sctp_skb2event(skb);
8211 if (event->asoc == assoc) {
8212 __skb_unlink(skb, &oldsk->sk_receive_queue);
8213 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8214 sctp_skb_set_owner_r_frag(skb, newsk);
8218 /* Clean up any messages pending delivery due to partial
8219 * delivery. Three cases:
8220 * 1) No partial deliver; no work.
8221 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8222 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8224 skb_queue_head_init(&newsp->pd_lobby);
8225 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8227 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8228 struct sk_buff_head *queue;
8230 /* Decide which queue to move pd_lobby skbs to. */
8231 if (assoc->ulpq.pd_mode) {
8232 queue = &newsp->pd_lobby;
8233 } else
8234 queue = &newsk->sk_receive_queue;
8236 /* Walk through the pd_lobby, looking for skbs that
8237 * need moved to the new socket.
8239 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8240 event = sctp_skb2event(skb);
8241 if (event->asoc == assoc) {
8242 __skb_unlink(skb, &oldsp->pd_lobby);
8243 __skb_queue_tail(queue, skb);
8244 sctp_skb_set_owner_r_frag(skb, newsk);
8248 /* Clear up any skbs waiting for the partial
8249 * delivery to finish.
8251 if (assoc->ulpq.pd_mode)
8252 sctp_clear_pd(oldsk, NULL);
8256 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
8257 sctp_skb_set_owner_r_frag(skb, newsk);
8259 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
8260 sctp_skb_set_owner_r_frag(skb, newsk);
8262 /* Set the type of socket to indicate that it is peeled off from the
8263 * original UDP-style socket or created with the accept() call on a
8264 * TCP-style socket..
8266 newsp->type = type;
8268 /* Mark the new socket "in-use" by the user so that any packets
8269 * that may arrive on the association after we've moved it are
8270 * queued to the backlog. This prevents a potential race between
8271 * backlog processing on the old socket and new-packet processing
8272 * on the new socket.
8274 * The caller has just allocated newsk so we can guarantee that other
8275 * paths won't try to lock it and then oldsk.
8277 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8278 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8279 sctp_assoc_migrate(assoc, newsk);
8280 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8282 /* If the association on the newsk is already closed before accept()
8283 * is called, set RCV_SHUTDOWN flag.
8285 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8286 newsk->sk_state = SCTP_SS_CLOSED;
8287 newsk->sk_shutdown |= RCV_SHUTDOWN;
8288 } else {
8289 newsk->sk_state = SCTP_SS_ESTABLISHED;
8292 release_sock(newsk);
8296 /* This proto struct describes the ULP interface for SCTP. */
8297 struct proto sctp_prot = {
8298 .name = "SCTP",
8299 .owner = THIS_MODULE,
8300 .close = sctp_close,
8301 .connect = sctp_connect,
8302 .disconnect = sctp_disconnect,
8303 .accept = sctp_accept,
8304 .ioctl = sctp_ioctl,
8305 .init = sctp_init_sock,
8306 .destroy = sctp_destroy_sock,
8307 .shutdown = sctp_shutdown,
8308 .setsockopt = sctp_setsockopt,
8309 .getsockopt = sctp_getsockopt,
8310 .sendmsg = sctp_sendmsg,
8311 .recvmsg = sctp_recvmsg,
8312 .bind = sctp_bind,
8313 .backlog_rcv = sctp_backlog_rcv,
8314 .hash = sctp_hash,
8315 .unhash = sctp_unhash,
8316 .get_port = sctp_get_port,
8317 .obj_size = sizeof(struct sctp_sock),
8318 .sysctl_mem = sysctl_sctp_mem,
8319 .sysctl_rmem = sysctl_sctp_rmem,
8320 .sysctl_wmem = sysctl_sctp_wmem,
8321 .memory_pressure = &sctp_memory_pressure,
8322 .enter_memory_pressure = sctp_enter_memory_pressure,
8323 .memory_allocated = &sctp_memory_allocated,
8324 .sockets_allocated = &sctp_sockets_allocated,
8327 #if IS_ENABLED(CONFIG_IPV6)
8329 #include <net/transp_v6.h>
8330 static void sctp_v6_destroy_sock(struct sock *sk)
8332 sctp_destroy_sock(sk);
8333 inet6_destroy_sock(sk);
8336 struct proto sctpv6_prot = {
8337 .name = "SCTPv6",
8338 .owner = THIS_MODULE,
8339 .close = sctp_close,
8340 .connect = sctp_connect,
8341 .disconnect = sctp_disconnect,
8342 .accept = sctp_accept,
8343 .ioctl = sctp_ioctl,
8344 .init = sctp_init_sock,
8345 .destroy = sctp_v6_destroy_sock,
8346 .shutdown = sctp_shutdown,
8347 .setsockopt = sctp_setsockopt,
8348 .getsockopt = sctp_getsockopt,
8349 .sendmsg = sctp_sendmsg,
8350 .recvmsg = sctp_recvmsg,
8351 .bind = sctp_bind,
8352 .backlog_rcv = sctp_backlog_rcv,
8353 .hash = sctp_hash,
8354 .unhash = sctp_unhash,
8355 .get_port = sctp_get_port,
8356 .obj_size = sizeof(struct sctp6_sock),
8357 .sysctl_mem = sysctl_sctp_mem,
8358 .sysctl_rmem = sysctl_sctp_rmem,
8359 .sysctl_wmem = sysctl_sctp_wmem,
8360 .memory_pressure = &sctp_memory_pressure,
8361 .enter_memory_pressure = sctp_enter_memory_pressure,
8362 .memory_allocated = &sctp_memory_allocated,
8363 .sockets_allocated = &sctp_sockets_allocated,
8365 #endif /* IS_ENABLED(CONFIG_IPV6) */