perf bpf: Move perf_event_output() from stdio.h to bpf.h
[linux/fpc-iii.git] / net / sctp / socket.c
blobb8cebd5a87e5c3571cbc184324ed483d3e6eb9bd
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>
69 #include <linux/rhashtable.h>
71 #include <net/ip.h>
72 #include <net/icmp.h>
73 #include <net/route.h>
74 #include <net/ipv6.h>
75 #include <net/inet_common.h>
76 #include <net/busy_poll.h>
78 #include <linux/socket.h> /* for sa_family_t */
79 #include <linux/export.h>
80 #include <net/sock.h>
81 #include <net/sctp/sctp.h>
82 #include <net/sctp/sm.h>
83 #include <net/sctp/stream_sched.h>
85 /* Forward declarations for internal helper functions. */
86 static bool sctp_writeable(struct sock *sk);
87 static void sctp_wfree(struct sk_buff *skb);
88 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
89 size_t msg_len);
90 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
91 static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
92 static int sctp_wait_for_accept(struct sock *sk, long timeo);
93 static void sctp_wait_for_close(struct sock *sk, long timeo);
94 static void sctp_destruct_sock(struct sock *sk);
95 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
96 union sctp_addr *addr, int len);
97 static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
98 static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
99 static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
100 static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
101 static int sctp_send_asconf(struct sctp_association *asoc,
102 struct sctp_chunk *chunk);
103 static int sctp_do_bind(struct sock *, union sctp_addr *, int);
104 static int sctp_autobind(struct sock *sk);
105 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
106 struct sctp_association *assoc,
107 enum sctp_socket_type type);
109 static unsigned long sctp_memory_pressure;
110 static atomic_long_t sctp_memory_allocated;
111 struct percpu_counter sctp_sockets_allocated;
113 static void sctp_enter_memory_pressure(struct sock *sk)
115 sctp_memory_pressure = 1;
119 /* Get the sndbuf space available at the time on the association. */
120 static inline int sctp_wspace(struct sctp_association *asoc)
122 struct sock *sk = asoc->base.sk;
124 return asoc->ep->sndbuf_policy ? sk->sk_sndbuf - asoc->sndbuf_used
125 : sk_stream_wspace(sk);
128 /* Increment the used sndbuf space count of the corresponding association by
129 * the size of the outgoing data chunk.
130 * Also, set the skb destructor for sndbuf accounting later.
132 * Since it is always 1-1 between chunk and skb, and also a new skb is always
133 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
134 * destructor in the data chunk skb for the purpose of the sndbuf space
135 * tracking.
137 static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
139 struct sctp_association *asoc = chunk->asoc;
140 struct sock *sk = asoc->base.sk;
142 /* The sndbuf space is tracked per association. */
143 sctp_association_hold(asoc);
145 if (chunk->shkey)
146 sctp_auth_shkey_hold(chunk->shkey);
148 skb_set_owner_w(chunk->skb, sk);
150 chunk->skb->destructor = sctp_wfree;
151 /* Save the chunk pointer in skb for sctp_wfree to use later. */
152 skb_shinfo(chunk->skb)->destructor_arg = chunk;
154 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
155 asoc->sndbuf_used += chunk->skb->truesize + sizeof(struct sctp_chunk);
156 sk->sk_wmem_queued += chunk->skb->truesize + sizeof(struct sctp_chunk);
157 sk_mem_charge(sk, chunk->skb->truesize);
160 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
162 skb_orphan(chunk->skb);
165 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
166 void (*cb)(struct sctp_chunk *))
169 struct sctp_outq *q = &asoc->outqueue;
170 struct sctp_transport *t;
171 struct sctp_chunk *chunk;
173 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
174 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
175 cb(chunk);
177 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
178 cb(chunk);
180 list_for_each_entry(chunk, &q->sacked, transmitted_list)
181 cb(chunk);
183 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
184 cb(chunk);
186 list_for_each_entry(chunk, &q->out_chunk_list, list)
187 cb(chunk);
190 static void sctp_for_each_rx_skb(struct sctp_association *asoc, struct sock *sk,
191 void (*cb)(struct sk_buff *, struct sock *))
194 struct sk_buff *skb, *tmp;
196 sctp_skb_for_each(skb, &asoc->ulpq.lobby, tmp)
197 cb(skb, sk);
199 sctp_skb_for_each(skb, &asoc->ulpq.reasm, tmp)
200 cb(skb, sk);
202 sctp_skb_for_each(skb, &asoc->ulpq.reasm_uo, tmp)
203 cb(skb, sk);
206 /* Verify that this is a valid address. */
207 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
208 int len)
210 struct sctp_af *af;
212 /* Verify basic sockaddr. */
213 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
214 if (!af)
215 return -EINVAL;
217 /* Is this a valid SCTP address? */
218 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
219 return -EINVAL;
221 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
222 return -EINVAL;
224 return 0;
227 /* Look up the association by its id. If this is not a UDP-style
228 * socket, the ID field is always ignored.
230 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
232 struct sctp_association *asoc = NULL;
234 /* If this is not a UDP-style socket, assoc id should be ignored. */
235 if (!sctp_style(sk, UDP)) {
236 /* Return NULL if the socket state is not ESTABLISHED. It
237 * could be a TCP-style listening socket or a socket which
238 * hasn't yet called connect() to establish an association.
240 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
241 return NULL;
243 /* Get the first and the only association from the list. */
244 if (!list_empty(&sctp_sk(sk)->ep->asocs))
245 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
246 struct sctp_association, asocs);
247 return asoc;
250 /* Otherwise this is a UDP-style socket. */
251 if (!id || (id == (sctp_assoc_t)-1))
252 return NULL;
254 spin_lock_bh(&sctp_assocs_id_lock);
255 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
256 if (asoc && (asoc->base.sk != sk || asoc->base.dead))
257 asoc = NULL;
258 spin_unlock_bh(&sctp_assocs_id_lock);
260 return asoc;
263 /* Look up the transport from an address and an assoc id. If both address and
264 * id are specified, the associations matching the address and the id should be
265 * the same.
267 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
268 struct sockaddr_storage *addr,
269 sctp_assoc_t id)
271 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
272 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
273 union sctp_addr *laddr = (union sctp_addr *)addr;
274 struct sctp_transport *transport;
276 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
277 return NULL;
279 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
280 laddr,
281 &transport);
283 if (!addr_asoc)
284 return NULL;
286 id_asoc = sctp_id2assoc(sk, id);
287 if (id_asoc && (id_asoc != addr_asoc))
288 return NULL;
290 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
291 (union sctp_addr *)addr);
293 return transport;
296 /* API 3.1.2 bind() - UDP Style Syntax
297 * The syntax of bind() is,
299 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
301 * sd - the socket descriptor returned by socket().
302 * addr - the address structure (struct sockaddr_in or struct
303 * sockaddr_in6 [RFC 2553]),
304 * addr_len - the size of the address structure.
306 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
308 int retval = 0;
310 lock_sock(sk);
312 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
313 addr, addr_len);
315 /* Disallow binding twice. */
316 if (!sctp_sk(sk)->ep->base.bind_addr.port)
317 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
318 addr_len);
319 else
320 retval = -EINVAL;
322 release_sock(sk);
324 return retval;
327 static long sctp_get_port_local(struct sock *, union sctp_addr *);
329 /* Verify this is a valid sockaddr. */
330 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
331 union sctp_addr *addr, int len)
333 struct sctp_af *af;
335 /* Check minimum size. */
336 if (len < sizeof (struct sockaddr))
337 return NULL;
339 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
340 return NULL;
342 if (addr->sa.sa_family == AF_INET6) {
343 if (len < SIN6_LEN_RFC2133)
344 return NULL;
345 /* V4 mapped address are really of AF_INET family */
346 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
347 !opt->pf->af_supported(AF_INET, opt))
348 return NULL;
351 /* If we get this far, af is valid. */
352 af = sctp_get_af_specific(addr->sa.sa_family);
354 if (len < af->sockaddr_len)
355 return NULL;
357 return af;
360 /* Bind a local address either to an endpoint or to an association. */
361 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
363 struct net *net = sock_net(sk);
364 struct sctp_sock *sp = sctp_sk(sk);
365 struct sctp_endpoint *ep = sp->ep;
366 struct sctp_bind_addr *bp = &ep->base.bind_addr;
367 struct sctp_af *af;
368 unsigned short snum;
369 int ret = 0;
371 /* Common sockaddr verification. */
372 af = sctp_sockaddr_af(sp, addr, len);
373 if (!af) {
374 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
375 __func__, sk, addr, len);
376 return -EINVAL;
379 snum = ntohs(addr->v4.sin_port);
381 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
382 __func__, sk, &addr->sa, bp->port, snum, len);
384 /* PF specific bind() address verification. */
385 if (!sp->pf->bind_verify(sp, addr))
386 return -EADDRNOTAVAIL;
388 /* We must either be unbound, or bind to the same port.
389 * It's OK to allow 0 ports if we are already bound.
390 * We'll just inhert an already bound port in this case
392 if (bp->port) {
393 if (!snum)
394 snum = bp->port;
395 else if (snum != bp->port) {
396 pr_debug("%s: new port %d doesn't match existing port "
397 "%d\n", __func__, snum, bp->port);
398 return -EINVAL;
402 if (snum && snum < inet_prot_sock(net) &&
403 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
404 return -EACCES;
406 /* See if the address matches any of the addresses we may have
407 * already bound before checking against other endpoints.
409 if (sctp_bind_addr_match(bp, addr, sp))
410 return -EINVAL;
412 /* Make sure we are allowed to bind here.
413 * The function sctp_get_port_local() does duplicate address
414 * detection.
416 addr->v4.sin_port = htons(snum);
417 if ((ret = sctp_get_port_local(sk, addr))) {
418 return -EADDRINUSE;
421 /* Refresh ephemeral port. */
422 if (!bp->port)
423 bp->port = inet_sk(sk)->inet_num;
425 /* Add the address to the bind address list.
426 * Use GFP_ATOMIC since BHs will be disabled.
428 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
429 SCTP_ADDR_SRC, GFP_ATOMIC);
431 /* Copy back into socket for getsockname() use. */
432 if (!ret) {
433 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
434 sp->pf->to_sk_saddr(addr, sk);
437 return ret;
440 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
442 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
443 * at any one time. If a sender, after sending an ASCONF chunk, decides
444 * it needs to transfer another ASCONF Chunk, it MUST wait until the
445 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
446 * subsequent ASCONF. Note this restriction binds each side, so at any
447 * time two ASCONF may be in-transit on any given association (one sent
448 * from each endpoint).
450 static int sctp_send_asconf(struct sctp_association *asoc,
451 struct sctp_chunk *chunk)
453 struct net *net = sock_net(asoc->base.sk);
454 int retval = 0;
456 /* If there is an outstanding ASCONF chunk, queue it for later
457 * transmission.
459 if (asoc->addip_last_asconf) {
460 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
461 goto out;
464 /* Hold the chunk until an ASCONF_ACK is received. */
465 sctp_chunk_hold(chunk);
466 retval = sctp_primitive_ASCONF(net, asoc, chunk);
467 if (retval)
468 sctp_chunk_free(chunk);
469 else
470 asoc->addip_last_asconf = chunk;
472 out:
473 return retval;
476 /* Add a list of addresses as bind addresses to local endpoint or
477 * association.
479 * Basically run through each address specified in the addrs/addrcnt
480 * array/length pair, determine if it is IPv6 or IPv4 and call
481 * sctp_do_bind() on it.
483 * If any of them fails, then the operation will be reversed and the
484 * ones that were added will be removed.
486 * Only sctp_setsockopt_bindx() is supposed to call this function.
488 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
490 int cnt;
491 int retval = 0;
492 void *addr_buf;
493 struct sockaddr *sa_addr;
494 struct sctp_af *af;
496 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
497 addrs, addrcnt);
499 addr_buf = addrs;
500 for (cnt = 0; cnt < addrcnt; cnt++) {
501 /* The list may contain either IPv4 or IPv6 address;
502 * determine the address length for walking thru the list.
504 sa_addr = addr_buf;
505 af = sctp_get_af_specific(sa_addr->sa_family);
506 if (!af) {
507 retval = -EINVAL;
508 goto err_bindx_add;
511 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
512 af->sockaddr_len);
514 addr_buf += af->sockaddr_len;
516 err_bindx_add:
517 if (retval < 0) {
518 /* Failed. Cleanup the ones that have been added */
519 if (cnt > 0)
520 sctp_bindx_rem(sk, addrs, cnt);
521 return retval;
525 return retval;
528 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
529 * associations that are part of the endpoint indicating that a list of local
530 * addresses are added to the endpoint.
532 * If any of the addresses is already in the bind address list of the
533 * association, we do not send the chunk for that association. But it will not
534 * affect other associations.
536 * Only sctp_setsockopt_bindx() is supposed to call this function.
538 static int sctp_send_asconf_add_ip(struct sock *sk,
539 struct sockaddr *addrs,
540 int addrcnt)
542 struct net *net = sock_net(sk);
543 struct sctp_sock *sp;
544 struct sctp_endpoint *ep;
545 struct sctp_association *asoc;
546 struct sctp_bind_addr *bp;
547 struct sctp_chunk *chunk;
548 struct sctp_sockaddr_entry *laddr;
549 union sctp_addr *addr;
550 union sctp_addr saveaddr;
551 void *addr_buf;
552 struct sctp_af *af;
553 struct list_head *p;
554 int i;
555 int retval = 0;
557 if (!net->sctp.addip_enable)
558 return retval;
560 sp = sctp_sk(sk);
561 ep = sp->ep;
563 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
564 __func__, sk, addrs, addrcnt);
566 list_for_each_entry(asoc, &ep->asocs, asocs) {
567 if (!asoc->peer.asconf_capable)
568 continue;
570 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
571 continue;
573 if (!sctp_state(asoc, ESTABLISHED))
574 continue;
576 /* Check if any address in the packed array of addresses is
577 * in the bind address list of the association. If so,
578 * do not send the asconf chunk to its peer, but continue with
579 * other associations.
581 addr_buf = addrs;
582 for (i = 0; i < addrcnt; i++) {
583 addr = addr_buf;
584 af = sctp_get_af_specific(addr->v4.sin_family);
585 if (!af) {
586 retval = -EINVAL;
587 goto out;
590 if (sctp_assoc_lookup_laddr(asoc, addr))
591 break;
593 addr_buf += af->sockaddr_len;
595 if (i < addrcnt)
596 continue;
598 /* Use the first valid address in bind addr list of
599 * association as Address Parameter of ASCONF CHUNK.
601 bp = &asoc->base.bind_addr;
602 p = bp->address_list.next;
603 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
604 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
605 addrcnt, SCTP_PARAM_ADD_IP);
606 if (!chunk) {
607 retval = -ENOMEM;
608 goto out;
611 /* Add the new addresses to the bind address list with
612 * use_as_src set to 0.
614 addr_buf = addrs;
615 for (i = 0; i < addrcnt; i++) {
616 addr = addr_buf;
617 af = sctp_get_af_specific(addr->v4.sin_family);
618 memcpy(&saveaddr, addr, af->sockaddr_len);
619 retval = sctp_add_bind_addr(bp, &saveaddr,
620 sizeof(saveaddr),
621 SCTP_ADDR_NEW, GFP_ATOMIC);
622 addr_buf += af->sockaddr_len;
624 if (asoc->src_out_of_asoc_ok) {
625 struct sctp_transport *trans;
627 list_for_each_entry(trans,
628 &asoc->peer.transport_addr_list, transports) {
629 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
630 2*asoc->pathmtu, 4380));
631 trans->ssthresh = asoc->peer.i.a_rwnd;
632 trans->rto = asoc->rto_initial;
633 sctp_max_rto(asoc, trans);
634 trans->rtt = trans->srtt = trans->rttvar = 0;
635 /* Clear the source and route cache */
636 sctp_transport_route(trans, NULL,
637 sctp_sk(asoc->base.sk));
640 retval = sctp_send_asconf(asoc, chunk);
643 out:
644 return retval;
647 /* Remove a list of addresses from bind addresses list. Do not remove the
648 * last address.
650 * Basically run through each address specified in the addrs/addrcnt
651 * array/length pair, determine if it is IPv6 or IPv4 and call
652 * sctp_del_bind() on it.
654 * If any of them fails, then the operation will be reversed and the
655 * ones that were removed will be added back.
657 * At least one address has to be left; if only one address is
658 * available, the operation will return -EBUSY.
660 * Only sctp_setsockopt_bindx() is supposed to call this function.
662 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
664 struct sctp_sock *sp = sctp_sk(sk);
665 struct sctp_endpoint *ep = sp->ep;
666 int cnt;
667 struct sctp_bind_addr *bp = &ep->base.bind_addr;
668 int retval = 0;
669 void *addr_buf;
670 union sctp_addr *sa_addr;
671 struct sctp_af *af;
673 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
674 __func__, sk, addrs, addrcnt);
676 addr_buf = addrs;
677 for (cnt = 0; cnt < addrcnt; cnt++) {
678 /* If the bind address list is empty or if there is only one
679 * bind address, there is nothing more to be removed (we need
680 * at least one address here).
682 if (list_empty(&bp->address_list) ||
683 (sctp_list_single_entry(&bp->address_list))) {
684 retval = -EBUSY;
685 goto err_bindx_rem;
688 sa_addr = addr_buf;
689 af = sctp_get_af_specific(sa_addr->sa.sa_family);
690 if (!af) {
691 retval = -EINVAL;
692 goto err_bindx_rem;
695 if (!af->addr_valid(sa_addr, sp, NULL)) {
696 retval = -EADDRNOTAVAIL;
697 goto err_bindx_rem;
700 if (sa_addr->v4.sin_port &&
701 sa_addr->v4.sin_port != htons(bp->port)) {
702 retval = -EINVAL;
703 goto err_bindx_rem;
706 if (!sa_addr->v4.sin_port)
707 sa_addr->v4.sin_port = htons(bp->port);
709 /* FIXME - There is probably a need to check if sk->sk_saddr and
710 * sk->sk_rcv_addr are currently set to one of the addresses to
711 * be removed. This is something which needs to be looked into
712 * when we are fixing the outstanding issues with multi-homing
713 * socket routing and failover schemes. Refer to comments in
714 * sctp_do_bind(). -daisy
716 retval = sctp_del_bind_addr(bp, sa_addr);
718 addr_buf += af->sockaddr_len;
719 err_bindx_rem:
720 if (retval < 0) {
721 /* Failed. Add the ones that has been removed back */
722 if (cnt > 0)
723 sctp_bindx_add(sk, addrs, cnt);
724 return retval;
728 return retval;
731 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
732 * the associations that are part of the endpoint indicating that a list of
733 * local addresses are removed from the endpoint.
735 * If any of the addresses is already in the bind address list of the
736 * association, we do not send the chunk for that association. But it will not
737 * affect other associations.
739 * Only sctp_setsockopt_bindx() is supposed to call this function.
741 static int sctp_send_asconf_del_ip(struct sock *sk,
742 struct sockaddr *addrs,
743 int addrcnt)
745 struct net *net = sock_net(sk);
746 struct sctp_sock *sp;
747 struct sctp_endpoint *ep;
748 struct sctp_association *asoc;
749 struct sctp_transport *transport;
750 struct sctp_bind_addr *bp;
751 struct sctp_chunk *chunk;
752 union sctp_addr *laddr;
753 void *addr_buf;
754 struct sctp_af *af;
755 struct sctp_sockaddr_entry *saddr;
756 int i;
757 int retval = 0;
758 int stored = 0;
760 chunk = NULL;
761 if (!net->sctp.addip_enable)
762 return retval;
764 sp = sctp_sk(sk);
765 ep = sp->ep;
767 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
768 __func__, sk, addrs, addrcnt);
770 list_for_each_entry(asoc, &ep->asocs, asocs) {
772 if (!asoc->peer.asconf_capable)
773 continue;
775 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
776 continue;
778 if (!sctp_state(asoc, ESTABLISHED))
779 continue;
781 /* Check if any address in the packed array of addresses is
782 * not present in the bind address list of the association.
783 * If so, do not send the asconf chunk to its peer, but
784 * continue with other associations.
786 addr_buf = addrs;
787 for (i = 0; i < addrcnt; i++) {
788 laddr = addr_buf;
789 af = sctp_get_af_specific(laddr->v4.sin_family);
790 if (!af) {
791 retval = -EINVAL;
792 goto out;
795 if (!sctp_assoc_lookup_laddr(asoc, laddr))
796 break;
798 addr_buf += af->sockaddr_len;
800 if (i < addrcnt)
801 continue;
803 /* Find one address in the association's bind address list
804 * that is not in the packed array of addresses. This is to
805 * make sure that we do not delete all the addresses in the
806 * association.
808 bp = &asoc->base.bind_addr;
809 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
810 addrcnt, sp);
811 if ((laddr == NULL) && (addrcnt == 1)) {
812 if (asoc->asconf_addr_del_pending)
813 continue;
814 asoc->asconf_addr_del_pending =
815 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
816 if (asoc->asconf_addr_del_pending == NULL) {
817 retval = -ENOMEM;
818 goto out;
820 asoc->asconf_addr_del_pending->sa.sa_family =
821 addrs->sa_family;
822 asoc->asconf_addr_del_pending->v4.sin_port =
823 htons(bp->port);
824 if (addrs->sa_family == AF_INET) {
825 struct sockaddr_in *sin;
827 sin = (struct sockaddr_in *)addrs;
828 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
829 } else if (addrs->sa_family == AF_INET6) {
830 struct sockaddr_in6 *sin6;
832 sin6 = (struct sockaddr_in6 *)addrs;
833 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
836 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
837 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
838 asoc->asconf_addr_del_pending);
840 asoc->src_out_of_asoc_ok = 1;
841 stored = 1;
842 goto skip_mkasconf;
845 if (laddr == NULL)
846 return -EINVAL;
848 /* We do not need RCU protection throughout this loop
849 * because this is done under a socket lock from the
850 * setsockopt call.
852 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
853 SCTP_PARAM_DEL_IP);
854 if (!chunk) {
855 retval = -ENOMEM;
856 goto out;
859 skip_mkasconf:
860 /* Reset use_as_src flag for the addresses in the bind address
861 * list that are to be deleted.
863 addr_buf = addrs;
864 for (i = 0; i < addrcnt; i++) {
865 laddr = addr_buf;
866 af = sctp_get_af_specific(laddr->v4.sin_family);
867 list_for_each_entry(saddr, &bp->address_list, list) {
868 if (sctp_cmp_addr_exact(&saddr->a, laddr))
869 saddr->state = SCTP_ADDR_DEL;
871 addr_buf += af->sockaddr_len;
874 /* Update the route and saddr entries for all the transports
875 * as some of the addresses in the bind address list are
876 * about to be deleted and cannot be used as source addresses.
878 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
879 transports) {
880 sctp_transport_route(transport, NULL,
881 sctp_sk(asoc->base.sk));
884 if (stored)
885 /* We don't need to transmit ASCONF */
886 continue;
887 retval = sctp_send_asconf(asoc, chunk);
889 out:
890 return retval;
893 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
894 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
896 struct sock *sk = sctp_opt2sk(sp);
897 union sctp_addr *addr;
898 struct sctp_af *af;
900 /* It is safe to write port space in caller. */
901 addr = &addrw->a;
902 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
903 af = sctp_get_af_specific(addr->sa.sa_family);
904 if (!af)
905 return -EINVAL;
906 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
907 return -EINVAL;
909 if (addrw->state == SCTP_ADDR_NEW)
910 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
911 else
912 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
915 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
917 * API 8.1
918 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
919 * int flags);
921 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
922 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
923 * or IPv6 addresses.
925 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
926 * Section 3.1.2 for this usage.
928 * addrs is a pointer to an array of one or more socket addresses. Each
929 * address is contained in its appropriate structure (i.e. struct
930 * sockaddr_in or struct sockaddr_in6) the family of the address type
931 * must be used to distinguish the address length (note that this
932 * representation is termed a "packed array" of addresses). The caller
933 * specifies the number of addresses in the array with addrcnt.
935 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
936 * -1, and sets errno to the appropriate error code.
938 * For SCTP, the port given in each socket address must be the same, or
939 * sctp_bindx() will fail, setting errno to EINVAL.
941 * The flags parameter is formed from the bitwise OR of zero or more of
942 * the following currently defined flags:
944 * SCTP_BINDX_ADD_ADDR
946 * SCTP_BINDX_REM_ADDR
948 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
949 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
950 * addresses from the association. The two flags are mutually exclusive;
951 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
952 * not remove all addresses from an association; sctp_bindx() will
953 * reject such an attempt with EINVAL.
955 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
956 * additional addresses with an endpoint after calling bind(). Or use
957 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
958 * socket is associated with so that no new association accepted will be
959 * associated with those addresses. If the endpoint supports dynamic
960 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
961 * endpoint to send the appropriate message to the peer to change the
962 * peers address lists.
964 * Adding and removing addresses from a connected association is
965 * optional functionality. Implementations that do not support this
966 * functionality should return EOPNOTSUPP.
968 * Basically do nothing but copying the addresses from user to kernel
969 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
970 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
971 * from userspace.
973 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
974 * it.
976 * sk The sk of the socket
977 * addrs The pointer to the addresses in user land
978 * addrssize Size of the addrs buffer
979 * op Operation to perform (add or remove, see the flags of
980 * sctp_bindx)
982 * Returns 0 if ok, <0 errno code on error.
984 static int sctp_setsockopt_bindx(struct sock *sk,
985 struct sockaddr __user *addrs,
986 int addrs_size, int op)
988 struct sockaddr *kaddrs;
989 int err;
990 int addrcnt = 0;
991 int walk_size = 0;
992 struct sockaddr *sa_addr;
993 void *addr_buf;
994 struct sctp_af *af;
996 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
997 __func__, sk, addrs, addrs_size, op);
999 if (unlikely(addrs_size <= 0))
1000 return -EINVAL;
1002 kaddrs = vmemdup_user(addrs, addrs_size);
1003 if (unlikely(IS_ERR(kaddrs)))
1004 return PTR_ERR(kaddrs);
1006 /* Walk through the addrs buffer and count the number of addresses. */
1007 addr_buf = kaddrs;
1008 while (walk_size < addrs_size) {
1009 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1010 kvfree(kaddrs);
1011 return -EINVAL;
1014 sa_addr = addr_buf;
1015 af = sctp_get_af_specific(sa_addr->sa_family);
1017 /* If the address family is not supported or if this address
1018 * causes the address buffer to overflow return EINVAL.
1020 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1021 kvfree(kaddrs);
1022 return -EINVAL;
1024 addrcnt++;
1025 addr_buf += af->sockaddr_len;
1026 walk_size += af->sockaddr_len;
1029 /* Do the work. */
1030 switch (op) {
1031 case SCTP_BINDX_ADD_ADDR:
1032 /* Allow security module to validate bindx addresses. */
1033 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_BINDX_ADD,
1034 (struct sockaddr *)kaddrs,
1035 addrs_size);
1036 if (err)
1037 goto out;
1038 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1039 if (err)
1040 goto out;
1041 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1042 break;
1044 case SCTP_BINDX_REM_ADDR:
1045 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1046 if (err)
1047 goto out;
1048 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1049 break;
1051 default:
1052 err = -EINVAL;
1053 break;
1056 out:
1057 kvfree(kaddrs);
1059 return err;
1062 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1064 * Common routine for handling connect() and sctp_connectx().
1065 * Connect will come in with just a single address.
1067 static int __sctp_connect(struct sock *sk,
1068 struct sockaddr *kaddrs,
1069 int addrs_size, int flags,
1070 sctp_assoc_t *assoc_id)
1072 struct net *net = sock_net(sk);
1073 struct sctp_sock *sp;
1074 struct sctp_endpoint *ep;
1075 struct sctp_association *asoc = NULL;
1076 struct sctp_association *asoc2;
1077 struct sctp_transport *transport;
1078 union sctp_addr to;
1079 enum sctp_scope scope;
1080 long timeo;
1081 int err = 0;
1082 int addrcnt = 0;
1083 int walk_size = 0;
1084 union sctp_addr *sa_addr = NULL;
1085 void *addr_buf;
1086 unsigned short port;
1088 sp = sctp_sk(sk);
1089 ep = sp->ep;
1091 /* connect() cannot be done on a socket that is already in ESTABLISHED
1092 * state - UDP-style peeled off socket or a TCP-style socket that
1093 * is already connected.
1094 * It cannot be done even on a TCP-style listening socket.
1096 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1097 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1098 err = -EISCONN;
1099 goto out_free;
1102 /* Walk through the addrs buffer and count the number of addresses. */
1103 addr_buf = kaddrs;
1104 while (walk_size < addrs_size) {
1105 struct sctp_af *af;
1107 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1108 err = -EINVAL;
1109 goto out_free;
1112 sa_addr = addr_buf;
1113 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1115 /* If the address family is not supported or if this address
1116 * causes the address buffer to overflow return EINVAL.
1118 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1119 err = -EINVAL;
1120 goto out_free;
1123 port = ntohs(sa_addr->v4.sin_port);
1125 /* Save current address so we can work with it */
1126 memcpy(&to, sa_addr, af->sockaddr_len);
1128 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1129 if (err)
1130 goto out_free;
1132 /* Make sure the destination port is correctly set
1133 * in all addresses.
1135 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1136 err = -EINVAL;
1137 goto out_free;
1140 /* Check if there already is a matching association on the
1141 * endpoint (other than the one created here).
1143 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1144 if (asoc2 && asoc2 != asoc) {
1145 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1146 err = -EISCONN;
1147 else
1148 err = -EALREADY;
1149 goto out_free;
1152 /* If we could not find a matching association on the endpoint,
1153 * make sure that there is no peeled-off association matching
1154 * the peer address even on another socket.
1156 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1157 err = -EADDRNOTAVAIL;
1158 goto out_free;
1161 if (!asoc) {
1162 /* If a bind() or sctp_bindx() is not called prior to
1163 * an sctp_connectx() call, the system picks an
1164 * ephemeral port and will choose an address set
1165 * equivalent to binding with a wildcard address.
1167 if (!ep->base.bind_addr.port) {
1168 if (sctp_autobind(sk)) {
1169 err = -EAGAIN;
1170 goto out_free;
1172 } else {
1174 * If an unprivileged user inherits a 1-many
1175 * style socket with open associations on a
1176 * privileged port, it MAY be permitted to
1177 * accept new associations, but it SHOULD NOT
1178 * be permitted to open new associations.
1180 if (ep->base.bind_addr.port <
1181 inet_prot_sock(net) &&
1182 !ns_capable(net->user_ns,
1183 CAP_NET_BIND_SERVICE)) {
1184 err = -EACCES;
1185 goto out_free;
1189 scope = sctp_scope(&to);
1190 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1191 if (!asoc) {
1192 err = -ENOMEM;
1193 goto out_free;
1196 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1197 GFP_KERNEL);
1198 if (err < 0) {
1199 goto out_free;
1204 /* Prime the peer's transport structures. */
1205 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1206 SCTP_UNKNOWN);
1207 if (!transport) {
1208 err = -ENOMEM;
1209 goto out_free;
1212 addrcnt++;
1213 addr_buf += af->sockaddr_len;
1214 walk_size += af->sockaddr_len;
1217 /* In case the user of sctp_connectx() wants an association
1218 * id back, assign one now.
1220 if (assoc_id) {
1221 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1222 if (err < 0)
1223 goto out_free;
1226 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1227 if (err < 0) {
1228 goto out_free;
1231 /* Initialize sk's dport and daddr for getpeername() */
1232 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1233 sp->pf->to_sk_daddr(sa_addr, sk);
1234 sk->sk_err = 0;
1236 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1238 if (assoc_id)
1239 *assoc_id = asoc->assoc_id;
1241 err = sctp_wait_for_connect(asoc, &timeo);
1242 /* Note: the asoc may be freed after the return of
1243 * sctp_wait_for_connect.
1246 /* Don't free association on exit. */
1247 asoc = NULL;
1249 out_free:
1250 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1251 __func__, asoc, kaddrs, err);
1253 if (asoc) {
1254 /* sctp_primitive_ASSOCIATE may have added this association
1255 * To the hash table, try to unhash it, just in case, its a noop
1256 * if it wasn't hashed so we're safe
1258 sctp_association_free(asoc);
1260 return err;
1263 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1265 * API 8.9
1266 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1267 * sctp_assoc_t *asoc);
1269 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1270 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1271 * or IPv6 addresses.
1273 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1274 * Section 3.1.2 for this usage.
1276 * addrs is a pointer to an array of one or more socket addresses. Each
1277 * address is contained in its appropriate structure (i.e. struct
1278 * sockaddr_in or struct sockaddr_in6) the family of the address type
1279 * must be used to distengish the address length (note that this
1280 * representation is termed a "packed array" of addresses). The caller
1281 * specifies the number of addresses in the array with addrcnt.
1283 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1284 * the association id of the new association. On failure, sctp_connectx()
1285 * returns -1, and sets errno to the appropriate error code. The assoc_id
1286 * is not touched by the kernel.
1288 * For SCTP, the port given in each socket address must be the same, or
1289 * sctp_connectx() will fail, setting errno to EINVAL.
1291 * An application can use sctp_connectx to initiate an association with
1292 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1293 * allows a caller to specify multiple addresses at which a peer can be
1294 * reached. The way the SCTP stack uses the list of addresses to set up
1295 * the association is implementation dependent. This function only
1296 * specifies that the stack will try to make use of all the addresses in
1297 * the list when needed.
1299 * Note that the list of addresses passed in is only used for setting up
1300 * the association. It does not necessarily equal the set of addresses
1301 * the peer uses for the resulting association. If the caller wants to
1302 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1303 * retrieve them after the association has been set up.
1305 * Basically do nothing but copying the addresses from user to kernel
1306 * land and invoking either sctp_connectx(). This is used for tunneling
1307 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1309 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1310 * it.
1312 * sk The sk of the socket
1313 * addrs The pointer to the addresses in user land
1314 * addrssize Size of the addrs buffer
1316 * Returns >=0 if ok, <0 errno code on error.
1318 static int __sctp_setsockopt_connectx(struct sock *sk,
1319 struct sockaddr __user *addrs,
1320 int addrs_size,
1321 sctp_assoc_t *assoc_id)
1323 struct sockaddr *kaddrs;
1324 int err = 0, flags = 0;
1326 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1327 __func__, sk, addrs, addrs_size);
1329 if (unlikely(addrs_size <= 0))
1330 return -EINVAL;
1332 kaddrs = vmemdup_user(addrs, addrs_size);
1333 if (unlikely(IS_ERR(kaddrs)))
1334 return PTR_ERR(kaddrs);
1336 /* Allow security module to validate connectx addresses. */
1337 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_CONNECTX,
1338 (struct sockaddr *)kaddrs,
1339 addrs_size);
1340 if (err)
1341 goto out_free;
1343 /* in-kernel sockets don't generally have a file allocated to them
1344 * if all they do is call sock_create_kern().
1346 if (sk->sk_socket->file)
1347 flags = sk->sk_socket->file->f_flags;
1349 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1351 out_free:
1352 kvfree(kaddrs);
1354 return err;
1358 * This is an older interface. It's kept for backward compatibility
1359 * to the option that doesn't provide association id.
1361 static int sctp_setsockopt_connectx_old(struct sock *sk,
1362 struct sockaddr __user *addrs,
1363 int addrs_size)
1365 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1369 * New interface for the API. The since the API is done with a socket
1370 * option, to make it simple we feed back the association id is as a return
1371 * indication to the call. Error is always negative and association id is
1372 * always positive.
1374 static int sctp_setsockopt_connectx(struct sock *sk,
1375 struct sockaddr __user *addrs,
1376 int addrs_size)
1378 sctp_assoc_t assoc_id = 0;
1379 int err = 0;
1381 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1383 if (err)
1384 return err;
1385 else
1386 return assoc_id;
1390 * New (hopefully final) interface for the API.
1391 * We use the sctp_getaddrs_old structure so that use-space library
1392 * can avoid any unnecessary allocations. The only different part
1393 * is that we store the actual length of the address buffer into the
1394 * addrs_num structure member. That way we can re-use the existing
1395 * code.
1397 #ifdef CONFIG_COMPAT
1398 struct compat_sctp_getaddrs_old {
1399 sctp_assoc_t assoc_id;
1400 s32 addr_num;
1401 compat_uptr_t addrs; /* struct sockaddr * */
1403 #endif
1405 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1406 char __user *optval,
1407 int __user *optlen)
1409 struct sctp_getaddrs_old param;
1410 sctp_assoc_t assoc_id = 0;
1411 int err = 0;
1413 #ifdef CONFIG_COMPAT
1414 if (in_compat_syscall()) {
1415 struct compat_sctp_getaddrs_old param32;
1417 if (len < sizeof(param32))
1418 return -EINVAL;
1419 if (copy_from_user(&param32, optval, sizeof(param32)))
1420 return -EFAULT;
1422 param.assoc_id = param32.assoc_id;
1423 param.addr_num = param32.addr_num;
1424 param.addrs = compat_ptr(param32.addrs);
1425 } else
1426 #endif
1428 if (len < sizeof(param))
1429 return -EINVAL;
1430 if (copy_from_user(&param, optval, sizeof(param)))
1431 return -EFAULT;
1434 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1435 param.addrs, param.addr_num,
1436 &assoc_id);
1437 if (err == 0 || err == -EINPROGRESS) {
1438 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1439 return -EFAULT;
1440 if (put_user(sizeof(assoc_id), optlen))
1441 return -EFAULT;
1444 return err;
1447 /* API 3.1.4 close() - UDP Style Syntax
1448 * Applications use close() to perform graceful shutdown (as described in
1449 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1450 * by a UDP-style socket.
1452 * The syntax is
1454 * ret = close(int sd);
1456 * sd - the socket descriptor of the associations to be closed.
1458 * To gracefully shutdown a specific association represented by the
1459 * UDP-style socket, an application should use the sendmsg() call,
1460 * passing no user data, but including the appropriate flag in the
1461 * ancillary data (see Section xxxx).
1463 * If sd in the close() call is a branched-off socket representing only
1464 * one association, the shutdown is performed on that association only.
1466 * 4.1.6 close() - TCP Style Syntax
1468 * Applications use close() to gracefully close down an association.
1470 * The syntax is:
1472 * int close(int sd);
1474 * sd - the socket descriptor of the association to be closed.
1476 * After an application calls close() on a socket descriptor, no further
1477 * socket operations will succeed on that descriptor.
1479 * API 7.1.4 SO_LINGER
1481 * An application using the TCP-style socket can use this option to
1482 * perform the SCTP ABORT primitive. The linger option structure is:
1484 * struct linger {
1485 * int l_onoff; // option on/off
1486 * int l_linger; // linger time
1487 * };
1489 * To enable the option, set l_onoff to 1. If the l_linger value is set
1490 * to 0, calling close() is the same as the ABORT primitive. If the
1491 * value is set to a negative value, the setsockopt() call will return
1492 * an error. If the value is set to a positive value linger_time, the
1493 * close() can be blocked for at most linger_time ms. If the graceful
1494 * shutdown phase does not finish during this period, close() will
1495 * return but the graceful shutdown phase continues in the system.
1497 static void sctp_close(struct sock *sk, long timeout)
1499 struct net *net = sock_net(sk);
1500 struct sctp_endpoint *ep;
1501 struct sctp_association *asoc;
1502 struct list_head *pos, *temp;
1503 unsigned int data_was_unread;
1505 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1507 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1508 sk->sk_shutdown = SHUTDOWN_MASK;
1509 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1511 ep = sctp_sk(sk)->ep;
1513 /* Clean up any skbs sitting on the receive queue. */
1514 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1515 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1517 /* Walk all associations on an endpoint. */
1518 list_for_each_safe(pos, temp, &ep->asocs) {
1519 asoc = list_entry(pos, struct sctp_association, asocs);
1521 if (sctp_style(sk, TCP)) {
1522 /* A closed association can still be in the list if
1523 * it belongs to a TCP-style listening socket that is
1524 * not yet accepted. If so, free it. If not, send an
1525 * ABORT or SHUTDOWN based on the linger options.
1527 if (sctp_state(asoc, CLOSED)) {
1528 sctp_association_free(asoc);
1529 continue;
1533 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1534 !skb_queue_empty(&asoc->ulpq.reasm) ||
1535 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1536 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1537 struct sctp_chunk *chunk;
1539 chunk = sctp_make_abort_user(asoc, NULL, 0);
1540 sctp_primitive_ABORT(net, asoc, chunk);
1541 } else
1542 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1545 /* On a TCP-style socket, block for at most linger_time if set. */
1546 if (sctp_style(sk, TCP) && timeout)
1547 sctp_wait_for_close(sk, timeout);
1549 /* This will run the backlog queue. */
1550 release_sock(sk);
1552 /* Supposedly, no process has access to the socket, but
1553 * the net layers still may.
1554 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1555 * held and that should be grabbed before socket lock.
1557 spin_lock_bh(&net->sctp.addr_wq_lock);
1558 bh_lock_sock_nested(sk);
1560 /* Hold the sock, since sk_common_release() will put sock_put()
1561 * and we have just a little more cleanup.
1563 sock_hold(sk);
1564 sk_common_release(sk);
1566 bh_unlock_sock(sk);
1567 spin_unlock_bh(&net->sctp.addr_wq_lock);
1569 sock_put(sk);
1571 SCTP_DBG_OBJCNT_DEC(sock);
1574 /* Handle EPIPE error. */
1575 static int sctp_error(struct sock *sk, int flags, int err)
1577 if (err == -EPIPE)
1578 err = sock_error(sk) ? : -EPIPE;
1579 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1580 send_sig(SIGPIPE, current, 0);
1581 return err;
1584 /* API 3.1.3 sendmsg() - UDP Style Syntax
1586 * An application uses sendmsg() and recvmsg() calls to transmit data to
1587 * and receive data from its peer.
1589 * ssize_t sendmsg(int socket, const struct msghdr *message,
1590 * int flags);
1592 * socket - the socket descriptor of the endpoint.
1593 * message - pointer to the msghdr structure which contains a single
1594 * user message and possibly some ancillary data.
1596 * See Section 5 for complete description of the data
1597 * structures.
1599 * flags - flags sent or received with the user message, see Section
1600 * 5 for complete description of the flags.
1602 * Note: This function could use a rewrite especially when explicit
1603 * connect support comes in.
1605 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1607 static int sctp_msghdr_parse(const struct msghdr *msg,
1608 struct sctp_cmsgs *cmsgs);
1610 static int sctp_sendmsg_parse(struct sock *sk, struct sctp_cmsgs *cmsgs,
1611 struct sctp_sndrcvinfo *srinfo,
1612 const struct msghdr *msg, size_t msg_len)
1614 __u16 sflags;
1615 int err;
1617 if (sctp_sstate(sk, LISTENING) && sctp_style(sk, TCP))
1618 return -EPIPE;
1620 if (msg_len > sk->sk_sndbuf)
1621 return -EMSGSIZE;
1623 memset(cmsgs, 0, sizeof(*cmsgs));
1624 err = sctp_msghdr_parse(msg, cmsgs);
1625 if (err) {
1626 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1627 return err;
1630 memset(srinfo, 0, sizeof(*srinfo));
1631 if (cmsgs->srinfo) {
1632 srinfo->sinfo_stream = cmsgs->srinfo->sinfo_stream;
1633 srinfo->sinfo_flags = cmsgs->srinfo->sinfo_flags;
1634 srinfo->sinfo_ppid = cmsgs->srinfo->sinfo_ppid;
1635 srinfo->sinfo_context = cmsgs->srinfo->sinfo_context;
1636 srinfo->sinfo_assoc_id = cmsgs->srinfo->sinfo_assoc_id;
1637 srinfo->sinfo_timetolive = cmsgs->srinfo->sinfo_timetolive;
1640 if (cmsgs->sinfo) {
1641 srinfo->sinfo_stream = cmsgs->sinfo->snd_sid;
1642 srinfo->sinfo_flags = cmsgs->sinfo->snd_flags;
1643 srinfo->sinfo_ppid = cmsgs->sinfo->snd_ppid;
1644 srinfo->sinfo_context = cmsgs->sinfo->snd_context;
1645 srinfo->sinfo_assoc_id = cmsgs->sinfo->snd_assoc_id;
1648 if (cmsgs->prinfo) {
1649 srinfo->sinfo_timetolive = cmsgs->prinfo->pr_value;
1650 SCTP_PR_SET_POLICY(srinfo->sinfo_flags,
1651 cmsgs->prinfo->pr_policy);
1654 sflags = srinfo->sinfo_flags;
1655 if (!sflags && msg_len)
1656 return 0;
1658 if (sctp_style(sk, TCP) && (sflags & (SCTP_EOF | SCTP_ABORT)))
1659 return -EINVAL;
1661 if (((sflags & SCTP_EOF) && msg_len > 0) ||
1662 (!(sflags & (SCTP_EOF | SCTP_ABORT)) && msg_len == 0))
1663 return -EINVAL;
1665 if ((sflags & SCTP_ADDR_OVER) && !msg->msg_name)
1666 return -EINVAL;
1668 return 0;
1671 static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
1672 struct sctp_cmsgs *cmsgs,
1673 union sctp_addr *daddr,
1674 struct sctp_transport **tp)
1676 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
1677 struct net *net = sock_net(sk);
1678 struct sctp_association *asoc;
1679 enum sctp_scope scope;
1680 struct cmsghdr *cmsg;
1681 __be32 flowinfo = 0;
1682 struct sctp_af *af;
1683 int err;
1685 *tp = NULL;
1687 if (sflags & (SCTP_EOF | SCTP_ABORT))
1688 return -EINVAL;
1690 if (sctp_style(sk, TCP) && (sctp_sstate(sk, ESTABLISHED) ||
1691 sctp_sstate(sk, CLOSING)))
1692 return -EADDRNOTAVAIL;
1694 if (sctp_endpoint_is_peeled_off(ep, daddr))
1695 return -EADDRNOTAVAIL;
1697 if (!ep->base.bind_addr.port) {
1698 if (sctp_autobind(sk))
1699 return -EAGAIN;
1700 } else {
1701 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1702 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1703 return -EACCES;
1706 scope = sctp_scope(daddr);
1708 /* Label connection socket for first association 1-to-many
1709 * style for client sequence socket()->sendmsg(). This
1710 * needs to be done before sctp_assoc_add_peer() as that will
1711 * set up the initial packet that needs to account for any
1712 * security ip options (CIPSO/CALIPSO) added to the packet.
1714 af = sctp_get_af_specific(daddr->sa.sa_family);
1715 if (!af)
1716 return -EINVAL;
1717 err = security_sctp_bind_connect(sk, SCTP_SENDMSG_CONNECT,
1718 (struct sockaddr *)daddr,
1719 af->sockaddr_len);
1720 if (err < 0)
1721 return err;
1723 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1724 if (!asoc)
1725 return -ENOMEM;
1727 if (sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL) < 0) {
1728 err = -ENOMEM;
1729 goto free;
1732 if (cmsgs->init) {
1733 struct sctp_initmsg *init = cmsgs->init;
1735 if (init->sinit_num_ostreams) {
1736 __u16 outcnt = init->sinit_num_ostreams;
1738 asoc->c.sinit_num_ostreams = outcnt;
1739 /* outcnt has been changed, need to re-init stream */
1740 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1741 GFP_KERNEL);
1742 if (err)
1743 goto free;
1746 if (init->sinit_max_instreams)
1747 asoc->c.sinit_max_instreams = init->sinit_max_instreams;
1749 if (init->sinit_max_attempts)
1750 asoc->max_init_attempts = init->sinit_max_attempts;
1752 if (init->sinit_max_init_timeo)
1753 asoc->max_init_timeo =
1754 msecs_to_jiffies(init->sinit_max_init_timeo);
1757 *tp = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL, SCTP_UNKNOWN);
1758 if (!*tp) {
1759 err = -ENOMEM;
1760 goto free;
1763 if (!cmsgs->addrs_msg)
1764 return 0;
1766 if (daddr->sa.sa_family == AF_INET6)
1767 flowinfo = daddr->v6.sin6_flowinfo;
1769 /* sendv addr list parse */
1770 for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
1771 struct sctp_transport *transport;
1772 struct sctp_association *old;
1773 union sctp_addr _daddr;
1774 int dlen;
1776 if (cmsg->cmsg_level != IPPROTO_SCTP ||
1777 (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
1778 cmsg->cmsg_type != SCTP_DSTADDRV6))
1779 continue;
1781 daddr = &_daddr;
1782 memset(daddr, 0, sizeof(*daddr));
1783 dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
1784 if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
1785 if (dlen < sizeof(struct in_addr)) {
1786 err = -EINVAL;
1787 goto free;
1790 dlen = sizeof(struct in_addr);
1791 daddr->v4.sin_family = AF_INET;
1792 daddr->v4.sin_port = htons(asoc->peer.port);
1793 memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
1794 } else {
1795 if (dlen < sizeof(struct in6_addr)) {
1796 err = -EINVAL;
1797 goto free;
1800 dlen = sizeof(struct in6_addr);
1801 daddr->v6.sin6_flowinfo = flowinfo;
1802 daddr->v6.sin6_family = AF_INET6;
1803 daddr->v6.sin6_port = htons(asoc->peer.port);
1804 memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
1806 err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
1807 if (err)
1808 goto free;
1810 old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
1811 if (old && old != asoc) {
1812 if (old->state >= SCTP_STATE_ESTABLISHED)
1813 err = -EISCONN;
1814 else
1815 err = -EALREADY;
1816 goto free;
1819 if (sctp_endpoint_is_peeled_off(ep, daddr)) {
1820 err = -EADDRNOTAVAIL;
1821 goto free;
1824 transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
1825 SCTP_UNKNOWN);
1826 if (!transport) {
1827 err = -ENOMEM;
1828 goto free;
1832 return 0;
1834 free:
1835 sctp_association_free(asoc);
1836 return err;
1839 static int sctp_sendmsg_check_sflags(struct sctp_association *asoc,
1840 __u16 sflags, struct msghdr *msg,
1841 size_t msg_len)
1843 struct sock *sk = asoc->base.sk;
1844 struct net *net = sock_net(sk);
1846 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP))
1847 return -EPIPE;
1849 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP) &&
1850 !sctp_state(asoc, ESTABLISHED))
1851 return 0;
1853 if (sflags & SCTP_EOF) {
1854 pr_debug("%s: shutting down association:%p\n", __func__, asoc);
1855 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1857 return 0;
1860 if (sflags & SCTP_ABORT) {
1861 struct sctp_chunk *chunk;
1863 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1864 if (!chunk)
1865 return -ENOMEM;
1867 pr_debug("%s: aborting association:%p\n", __func__, asoc);
1868 sctp_primitive_ABORT(net, asoc, chunk);
1870 return 0;
1873 return 1;
1876 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1877 struct msghdr *msg, size_t msg_len,
1878 struct sctp_transport *transport,
1879 struct sctp_sndrcvinfo *sinfo)
1881 struct sock *sk = asoc->base.sk;
1882 struct sctp_sock *sp = sctp_sk(sk);
1883 struct net *net = sock_net(sk);
1884 struct sctp_datamsg *datamsg;
1885 bool wait_connect = false;
1886 struct sctp_chunk *chunk;
1887 long timeo;
1888 int err;
1890 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1891 err = -EINVAL;
1892 goto err;
1895 if (unlikely(!SCTP_SO(&asoc->stream, sinfo->sinfo_stream)->ext)) {
1896 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1897 if (err)
1898 goto err;
1901 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1902 err = -EMSGSIZE;
1903 goto err;
1906 if (asoc->pmtu_pending) {
1907 if (sp->param_flags & SPP_PMTUD_ENABLE)
1908 sctp_assoc_sync_pmtu(asoc);
1909 asoc->pmtu_pending = 0;
1912 if (sctp_wspace(asoc) < (int)msg_len)
1913 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1915 if (sctp_wspace(asoc) <= 0) {
1916 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1917 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1918 if (err)
1919 goto err;
1922 if (sctp_state(asoc, CLOSED)) {
1923 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1924 if (err)
1925 goto err;
1927 if (sp->strm_interleave) {
1928 timeo = sock_sndtimeo(sk, 0);
1929 err = sctp_wait_for_connect(asoc, &timeo);
1930 if (err) {
1931 err = -ESRCH;
1932 goto err;
1934 } else {
1935 wait_connect = true;
1938 pr_debug("%s: we associated primitively\n", __func__);
1941 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1942 if (IS_ERR(datamsg)) {
1943 err = PTR_ERR(datamsg);
1944 goto err;
1947 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1949 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1950 sctp_chunk_hold(chunk);
1951 sctp_set_owner_w(chunk);
1952 chunk->transport = transport;
1955 err = sctp_primitive_SEND(net, asoc, datamsg);
1956 if (err) {
1957 sctp_datamsg_free(datamsg);
1958 goto err;
1961 pr_debug("%s: we sent primitively\n", __func__);
1963 sctp_datamsg_put(datamsg);
1965 if (unlikely(wait_connect)) {
1966 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1967 sctp_wait_for_connect(asoc, &timeo);
1970 err = msg_len;
1972 err:
1973 return err;
1976 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1977 const struct msghdr *msg,
1978 struct sctp_cmsgs *cmsgs)
1980 union sctp_addr *daddr = NULL;
1981 int err;
1983 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1984 int len = msg->msg_namelen;
1986 if (len > sizeof(*daddr))
1987 len = sizeof(*daddr);
1989 daddr = (union sctp_addr *)msg->msg_name;
1991 err = sctp_verify_addr(sk, daddr, len);
1992 if (err)
1993 return ERR_PTR(err);
1996 return daddr;
1999 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2000 struct sctp_sndrcvinfo *sinfo,
2001 struct sctp_cmsgs *cmsgs)
2003 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2004 sinfo->sinfo_stream = asoc->default_stream;
2005 sinfo->sinfo_ppid = asoc->default_ppid;
2006 sinfo->sinfo_context = asoc->default_context;
2007 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2009 if (!cmsgs->prinfo)
2010 sinfo->sinfo_flags = asoc->default_flags;
2013 if (!cmsgs->srinfo && !cmsgs->prinfo)
2014 sinfo->sinfo_timetolive = asoc->default_timetolive;
2016 if (cmsgs->authinfo) {
2017 /* Reuse sinfo_tsn to indicate that authinfo was set and
2018 * sinfo_ssn to save the keyid on tx path.
2020 sinfo->sinfo_tsn = 1;
2021 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2025 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2027 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2028 struct sctp_transport *transport = NULL;
2029 struct sctp_sndrcvinfo _sinfo, *sinfo;
2030 struct sctp_association *asoc;
2031 struct sctp_cmsgs cmsgs;
2032 union sctp_addr *daddr;
2033 bool new = false;
2034 __u16 sflags;
2035 int err;
2037 /* Parse and get snd_info */
2038 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2039 if (err)
2040 goto out;
2042 sinfo = &_sinfo;
2043 sflags = sinfo->sinfo_flags;
2045 /* Get daddr from msg */
2046 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2047 if (IS_ERR(daddr)) {
2048 err = PTR_ERR(daddr);
2049 goto out;
2052 lock_sock(sk);
2054 /* SCTP_SENDALL process */
2055 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2056 list_for_each_entry(asoc, &ep->asocs, asocs) {
2057 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2058 msg_len);
2059 if (err == 0)
2060 continue;
2061 if (err < 0)
2062 goto out_unlock;
2064 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2066 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2067 NULL, sinfo);
2068 if (err < 0)
2069 goto out_unlock;
2071 iov_iter_revert(&msg->msg_iter, err);
2074 goto out_unlock;
2077 /* Get and check or create asoc */
2078 if (daddr) {
2079 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2080 if (asoc) {
2081 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2082 msg_len);
2083 if (err <= 0)
2084 goto out_unlock;
2085 } else {
2086 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2087 &transport);
2088 if (err)
2089 goto out_unlock;
2091 asoc = transport->asoc;
2092 new = true;
2095 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2096 transport = NULL;
2097 } else {
2098 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2099 if (!asoc) {
2100 err = -EPIPE;
2101 goto out_unlock;
2104 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2105 if (err <= 0)
2106 goto out_unlock;
2109 /* Update snd_info with the asoc */
2110 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2112 /* Send msg to the asoc */
2113 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2114 if (err < 0 && err != -ESRCH && new)
2115 sctp_association_free(asoc);
2117 out_unlock:
2118 release_sock(sk);
2119 out:
2120 return sctp_error(sk, msg->msg_flags, err);
2123 /* This is an extended version of skb_pull() that removes the data from the
2124 * start of a skb even when data is spread across the list of skb's in the
2125 * frag_list. len specifies the total amount of data that needs to be removed.
2126 * when 'len' bytes could be removed from the skb, it returns 0.
2127 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2128 * could not be removed.
2130 static int sctp_skb_pull(struct sk_buff *skb, int len)
2132 struct sk_buff *list;
2133 int skb_len = skb_headlen(skb);
2134 int rlen;
2136 if (len <= skb_len) {
2137 __skb_pull(skb, len);
2138 return 0;
2140 len -= skb_len;
2141 __skb_pull(skb, skb_len);
2143 skb_walk_frags(skb, list) {
2144 rlen = sctp_skb_pull(list, len);
2145 skb->len -= (len-rlen);
2146 skb->data_len -= (len-rlen);
2148 if (!rlen)
2149 return 0;
2151 len = rlen;
2154 return len;
2157 /* API 3.1.3 recvmsg() - UDP Style Syntax
2159 * ssize_t recvmsg(int socket, struct msghdr *message,
2160 * int flags);
2162 * socket - the socket descriptor of the endpoint.
2163 * message - pointer to the msghdr structure which contains a single
2164 * user message and possibly some ancillary data.
2166 * See Section 5 for complete description of the data
2167 * structures.
2169 * flags - flags sent or received with the user message, see Section
2170 * 5 for complete description of the flags.
2172 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2173 int noblock, int flags, int *addr_len)
2175 struct sctp_ulpevent *event = NULL;
2176 struct sctp_sock *sp = sctp_sk(sk);
2177 struct sk_buff *skb, *head_skb;
2178 int copied;
2179 int err = 0;
2180 int skb_len;
2182 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2183 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2184 addr_len);
2186 lock_sock(sk);
2188 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2189 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2190 err = -ENOTCONN;
2191 goto out;
2194 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2195 if (!skb)
2196 goto out;
2198 /* Get the total length of the skb including any skb's in the
2199 * frag_list.
2201 skb_len = skb->len;
2203 copied = skb_len;
2204 if (copied > len)
2205 copied = len;
2207 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2209 event = sctp_skb2event(skb);
2211 if (err)
2212 goto out_free;
2214 if (event->chunk && event->chunk->head_skb)
2215 head_skb = event->chunk->head_skb;
2216 else
2217 head_skb = skb;
2218 sock_recv_ts_and_drops(msg, sk, head_skb);
2219 if (sctp_ulpevent_is_notification(event)) {
2220 msg->msg_flags |= MSG_NOTIFICATION;
2221 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2222 } else {
2223 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2226 /* Check if we allow SCTP_NXTINFO. */
2227 if (sp->recvnxtinfo)
2228 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2229 /* Check if we allow SCTP_RCVINFO. */
2230 if (sp->recvrcvinfo)
2231 sctp_ulpevent_read_rcvinfo(event, msg);
2232 /* Check if we allow SCTP_SNDRCVINFO. */
2233 if (sp->subscribe.sctp_data_io_event)
2234 sctp_ulpevent_read_sndrcvinfo(event, msg);
2236 err = copied;
2238 /* If skb's length exceeds the user's buffer, update the skb and
2239 * push it back to the receive_queue so that the next call to
2240 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2242 if (skb_len > copied) {
2243 msg->msg_flags &= ~MSG_EOR;
2244 if (flags & MSG_PEEK)
2245 goto out_free;
2246 sctp_skb_pull(skb, copied);
2247 skb_queue_head(&sk->sk_receive_queue, skb);
2249 /* When only partial message is copied to the user, increase
2250 * rwnd by that amount. If all the data in the skb is read,
2251 * rwnd is updated when the event is freed.
2253 if (!sctp_ulpevent_is_notification(event))
2254 sctp_assoc_rwnd_increase(event->asoc, copied);
2255 goto out;
2256 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2257 (event->msg_flags & MSG_EOR))
2258 msg->msg_flags |= MSG_EOR;
2259 else
2260 msg->msg_flags &= ~MSG_EOR;
2262 out_free:
2263 if (flags & MSG_PEEK) {
2264 /* Release the skb reference acquired after peeking the skb in
2265 * sctp_skb_recv_datagram().
2267 kfree_skb(skb);
2268 } else {
2269 /* Free the event which includes releasing the reference to
2270 * the owner of the skb, freeing the skb and updating the
2271 * rwnd.
2273 sctp_ulpevent_free(event);
2275 out:
2276 release_sock(sk);
2277 return err;
2280 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2282 * This option is a on/off flag. If enabled no SCTP message
2283 * fragmentation will be performed. Instead if a message being sent
2284 * exceeds the current PMTU size, the message will NOT be sent and
2285 * instead a error will be indicated to the user.
2287 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2288 char __user *optval,
2289 unsigned int optlen)
2291 int val;
2293 if (optlen < sizeof(int))
2294 return -EINVAL;
2296 if (get_user(val, (int __user *)optval))
2297 return -EFAULT;
2299 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2301 return 0;
2304 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2305 unsigned int optlen)
2307 struct sctp_association *asoc;
2308 struct sctp_ulpevent *event;
2310 if (optlen > sizeof(struct sctp_event_subscribe))
2311 return -EINVAL;
2312 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2313 return -EFAULT;
2315 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2316 * if there is no data to be sent or retransmit, the stack will
2317 * immediately send up this notification.
2319 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2320 &sctp_sk(sk)->subscribe)) {
2321 asoc = sctp_id2assoc(sk, 0);
2323 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2324 event = sctp_ulpevent_make_sender_dry_event(asoc,
2325 GFP_USER | __GFP_NOWARN);
2326 if (!event)
2327 return -ENOMEM;
2329 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2333 return 0;
2336 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2338 * This socket option is applicable to the UDP-style socket only. When
2339 * set it will cause associations that are idle for more than the
2340 * specified number of seconds to automatically close. An association
2341 * being idle is defined an association that has NOT sent or received
2342 * user data. The special value of '0' indicates that no automatic
2343 * close of any associations should be performed. The option expects an
2344 * integer defining the number of seconds of idle time before an
2345 * association is closed.
2347 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2348 unsigned int optlen)
2350 struct sctp_sock *sp = sctp_sk(sk);
2351 struct net *net = sock_net(sk);
2353 /* Applicable to UDP-style socket only */
2354 if (sctp_style(sk, TCP))
2355 return -EOPNOTSUPP;
2356 if (optlen != sizeof(int))
2357 return -EINVAL;
2358 if (copy_from_user(&sp->autoclose, optval, optlen))
2359 return -EFAULT;
2361 if (sp->autoclose > net->sctp.max_autoclose)
2362 sp->autoclose = net->sctp.max_autoclose;
2364 return 0;
2367 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2369 * Applications can enable or disable heartbeats for any peer address of
2370 * an association, modify an address's heartbeat interval, force a
2371 * heartbeat to be sent immediately, and adjust the address's maximum
2372 * number of retransmissions sent before an address is considered
2373 * unreachable. The following structure is used to access and modify an
2374 * address's parameters:
2376 * struct sctp_paddrparams {
2377 * sctp_assoc_t spp_assoc_id;
2378 * struct sockaddr_storage spp_address;
2379 * uint32_t spp_hbinterval;
2380 * uint16_t spp_pathmaxrxt;
2381 * uint32_t spp_pathmtu;
2382 * uint32_t spp_sackdelay;
2383 * uint32_t spp_flags;
2384 * uint32_t spp_ipv6_flowlabel;
2385 * uint8_t spp_dscp;
2386 * };
2388 * spp_assoc_id - (one-to-many style socket) This is filled in the
2389 * application, and identifies the association for
2390 * this query.
2391 * spp_address - This specifies which address is of interest.
2392 * spp_hbinterval - This contains the value of the heartbeat interval,
2393 * in milliseconds. If a value of zero
2394 * is present in this field then no changes are to
2395 * be made to this parameter.
2396 * spp_pathmaxrxt - This contains the maximum number of
2397 * retransmissions before this address shall be
2398 * considered unreachable. If a value of zero
2399 * is present in this field then no changes are to
2400 * be made to this parameter.
2401 * spp_pathmtu - When Path MTU discovery is disabled the value
2402 * specified here will be the "fixed" path mtu.
2403 * Note that if the spp_address field is empty
2404 * then all associations on this address will
2405 * have this fixed path mtu set upon them.
2407 * spp_sackdelay - When delayed sack is enabled, this value specifies
2408 * the number of milliseconds that sacks will be delayed
2409 * for. This value will apply to all addresses of an
2410 * association if the spp_address field is empty. Note
2411 * also, that if delayed sack is enabled and this
2412 * value is set to 0, no change is made to the last
2413 * recorded delayed sack timer value.
2415 * spp_flags - These flags are used to control various features
2416 * on an association. The flag field may contain
2417 * zero or more of the following options.
2419 * SPP_HB_ENABLE - Enable heartbeats on the
2420 * specified address. Note that if the address
2421 * field is empty all addresses for the association
2422 * have heartbeats enabled upon them.
2424 * SPP_HB_DISABLE - Disable heartbeats on the
2425 * speicifed address. Note that if the address
2426 * field is empty all addresses for the association
2427 * will have their heartbeats disabled. Note also
2428 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2429 * mutually exclusive, only one of these two should
2430 * be specified. Enabling both fields will have
2431 * undetermined results.
2433 * SPP_HB_DEMAND - Request a user initiated heartbeat
2434 * to be made immediately.
2436 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2437 * heartbeat delayis to be set to the value of 0
2438 * milliseconds.
2440 * SPP_PMTUD_ENABLE - This field will enable PMTU
2441 * discovery upon the specified address. Note that
2442 * if the address feild is empty then all addresses
2443 * on the association are effected.
2445 * SPP_PMTUD_DISABLE - This field will disable PMTU
2446 * discovery upon the specified address. Note that
2447 * if the address feild is empty then all addresses
2448 * on the association are effected. Not also that
2449 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2450 * exclusive. Enabling both will have undetermined
2451 * results.
2453 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2454 * on delayed sack. The time specified in spp_sackdelay
2455 * is used to specify the sack delay for this address. Note
2456 * that if spp_address is empty then all addresses will
2457 * enable delayed sack and take on the sack delay
2458 * value specified in spp_sackdelay.
2459 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2460 * off delayed sack. If the spp_address field is blank then
2461 * delayed sack is disabled for the entire association. Note
2462 * also that this field is mutually exclusive to
2463 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2464 * results.
2466 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
2467 * setting of the IPV6 flow label value. The value is
2468 * contained in the spp_ipv6_flowlabel field.
2469 * Upon retrieval, this flag will be set to indicate that
2470 * the spp_ipv6_flowlabel field has a valid value returned.
2471 * If a specific destination address is set (in the
2472 * spp_address field), then the value returned is that of
2473 * the address. If just an association is specified (and
2474 * no address), then the association's default flow label
2475 * is returned. If neither an association nor a destination
2476 * is specified, then the socket's default flow label is
2477 * returned. For non-IPv6 sockets, this flag will be left
2478 * cleared.
2480 * SPP_DSCP: Setting this flag enables the setting of the
2481 * Differentiated Services Code Point (DSCP) value
2482 * associated with either the association or a specific
2483 * address. The value is obtained in the spp_dscp field.
2484 * Upon retrieval, this flag will be set to indicate that
2485 * the spp_dscp field has a valid value returned. If a
2486 * specific destination address is set when called (in the
2487 * spp_address field), then that specific destination
2488 * address's DSCP value is returned. If just an association
2489 * is specified, then the association's default DSCP is
2490 * returned. If neither an association nor a destination is
2491 * specified, then the socket's default DSCP is returned.
2493 * spp_ipv6_flowlabel
2494 * - This field is used in conjunction with the
2495 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
2496 * The 20 least significant bits are used for the flow
2497 * label. This setting has precedence over any IPv6-layer
2498 * setting.
2500 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
2501 * and contains the DSCP. The 6 most significant bits are
2502 * used for the DSCP. This setting has precedence over any
2503 * IPv4- or IPv6- layer setting.
2505 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2506 struct sctp_transport *trans,
2507 struct sctp_association *asoc,
2508 struct sctp_sock *sp,
2509 int hb_change,
2510 int pmtud_change,
2511 int sackdelay_change)
2513 int error;
2515 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2516 struct net *net = sock_net(trans->asoc->base.sk);
2518 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2519 if (error)
2520 return error;
2523 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2524 * this field is ignored. Note also that a value of zero indicates
2525 * the current setting should be left unchanged.
2527 if (params->spp_flags & SPP_HB_ENABLE) {
2529 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2530 * set. This lets us use 0 value when this flag
2531 * is set.
2533 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2534 params->spp_hbinterval = 0;
2536 if (params->spp_hbinterval ||
2537 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2538 if (trans) {
2539 trans->hbinterval =
2540 msecs_to_jiffies(params->spp_hbinterval);
2541 } else if (asoc) {
2542 asoc->hbinterval =
2543 msecs_to_jiffies(params->spp_hbinterval);
2544 } else {
2545 sp->hbinterval = params->spp_hbinterval;
2550 if (hb_change) {
2551 if (trans) {
2552 trans->param_flags =
2553 (trans->param_flags & ~SPP_HB) | hb_change;
2554 } else if (asoc) {
2555 asoc->param_flags =
2556 (asoc->param_flags & ~SPP_HB) | hb_change;
2557 } else {
2558 sp->param_flags =
2559 (sp->param_flags & ~SPP_HB) | hb_change;
2563 /* When Path MTU discovery is disabled the value specified here will
2564 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2565 * include the flag SPP_PMTUD_DISABLE for this field to have any
2566 * effect).
2568 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2569 if (trans) {
2570 trans->pathmtu = params->spp_pathmtu;
2571 sctp_assoc_sync_pmtu(asoc);
2572 } else if (asoc) {
2573 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2574 } else {
2575 sp->pathmtu = params->spp_pathmtu;
2579 if (pmtud_change) {
2580 if (trans) {
2581 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2582 (params->spp_flags & SPP_PMTUD_ENABLE);
2583 trans->param_flags =
2584 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2585 if (update) {
2586 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2587 sctp_assoc_sync_pmtu(asoc);
2589 } else if (asoc) {
2590 asoc->param_flags =
2591 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2592 } else {
2593 sp->param_flags =
2594 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2598 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2599 * value of this field is ignored. Note also that a value of zero
2600 * indicates the current setting should be left unchanged.
2602 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2603 if (trans) {
2604 trans->sackdelay =
2605 msecs_to_jiffies(params->spp_sackdelay);
2606 } else if (asoc) {
2607 asoc->sackdelay =
2608 msecs_to_jiffies(params->spp_sackdelay);
2609 } else {
2610 sp->sackdelay = params->spp_sackdelay;
2614 if (sackdelay_change) {
2615 if (trans) {
2616 trans->param_flags =
2617 (trans->param_flags & ~SPP_SACKDELAY) |
2618 sackdelay_change;
2619 } else if (asoc) {
2620 asoc->param_flags =
2621 (asoc->param_flags & ~SPP_SACKDELAY) |
2622 sackdelay_change;
2623 } else {
2624 sp->param_flags =
2625 (sp->param_flags & ~SPP_SACKDELAY) |
2626 sackdelay_change;
2630 /* Note that a value of zero indicates the current setting should be
2631 left unchanged.
2633 if (params->spp_pathmaxrxt) {
2634 if (trans) {
2635 trans->pathmaxrxt = params->spp_pathmaxrxt;
2636 } else if (asoc) {
2637 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2638 } else {
2639 sp->pathmaxrxt = params->spp_pathmaxrxt;
2643 if (params->spp_flags & SPP_IPV6_FLOWLABEL) {
2644 if (trans) {
2645 if (trans->ipaddr.sa.sa_family == AF_INET6) {
2646 trans->flowlabel = params->spp_ipv6_flowlabel &
2647 SCTP_FLOWLABEL_VAL_MASK;
2648 trans->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2650 } else if (asoc) {
2651 struct sctp_transport *t;
2653 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2654 transports) {
2655 if (t->ipaddr.sa.sa_family != AF_INET6)
2656 continue;
2657 t->flowlabel = params->spp_ipv6_flowlabel &
2658 SCTP_FLOWLABEL_VAL_MASK;
2659 t->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2661 asoc->flowlabel = params->spp_ipv6_flowlabel &
2662 SCTP_FLOWLABEL_VAL_MASK;
2663 asoc->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2664 } else if (sctp_opt2sk(sp)->sk_family == AF_INET6) {
2665 sp->flowlabel = params->spp_ipv6_flowlabel &
2666 SCTP_FLOWLABEL_VAL_MASK;
2667 sp->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2671 if (params->spp_flags & SPP_DSCP) {
2672 if (trans) {
2673 trans->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2674 trans->dscp |= SCTP_DSCP_SET_MASK;
2675 } else if (asoc) {
2676 struct sctp_transport *t;
2678 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2679 transports) {
2680 t->dscp = params->spp_dscp &
2681 SCTP_DSCP_VAL_MASK;
2682 t->dscp |= SCTP_DSCP_SET_MASK;
2684 asoc->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2685 asoc->dscp |= SCTP_DSCP_SET_MASK;
2686 } else {
2687 sp->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2688 sp->dscp |= SCTP_DSCP_SET_MASK;
2692 return 0;
2695 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2696 char __user *optval,
2697 unsigned int optlen)
2699 struct sctp_paddrparams params;
2700 struct sctp_transport *trans = NULL;
2701 struct sctp_association *asoc = NULL;
2702 struct sctp_sock *sp = sctp_sk(sk);
2703 int error;
2704 int hb_change, pmtud_change, sackdelay_change;
2706 if (optlen == sizeof(params)) {
2707 if (copy_from_user(&params, optval, optlen))
2708 return -EFAULT;
2709 } else if (optlen == ALIGN(offsetof(struct sctp_paddrparams,
2710 spp_ipv6_flowlabel), 4)) {
2711 if (copy_from_user(&params, optval, optlen))
2712 return -EFAULT;
2713 if (params.spp_flags & (SPP_DSCP | SPP_IPV6_FLOWLABEL))
2714 return -EINVAL;
2715 } else {
2716 return -EINVAL;
2719 /* Validate flags and value parameters. */
2720 hb_change = params.spp_flags & SPP_HB;
2721 pmtud_change = params.spp_flags & SPP_PMTUD;
2722 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2724 if (hb_change == SPP_HB ||
2725 pmtud_change == SPP_PMTUD ||
2726 sackdelay_change == SPP_SACKDELAY ||
2727 params.spp_sackdelay > 500 ||
2728 (params.spp_pathmtu &&
2729 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2730 return -EINVAL;
2732 /* If an address other than INADDR_ANY is specified, and
2733 * no transport is found, then the request is invalid.
2735 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2736 trans = sctp_addr_id2transport(sk, &params.spp_address,
2737 params.spp_assoc_id);
2738 if (!trans)
2739 return -EINVAL;
2742 /* Get association, if assoc_id != 0 and the socket is a one
2743 * to many style socket, and an association was not found, then
2744 * the id was invalid.
2746 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2747 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2748 return -EINVAL;
2750 /* Heartbeat demand can only be sent on a transport or
2751 * association, but not a socket.
2753 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2754 return -EINVAL;
2756 /* Process parameters. */
2757 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2758 hb_change, pmtud_change,
2759 sackdelay_change);
2761 if (error)
2762 return error;
2764 /* If changes are for association, also apply parameters to each
2765 * transport.
2767 if (!trans && asoc) {
2768 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2769 transports) {
2770 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2771 hb_change, pmtud_change,
2772 sackdelay_change);
2776 return 0;
2779 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2781 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2784 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2786 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2790 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2792 * This option will effect the way delayed acks are performed. This
2793 * option allows you to get or set the delayed ack time, in
2794 * milliseconds. It also allows changing the delayed ack frequency.
2795 * Changing the frequency to 1 disables the delayed sack algorithm. If
2796 * the assoc_id is 0, then this sets or gets the endpoints default
2797 * values. If the assoc_id field is non-zero, then the set or get
2798 * effects the specified association for the one to many model (the
2799 * assoc_id field is ignored by the one to one model). Note that if
2800 * sack_delay or sack_freq are 0 when setting this option, then the
2801 * current values will remain unchanged.
2803 * struct sctp_sack_info {
2804 * sctp_assoc_t sack_assoc_id;
2805 * uint32_t sack_delay;
2806 * uint32_t sack_freq;
2807 * };
2809 * sack_assoc_id - This parameter, indicates which association the user
2810 * is performing an action upon. Note that if this field's value is
2811 * zero then the endpoints default value is changed (effecting future
2812 * associations only).
2814 * sack_delay - This parameter contains the number of milliseconds that
2815 * the user is requesting the delayed ACK timer be set to. Note that
2816 * this value is defined in the standard to be between 200 and 500
2817 * milliseconds.
2819 * sack_freq - This parameter contains the number of packets that must
2820 * be received before a sack is sent without waiting for the delay
2821 * timer to expire. The default value for this is 2, setting this
2822 * value to 1 will disable the delayed sack algorithm.
2825 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2826 char __user *optval, unsigned int optlen)
2828 struct sctp_sack_info params;
2829 struct sctp_transport *trans = NULL;
2830 struct sctp_association *asoc = NULL;
2831 struct sctp_sock *sp = sctp_sk(sk);
2833 if (optlen == sizeof(struct sctp_sack_info)) {
2834 if (copy_from_user(&params, optval, optlen))
2835 return -EFAULT;
2837 if (params.sack_delay == 0 && params.sack_freq == 0)
2838 return 0;
2839 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2840 pr_warn_ratelimited(DEPRECATED
2841 "%s (pid %d) "
2842 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2843 "Use struct sctp_sack_info instead\n",
2844 current->comm, task_pid_nr(current));
2845 if (copy_from_user(&params, optval, optlen))
2846 return -EFAULT;
2848 if (params.sack_delay == 0)
2849 params.sack_freq = 1;
2850 else
2851 params.sack_freq = 0;
2852 } else
2853 return -EINVAL;
2855 /* Validate value parameter. */
2856 if (params.sack_delay > 500)
2857 return -EINVAL;
2859 /* Get association, if sack_assoc_id != 0 and the socket is a one
2860 * to many style socket, and an association was not found, then
2861 * the id was invalid.
2863 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2864 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2865 return -EINVAL;
2867 if (params.sack_delay) {
2868 if (asoc) {
2869 asoc->sackdelay =
2870 msecs_to_jiffies(params.sack_delay);
2871 asoc->param_flags =
2872 sctp_spp_sackdelay_enable(asoc->param_flags);
2873 } else {
2874 sp->sackdelay = params.sack_delay;
2875 sp->param_flags =
2876 sctp_spp_sackdelay_enable(sp->param_flags);
2880 if (params.sack_freq == 1) {
2881 if (asoc) {
2882 asoc->param_flags =
2883 sctp_spp_sackdelay_disable(asoc->param_flags);
2884 } else {
2885 sp->param_flags =
2886 sctp_spp_sackdelay_disable(sp->param_flags);
2888 } else if (params.sack_freq > 1) {
2889 if (asoc) {
2890 asoc->sackfreq = params.sack_freq;
2891 asoc->param_flags =
2892 sctp_spp_sackdelay_enable(asoc->param_flags);
2893 } else {
2894 sp->sackfreq = params.sack_freq;
2895 sp->param_flags =
2896 sctp_spp_sackdelay_enable(sp->param_flags);
2900 /* If change is for association, also apply to each transport. */
2901 if (asoc) {
2902 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2903 transports) {
2904 if (params.sack_delay) {
2905 trans->sackdelay =
2906 msecs_to_jiffies(params.sack_delay);
2907 trans->param_flags =
2908 sctp_spp_sackdelay_enable(trans->param_flags);
2910 if (params.sack_freq == 1) {
2911 trans->param_flags =
2912 sctp_spp_sackdelay_disable(trans->param_flags);
2913 } else if (params.sack_freq > 1) {
2914 trans->sackfreq = params.sack_freq;
2915 trans->param_flags =
2916 sctp_spp_sackdelay_enable(trans->param_flags);
2921 return 0;
2924 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2926 * Applications can specify protocol parameters for the default association
2927 * initialization. The option name argument to setsockopt() and getsockopt()
2928 * is SCTP_INITMSG.
2930 * Setting initialization parameters is effective only on an unconnected
2931 * socket (for UDP-style sockets only future associations are effected
2932 * by the change). With TCP-style sockets, this option is inherited by
2933 * sockets derived from a listener socket.
2935 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2937 struct sctp_initmsg sinit;
2938 struct sctp_sock *sp = sctp_sk(sk);
2940 if (optlen != sizeof(struct sctp_initmsg))
2941 return -EINVAL;
2942 if (copy_from_user(&sinit, optval, optlen))
2943 return -EFAULT;
2945 if (sinit.sinit_num_ostreams)
2946 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2947 if (sinit.sinit_max_instreams)
2948 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2949 if (sinit.sinit_max_attempts)
2950 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2951 if (sinit.sinit_max_init_timeo)
2952 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2954 return 0;
2958 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2960 * Applications that wish to use the sendto() system call may wish to
2961 * specify a default set of parameters that would normally be supplied
2962 * through the inclusion of ancillary data. This socket option allows
2963 * such an application to set the default sctp_sndrcvinfo structure.
2964 * The application that wishes to use this socket option simply passes
2965 * in to this call the sctp_sndrcvinfo structure defined in Section
2966 * 5.2.2) The input parameters accepted by this call include
2967 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2968 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2969 * to this call if the caller is using the UDP model.
2971 static int sctp_setsockopt_default_send_param(struct sock *sk,
2972 char __user *optval,
2973 unsigned int optlen)
2975 struct sctp_sock *sp = sctp_sk(sk);
2976 struct sctp_association *asoc;
2977 struct sctp_sndrcvinfo info;
2979 if (optlen != sizeof(info))
2980 return -EINVAL;
2981 if (copy_from_user(&info, optval, optlen))
2982 return -EFAULT;
2983 if (info.sinfo_flags &
2984 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
2985 SCTP_ABORT | SCTP_EOF))
2986 return -EINVAL;
2988 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2989 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2990 return -EINVAL;
2991 if (asoc) {
2992 asoc->default_stream = info.sinfo_stream;
2993 asoc->default_flags = info.sinfo_flags;
2994 asoc->default_ppid = info.sinfo_ppid;
2995 asoc->default_context = info.sinfo_context;
2996 asoc->default_timetolive = info.sinfo_timetolive;
2997 } else {
2998 sp->default_stream = info.sinfo_stream;
2999 sp->default_flags = info.sinfo_flags;
3000 sp->default_ppid = info.sinfo_ppid;
3001 sp->default_context = info.sinfo_context;
3002 sp->default_timetolive = info.sinfo_timetolive;
3005 return 0;
3008 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
3009 * (SCTP_DEFAULT_SNDINFO)
3011 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
3012 char __user *optval,
3013 unsigned int optlen)
3015 struct sctp_sock *sp = sctp_sk(sk);
3016 struct sctp_association *asoc;
3017 struct sctp_sndinfo info;
3019 if (optlen != sizeof(info))
3020 return -EINVAL;
3021 if (copy_from_user(&info, optval, optlen))
3022 return -EFAULT;
3023 if (info.snd_flags &
3024 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3025 SCTP_ABORT | SCTP_EOF))
3026 return -EINVAL;
3028 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
3029 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
3030 return -EINVAL;
3031 if (asoc) {
3032 asoc->default_stream = info.snd_sid;
3033 asoc->default_flags = info.snd_flags;
3034 asoc->default_ppid = info.snd_ppid;
3035 asoc->default_context = info.snd_context;
3036 } else {
3037 sp->default_stream = info.snd_sid;
3038 sp->default_flags = info.snd_flags;
3039 sp->default_ppid = info.snd_ppid;
3040 sp->default_context = info.snd_context;
3043 return 0;
3046 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3048 * Requests that the local SCTP stack use the enclosed peer address as
3049 * the association primary. The enclosed address must be one of the
3050 * association peer's addresses.
3052 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
3053 unsigned int optlen)
3055 struct sctp_prim prim;
3056 struct sctp_transport *trans;
3057 struct sctp_af *af;
3058 int err;
3060 if (optlen != sizeof(struct sctp_prim))
3061 return -EINVAL;
3063 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3064 return -EFAULT;
3066 /* Allow security module to validate address but need address len. */
3067 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
3068 if (!af)
3069 return -EINVAL;
3071 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
3072 (struct sockaddr *)&prim.ssp_addr,
3073 af->sockaddr_len);
3074 if (err)
3075 return err;
3077 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
3078 if (!trans)
3079 return -EINVAL;
3081 sctp_assoc_set_primary(trans->asoc, trans);
3083 return 0;
3087 * 7.1.5 SCTP_NODELAY
3089 * Turn on/off any Nagle-like algorithm. This means that packets are
3090 * generally sent as soon as possible and no unnecessary delays are
3091 * introduced, at the cost of more packets in the network. Expects an
3092 * integer boolean flag.
3094 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3095 unsigned int optlen)
3097 int val;
3099 if (optlen < sizeof(int))
3100 return -EINVAL;
3101 if (get_user(val, (int __user *)optval))
3102 return -EFAULT;
3104 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3105 return 0;
3110 * 7.1.1 SCTP_RTOINFO
3112 * The protocol parameters used to initialize and bound retransmission
3113 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3114 * and modify these parameters.
3115 * All parameters are time values, in milliseconds. A value of 0, when
3116 * modifying the parameters, indicates that the current value should not
3117 * be changed.
3120 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3122 struct sctp_rtoinfo rtoinfo;
3123 struct sctp_association *asoc;
3124 unsigned long rto_min, rto_max;
3125 struct sctp_sock *sp = sctp_sk(sk);
3127 if (optlen != sizeof (struct sctp_rtoinfo))
3128 return -EINVAL;
3130 if (copy_from_user(&rtoinfo, optval, optlen))
3131 return -EFAULT;
3133 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3135 /* Set the values to the specific association */
3136 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3137 return -EINVAL;
3139 rto_max = rtoinfo.srto_max;
3140 rto_min = rtoinfo.srto_min;
3142 if (rto_max)
3143 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3144 else
3145 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3147 if (rto_min)
3148 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3149 else
3150 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3152 if (rto_min > rto_max)
3153 return -EINVAL;
3155 if (asoc) {
3156 if (rtoinfo.srto_initial != 0)
3157 asoc->rto_initial =
3158 msecs_to_jiffies(rtoinfo.srto_initial);
3159 asoc->rto_max = rto_max;
3160 asoc->rto_min = rto_min;
3161 } else {
3162 /* If there is no association or the association-id = 0
3163 * set the values to the endpoint.
3165 if (rtoinfo.srto_initial != 0)
3166 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3167 sp->rtoinfo.srto_max = rto_max;
3168 sp->rtoinfo.srto_min = rto_min;
3171 return 0;
3176 * 7.1.2 SCTP_ASSOCINFO
3178 * This option is used to tune the maximum retransmission attempts
3179 * of the association.
3180 * Returns an error if the new association retransmission value is
3181 * greater than the sum of the retransmission value of the peer.
3182 * See [SCTP] for more information.
3185 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3188 struct sctp_assocparams assocparams;
3189 struct sctp_association *asoc;
3191 if (optlen != sizeof(struct sctp_assocparams))
3192 return -EINVAL;
3193 if (copy_from_user(&assocparams, optval, optlen))
3194 return -EFAULT;
3196 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3198 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3199 return -EINVAL;
3201 /* Set the values to the specific association */
3202 if (asoc) {
3203 if (assocparams.sasoc_asocmaxrxt != 0) {
3204 __u32 path_sum = 0;
3205 int paths = 0;
3206 struct sctp_transport *peer_addr;
3208 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3209 transports) {
3210 path_sum += peer_addr->pathmaxrxt;
3211 paths++;
3214 /* Only validate asocmaxrxt if we have more than
3215 * one path/transport. We do this because path
3216 * retransmissions are only counted when we have more
3217 * then one path.
3219 if (paths > 1 &&
3220 assocparams.sasoc_asocmaxrxt > path_sum)
3221 return -EINVAL;
3223 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3226 if (assocparams.sasoc_cookie_life != 0)
3227 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3228 } else {
3229 /* Set the values to the endpoint */
3230 struct sctp_sock *sp = sctp_sk(sk);
3232 if (assocparams.sasoc_asocmaxrxt != 0)
3233 sp->assocparams.sasoc_asocmaxrxt =
3234 assocparams.sasoc_asocmaxrxt;
3235 if (assocparams.sasoc_cookie_life != 0)
3236 sp->assocparams.sasoc_cookie_life =
3237 assocparams.sasoc_cookie_life;
3239 return 0;
3243 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3245 * This socket option is a boolean flag which turns on or off mapped V4
3246 * addresses. If this option is turned on and the socket is type
3247 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3248 * If this option is turned off, then no mapping will be done of V4
3249 * addresses and a user will receive both PF_INET6 and PF_INET type
3250 * addresses on the socket.
3252 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3254 int val;
3255 struct sctp_sock *sp = sctp_sk(sk);
3257 if (optlen < sizeof(int))
3258 return -EINVAL;
3259 if (get_user(val, (int __user *)optval))
3260 return -EFAULT;
3261 if (val)
3262 sp->v4mapped = 1;
3263 else
3264 sp->v4mapped = 0;
3266 return 0;
3270 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3271 * This option will get or set the maximum size to put in any outgoing
3272 * SCTP DATA chunk. If a message is larger than this size it will be
3273 * fragmented by SCTP into the specified size. Note that the underlying
3274 * SCTP implementation may fragment into smaller sized chunks when the
3275 * PMTU of the underlying association is smaller than the value set by
3276 * the user. The default value for this option is '0' which indicates
3277 * the user is NOT limiting fragmentation and only the PMTU will effect
3278 * SCTP's choice of DATA chunk size. Note also that values set larger
3279 * than the maximum size of an IP datagram will effectively let SCTP
3280 * control fragmentation (i.e. the same as setting this option to 0).
3282 * The following structure is used to access and modify this parameter:
3284 * struct sctp_assoc_value {
3285 * sctp_assoc_t assoc_id;
3286 * uint32_t assoc_value;
3287 * };
3289 * assoc_id: This parameter is ignored for one-to-one style sockets.
3290 * For one-to-many style sockets this parameter indicates which
3291 * association the user is performing an action upon. Note that if
3292 * this field's value is zero then the endpoints default value is
3293 * changed (effecting future associations only).
3294 * assoc_value: This parameter specifies the maximum size in bytes.
3296 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3298 struct sctp_sock *sp = sctp_sk(sk);
3299 struct sctp_assoc_value params;
3300 struct sctp_association *asoc;
3301 int val;
3303 if (optlen == sizeof(int)) {
3304 pr_warn_ratelimited(DEPRECATED
3305 "%s (pid %d) "
3306 "Use of int in maxseg socket option.\n"
3307 "Use struct sctp_assoc_value instead\n",
3308 current->comm, task_pid_nr(current));
3309 if (copy_from_user(&val, optval, optlen))
3310 return -EFAULT;
3311 params.assoc_id = 0;
3312 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3313 if (copy_from_user(&params, optval, optlen))
3314 return -EFAULT;
3315 val = params.assoc_value;
3316 } else {
3317 return -EINVAL;
3320 asoc = sctp_id2assoc(sk, params.assoc_id);
3322 if (val) {
3323 int min_len, max_len;
3324 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3325 sizeof(struct sctp_data_chunk);
3327 min_len = sctp_min_frag_point(sp, datasize);
3328 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3330 if (val < min_len || val > max_len)
3331 return -EINVAL;
3334 if (asoc) {
3335 asoc->user_frag = val;
3336 sctp_assoc_update_frag_point(asoc);
3337 } else {
3338 if (params.assoc_id && sctp_style(sk, UDP))
3339 return -EINVAL;
3340 sp->user_frag = val;
3343 return 0;
3348 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3350 * Requests that the peer mark the enclosed address as the association
3351 * primary. The enclosed address must be one of the association's
3352 * locally bound addresses. The following structure is used to make a
3353 * set primary request:
3355 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3356 unsigned int optlen)
3358 struct net *net = sock_net(sk);
3359 struct sctp_sock *sp;
3360 struct sctp_association *asoc = NULL;
3361 struct sctp_setpeerprim prim;
3362 struct sctp_chunk *chunk;
3363 struct sctp_af *af;
3364 int err;
3366 sp = sctp_sk(sk);
3368 if (!net->sctp.addip_enable)
3369 return -EPERM;
3371 if (optlen != sizeof(struct sctp_setpeerprim))
3372 return -EINVAL;
3374 if (copy_from_user(&prim, optval, optlen))
3375 return -EFAULT;
3377 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3378 if (!asoc)
3379 return -EINVAL;
3381 if (!asoc->peer.asconf_capable)
3382 return -EPERM;
3384 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3385 return -EPERM;
3387 if (!sctp_state(asoc, ESTABLISHED))
3388 return -ENOTCONN;
3390 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3391 if (!af)
3392 return -EINVAL;
3394 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3395 return -EADDRNOTAVAIL;
3397 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3398 return -EADDRNOTAVAIL;
3400 /* Allow security module to validate address. */
3401 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3402 (struct sockaddr *)&prim.sspp_addr,
3403 af->sockaddr_len);
3404 if (err)
3405 return err;
3407 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3408 chunk = sctp_make_asconf_set_prim(asoc,
3409 (union sctp_addr *)&prim.sspp_addr);
3410 if (!chunk)
3411 return -ENOMEM;
3413 err = sctp_send_asconf(asoc, chunk);
3415 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3417 return err;
3420 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3421 unsigned int optlen)
3423 struct sctp_setadaptation adaptation;
3425 if (optlen != sizeof(struct sctp_setadaptation))
3426 return -EINVAL;
3427 if (copy_from_user(&adaptation, optval, optlen))
3428 return -EFAULT;
3430 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3432 return 0;
3436 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3438 * The context field in the sctp_sndrcvinfo structure is normally only
3439 * used when a failed message is retrieved holding the value that was
3440 * sent down on the actual send call. This option allows the setting of
3441 * a default context on an association basis that will be received on
3442 * reading messages from the peer. This is especially helpful in the
3443 * one-2-many model for an application to keep some reference to an
3444 * internal state machine that is processing messages on the
3445 * association. Note that the setting of this value only effects
3446 * received messages from the peer and does not effect the value that is
3447 * saved with outbound messages.
3449 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3450 unsigned int optlen)
3452 struct sctp_assoc_value params;
3453 struct sctp_sock *sp;
3454 struct sctp_association *asoc;
3456 if (optlen != sizeof(struct sctp_assoc_value))
3457 return -EINVAL;
3458 if (copy_from_user(&params, optval, optlen))
3459 return -EFAULT;
3461 sp = sctp_sk(sk);
3463 if (params.assoc_id != 0) {
3464 asoc = sctp_id2assoc(sk, params.assoc_id);
3465 if (!asoc)
3466 return -EINVAL;
3467 asoc->default_rcv_context = params.assoc_value;
3468 } else {
3469 sp->default_rcv_context = params.assoc_value;
3472 return 0;
3476 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3478 * This options will at a minimum specify if the implementation is doing
3479 * fragmented interleave. Fragmented interleave, for a one to many
3480 * socket, is when subsequent calls to receive a message may return
3481 * parts of messages from different associations. Some implementations
3482 * may allow you to turn this value on or off. If so, when turned off,
3483 * no fragment interleave will occur (which will cause a head of line
3484 * blocking amongst multiple associations sharing the same one to many
3485 * socket). When this option is turned on, then each receive call may
3486 * come from a different association (thus the user must receive data
3487 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3488 * association each receive belongs to.
3490 * This option takes a boolean value. A non-zero value indicates that
3491 * fragmented interleave is on. A value of zero indicates that
3492 * fragmented interleave is off.
3494 * Note that it is important that an implementation that allows this
3495 * option to be turned on, have it off by default. Otherwise an unaware
3496 * application using the one to many model may become confused and act
3497 * incorrectly.
3499 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3500 char __user *optval,
3501 unsigned int optlen)
3503 int val;
3505 if (optlen != sizeof(int))
3506 return -EINVAL;
3507 if (get_user(val, (int __user *)optval))
3508 return -EFAULT;
3510 sctp_sk(sk)->frag_interleave = !!val;
3512 if (!sctp_sk(sk)->frag_interleave)
3513 sctp_sk(sk)->strm_interleave = 0;
3515 return 0;
3519 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3520 * (SCTP_PARTIAL_DELIVERY_POINT)
3522 * This option will set or get the SCTP partial delivery point. This
3523 * point is the size of a message where the partial delivery API will be
3524 * invoked to help free up rwnd space for the peer. Setting this to a
3525 * lower value will cause partial deliveries to happen more often. The
3526 * calls argument is an integer that sets or gets the partial delivery
3527 * point. Note also that the call will fail if the user attempts to set
3528 * this value larger than the socket receive buffer size.
3530 * Note that any single message having a length smaller than or equal to
3531 * the SCTP partial delivery point will be delivered in one single read
3532 * call as long as the user provided buffer is large enough to hold the
3533 * message.
3535 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3536 char __user *optval,
3537 unsigned int optlen)
3539 u32 val;
3541 if (optlen != sizeof(u32))
3542 return -EINVAL;
3543 if (get_user(val, (int __user *)optval))
3544 return -EFAULT;
3546 /* Note: We double the receive buffer from what the user sets
3547 * it to be, also initial rwnd is based on rcvbuf/2.
3549 if (val > (sk->sk_rcvbuf >> 1))
3550 return -EINVAL;
3552 sctp_sk(sk)->pd_point = val;
3554 return 0; /* is this the right error code? */
3558 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3560 * This option will allow a user to change the maximum burst of packets
3561 * that can be emitted by this association. Note that the default value
3562 * is 4, and some implementations may restrict this setting so that it
3563 * can only be lowered.
3565 * NOTE: This text doesn't seem right. Do this on a socket basis with
3566 * future associations inheriting the socket value.
3568 static int sctp_setsockopt_maxburst(struct sock *sk,
3569 char __user *optval,
3570 unsigned int optlen)
3572 struct sctp_assoc_value params;
3573 struct sctp_sock *sp;
3574 struct sctp_association *asoc;
3575 int val;
3576 int assoc_id = 0;
3578 if (optlen == sizeof(int)) {
3579 pr_warn_ratelimited(DEPRECATED
3580 "%s (pid %d) "
3581 "Use of int in max_burst socket option deprecated.\n"
3582 "Use struct sctp_assoc_value instead\n",
3583 current->comm, task_pid_nr(current));
3584 if (copy_from_user(&val, optval, optlen))
3585 return -EFAULT;
3586 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3587 if (copy_from_user(&params, optval, optlen))
3588 return -EFAULT;
3589 val = params.assoc_value;
3590 assoc_id = params.assoc_id;
3591 } else
3592 return -EINVAL;
3594 sp = sctp_sk(sk);
3596 if (assoc_id != 0) {
3597 asoc = sctp_id2assoc(sk, assoc_id);
3598 if (!asoc)
3599 return -EINVAL;
3600 asoc->max_burst = val;
3601 } else
3602 sp->max_burst = val;
3604 return 0;
3608 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3610 * This set option adds a chunk type that the user is requesting to be
3611 * received only in an authenticated way. Changes to the list of chunks
3612 * will only effect future associations on the socket.
3614 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3615 char __user *optval,
3616 unsigned int optlen)
3618 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3619 struct sctp_authchunk val;
3621 if (!ep->auth_enable)
3622 return -EACCES;
3624 if (optlen != sizeof(struct sctp_authchunk))
3625 return -EINVAL;
3626 if (copy_from_user(&val, optval, optlen))
3627 return -EFAULT;
3629 switch (val.sauth_chunk) {
3630 case SCTP_CID_INIT:
3631 case SCTP_CID_INIT_ACK:
3632 case SCTP_CID_SHUTDOWN_COMPLETE:
3633 case SCTP_CID_AUTH:
3634 return -EINVAL;
3637 /* add this chunk id to the endpoint */
3638 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3642 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3644 * This option gets or sets the list of HMAC algorithms that the local
3645 * endpoint requires the peer to use.
3647 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3648 char __user *optval,
3649 unsigned int optlen)
3651 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3652 struct sctp_hmacalgo *hmacs;
3653 u32 idents;
3654 int err;
3656 if (!ep->auth_enable)
3657 return -EACCES;
3659 if (optlen < sizeof(struct sctp_hmacalgo))
3660 return -EINVAL;
3661 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3662 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3664 hmacs = memdup_user(optval, optlen);
3665 if (IS_ERR(hmacs))
3666 return PTR_ERR(hmacs);
3668 idents = hmacs->shmac_num_idents;
3669 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3670 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3671 err = -EINVAL;
3672 goto out;
3675 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3676 out:
3677 kfree(hmacs);
3678 return err;
3682 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3684 * This option will set a shared secret key which is used to build an
3685 * association shared key.
3687 static int sctp_setsockopt_auth_key(struct sock *sk,
3688 char __user *optval,
3689 unsigned int optlen)
3691 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3692 struct sctp_authkey *authkey;
3693 struct sctp_association *asoc;
3694 int ret;
3696 if (!ep->auth_enable)
3697 return -EACCES;
3699 if (optlen <= sizeof(struct sctp_authkey))
3700 return -EINVAL;
3701 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3702 * this.
3704 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3705 sizeof(struct sctp_authkey));
3707 authkey = memdup_user(optval, optlen);
3708 if (IS_ERR(authkey))
3709 return PTR_ERR(authkey);
3711 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3712 ret = -EINVAL;
3713 goto out;
3716 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3717 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3718 ret = -EINVAL;
3719 goto out;
3722 ret = sctp_auth_set_key(ep, asoc, authkey);
3723 out:
3724 kzfree(authkey);
3725 return ret;
3729 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3731 * This option will get or set the active shared key to be used to build
3732 * the association shared key.
3734 static int sctp_setsockopt_active_key(struct sock *sk,
3735 char __user *optval,
3736 unsigned int optlen)
3738 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3739 struct sctp_authkeyid val;
3740 struct sctp_association *asoc;
3742 if (!ep->auth_enable)
3743 return -EACCES;
3745 if (optlen != sizeof(struct sctp_authkeyid))
3746 return -EINVAL;
3747 if (copy_from_user(&val, optval, optlen))
3748 return -EFAULT;
3750 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3751 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3752 return -EINVAL;
3754 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3758 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3760 * This set option will delete a shared secret key from use.
3762 static int sctp_setsockopt_del_key(struct sock *sk,
3763 char __user *optval,
3764 unsigned int optlen)
3766 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3767 struct sctp_authkeyid val;
3768 struct sctp_association *asoc;
3770 if (!ep->auth_enable)
3771 return -EACCES;
3773 if (optlen != sizeof(struct sctp_authkeyid))
3774 return -EINVAL;
3775 if (copy_from_user(&val, optval, optlen))
3776 return -EFAULT;
3778 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3779 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3780 return -EINVAL;
3782 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3787 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3789 * This set option will deactivate a shared secret key.
3791 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3792 unsigned int optlen)
3794 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3795 struct sctp_authkeyid val;
3796 struct sctp_association *asoc;
3798 if (!ep->auth_enable)
3799 return -EACCES;
3801 if (optlen != sizeof(struct sctp_authkeyid))
3802 return -EINVAL;
3803 if (copy_from_user(&val, optval, optlen))
3804 return -EFAULT;
3806 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3807 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3808 return -EINVAL;
3810 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3814 * 8.1.23 SCTP_AUTO_ASCONF
3816 * This option will enable or disable the use of the automatic generation of
3817 * ASCONF chunks to add and delete addresses to an existing association. Note
3818 * that this option has two caveats namely: a) it only affects sockets that
3819 * are bound to all addresses available to the SCTP stack, and b) the system
3820 * administrator may have an overriding control that turns the ASCONF feature
3821 * off no matter what setting the socket option may have.
3822 * This option expects an integer boolean flag, where a non-zero value turns on
3823 * the option, and a zero value turns off the option.
3824 * Note. In this implementation, socket operation overrides default parameter
3825 * being set by sysctl as well as FreeBSD implementation
3827 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3828 unsigned int optlen)
3830 int val;
3831 struct sctp_sock *sp = sctp_sk(sk);
3833 if (optlen < sizeof(int))
3834 return -EINVAL;
3835 if (get_user(val, (int __user *)optval))
3836 return -EFAULT;
3837 if (!sctp_is_ep_boundall(sk) && val)
3838 return -EINVAL;
3839 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3840 return 0;
3842 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3843 if (val == 0 && sp->do_auto_asconf) {
3844 list_del(&sp->auto_asconf_list);
3845 sp->do_auto_asconf = 0;
3846 } else if (val && !sp->do_auto_asconf) {
3847 list_add_tail(&sp->auto_asconf_list,
3848 &sock_net(sk)->sctp.auto_asconf_splist);
3849 sp->do_auto_asconf = 1;
3851 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3852 return 0;
3856 * SCTP_PEER_ADDR_THLDS
3858 * This option allows us to alter the partially failed threshold for one or all
3859 * transports in an association. See Section 6.1 of:
3860 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3862 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3863 char __user *optval,
3864 unsigned int optlen)
3866 struct sctp_paddrthlds val;
3867 struct sctp_transport *trans;
3868 struct sctp_association *asoc;
3870 if (optlen < sizeof(struct sctp_paddrthlds))
3871 return -EINVAL;
3872 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3873 sizeof(struct sctp_paddrthlds)))
3874 return -EFAULT;
3877 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3878 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3879 if (!asoc)
3880 return -ENOENT;
3881 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3882 transports) {
3883 if (val.spt_pathmaxrxt)
3884 trans->pathmaxrxt = val.spt_pathmaxrxt;
3885 trans->pf_retrans = val.spt_pathpfthld;
3888 if (val.spt_pathmaxrxt)
3889 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3890 asoc->pf_retrans = val.spt_pathpfthld;
3891 } else {
3892 trans = sctp_addr_id2transport(sk, &val.spt_address,
3893 val.spt_assoc_id);
3894 if (!trans)
3895 return -ENOENT;
3897 if (val.spt_pathmaxrxt)
3898 trans->pathmaxrxt = val.spt_pathmaxrxt;
3899 trans->pf_retrans = val.spt_pathpfthld;
3902 return 0;
3905 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3906 char __user *optval,
3907 unsigned int optlen)
3909 int val;
3911 if (optlen < sizeof(int))
3912 return -EINVAL;
3913 if (get_user(val, (int __user *) optval))
3914 return -EFAULT;
3916 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3918 return 0;
3921 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3922 char __user *optval,
3923 unsigned int optlen)
3925 int val;
3927 if (optlen < sizeof(int))
3928 return -EINVAL;
3929 if (get_user(val, (int __user *) optval))
3930 return -EFAULT;
3932 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3934 return 0;
3937 static int sctp_setsockopt_pr_supported(struct sock *sk,
3938 char __user *optval,
3939 unsigned int optlen)
3941 struct sctp_assoc_value params;
3943 if (optlen != sizeof(params))
3944 return -EINVAL;
3946 if (copy_from_user(&params, optval, optlen))
3947 return -EFAULT;
3949 sctp_sk(sk)->ep->prsctp_enable = !!params.assoc_value;
3951 return 0;
3954 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3955 char __user *optval,
3956 unsigned int optlen)
3958 struct sctp_default_prinfo info;
3959 struct sctp_association *asoc;
3960 int retval = -EINVAL;
3962 if (optlen != sizeof(info))
3963 goto out;
3965 if (copy_from_user(&info, optval, sizeof(info))) {
3966 retval = -EFAULT;
3967 goto out;
3970 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3971 goto out;
3973 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3974 info.pr_value = 0;
3976 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3977 if (asoc) {
3978 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3979 asoc->default_timetolive = info.pr_value;
3980 } else if (!info.pr_assoc_id) {
3981 struct sctp_sock *sp = sctp_sk(sk);
3983 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
3984 sp->default_timetolive = info.pr_value;
3985 } else {
3986 goto out;
3989 retval = 0;
3991 out:
3992 return retval;
3995 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
3996 char __user *optval,
3997 unsigned int optlen)
3999 struct sctp_assoc_value params;
4000 struct sctp_association *asoc;
4001 int retval = -EINVAL;
4003 if (optlen != sizeof(params))
4004 goto out;
4006 if (copy_from_user(&params, optval, optlen)) {
4007 retval = -EFAULT;
4008 goto out;
4011 asoc = sctp_id2assoc(sk, params.assoc_id);
4012 if (asoc) {
4013 asoc->reconf_enable = !!params.assoc_value;
4014 } else if (!params.assoc_id) {
4015 struct sctp_sock *sp = sctp_sk(sk);
4017 sp->ep->reconf_enable = !!params.assoc_value;
4018 } else {
4019 goto out;
4022 retval = 0;
4024 out:
4025 return retval;
4028 static int sctp_setsockopt_enable_strreset(struct sock *sk,
4029 char __user *optval,
4030 unsigned int optlen)
4032 struct sctp_assoc_value params;
4033 struct sctp_association *asoc;
4034 int retval = -EINVAL;
4036 if (optlen != sizeof(params))
4037 goto out;
4039 if (copy_from_user(&params, optval, optlen)) {
4040 retval = -EFAULT;
4041 goto out;
4044 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
4045 goto out;
4047 asoc = sctp_id2assoc(sk, params.assoc_id);
4048 if (asoc) {
4049 asoc->strreset_enable = params.assoc_value;
4050 } else if (!params.assoc_id) {
4051 struct sctp_sock *sp = sctp_sk(sk);
4053 sp->ep->strreset_enable = params.assoc_value;
4054 } else {
4055 goto out;
4058 retval = 0;
4060 out:
4061 return retval;
4064 static int sctp_setsockopt_reset_streams(struct sock *sk,
4065 char __user *optval,
4066 unsigned int optlen)
4068 struct sctp_reset_streams *params;
4069 struct sctp_association *asoc;
4070 int retval = -EINVAL;
4072 if (optlen < sizeof(*params))
4073 return -EINVAL;
4074 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4075 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4076 sizeof(__u16) * sizeof(*params));
4078 params = memdup_user(optval, optlen);
4079 if (IS_ERR(params))
4080 return PTR_ERR(params);
4082 if (params->srs_number_streams * sizeof(__u16) >
4083 optlen - sizeof(*params))
4084 goto out;
4086 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4087 if (!asoc)
4088 goto out;
4090 retval = sctp_send_reset_streams(asoc, params);
4092 out:
4093 kfree(params);
4094 return retval;
4097 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4098 char __user *optval,
4099 unsigned int optlen)
4101 struct sctp_association *asoc;
4102 sctp_assoc_t associd;
4103 int retval = -EINVAL;
4105 if (optlen != sizeof(associd))
4106 goto out;
4108 if (copy_from_user(&associd, optval, optlen)) {
4109 retval = -EFAULT;
4110 goto out;
4113 asoc = sctp_id2assoc(sk, associd);
4114 if (!asoc)
4115 goto out;
4117 retval = sctp_send_reset_assoc(asoc);
4119 out:
4120 return retval;
4123 static int sctp_setsockopt_add_streams(struct sock *sk,
4124 char __user *optval,
4125 unsigned int optlen)
4127 struct sctp_association *asoc;
4128 struct sctp_add_streams params;
4129 int retval = -EINVAL;
4131 if (optlen != sizeof(params))
4132 goto out;
4134 if (copy_from_user(&params, optval, optlen)) {
4135 retval = -EFAULT;
4136 goto out;
4139 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4140 if (!asoc)
4141 goto out;
4143 retval = sctp_send_add_streams(asoc, &params);
4145 out:
4146 return retval;
4149 static int sctp_setsockopt_scheduler(struct sock *sk,
4150 char __user *optval,
4151 unsigned int optlen)
4153 struct sctp_association *asoc;
4154 struct sctp_assoc_value params;
4155 int retval = -EINVAL;
4157 if (optlen < sizeof(params))
4158 goto out;
4160 optlen = sizeof(params);
4161 if (copy_from_user(&params, optval, optlen)) {
4162 retval = -EFAULT;
4163 goto out;
4166 if (params.assoc_value > SCTP_SS_MAX)
4167 goto out;
4169 asoc = sctp_id2assoc(sk, params.assoc_id);
4170 if (!asoc)
4171 goto out;
4173 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4175 out:
4176 return retval;
4179 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4180 char __user *optval,
4181 unsigned int optlen)
4183 struct sctp_association *asoc;
4184 struct sctp_stream_value params;
4185 int retval = -EINVAL;
4187 if (optlen < sizeof(params))
4188 goto out;
4190 optlen = sizeof(params);
4191 if (copy_from_user(&params, optval, optlen)) {
4192 retval = -EFAULT;
4193 goto out;
4196 asoc = sctp_id2assoc(sk, params.assoc_id);
4197 if (!asoc)
4198 goto out;
4200 retval = sctp_sched_set_value(asoc, params.stream_id,
4201 params.stream_value, GFP_KERNEL);
4203 out:
4204 return retval;
4207 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4208 char __user *optval,
4209 unsigned int optlen)
4211 struct sctp_sock *sp = sctp_sk(sk);
4212 struct net *net = sock_net(sk);
4213 struct sctp_assoc_value params;
4214 int retval = -EINVAL;
4216 if (optlen < sizeof(params))
4217 goto out;
4219 optlen = sizeof(params);
4220 if (copy_from_user(&params, optval, optlen)) {
4221 retval = -EFAULT;
4222 goto out;
4225 if (params.assoc_id)
4226 goto out;
4228 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4229 retval = -EPERM;
4230 goto out;
4233 sp->strm_interleave = !!params.assoc_value;
4235 retval = 0;
4237 out:
4238 return retval;
4241 static int sctp_setsockopt_reuse_port(struct sock *sk, char __user *optval,
4242 unsigned int optlen)
4244 int val;
4246 if (!sctp_style(sk, TCP))
4247 return -EOPNOTSUPP;
4249 if (sctp_sk(sk)->ep->base.bind_addr.port)
4250 return -EFAULT;
4252 if (optlen < sizeof(int))
4253 return -EINVAL;
4255 if (get_user(val, (int __user *)optval))
4256 return -EFAULT;
4258 sctp_sk(sk)->reuse = !!val;
4260 return 0;
4263 /* API 6.2 setsockopt(), getsockopt()
4265 * Applications use setsockopt() and getsockopt() to set or retrieve
4266 * socket options. Socket options are used to change the default
4267 * behavior of sockets calls. They are described in Section 7.
4269 * The syntax is:
4271 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4272 * int __user *optlen);
4273 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4274 * int optlen);
4276 * sd - the socket descript.
4277 * level - set to IPPROTO_SCTP for all SCTP options.
4278 * optname - the option name.
4279 * optval - the buffer to store the value of the option.
4280 * optlen - the size of the buffer.
4282 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4283 char __user *optval, unsigned int optlen)
4285 int retval = 0;
4287 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4289 /* I can hardly begin to describe how wrong this is. This is
4290 * so broken as to be worse than useless. The API draft
4291 * REALLY is NOT helpful here... I am not convinced that the
4292 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4293 * are at all well-founded.
4295 if (level != SOL_SCTP) {
4296 struct sctp_af *af = sctp_sk(sk)->pf->af;
4297 retval = af->setsockopt(sk, level, optname, optval, optlen);
4298 goto out_nounlock;
4301 lock_sock(sk);
4303 switch (optname) {
4304 case SCTP_SOCKOPT_BINDX_ADD:
4305 /* 'optlen' is the size of the addresses buffer. */
4306 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4307 optlen, SCTP_BINDX_ADD_ADDR);
4308 break;
4310 case SCTP_SOCKOPT_BINDX_REM:
4311 /* 'optlen' is the size of the addresses buffer. */
4312 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4313 optlen, SCTP_BINDX_REM_ADDR);
4314 break;
4316 case SCTP_SOCKOPT_CONNECTX_OLD:
4317 /* 'optlen' is the size of the addresses buffer. */
4318 retval = sctp_setsockopt_connectx_old(sk,
4319 (struct sockaddr __user *)optval,
4320 optlen);
4321 break;
4323 case SCTP_SOCKOPT_CONNECTX:
4324 /* 'optlen' is the size of the addresses buffer. */
4325 retval = sctp_setsockopt_connectx(sk,
4326 (struct sockaddr __user *)optval,
4327 optlen);
4328 break;
4330 case SCTP_DISABLE_FRAGMENTS:
4331 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4332 break;
4334 case SCTP_EVENTS:
4335 retval = sctp_setsockopt_events(sk, optval, optlen);
4336 break;
4338 case SCTP_AUTOCLOSE:
4339 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4340 break;
4342 case SCTP_PEER_ADDR_PARAMS:
4343 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4344 break;
4346 case SCTP_DELAYED_SACK:
4347 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4348 break;
4349 case SCTP_PARTIAL_DELIVERY_POINT:
4350 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4351 break;
4353 case SCTP_INITMSG:
4354 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4355 break;
4356 case SCTP_DEFAULT_SEND_PARAM:
4357 retval = sctp_setsockopt_default_send_param(sk, optval,
4358 optlen);
4359 break;
4360 case SCTP_DEFAULT_SNDINFO:
4361 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4362 break;
4363 case SCTP_PRIMARY_ADDR:
4364 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4365 break;
4366 case SCTP_SET_PEER_PRIMARY_ADDR:
4367 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4368 break;
4369 case SCTP_NODELAY:
4370 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4371 break;
4372 case SCTP_RTOINFO:
4373 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4374 break;
4375 case SCTP_ASSOCINFO:
4376 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4377 break;
4378 case SCTP_I_WANT_MAPPED_V4_ADDR:
4379 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4380 break;
4381 case SCTP_MAXSEG:
4382 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4383 break;
4384 case SCTP_ADAPTATION_LAYER:
4385 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4386 break;
4387 case SCTP_CONTEXT:
4388 retval = sctp_setsockopt_context(sk, optval, optlen);
4389 break;
4390 case SCTP_FRAGMENT_INTERLEAVE:
4391 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4392 break;
4393 case SCTP_MAX_BURST:
4394 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4395 break;
4396 case SCTP_AUTH_CHUNK:
4397 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4398 break;
4399 case SCTP_HMAC_IDENT:
4400 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4401 break;
4402 case SCTP_AUTH_KEY:
4403 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4404 break;
4405 case SCTP_AUTH_ACTIVE_KEY:
4406 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4407 break;
4408 case SCTP_AUTH_DELETE_KEY:
4409 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4410 break;
4411 case SCTP_AUTH_DEACTIVATE_KEY:
4412 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4413 break;
4414 case SCTP_AUTO_ASCONF:
4415 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4416 break;
4417 case SCTP_PEER_ADDR_THLDS:
4418 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4419 break;
4420 case SCTP_RECVRCVINFO:
4421 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4422 break;
4423 case SCTP_RECVNXTINFO:
4424 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4425 break;
4426 case SCTP_PR_SUPPORTED:
4427 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4428 break;
4429 case SCTP_DEFAULT_PRINFO:
4430 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4431 break;
4432 case SCTP_RECONFIG_SUPPORTED:
4433 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4434 break;
4435 case SCTP_ENABLE_STREAM_RESET:
4436 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4437 break;
4438 case SCTP_RESET_STREAMS:
4439 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4440 break;
4441 case SCTP_RESET_ASSOC:
4442 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4443 break;
4444 case SCTP_ADD_STREAMS:
4445 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4446 break;
4447 case SCTP_STREAM_SCHEDULER:
4448 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4449 break;
4450 case SCTP_STREAM_SCHEDULER_VALUE:
4451 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4452 break;
4453 case SCTP_INTERLEAVING_SUPPORTED:
4454 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4455 optlen);
4456 break;
4457 case SCTP_REUSE_PORT:
4458 retval = sctp_setsockopt_reuse_port(sk, optval, optlen);
4459 break;
4460 default:
4461 retval = -ENOPROTOOPT;
4462 break;
4465 release_sock(sk);
4467 out_nounlock:
4468 return retval;
4471 /* API 3.1.6 connect() - UDP Style Syntax
4473 * An application may use the connect() call in the UDP model to initiate an
4474 * association without sending data.
4476 * The syntax is:
4478 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4480 * sd: the socket descriptor to have a new association added to.
4482 * nam: the address structure (either struct sockaddr_in or struct
4483 * sockaddr_in6 defined in RFC2553 [7]).
4485 * len: the size of the address.
4487 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4488 int addr_len, int flags)
4490 struct inet_sock *inet = inet_sk(sk);
4491 struct sctp_af *af;
4492 int err = 0;
4494 lock_sock(sk);
4496 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4497 addr, addr_len);
4499 /* We may need to bind the socket. */
4500 if (!inet->inet_num) {
4501 if (sk->sk_prot->get_port(sk, 0)) {
4502 release_sock(sk);
4503 return -EAGAIN;
4505 inet->inet_sport = htons(inet->inet_num);
4508 /* Validate addr_len before calling common connect/connectx routine. */
4509 af = sctp_get_af_specific(addr->sa_family);
4510 if (!af || addr_len < af->sockaddr_len) {
4511 err = -EINVAL;
4512 } else {
4513 /* Pass correct addr len to common routine (so it knows there
4514 * is only one address being passed.
4516 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4519 release_sock(sk);
4520 return err;
4523 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4524 int addr_len, int flags)
4526 if (addr_len < sizeof(uaddr->sa_family))
4527 return -EINVAL;
4529 if (uaddr->sa_family == AF_UNSPEC)
4530 return -EOPNOTSUPP;
4532 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4535 /* FIXME: Write comments. */
4536 static int sctp_disconnect(struct sock *sk, int flags)
4538 return -EOPNOTSUPP; /* STUB */
4541 /* 4.1.4 accept() - TCP Style Syntax
4543 * Applications use accept() call to remove an established SCTP
4544 * association from the accept queue of the endpoint. A new socket
4545 * descriptor will be returned from accept() to represent the newly
4546 * formed association.
4548 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4550 struct sctp_sock *sp;
4551 struct sctp_endpoint *ep;
4552 struct sock *newsk = NULL;
4553 struct sctp_association *asoc;
4554 long timeo;
4555 int error = 0;
4557 lock_sock(sk);
4559 sp = sctp_sk(sk);
4560 ep = sp->ep;
4562 if (!sctp_style(sk, TCP)) {
4563 error = -EOPNOTSUPP;
4564 goto out;
4567 if (!sctp_sstate(sk, LISTENING)) {
4568 error = -EINVAL;
4569 goto out;
4572 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4574 error = sctp_wait_for_accept(sk, timeo);
4575 if (error)
4576 goto out;
4578 /* We treat the list of associations on the endpoint as the accept
4579 * queue and pick the first association on the list.
4581 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4583 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4584 if (!newsk) {
4585 error = -ENOMEM;
4586 goto out;
4589 /* Populate the fields of the newsk from the oldsk and migrate the
4590 * asoc to the newsk.
4592 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4594 out:
4595 release_sock(sk);
4596 *err = error;
4597 return newsk;
4600 /* The SCTP ioctl handler. */
4601 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4603 int rc = -ENOTCONN;
4605 lock_sock(sk);
4608 * SEQPACKET-style sockets in LISTENING state are valid, for
4609 * SCTP, so only discard TCP-style sockets in LISTENING state.
4611 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4612 goto out;
4614 switch (cmd) {
4615 case SIOCINQ: {
4616 struct sk_buff *skb;
4617 unsigned int amount = 0;
4619 skb = skb_peek(&sk->sk_receive_queue);
4620 if (skb != NULL) {
4622 * We will only return the amount of this packet since
4623 * that is all that will be read.
4625 amount = skb->len;
4627 rc = put_user(amount, (int __user *)arg);
4628 break;
4630 default:
4631 rc = -ENOIOCTLCMD;
4632 break;
4634 out:
4635 release_sock(sk);
4636 return rc;
4639 /* This is the function which gets called during socket creation to
4640 * initialized the SCTP-specific portion of the sock.
4641 * The sock structure should already be zero-filled memory.
4643 static int sctp_init_sock(struct sock *sk)
4645 struct net *net = sock_net(sk);
4646 struct sctp_sock *sp;
4648 pr_debug("%s: sk:%p\n", __func__, sk);
4650 sp = sctp_sk(sk);
4652 /* Initialize the SCTP per socket area. */
4653 switch (sk->sk_type) {
4654 case SOCK_SEQPACKET:
4655 sp->type = SCTP_SOCKET_UDP;
4656 break;
4657 case SOCK_STREAM:
4658 sp->type = SCTP_SOCKET_TCP;
4659 break;
4660 default:
4661 return -ESOCKTNOSUPPORT;
4664 sk->sk_gso_type = SKB_GSO_SCTP;
4666 /* Initialize default send parameters. These parameters can be
4667 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4669 sp->default_stream = 0;
4670 sp->default_ppid = 0;
4671 sp->default_flags = 0;
4672 sp->default_context = 0;
4673 sp->default_timetolive = 0;
4675 sp->default_rcv_context = 0;
4676 sp->max_burst = net->sctp.max_burst;
4678 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4680 /* Initialize default setup parameters. These parameters
4681 * can be modified with the SCTP_INITMSG socket option or
4682 * overridden by the SCTP_INIT CMSG.
4684 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4685 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4686 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4687 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4689 /* Initialize default RTO related parameters. These parameters can
4690 * be modified for with the SCTP_RTOINFO socket option.
4692 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4693 sp->rtoinfo.srto_max = net->sctp.rto_max;
4694 sp->rtoinfo.srto_min = net->sctp.rto_min;
4696 /* Initialize default association related parameters. These parameters
4697 * can be modified with the SCTP_ASSOCINFO socket option.
4699 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4700 sp->assocparams.sasoc_number_peer_destinations = 0;
4701 sp->assocparams.sasoc_peer_rwnd = 0;
4702 sp->assocparams.sasoc_local_rwnd = 0;
4703 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4705 /* Initialize default event subscriptions. By default, all the
4706 * options are off.
4708 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4710 /* Default Peer Address Parameters. These defaults can
4711 * be modified via SCTP_PEER_ADDR_PARAMS
4713 sp->hbinterval = net->sctp.hb_interval;
4714 sp->pathmaxrxt = net->sctp.max_retrans_path;
4715 sp->pathmtu = 0; /* allow default discovery */
4716 sp->sackdelay = net->sctp.sack_timeout;
4717 sp->sackfreq = 2;
4718 sp->param_flags = SPP_HB_ENABLE |
4719 SPP_PMTUD_ENABLE |
4720 SPP_SACKDELAY_ENABLE;
4722 /* If enabled no SCTP message fragmentation will be performed.
4723 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4725 sp->disable_fragments = 0;
4727 /* Enable Nagle algorithm by default. */
4728 sp->nodelay = 0;
4730 sp->recvrcvinfo = 0;
4731 sp->recvnxtinfo = 0;
4733 /* Enable by default. */
4734 sp->v4mapped = 1;
4736 /* Auto-close idle associations after the configured
4737 * number of seconds. A value of 0 disables this
4738 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4739 * for UDP-style sockets only.
4741 sp->autoclose = 0;
4743 /* User specified fragmentation limit. */
4744 sp->user_frag = 0;
4746 sp->adaptation_ind = 0;
4748 sp->pf = sctp_get_pf_specific(sk->sk_family);
4750 /* Control variables for partial data delivery. */
4751 atomic_set(&sp->pd_mode, 0);
4752 skb_queue_head_init(&sp->pd_lobby);
4753 sp->frag_interleave = 0;
4755 /* Create a per socket endpoint structure. Even if we
4756 * change the data structure relationships, this may still
4757 * be useful for storing pre-connect address information.
4759 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4760 if (!sp->ep)
4761 return -ENOMEM;
4763 sp->hmac = NULL;
4765 sk->sk_destruct = sctp_destruct_sock;
4767 SCTP_DBG_OBJCNT_INC(sock);
4769 local_bh_disable();
4770 sk_sockets_allocated_inc(sk);
4771 sock_prot_inuse_add(net, sk->sk_prot, 1);
4773 /* Nothing can fail after this block, otherwise
4774 * sctp_destroy_sock() will be called without addr_wq_lock held
4776 if (net->sctp.default_auto_asconf) {
4777 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4778 list_add_tail(&sp->auto_asconf_list,
4779 &net->sctp.auto_asconf_splist);
4780 sp->do_auto_asconf = 1;
4781 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4782 } else {
4783 sp->do_auto_asconf = 0;
4786 local_bh_enable();
4788 return 0;
4791 /* Cleanup any SCTP per socket resources. Must be called with
4792 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4794 static void sctp_destroy_sock(struct sock *sk)
4796 struct sctp_sock *sp;
4798 pr_debug("%s: sk:%p\n", __func__, sk);
4800 /* Release our hold on the endpoint. */
4801 sp = sctp_sk(sk);
4802 /* This could happen during socket init, thus we bail out
4803 * early, since the rest of the below is not setup either.
4805 if (sp->ep == NULL)
4806 return;
4808 if (sp->do_auto_asconf) {
4809 sp->do_auto_asconf = 0;
4810 list_del(&sp->auto_asconf_list);
4812 sctp_endpoint_free(sp->ep);
4813 local_bh_disable();
4814 sk_sockets_allocated_dec(sk);
4815 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4816 local_bh_enable();
4819 /* Triggered when there are no references on the socket anymore */
4820 static void sctp_destruct_sock(struct sock *sk)
4822 struct sctp_sock *sp = sctp_sk(sk);
4824 /* Free up the HMAC transform. */
4825 crypto_free_shash(sp->hmac);
4827 inet_sock_destruct(sk);
4830 /* API 4.1.7 shutdown() - TCP Style Syntax
4831 * int shutdown(int socket, int how);
4833 * sd - the socket descriptor of the association to be closed.
4834 * how - Specifies the type of shutdown. The values are
4835 * as follows:
4836 * SHUT_RD
4837 * Disables further receive operations. No SCTP
4838 * protocol action is taken.
4839 * SHUT_WR
4840 * Disables further send operations, and initiates
4841 * the SCTP shutdown sequence.
4842 * SHUT_RDWR
4843 * Disables further send and receive operations
4844 * and initiates the SCTP shutdown sequence.
4846 static void sctp_shutdown(struct sock *sk, int how)
4848 struct net *net = sock_net(sk);
4849 struct sctp_endpoint *ep;
4851 if (!sctp_style(sk, TCP))
4852 return;
4854 ep = sctp_sk(sk)->ep;
4855 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4856 struct sctp_association *asoc;
4858 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4859 asoc = list_entry(ep->asocs.next,
4860 struct sctp_association, asocs);
4861 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4865 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4866 struct sctp_info *info)
4868 struct sctp_transport *prim;
4869 struct list_head *pos;
4870 int mask;
4872 memset(info, 0, sizeof(*info));
4873 if (!asoc) {
4874 struct sctp_sock *sp = sctp_sk(sk);
4876 info->sctpi_s_autoclose = sp->autoclose;
4877 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4878 info->sctpi_s_pd_point = sp->pd_point;
4879 info->sctpi_s_nodelay = sp->nodelay;
4880 info->sctpi_s_disable_fragments = sp->disable_fragments;
4881 info->sctpi_s_v4mapped = sp->v4mapped;
4882 info->sctpi_s_frag_interleave = sp->frag_interleave;
4883 info->sctpi_s_type = sp->type;
4885 return 0;
4888 info->sctpi_tag = asoc->c.my_vtag;
4889 info->sctpi_state = asoc->state;
4890 info->sctpi_rwnd = asoc->a_rwnd;
4891 info->sctpi_unackdata = asoc->unack_data;
4892 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4893 info->sctpi_instrms = asoc->stream.incnt;
4894 info->sctpi_outstrms = asoc->stream.outcnt;
4895 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4896 info->sctpi_inqueue++;
4897 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4898 info->sctpi_outqueue++;
4899 info->sctpi_overall_error = asoc->overall_error_count;
4900 info->sctpi_max_burst = asoc->max_burst;
4901 info->sctpi_maxseg = asoc->frag_point;
4902 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4903 info->sctpi_peer_tag = asoc->c.peer_vtag;
4905 mask = asoc->peer.ecn_capable << 1;
4906 mask = (mask | asoc->peer.ipv4_address) << 1;
4907 mask = (mask | asoc->peer.ipv6_address) << 1;
4908 mask = (mask | asoc->peer.hostname_address) << 1;
4909 mask = (mask | asoc->peer.asconf_capable) << 1;
4910 mask = (mask | asoc->peer.prsctp_capable) << 1;
4911 mask = (mask | asoc->peer.auth_capable);
4912 info->sctpi_peer_capable = mask;
4913 mask = asoc->peer.sack_needed << 1;
4914 mask = (mask | asoc->peer.sack_generation) << 1;
4915 mask = (mask | asoc->peer.zero_window_announced);
4916 info->sctpi_peer_sack = mask;
4918 info->sctpi_isacks = asoc->stats.isacks;
4919 info->sctpi_osacks = asoc->stats.osacks;
4920 info->sctpi_opackets = asoc->stats.opackets;
4921 info->sctpi_ipackets = asoc->stats.ipackets;
4922 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4923 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4924 info->sctpi_idupchunks = asoc->stats.idupchunks;
4925 info->sctpi_gapcnt = asoc->stats.gapcnt;
4926 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4927 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4928 info->sctpi_oodchunks = asoc->stats.oodchunks;
4929 info->sctpi_iodchunks = asoc->stats.iodchunks;
4930 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4931 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4933 prim = asoc->peer.primary_path;
4934 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4935 info->sctpi_p_state = prim->state;
4936 info->sctpi_p_cwnd = prim->cwnd;
4937 info->sctpi_p_srtt = prim->srtt;
4938 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4939 info->sctpi_p_hbinterval = prim->hbinterval;
4940 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4941 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4942 info->sctpi_p_ssthresh = prim->ssthresh;
4943 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4944 info->sctpi_p_flight_size = prim->flight_size;
4945 info->sctpi_p_error = prim->error_count;
4947 return 0;
4949 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4951 /* use callback to avoid exporting the core structure */
4952 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4954 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4956 rhashtable_walk_start(iter);
4959 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4961 rhashtable_walk_stop(iter);
4962 rhashtable_walk_exit(iter);
4965 struct sctp_transport *sctp_transport_get_next(struct net *net,
4966 struct rhashtable_iter *iter)
4968 struct sctp_transport *t;
4970 t = rhashtable_walk_next(iter);
4971 for (; t; t = rhashtable_walk_next(iter)) {
4972 if (IS_ERR(t)) {
4973 if (PTR_ERR(t) == -EAGAIN)
4974 continue;
4975 break;
4978 if (!sctp_transport_hold(t))
4979 continue;
4981 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4982 t->asoc->peer.primary_path == t)
4983 break;
4985 sctp_transport_put(t);
4988 return t;
4991 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4992 struct rhashtable_iter *iter,
4993 int pos)
4995 struct sctp_transport *t;
4997 if (!pos)
4998 return SEQ_START_TOKEN;
5000 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
5001 if (!--pos)
5002 break;
5003 sctp_transport_put(t);
5006 return t;
5009 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
5010 void *p) {
5011 int err = 0;
5012 int hash = 0;
5013 struct sctp_ep_common *epb;
5014 struct sctp_hashbucket *head;
5016 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
5017 hash++, head++) {
5018 read_lock_bh(&head->lock);
5019 sctp_for_each_hentry(epb, &head->chain) {
5020 err = cb(sctp_ep(epb), p);
5021 if (err)
5022 break;
5024 read_unlock_bh(&head->lock);
5027 return err;
5029 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
5031 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
5032 struct net *net,
5033 const union sctp_addr *laddr,
5034 const union sctp_addr *paddr, void *p)
5036 struct sctp_transport *transport;
5037 int err;
5039 rcu_read_lock();
5040 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
5041 rcu_read_unlock();
5042 if (!transport)
5043 return -ENOENT;
5045 err = cb(transport, p);
5046 sctp_transport_put(transport);
5048 return err;
5050 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
5052 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
5053 int (*cb_done)(struct sctp_transport *, void *),
5054 struct net *net, int *pos, void *p) {
5055 struct rhashtable_iter hti;
5056 struct sctp_transport *tsp;
5057 int ret;
5059 again:
5060 ret = 0;
5061 sctp_transport_walk_start(&hti);
5063 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
5064 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
5065 ret = cb(tsp, p);
5066 if (ret)
5067 break;
5068 (*pos)++;
5069 sctp_transport_put(tsp);
5071 sctp_transport_walk_stop(&hti);
5073 if (ret) {
5074 if (cb_done && !cb_done(tsp, p)) {
5075 (*pos)++;
5076 sctp_transport_put(tsp);
5077 goto again;
5079 sctp_transport_put(tsp);
5082 return ret;
5084 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
5086 /* 7.2.1 Association Status (SCTP_STATUS)
5088 * Applications can retrieve current status information about an
5089 * association, including association state, peer receiver window size,
5090 * number of unacked data chunks, and number of data chunks pending
5091 * receipt. This information is read-only.
5093 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
5094 char __user *optval,
5095 int __user *optlen)
5097 struct sctp_status status;
5098 struct sctp_association *asoc = NULL;
5099 struct sctp_transport *transport;
5100 sctp_assoc_t associd;
5101 int retval = 0;
5103 if (len < sizeof(status)) {
5104 retval = -EINVAL;
5105 goto out;
5108 len = sizeof(status);
5109 if (copy_from_user(&status, optval, len)) {
5110 retval = -EFAULT;
5111 goto out;
5114 associd = status.sstat_assoc_id;
5115 asoc = sctp_id2assoc(sk, associd);
5116 if (!asoc) {
5117 retval = -EINVAL;
5118 goto out;
5121 transport = asoc->peer.primary_path;
5123 status.sstat_assoc_id = sctp_assoc2id(asoc);
5124 status.sstat_state = sctp_assoc_to_state(asoc);
5125 status.sstat_rwnd = asoc->peer.rwnd;
5126 status.sstat_unackdata = asoc->unack_data;
5128 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5129 status.sstat_instrms = asoc->stream.incnt;
5130 status.sstat_outstrms = asoc->stream.outcnt;
5131 status.sstat_fragmentation_point = asoc->frag_point;
5132 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5133 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5134 transport->af_specific->sockaddr_len);
5135 /* Map ipv4 address into v4-mapped-on-v6 address. */
5136 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5137 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5138 status.sstat_primary.spinfo_state = transport->state;
5139 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5140 status.sstat_primary.spinfo_srtt = transport->srtt;
5141 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5142 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5144 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5145 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5147 if (put_user(len, optlen)) {
5148 retval = -EFAULT;
5149 goto out;
5152 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5153 __func__, len, status.sstat_state, status.sstat_rwnd,
5154 status.sstat_assoc_id);
5156 if (copy_to_user(optval, &status, len)) {
5157 retval = -EFAULT;
5158 goto out;
5161 out:
5162 return retval;
5166 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5168 * Applications can retrieve information about a specific peer address
5169 * of an association, including its reachability state, congestion
5170 * window, and retransmission timer values. This information is
5171 * read-only.
5173 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5174 char __user *optval,
5175 int __user *optlen)
5177 struct sctp_paddrinfo pinfo;
5178 struct sctp_transport *transport;
5179 int retval = 0;
5181 if (len < sizeof(pinfo)) {
5182 retval = -EINVAL;
5183 goto out;
5186 len = sizeof(pinfo);
5187 if (copy_from_user(&pinfo, optval, len)) {
5188 retval = -EFAULT;
5189 goto out;
5192 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5193 pinfo.spinfo_assoc_id);
5194 if (!transport)
5195 return -EINVAL;
5197 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5198 pinfo.spinfo_state = transport->state;
5199 pinfo.spinfo_cwnd = transport->cwnd;
5200 pinfo.spinfo_srtt = transport->srtt;
5201 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5202 pinfo.spinfo_mtu = transport->pathmtu;
5204 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5205 pinfo.spinfo_state = SCTP_ACTIVE;
5207 if (put_user(len, optlen)) {
5208 retval = -EFAULT;
5209 goto out;
5212 if (copy_to_user(optval, &pinfo, len)) {
5213 retval = -EFAULT;
5214 goto out;
5217 out:
5218 return retval;
5221 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5223 * This option is a on/off flag. If enabled no SCTP message
5224 * fragmentation will be performed. Instead if a message being sent
5225 * exceeds the current PMTU size, the message will NOT be sent and
5226 * instead a error will be indicated to the user.
5228 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5229 char __user *optval, int __user *optlen)
5231 int val;
5233 if (len < sizeof(int))
5234 return -EINVAL;
5236 len = sizeof(int);
5237 val = (sctp_sk(sk)->disable_fragments == 1);
5238 if (put_user(len, optlen))
5239 return -EFAULT;
5240 if (copy_to_user(optval, &val, len))
5241 return -EFAULT;
5242 return 0;
5245 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5247 * This socket option is used to specify various notifications and
5248 * ancillary data the user wishes to receive.
5250 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5251 int __user *optlen)
5253 if (len == 0)
5254 return -EINVAL;
5255 if (len > sizeof(struct sctp_event_subscribe))
5256 len = sizeof(struct sctp_event_subscribe);
5257 if (put_user(len, optlen))
5258 return -EFAULT;
5259 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5260 return -EFAULT;
5261 return 0;
5264 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5266 * This socket option is applicable to the UDP-style socket only. When
5267 * set it will cause associations that are idle for more than the
5268 * specified number of seconds to automatically close. An association
5269 * being idle is defined an association that has NOT sent or received
5270 * user data. The special value of '0' indicates that no automatic
5271 * close of any associations should be performed. The option expects an
5272 * integer defining the number of seconds of idle time before an
5273 * association is closed.
5275 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5277 /* Applicable to UDP-style socket only */
5278 if (sctp_style(sk, TCP))
5279 return -EOPNOTSUPP;
5280 if (len < sizeof(int))
5281 return -EINVAL;
5282 len = sizeof(int);
5283 if (put_user(len, optlen))
5284 return -EFAULT;
5285 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5286 return -EFAULT;
5287 return 0;
5290 /* Helper routine to branch off an association to a new socket. */
5291 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5293 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5294 struct sctp_sock *sp = sctp_sk(sk);
5295 struct socket *sock;
5296 int err = 0;
5298 /* Do not peel off from one netns to another one. */
5299 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5300 return -EINVAL;
5302 if (!asoc)
5303 return -EINVAL;
5305 /* An association cannot be branched off from an already peeled-off
5306 * socket, nor is this supported for tcp style sockets.
5308 if (!sctp_style(sk, UDP))
5309 return -EINVAL;
5311 /* Create a new socket. */
5312 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5313 if (err < 0)
5314 return err;
5316 sctp_copy_sock(sock->sk, sk, asoc);
5318 /* Make peeled-off sockets more like 1-1 accepted sockets.
5319 * Set the daddr and initialize id to something more random and also
5320 * copy over any ip options.
5322 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5323 sp->pf->copy_ip_options(sk, sock->sk);
5325 /* Populate the fields of the newsk from the oldsk and migrate the
5326 * asoc to the newsk.
5328 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5330 *sockp = sock;
5332 return err;
5334 EXPORT_SYMBOL(sctp_do_peeloff);
5336 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5337 struct file **newfile, unsigned flags)
5339 struct socket *newsock;
5340 int retval;
5342 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5343 if (retval < 0)
5344 goto out;
5346 /* Map the socket to an unused fd that can be returned to the user. */
5347 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5348 if (retval < 0) {
5349 sock_release(newsock);
5350 goto out;
5353 *newfile = sock_alloc_file(newsock, 0, NULL);
5354 if (IS_ERR(*newfile)) {
5355 put_unused_fd(retval);
5356 retval = PTR_ERR(*newfile);
5357 *newfile = NULL;
5358 return retval;
5361 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5362 retval);
5364 peeloff->sd = retval;
5366 if (flags & SOCK_NONBLOCK)
5367 (*newfile)->f_flags |= O_NONBLOCK;
5368 out:
5369 return retval;
5372 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5374 sctp_peeloff_arg_t peeloff;
5375 struct file *newfile = NULL;
5376 int retval = 0;
5378 if (len < sizeof(sctp_peeloff_arg_t))
5379 return -EINVAL;
5380 len = sizeof(sctp_peeloff_arg_t);
5381 if (copy_from_user(&peeloff, optval, len))
5382 return -EFAULT;
5384 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5385 if (retval < 0)
5386 goto out;
5388 /* Return the fd mapped to the new socket. */
5389 if (put_user(len, optlen)) {
5390 fput(newfile);
5391 put_unused_fd(retval);
5392 return -EFAULT;
5395 if (copy_to_user(optval, &peeloff, len)) {
5396 fput(newfile);
5397 put_unused_fd(retval);
5398 return -EFAULT;
5400 fd_install(retval, newfile);
5401 out:
5402 return retval;
5405 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5406 char __user *optval, int __user *optlen)
5408 sctp_peeloff_flags_arg_t peeloff;
5409 struct file *newfile = NULL;
5410 int retval = 0;
5412 if (len < sizeof(sctp_peeloff_flags_arg_t))
5413 return -EINVAL;
5414 len = sizeof(sctp_peeloff_flags_arg_t);
5415 if (copy_from_user(&peeloff, optval, len))
5416 return -EFAULT;
5418 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5419 &newfile, peeloff.flags);
5420 if (retval < 0)
5421 goto out;
5423 /* Return the fd mapped to the new socket. */
5424 if (put_user(len, optlen)) {
5425 fput(newfile);
5426 put_unused_fd(retval);
5427 return -EFAULT;
5430 if (copy_to_user(optval, &peeloff, len)) {
5431 fput(newfile);
5432 put_unused_fd(retval);
5433 return -EFAULT;
5435 fd_install(retval, newfile);
5436 out:
5437 return retval;
5440 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5442 * Applications can enable or disable heartbeats for any peer address of
5443 * an association, modify an address's heartbeat interval, force a
5444 * heartbeat to be sent immediately, and adjust the address's maximum
5445 * number of retransmissions sent before an address is considered
5446 * unreachable. The following structure is used to access and modify an
5447 * address's parameters:
5449 * struct sctp_paddrparams {
5450 * sctp_assoc_t spp_assoc_id;
5451 * struct sockaddr_storage spp_address;
5452 * uint32_t spp_hbinterval;
5453 * uint16_t spp_pathmaxrxt;
5454 * uint32_t spp_pathmtu;
5455 * uint32_t spp_sackdelay;
5456 * uint32_t spp_flags;
5457 * };
5459 * spp_assoc_id - (one-to-many style socket) This is filled in the
5460 * application, and identifies the association for
5461 * this query.
5462 * spp_address - This specifies which address is of interest.
5463 * spp_hbinterval - This contains the value of the heartbeat interval,
5464 * in milliseconds. If a value of zero
5465 * is present in this field then no changes are to
5466 * be made to this parameter.
5467 * spp_pathmaxrxt - This contains the maximum number of
5468 * retransmissions before this address shall be
5469 * considered unreachable. If a value of zero
5470 * is present in this field then no changes are to
5471 * be made to this parameter.
5472 * spp_pathmtu - When Path MTU discovery is disabled the value
5473 * specified here will be the "fixed" path mtu.
5474 * Note that if the spp_address field is empty
5475 * then all associations on this address will
5476 * have this fixed path mtu set upon them.
5478 * spp_sackdelay - When delayed sack is enabled, this value specifies
5479 * the number of milliseconds that sacks will be delayed
5480 * for. This value will apply to all addresses of an
5481 * association if the spp_address field is empty. Note
5482 * also, that if delayed sack is enabled and this
5483 * value is set to 0, no change is made to the last
5484 * recorded delayed sack timer value.
5486 * spp_flags - These flags are used to control various features
5487 * on an association. The flag field may contain
5488 * zero or more of the following options.
5490 * SPP_HB_ENABLE - Enable heartbeats on the
5491 * specified address. Note that if the address
5492 * field is empty all addresses for the association
5493 * have heartbeats enabled upon them.
5495 * SPP_HB_DISABLE - Disable heartbeats on the
5496 * speicifed address. Note that if the address
5497 * field is empty all addresses for the association
5498 * will have their heartbeats disabled. Note also
5499 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5500 * mutually exclusive, only one of these two should
5501 * be specified. Enabling both fields will have
5502 * undetermined results.
5504 * SPP_HB_DEMAND - Request a user initiated heartbeat
5505 * to be made immediately.
5507 * SPP_PMTUD_ENABLE - This field will enable PMTU
5508 * discovery upon the specified address. Note that
5509 * if the address feild is empty then all addresses
5510 * on the association are effected.
5512 * SPP_PMTUD_DISABLE - This field will disable PMTU
5513 * discovery upon the specified address. Note that
5514 * if the address feild is empty then all addresses
5515 * on the association are effected. Not also that
5516 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5517 * exclusive. Enabling both will have undetermined
5518 * results.
5520 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5521 * on delayed sack. The time specified in spp_sackdelay
5522 * is used to specify the sack delay for this address. Note
5523 * that if spp_address is empty then all addresses will
5524 * enable delayed sack and take on the sack delay
5525 * value specified in spp_sackdelay.
5526 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5527 * off delayed sack. If the spp_address field is blank then
5528 * delayed sack is disabled for the entire association. Note
5529 * also that this field is mutually exclusive to
5530 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5531 * results.
5533 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
5534 * setting of the IPV6 flow label value. The value is
5535 * contained in the spp_ipv6_flowlabel field.
5536 * Upon retrieval, this flag will be set to indicate that
5537 * the spp_ipv6_flowlabel field has a valid value returned.
5538 * If a specific destination address is set (in the
5539 * spp_address field), then the value returned is that of
5540 * the address. If just an association is specified (and
5541 * no address), then the association's default flow label
5542 * is returned. If neither an association nor a destination
5543 * is specified, then the socket's default flow label is
5544 * returned. For non-IPv6 sockets, this flag will be left
5545 * cleared.
5547 * SPP_DSCP: Setting this flag enables the setting of the
5548 * Differentiated Services Code Point (DSCP) value
5549 * associated with either the association or a specific
5550 * address. The value is obtained in the spp_dscp field.
5551 * Upon retrieval, this flag will be set to indicate that
5552 * the spp_dscp field has a valid value returned. If a
5553 * specific destination address is set when called (in the
5554 * spp_address field), then that specific destination
5555 * address's DSCP value is returned. If just an association
5556 * is specified, then the association's default DSCP is
5557 * returned. If neither an association nor a destination is
5558 * specified, then the socket's default DSCP is returned.
5560 * spp_ipv6_flowlabel
5561 * - This field is used in conjunction with the
5562 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
5563 * The 20 least significant bits are used for the flow
5564 * label. This setting has precedence over any IPv6-layer
5565 * setting.
5567 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
5568 * and contains the DSCP. The 6 most significant bits are
5569 * used for the DSCP. This setting has precedence over any
5570 * IPv4- or IPv6- layer setting.
5572 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5573 char __user *optval, int __user *optlen)
5575 struct sctp_paddrparams params;
5576 struct sctp_transport *trans = NULL;
5577 struct sctp_association *asoc = NULL;
5578 struct sctp_sock *sp = sctp_sk(sk);
5580 if (len >= sizeof(params))
5581 len = sizeof(params);
5582 else if (len >= ALIGN(offsetof(struct sctp_paddrparams,
5583 spp_ipv6_flowlabel), 4))
5584 len = ALIGN(offsetof(struct sctp_paddrparams,
5585 spp_ipv6_flowlabel), 4);
5586 else
5587 return -EINVAL;
5589 if (copy_from_user(&params, optval, len))
5590 return -EFAULT;
5592 /* If an address other than INADDR_ANY is specified, and
5593 * no transport is found, then the request is invalid.
5595 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
5596 trans = sctp_addr_id2transport(sk, &params.spp_address,
5597 params.spp_assoc_id);
5598 if (!trans) {
5599 pr_debug("%s: failed no transport\n", __func__);
5600 return -EINVAL;
5604 /* Get association, if assoc_id != 0 and the socket is a one
5605 * to many style socket, and an association was not found, then
5606 * the id was invalid.
5608 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5609 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5610 pr_debug("%s: failed no association\n", __func__);
5611 return -EINVAL;
5614 if (trans) {
5615 /* Fetch transport values. */
5616 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5617 params.spp_pathmtu = trans->pathmtu;
5618 params.spp_pathmaxrxt = trans->pathmaxrxt;
5619 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5621 /*draft-11 doesn't say what to return in spp_flags*/
5622 params.spp_flags = trans->param_flags;
5623 if (trans->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5624 params.spp_ipv6_flowlabel = trans->flowlabel &
5625 SCTP_FLOWLABEL_VAL_MASK;
5626 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5628 if (trans->dscp & SCTP_DSCP_SET_MASK) {
5629 params.spp_dscp = trans->dscp & SCTP_DSCP_VAL_MASK;
5630 params.spp_flags |= SPP_DSCP;
5632 } else if (asoc) {
5633 /* Fetch association values. */
5634 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5635 params.spp_pathmtu = asoc->pathmtu;
5636 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5637 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5639 /*draft-11 doesn't say what to return in spp_flags*/
5640 params.spp_flags = asoc->param_flags;
5641 if (asoc->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5642 params.spp_ipv6_flowlabel = asoc->flowlabel &
5643 SCTP_FLOWLABEL_VAL_MASK;
5644 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5646 if (asoc->dscp & SCTP_DSCP_SET_MASK) {
5647 params.spp_dscp = asoc->dscp & SCTP_DSCP_VAL_MASK;
5648 params.spp_flags |= SPP_DSCP;
5650 } else {
5651 /* Fetch socket values. */
5652 params.spp_hbinterval = sp->hbinterval;
5653 params.spp_pathmtu = sp->pathmtu;
5654 params.spp_sackdelay = sp->sackdelay;
5655 params.spp_pathmaxrxt = sp->pathmaxrxt;
5657 /*draft-11 doesn't say what to return in spp_flags*/
5658 params.spp_flags = sp->param_flags;
5659 if (sp->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5660 params.spp_ipv6_flowlabel = sp->flowlabel &
5661 SCTP_FLOWLABEL_VAL_MASK;
5662 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5664 if (sp->dscp & SCTP_DSCP_SET_MASK) {
5665 params.spp_dscp = sp->dscp & SCTP_DSCP_VAL_MASK;
5666 params.spp_flags |= SPP_DSCP;
5670 if (copy_to_user(optval, &params, len))
5671 return -EFAULT;
5673 if (put_user(len, optlen))
5674 return -EFAULT;
5676 return 0;
5680 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5682 * This option will effect the way delayed acks are performed. This
5683 * option allows you to get or set the delayed ack time, in
5684 * milliseconds. It also allows changing the delayed ack frequency.
5685 * Changing the frequency to 1 disables the delayed sack algorithm. If
5686 * the assoc_id is 0, then this sets or gets the endpoints default
5687 * values. If the assoc_id field is non-zero, then the set or get
5688 * effects the specified association for the one to many model (the
5689 * assoc_id field is ignored by the one to one model). Note that if
5690 * sack_delay or sack_freq are 0 when setting this option, then the
5691 * current values will remain unchanged.
5693 * struct sctp_sack_info {
5694 * sctp_assoc_t sack_assoc_id;
5695 * uint32_t sack_delay;
5696 * uint32_t sack_freq;
5697 * };
5699 * sack_assoc_id - This parameter, indicates which association the user
5700 * is performing an action upon. Note that if this field's value is
5701 * zero then the endpoints default value is changed (effecting future
5702 * associations only).
5704 * sack_delay - This parameter contains the number of milliseconds that
5705 * the user is requesting the delayed ACK timer be set to. Note that
5706 * this value is defined in the standard to be between 200 and 500
5707 * milliseconds.
5709 * sack_freq - This parameter contains the number of packets that must
5710 * be received before a sack is sent without waiting for the delay
5711 * timer to expire. The default value for this is 2, setting this
5712 * value to 1 will disable the delayed sack algorithm.
5714 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5715 char __user *optval,
5716 int __user *optlen)
5718 struct sctp_sack_info params;
5719 struct sctp_association *asoc = NULL;
5720 struct sctp_sock *sp = sctp_sk(sk);
5722 if (len >= sizeof(struct sctp_sack_info)) {
5723 len = sizeof(struct sctp_sack_info);
5725 if (copy_from_user(&params, optval, len))
5726 return -EFAULT;
5727 } else if (len == sizeof(struct sctp_assoc_value)) {
5728 pr_warn_ratelimited(DEPRECATED
5729 "%s (pid %d) "
5730 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5731 "Use struct sctp_sack_info instead\n",
5732 current->comm, task_pid_nr(current));
5733 if (copy_from_user(&params, optval, len))
5734 return -EFAULT;
5735 } else
5736 return -EINVAL;
5738 /* Get association, if sack_assoc_id != 0 and the socket is a one
5739 * to many style socket, and an association was not found, then
5740 * the id was invalid.
5742 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5743 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5744 return -EINVAL;
5746 if (asoc) {
5747 /* Fetch association values. */
5748 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5749 params.sack_delay = jiffies_to_msecs(
5750 asoc->sackdelay);
5751 params.sack_freq = asoc->sackfreq;
5753 } else {
5754 params.sack_delay = 0;
5755 params.sack_freq = 1;
5757 } else {
5758 /* Fetch socket values. */
5759 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5760 params.sack_delay = sp->sackdelay;
5761 params.sack_freq = sp->sackfreq;
5762 } else {
5763 params.sack_delay = 0;
5764 params.sack_freq = 1;
5768 if (copy_to_user(optval, &params, len))
5769 return -EFAULT;
5771 if (put_user(len, optlen))
5772 return -EFAULT;
5774 return 0;
5777 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5779 * Applications can specify protocol parameters for the default association
5780 * initialization. The option name argument to setsockopt() and getsockopt()
5781 * is SCTP_INITMSG.
5783 * Setting initialization parameters is effective only on an unconnected
5784 * socket (for UDP-style sockets only future associations are effected
5785 * by the change). With TCP-style sockets, this option is inherited by
5786 * sockets derived from a listener socket.
5788 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5790 if (len < sizeof(struct sctp_initmsg))
5791 return -EINVAL;
5792 len = sizeof(struct sctp_initmsg);
5793 if (put_user(len, optlen))
5794 return -EFAULT;
5795 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5796 return -EFAULT;
5797 return 0;
5801 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5802 char __user *optval, int __user *optlen)
5804 struct sctp_association *asoc;
5805 int cnt = 0;
5806 struct sctp_getaddrs getaddrs;
5807 struct sctp_transport *from;
5808 void __user *to;
5809 union sctp_addr temp;
5810 struct sctp_sock *sp = sctp_sk(sk);
5811 int addrlen;
5812 size_t space_left;
5813 int bytes_copied;
5815 if (len < sizeof(struct sctp_getaddrs))
5816 return -EINVAL;
5818 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5819 return -EFAULT;
5821 /* For UDP-style sockets, id specifies the association to query. */
5822 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5823 if (!asoc)
5824 return -EINVAL;
5826 to = optval + offsetof(struct sctp_getaddrs, addrs);
5827 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5829 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5830 transports) {
5831 memcpy(&temp, &from->ipaddr, sizeof(temp));
5832 addrlen = sctp_get_pf_specific(sk->sk_family)
5833 ->addr_to_user(sp, &temp);
5834 if (space_left < addrlen)
5835 return -ENOMEM;
5836 if (copy_to_user(to, &temp, addrlen))
5837 return -EFAULT;
5838 to += addrlen;
5839 cnt++;
5840 space_left -= addrlen;
5843 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5844 return -EFAULT;
5845 bytes_copied = ((char __user *)to) - optval;
5846 if (put_user(bytes_copied, optlen))
5847 return -EFAULT;
5849 return 0;
5852 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5853 size_t space_left, int *bytes_copied)
5855 struct sctp_sockaddr_entry *addr;
5856 union sctp_addr temp;
5857 int cnt = 0;
5858 int addrlen;
5859 struct net *net = sock_net(sk);
5861 rcu_read_lock();
5862 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5863 if (!addr->valid)
5864 continue;
5866 if ((PF_INET == sk->sk_family) &&
5867 (AF_INET6 == addr->a.sa.sa_family))
5868 continue;
5869 if ((PF_INET6 == sk->sk_family) &&
5870 inet_v6_ipv6only(sk) &&
5871 (AF_INET == addr->a.sa.sa_family))
5872 continue;
5873 memcpy(&temp, &addr->a, sizeof(temp));
5874 if (!temp.v4.sin_port)
5875 temp.v4.sin_port = htons(port);
5877 addrlen = sctp_get_pf_specific(sk->sk_family)
5878 ->addr_to_user(sctp_sk(sk), &temp);
5880 if (space_left < addrlen) {
5881 cnt = -ENOMEM;
5882 break;
5884 memcpy(to, &temp, addrlen);
5886 to += addrlen;
5887 cnt++;
5888 space_left -= addrlen;
5889 *bytes_copied += addrlen;
5891 rcu_read_unlock();
5893 return cnt;
5897 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5898 char __user *optval, int __user *optlen)
5900 struct sctp_bind_addr *bp;
5901 struct sctp_association *asoc;
5902 int cnt = 0;
5903 struct sctp_getaddrs getaddrs;
5904 struct sctp_sockaddr_entry *addr;
5905 void __user *to;
5906 union sctp_addr temp;
5907 struct sctp_sock *sp = sctp_sk(sk);
5908 int addrlen;
5909 int err = 0;
5910 size_t space_left;
5911 int bytes_copied = 0;
5912 void *addrs;
5913 void *buf;
5915 if (len < sizeof(struct sctp_getaddrs))
5916 return -EINVAL;
5918 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5919 return -EFAULT;
5922 * For UDP-style sockets, id specifies the association to query.
5923 * If the id field is set to the value '0' then the locally bound
5924 * addresses are returned without regard to any particular
5925 * association.
5927 if (0 == getaddrs.assoc_id) {
5928 bp = &sctp_sk(sk)->ep->base.bind_addr;
5929 } else {
5930 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5931 if (!asoc)
5932 return -EINVAL;
5933 bp = &asoc->base.bind_addr;
5936 to = optval + offsetof(struct sctp_getaddrs, addrs);
5937 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5939 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5940 if (!addrs)
5941 return -ENOMEM;
5943 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5944 * addresses from the global local address list.
5946 if (sctp_list_single_entry(&bp->address_list)) {
5947 addr = list_entry(bp->address_list.next,
5948 struct sctp_sockaddr_entry, list);
5949 if (sctp_is_any(sk, &addr->a)) {
5950 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5951 space_left, &bytes_copied);
5952 if (cnt < 0) {
5953 err = cnt;
5954 goto out;
5956 goto copy_getaddrs;
5960 buf = addrs;
5961 /* Protection on the bound address list is not needed since
5962 * in the socket option context we hold a socket lock and
5963 * thus the bound address list can't change.
5965 list_for_each_entry(addr, &bp->address_list, list) {
5966 memcpy(&temp, &addr->a, sizeof(temp));
5967 addrlen = sctp_get_pf_specific(sk->sk_family)
5968 ->addr_to_user(sp, &temp);
5969 if (space_left < addrlen) {
5970 err = -ENOMEM; /*fixme: right error?*/
5971 goto out;
5973 memcpy(buf, &temp, addrlen);
5974 buf += addrlen;
5975 bytes_copied += addrlen;
5976 cnt++;
5977 space_left -= addrlen;
5980 copy_getaddrs:
5981 if (copy_to_user(to, addrs, bytes_copied)) {
5982 err = -EFAULT;
5983 goto out;
5985 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5986 err = -EFAULT;
5987 goto out;
5989 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5990 * but we can't change it anymore.
5992 if (put_user(bytes_copied, optlen))
5993 err = -EFAULT;
5994 out:
5995 kfree(addrs);
5996 return err;
5999 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
6001 * Requests that the local SCTP stack use the enclosed peer address as
6002 * the association primary. The enclosed address must be one of the
6003 * association peer's addresses.
6005 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
6006 char __user *optval, int __user *optlen)
6008 struct sctp_prim prim;
6009 struct sctp_association *asoc;
6010 struct sctp_sock *sp = sctp_sk(sk);
6012 if (len < sizeof(struct sctp_prim))
6013 return -EINVAL;
6015 len = sizeof(struct sctp_prim);
6017 if (copy_from_user(&prim, optval, len))
6018 return -EFAULT;
6020 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
6021 if (!asoc)
6022 return -EINVAL;
6024 if (!asoc->peer.primary_path)
6025 return -ENOTCONN;
6027 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
6028 asoc->peer.primary_path->af_specific->sockaddr_len);
6030 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
6031 (union sctp_addr *)&prim.ssp_addr);
6033 if (put_user(len, optlen))
6034 return -EFAULT;
6035 if (copy_to_user(optval, &prim, len))
6036 return -EFAULT;
6038 return 0;
6042 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
6044 * Requests that the local endpoint set the specified Adaptation Layer
6045 * Indication parameter for all future INIT and INIT-ACK exchanges.
6047 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
6048 char __user *optval, int __user *optlen)
6050 struct sctp_setadaptation adaptation;
6052 if (len < sizeof(struct sctp_setadaptation))
6053 return -EINVAL;
6055 len = sizeof(struct sctp_setadaptation);
6057 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
6059 if (put_user(len, optlen))
6060 return -EFAULT;
6061 if (copy_to_user(optval, &adaptation, len))
6062 return -EFAULT;
6064 return 0;
6069 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
6071 * Applications that wish to use the sendto() system call may wish to
6072 * specify a default set of parameters that would normally be supplied
6073 * through the inclusion of ancillary data. This socket option allows
6074 * such an application to set the default sctp_sndrcvinfo structure.
6077 * The application that wishes to use this socket option simply passes
6078 * in to this call the sctp_sndrcvinfo structure defined in Section
6079 * 5.2.2) The input parameters accepted by this call include
6080 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
6081 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
6082 * to this call if the caller is using the UDP model.
6084 * For getsockopt, it get the default sctp_sndrcvinfo structure.
6086 static int sctp_getsockopt_default_send_param(struct sock *sk,
6087 int len, char __user *optval,
6088 int __user *optlen)
6090 struct sctp_sock *sp = sctp_sk(sk);
6091 struct sctp_association *asoc;
6092 struct sctp_sndrcvinfo info;
6094 if (len < sizeof(info))
6095 return -EINVAL;
6097 len = sizeof(info);
6099 if (copy_from_user(&info, optval, len))
6100 return -EFAULT;
6102 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
6103 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
6104 return -EINVAL;
6105 if (asoc) {
6106 info.sinfo_stream = asoc->default_stream;
6107 info.sinfo_flags = asoc->default_flags;
6108 info.sinfo_ppid = asoc->default_ppid;
6109 info.sinfo_context = asoc->default_context;
6110 info.sinfo_timetolive = asoc->default_timetolive;
6111 } else {
6112 info.sinfo_stream = sp->default_stream;
6113 info.sinfo_flags = sp->default_flags;
6114 info.sinfo_ppid = sp->default_ppid;
6115 info.sinfo_context = sp->default_context;
6116 info.sinfo_timetolive = sp->default_timetolive;
6119 if (put_user(len, optlen))
6120 return -EFAULT;
6121 if (copy_to_user(optval, &info, len))
6122 return -EFAULT;
6124 return 0;
6127 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
6128 * (SCTP_DEFAULT_SNDINFO)
6130 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
6131 char __user *optval,
6132 int __user *optlen)
6134 struct sctp_sock *sp = sctp_sk(sk);
6135 struct sctp_association *asoc;
6136 struct sctp_sndinfo info;
6138 if (len < sizeof(info))
6139 return -EINVAL;
6141 len = sizeof(info);
6143 if (copy_from_user(&info, optval, len))
6144 return -EFAULT;
6146 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
6147 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
6148 return -EINVAL;
6149 if (asoc) {
6150 info.snd_sid = asoc->default_stream;
6151 info.snd_flags = asoc->default_flags;
6152 info.snd_ppid = asoc->default_ppid;
6153 info.snd_context = asoc->default_context;
6154 } else {
6155 info.snd_sid = sp->default_stream;
6156 info.snd_flags = sp->default_flags;
6157 info.snd_ppid = sp->default_ppid;
6158 info.snd_context = sp->default_context;
6161 if (put_user(len, optlen))
6162 return -EFAULT;
6163 if (copy_to_user(optval, &info, len))
6164 return -EFAULT;
6166 return 0;
6171 * 7.1.5 SCTP_NODELAY
6173 * Turn on/off any Nagle-like algorithm. This means that packets are
6174 * generally sent as soon as possible and no unnecessary delays are
6175 * introduced, at the cost of more packets in the network. Expects an
6176 * integer boolean flag.
6179 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
6180 char __user *optval, int __user *optlen)
6182 int val;
6184 if (len < sizeof(int))
6185 return -EINVAL;
6187 len = sizeof(int);
6188 val = (sctp_sk(sk)->nodelay == 1);
6189 if (put_user(len, optlen))
6190 return -EFAULT;
6191 if (copy_to_user(optval, &val, len))
6192 return -EFAULT;
6193 return 0;
6198 * 7.1.1 SCTP_RTOINFO
6200 * The protocol parameters used to initialize and bound retransmission
6201 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6202 * and modify these parameters.
6203 * All parameters are time values, in milliseconds. A value of 0, when
6204 * modifying the parameters, indicates that the current value should not
6205 * be changed.
6208 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6209 char __user *optval,
6210 int __user *optlen) {
6211 struct sctp_rtoinfo rtoinfo;
6212 struct sctp_association *asoc;
6214 if (len < sizeof (struct sctp_rtoinfo))
6215 return -EINVAL;
6217 len = sizeof(struct sctp_rtoinfo);
6219 if (copy_from_user(&rtoinfo, optval, len))
6220 return -EFAULT;
6222 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6224 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
6225 return -EINVAL;
6227 /* Values corresponding to the specific association. */
6228 if (asoc) {
6229 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6230 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6231 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6232 } else {
6233 /* Values corresponding to the endpoint. */
6234 struct sctp_sock *sp = sctp_sk(sk);
6236 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6237 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6238 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6241 if (put_user(len, optlen))
6242 return -EFAULT;
6244 if (copy_to_user(optval, &rtoinfo, len))
6245 return -EFAULT;
6247 return 0;
6252 * 7.1.2 SCTP_ASSOCINFO
6254 * This option is used to tune the maximum retransmission attempts
6255 * of the association.
6256 * Returns an error if the new association retransmission value is
6257 * greater than the sum of the retransmission value of the peer.
6258 * See [SCTP] for more information.
6261 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6262 char __user *optval,
6263 int __user *optlen)
6266 struct sctp_assocparams assocparams;
6267 struct sctp_association *asoc;
6268 struct list_head *pos;
6269 int cnt = 0;
6271 if (len < sizeof (struct sctp_assocparams))
6272 return -EINVAL;
6274 len = sizeof(struct sctp_assocparams);
6276 if (copy_from_user(&assocparams, optval, len))
6277 return -EFAULT;
6279 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6281 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
6282 return -EINVAL;
6284 /* Values correspoinding to the specific association */
6285 if (asoc) {
6286 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6287 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6288 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6289 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6291 list_for_each(pos, &asoc->peer.transport_addr_list) {
6292 cnt++;
6295 assocparams.sasoc_number_peer_destinations = cnt;
6296 } else {
6297 /* Values corresponding to the endpoint */
6298 struct sctp_sock *sp = sctp_sk(sk);
6300 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6301 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6302 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6303 assocparams.sasoc_cookie_life =
6304 sp->assocparams.sasoc_cookie_life;
6305 assocparams.sasoc_number_peer_destinations =
6306 sp->assocparams.
6307 sasoc_number_peer_destinations;
6310 if (put_user(len, optlen))
6311 return -EFAULT;
6313 if (copy_to_user(optval, &assocparams, len))
6314 return -EFAULT;
6316 return 0;
6320 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6322 * This socket option is a boolean flag which turns on or off mapped V4
6323 * addresses. If this option is turned on and the socket is type
6324 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6325 * If this option is turned off, then no mapping will be done of V4
6326 * addresses and a user will receive both PF_INET6 and PF_INET type
6327 * addresses on the socket.
6329 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6330 char __user *optval, int __user *optlen)
6332 int val;
6333 struct sctp_sock *sp = sctp_sk(sk);
6335 if (len < sizeof(int))
6336 return -EINVAL;
6338 len = sizeof(int);
6339 val = sp->v4mapped;
6340 if (put_user(len, optlen))
6341 return -EFAULT;
6342 if (copy_to_user(optval, &val, len))
6343 return -EFAULT;
6345 return 0;
6349 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6350 * (chapter and verse is quoted at sctp_setsockopt_context())
6352 static int sctp_getsockopt_context(struct sock *sk, int len,
6353 char __user *optval, int __user *optlen)
6355 struct sctp_assoc_value params;
6356 struct sctp_sock *sp;
6357 struct sctp_association *asoc;
6359 if (len < sizeof(struct sctp_assoc_value))
6360 return -EINVAL;
6362 len = sizeof(struct sctp_assoc_value);
6364 if (copy_from_user(&params, optval, len))
6365 return -EFAULT;
6367 sp = sctp_sk(sk);
6369 if (params.assoc_id != 0) {
6370 asoc = sctp_id2assoc(sk, params.assoc_id);
6371 if (!asoc)
6372 return -EINVAL;
6373 params.assoc_value = asoc->default_rcv_context;
6374 } else {
6375 params.assoc_value = sp->default_rcv_context;
6378 if (put_user(len, optlen))
6379 return -EFAULT;
6380 if (copy_to_user(optval, &params, len))
6381 return -EFAULT;
6383 return 0;
6387 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6388 * This option will get or set the maximum size to put in any outgoing
6389 * SCTP DATA chunk. If a message is larger than this size it will be
6390 * fragmented by SCTP into the specified size. Note that the underlying
6391 * SCTP implementation may fragment into smaller sized chunks when the
6392 * PMTU of the underlying association is smaller than the value set by
6393 * the user. The default value for this option is '0' which indicates
6394 * the user is NOT limiting fragmentation and only the PMTU will effect
6395 * SCTP's choice of DATA chunk size. Note also that values set larger
6396 * than the maximum size of an IP datagram will effectively let SCTP
6397 * control fragmentation (i.e. the same as setting this option to 0).
6399 * The following structure is used to access and modify this parameter:
6401 * struct sctp_assoc_value {
6402 * sctp_assoc_t assoc_id;
6403 * uint32_t assoc_value;
6404 * };
6406 * assoc_id: This parameter is ignored for one-to-one style sockets.
6407 * For one-to-many style sockets this parameter indicates which
6408 * association the user is performing an action upon. Note that if
6409 * this field's value is zero then the endpoints default value is
6410 * changed (effecting future associations only).
6411 * assoc_value: This parameter specifies the maximum size in bytes.
6413 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6414 char __user *optval, int __user *optlen)
6416 struct sctp_assoc_value params;
6417 struct sctp_association *asoc;
6419 if (len == sizeof(int)) {
6420 pr_warn_ratelimited(DEPRECATED
6421 "%s (pid %d) "
6422 "Use of int in maxseg socket option.\n"
6423 "Use struct sctp_assoc_value instead\n",
6424 current->comm, task_pid_nr(current));
6425 params.assoc_id = 0;
6426 } else if (len >= sizeof(struct sctp_assoc_value)) {
6427 len = sizeof(struct sctp_assoc_value);
6428 if (copy_from_user(&params, optval, len))
6429 return -EFAULT;
6430 } else
6431 return -EINVAL;
6433 asoc = sctp_id2assoc(sk, params.assoc_id);
6434 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6435 return -EINVAL;
6437 if (asoc)
6438 params.assoc_value = asoc->frag_point;
6439 else
6440 params.assoc_value = sctp_sk(sk)->user_frag;
6442 if (put_user(len, optlen))
6443 return -EFAULT;
6444 if (len == sizeof(int)) {
6445 if (copy_to_user(optval, &params.assoc_value, len))
6446 return -EFAULT;
6447 } else {
6448 if (copy_to_user(optval, &params, len))
6449 return -EFAULT;
6452 return 0;
6456 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6457 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6459 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6460 char __user *optval, int __user *optlen)
6462 int val;
6464 if (len < sizeof(int))
6465 return -EINVAL;
6467 len = sizeof(int);
6469 val = sctp_sk(sk)->frag_interleave;
6470 if (put_user(len, optlen))
6471 return -EFAULT;
6472 if (copy_to_user(optval, &val, len))
6473 return -EFAULT;
6475 return 0;
6479 * 7.1.25. Set or Get the sctp partial delivery point
6480 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6482 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6483 char __user *optval,
6484 int __user *optlen)
6486 u32 val;
6488 if (len < sizeof(u32))
6489 return -EINVAL;
6491 len = sizeof(u32);
6493 val = sctp_sk(sk)->pd_point;
6494 if (put_user(len, optlen))
6495 return -EFAULT;
6496 if (copy_to_user(optval, &val, len))
6497 return -EFAULT;
6499 return 0;
6503 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6504 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6506 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6507 char __user *optval,
6508 int __user *optlen)
6510 struct sctp_assoc_value params;
6511 struct sctp_sock *sp;
6512 struct sctp_association *asoc;
6514 if (len == sizeof(int)) {
6515 pr_warn_ratelimited(DEPRECATED
6516 "%s (pid %d) "
6517 "Use of int in max_burst socket option.\n"
6518 "Use struct sctp_assoc_value instead\n",
6519 current->comm, task_pid_nr(current));
6520 params.assoc_id = 0;
6521 } else if (len >= sizeof(struct sctp_assoc_value)) {
6522 len = sizeof(struct sctp_assoc_value);
6523 if (copy_from_user(&params, optval, len))
6524 return -EFAULT;
6525 } else
6526 return -EINVAL;
6528 sp = sctp_sk(sk);
6530 if (params.assoc_id != 0) {
6531 asoc = sctp_id2assoc(sk, params.assoc_id);
6532 if (!asoc)
6533 return -EINVAL;
6534 params.assoc_value = asoc->max_burst;
6535 } else
6536 params.assoc_value = sp->max_burst;
6538 if (len == sizeof(int)) {
6539 if (copy_to_user(optval, &params.assoc_value, len))
6540 return -EFAULT;
6541 } else {
6542 if (copy_to_user(optval, &params, len))
6543 return -EFAULT;
6546 return 0;
6550 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6551 char __user *optval, int __user *optlen)
6553 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6554 struct sctp_hmacalgo __user *p = (void __user *)optval;
6555 struct sctp_hmac_algo_param *hmacs;
6556 __u16 data_len = 0;
6557 u32 num_idents;
6558 int i;
6560 if (!ep->auth_enable)
6561 return -EACCES;
6563 hmacs = ep->auth_hmacs_list;
6564 data_len = ntohs(hmacs->param_hdr.length) -
6565 sizeof(struct sctp_paramhdr);
6567 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6568 return -EINVAL;
6570 len = sizeof(struct sctp_hmacalgo) + data_len;
6571 num_idents = data_len / sizeof(u16);
6573 if (put_user(len, optlen))
6574 return -EFAULT;
6575 if (put_user(num_idents, &p->shmac_num_idents))
6576 return -EFAULT;
6577 for (i = 0; i < num_idents; i++) {
6578 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6580 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6581 return -EFAULT;
6583 return 0;
6586 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6587 char __user *optval, int __user *optlen)
6589 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6590 struct sctp_authkeyid val;
6591 struct sctp_association *asoc;
6593 if (!ep->auth_enable)
6594 return -EACCES;
6596 if (len < sizeof(struct sctp_authkeyid))
6597 return -EINVAL;
6599 len = sizeof(struct sctp_authkeyid);
6600 if (copy_from_user(&val, optval, len))
6601 return -EFAULT;
6603 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6604 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6605 return -EINVAL;
6607 if (asoc)
6608 val.scact_keynumber = asoc->active_key_id;
6609 else
6610 val.scact_keynumber = ep->active_key_id;
6612 if (put_user(len, optlen))
6613 return -EFAULT;
6614 if (copy_to_user(optval, &val, len))
6615 return -EFAULT;
6617 return 0;
6620 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6621 char __user *optval, int __user *optlen)
6623 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6624 struct sctp_authchunks __user *p = (void __user *)optval;
6625 struct sctp_authchunks val;
6626 struct sctp_association *asoc;
6627 struct sctp_chunks_param *ch;
6628 u32 num_chunks = 0;
6629 char __user *to;
6631 if (!ep->auth_enable)
6632 return -EACCES;
6634 if (len < sizeof(struct sctp_authchunks))
6635 return -EINVAL;
6637 if (copy_from_user(&val, optval, sizeof(val)))
6638 return -EFAULT;
6640 to = p->gauth_chunks;
6641 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6642 if (!asoc)
6643 return -EINVAL;
6645 ch = asoc->peer.peer_chunks;
6646 if (!ch)
6647 goto num;
6649 /* See if the user provided enough room for all the data */
6650 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6651 if (len < num_chunks)
6652 return -EINVAL;
6654 if (copy_to_user(to, ch->chunks, num_chunks))
6655 return -EFAULT;
6656 num:
6657 len = sizeof(struct sctp_authchunks) + num_chunks;
6658 if (put_user(len, optlen))
6659 return -EFAULT;
6660 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6661 return -EFAULT;
6662 return 0;
6665 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6666 char __user *optval, int __user *optlen)
6668 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6669 struct sctp_authchunks __user *p = (void __user *)optval;
6670 struct sctp_authchunks val;
6671 struct sctp_association *asoc;
6672 struct sctp_chunks_param *ch;
6673 u32 num_chunks = 0;
6674 char __user *to;
6676 if (!ep->auth_enable)
6677 return -EACCES;
6679 if (len < sizeof(struct sctp_authchunks))
6680 return -EINVAL;
6682 if (copy_from_user(&val, optval, sizeof(val)))
6683 return -EFAULT;
6685 to = p->gauth_chunks;
6686 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6687 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6688 return -EINVAL;
6690 if (asoc)
6691 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6692 else
6693 ch = ep->auth_chunk_list;
6695 if (!ch)
6696 goto num;
6698 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6699 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6700 return -EINVAL;
6702 if (copy_to_user(to, ch->chunks, num_chunks))
6703 return -EFAULT;
6704 num:
6705 len = sizeof(struct sctp_authchunks) + num_chunks;
6706 if (put_user(len, optlen))
6707 return -EFAULT;
6708 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6709 return -EFAULT;
6711 return 0;
6715 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6716 * This option gets the current number of associations that are attached
6717 * to a one-to-many style socket. The option value is an uint32_t.
6719 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6720 char __user *optval, int __user *optlen)
6722 struct sctp_sock *sp = sctp_sk(sk);
6723 struct sctp_association *asoc;
6724 u32 val = 0;
6726 if (sctp_style(sk, TCP))
6727 return -EOPNOTSUPP;
6729 if (len < sizeof(u32))
6730 return -EINVAL;
6732 len = sizeof(u32);
6734 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6735 val++;
6738 if (put_user(len, optlen))
6739 return -EFAULT;
6740 if (copy_to_user(optval, &val, len))
6741 return -EFAULT;
6743 return 0;
6747 * 8.1.23 SCTP_AUTO_ASCONF
6748 * See the corresponding setsockopt entry as description
6750 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6751 char __user *optval, int __user *optlen)
6753 int val = 0;
6755 if (len < sizeof(int))
6756 return -EINVAL;
6758 len = sizeof(int);
6759 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6760 val = 1;
6761 if (put_user(len, optlen))
6762 return -EFAULT;
6763 if (copy_to_user(optval, &val, len))
6764 return -EFAULT;
6765 return 0;
6769 * 8.2.6. Get the Current Identifiers of Associations
6770 * (SCTP_GET_ASSOC_ID_LIST)
6772 * This option gets the current list of SCTP association identifiers of
6773 * the SCTP associations handled by a one-to-many style socket.
6775 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6776 char __user *optval, int __user *optlen)
6778 struct sctp_sock *sp = sctp_sk(sk);
6779 struct sctp_association *asoc;
6780 struct sctp_assoc_ids *ids;
6781 u32 num = 0;
6783 if (sctp_style(sk, TCP))
6784 return -EOPNOTSUPP;
6786 if (len < sizeof(struct sctp_assoc_ids))
6787 return -EINVAL;
6789 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6790 num++;
6793 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6794 return -EINVAL;
6796 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6798 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6799 if (unlikely(!ids))
6800 return -ENOMEM;
6802 ids->gaids_number_of_ids = num;
6803 num = 0;
6804 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6805 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6808 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6809 kfree(ids);
6810 return -EFAULT;
6813 kfree(ids);
6814 return 0;
6818 * SCTP_PEER_ADDR_THLDS
6820 * This option allows us to fetch the partially failed threshold for one or all
6821 * transports in an association. See Section 6.1 of:
6822 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6824 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6825 char __user *optval,
6826 int len,
6827 int __user *optlen)
6829 struct sctp_paddrthlds val;
6830 struct sctp_transport *trans;
6831 struct sctp_association *asoc;
6833 if (len < sizeof(struct sctp_paddrthlds))
6834 return -EINVAL;
6835 len = sizeof(struct sctp_paddrthlds);
6836 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6837 return -EFAULT;
6839 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6840 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6841 if (!asoc)
6842 return -ENOENT;
6844 val.spt_pathpfthld = asoc->pf_retrans;
6845 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6846 } else {
6847 trans = sctp_addr_id2transport(sk, &val.spt_address,
6848 val.spt_assoc_id);
6849 if (!trans)
6850 return -ENOENT;
6852 val.spt_pathmaxrxt = trans->pathmaxrxt;
6853 val.spt_pathpfthld = trans->pf_retrans;
6856 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6857 return -EFAULT;
6859 return 0;
6863 * SCTP_GET_ASSOC_STATS
6865 * This option retrieves local per endpoint statistics. It is modeled
6866 * after OpenSolaris' implementation
6868 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6869 char __user *optval,
6870 int __user *optlen)
6872 struct sctp_assoc_stats sas;
6873 struct sctp_association *asoc = NULL;
6875 /* User must provide at least the assoc id */
6876 if (len < sizeof(sctp_assoc_t))
6877 return -EINVAL;
6879 /* Allow the struct to grow and fill in as much as possible */
6880 len = min_t(size_t, len, sizeof(sas));
6882 if (copy_from_user(&sas, optval, len))
6883 return -EFAULT;
6885 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6886 if (!asoc)
6887 return -EINVAL;
6889 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6890 sas.sas_gapcnt = asoc->stats.gapcnt;
6891 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6892 sas.sas_osacks = asoc->stats.osacks;
6893 sas.sas_isacks = asoc->stats.isacks;
6894 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6895 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6896 sas.sas_oodchunks = asoc->stats.oodchunks;
6897 sas.sas_iodchunks = asoc->stats.iodchunks;
6898 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6899 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6900 sas.sas_idupchunks = asoc->stats.idupchunks;
6901 sas.sas_opackets = asoc->stats.opackets;
6902 sas.sas_ipackets = asoc->stats.ipackets;
6904 /* New high max rto observed, will return 0 if not a single
6905 * RTO update took place. obs_rto_ipaddr will be bogus
6906 * in such a case
6908 sas.sas_maxrto = asoc->stats.max_obs_rto;
6909 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6910 sizeof(struct sockaddr_storage));
6912 /* Mark beginning of a new observation period */
6913 asoc->stats.max_obs_rto = asoc->rto_min;
6915 if (put_user(len, optlen))
6916 return -EFAULT;
6918 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6920 if (copy_to_user(optval, &sas, len))
6921 return -EFAULT;
6923 return 0;
6926 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6927 char __user *optval,
6928 int __user *optlen)
6930 int val = 0;
6932 if (len < sizeof(int))
6933 return -EINVAL;
6935 len = sizeof(int);
6936 if (sctp_sk(sk)->recvrcvinfo)
6937 val = 1;
6938 if (put_user(len, optlen))
6939 return -EFAULT;
6940 if (copy_to_user(optval, &val, len))
6941 return -EFAULT;
6943 return 0;
6946 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6947 char __user *optval,
6948 int __user *optlen)
6950 int val = 0;
6952 if (len < sizeof(int))
6953 return -EINVAL;
6955 len = sizeof(int);
6956 if (sctp_sk(sk)->recvnxtinfo)
6957 val = 1;
6958 if (put_user(len, optlen))
6959 return -EFAULT;
6960 if (copy_to_user(optval, &val, len))
6961 return -EFAULT;
6963 return 0;
6966 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6967 char __user *optval,
6968 int __user *optlen)
6970 struct sctp_assoc_value params;
6971 struct sctp_association *asoc;
6972 int retval = -EFAULT;
6974 if (len < sizeof(params)) {
6975 retval = -EINVAL;
6976 goto out;
6979 len = sizeof(params);
6980 if (copy_from_user(&params, optval, len))
6981 goto out;
6983 asoc = sctp_id2assoc(sk, params.assoc_id);
6984 if (asoc) {
6985 params.assoc_value = asoc->prsctp_enable;
6986 } else if (!params.assoc_id) {
6987 struct sctp_sock *sp = sctp_sk(sk);
6989 params.assoc_value = sp->ep->prsctp_enable;
6990 } else {
6991 retval = -EINVAL;
6992 goto out;
6995 if (put_user(len, optlen))
6996 goto out;
6998 if (copy_to_user(optval, &params, len))
6999 goto out;
7001 retval = 0;
7003 out:
7004 return retval;
7007 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
7008 char __user *optval,
7009 int __user *optlen)
7011 struct sctp_default_prinfo info;
7012 struct sctp_association *asoc;
7013 int retval = -EFAULT;
7015 if (len < sizeof(info)) {
7016 retval = -EINVAL;
7017 goto out;
7020 len = sizeof(info);
7021 if (copy_from_user(&info, optval, len))
7022 goto out;
7024 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
7025 if (asoc) {
7026 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
7027 info.pr_value = asoc->default_timetolive;
7028 } else if (!info.pr_assoc_id) {
7029 struct sctp_sock *sp = sctp_sk(sk);
7031 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
7032 info.pr_value = sp->default_timetolive;
7033 } else {
7034 retval = -EINVAL;
7035 goto out;
7038 if (put_user(len, optlen))
7039 goto out;
7041 if (copy_to_user(optval, &info, len))
7042 goto out;
7044 retval = 0;
7046 out:
7047 return retval;
7050 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
7051 char __user *optval,
7052 int __user *optlen)
7054 struct sctp_prstatus params;
7055 struct sctp_association *asoc;
7056 int policy;
7057 int retval = -EINVAL;
7059 if (len < sizeof(params))
7060 goto out;
7062 len = sizeof(params);
7063 if (copy_from_user(&params, optval, len)) {
7064 retval = -EFAULT;
7065 goto out;
7068 policy = params.sprstat_policy;
7069 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7070 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7071 goto out;
7073 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7074 if (!asoc)
7075 goto out;
7077 if (policy == SCTP_PR_SCTP_ALL) {
7078 params.sprstat_abandoned_unsent = 0;
7079 params.sprstat_abandoned_sent = 0;
7080 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7081 params.sprstat_abandoned_unsent +=
7082 asoc->abandoned_unsent[policy];
7083 params.sprstat_abandoned_sent +=
7084 asoc->abandoned_sent[policy];
7086 } else {
7087 params.sprstat_abandoned_unsent =
7088 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7089 params.sprstat_abandoned_sent =
7090 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
7093 if (put_user(len, optlen)) {
7094 retval = -EFAULT;
7095 goto out;
7098 if (copy_to_user(optval, &params, len)) {
7099 retval = -EFAULT;
7100 goto out;
7103 retval = 0;
7105 out:
7106 return retval;
7109 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
7110 char __user *optval,
7111 int __user *optlen)
7113 struct sctp_stream_out_ext *streamoute;
7114 struct sctp_association *asoc;
7115 struct sctp_prstatus params;
7116 int retval = -EINVAL;
7117 int policy;
7119 if (len < sizeof(params))
7120 goto out;
7122 len = sizeof(params);
7123 if (copy_from_user(&params, optval, len)) {
7124 retval = -EFAULT;
7125 goto out;
7128 policy = params.sprstat_policy;
7129 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7130 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7131 goto out;
7133 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7134 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
7135 goto out;
7137 streamoute = SCTP_SO(&asoc->stream, params.sprstat_sid)->ext;
7138 if (!streamoute) {
7139 /* Not allocated yet, means all stats are 0 */
7140 params.sprstat_abandoned_unsent = 0;
7141 params.sprstat_abandoned_sent = 0;
7142 retval = 0;
7143 goto out;
7146 if (policy == SCTP_PR_SCTP_ALL) {
7147 params.sprstat_abandoned_unsent = 0;
7148 params.sprstat_abandoned_sent = 0;
7149 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7150 params.sprstat_abandoned_unsent +=
7151 streamoute->abandoned_unsent[policy];
7152 params.sprstat_abandoned_sent +=
7153 streamoute->abandoned_sent[policy];
7155 } else {
7156 params.sprstat_abandoned_unsent =
7157 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7158 params.sprstat_abandoned_sent =
7159 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
7162 if (put_user(len, optlen) || copy_to_user(optval, &params, len)) {
7163 retval = -EFAULT;
7164 goto out;
7167 retval = 0;
7169 out:
7170 return retval;
7173 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
7174 char __user *optval,
7175 int __user *optlen)
7177 struct sctp_assoc_value params;
7178 struct sctp_association *asoc;
7179 int retval = -EFAULT;
7181 if (len < sizeof(params)) {
7182 retval = -EINVAL;
7183 goto out;
7186 len = sizeof(params);
7187 if (copy_from_user(&params, optval, len))
7188 goto out;
7190 asoc = sctp_id2assoc(sk, params.assoc_id);
7191 if (asoc) {
7192 params.assoc_value = asoc->reconf_enable;
7193 } else if (!params.assoc_id) {
7194 struct sctp_sock *sp = sctp_sk(sk);
7196 params.assoc_value = sp->ep->reconf_enable;
7197 } else {
7198 retval = -EINVAL;
7199 goto out;
7202 if (put_user(len, optlen))
7203 goto out;
7205 if (copy_to_user(optval, &params, len))
7206 goto out;
7208 retval = 0;
7210 out:
7211 return retval;
7214 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7215 char __user *optval,
7216 int __user *optlen)
7218 struct sctp_assoc_value params;
7219 struct sctp_association *asoc;
7220 int retval = -EFAULT;
7222 if (len < sizeof(params)) {
7223 retval = -EINVAL;
7224 goto out;
7227 len = sizeof(params);
7228 if (copy_from_user(&params, optval, len))
7229 goto out;
7231 asoc = sctp_id2assoc(sk, params.assoc_id);
7232 if (asoc) {
7233 params.assoc_value = asoc->strreset_enable;
7234 } else if (!params.assoc_id) {
7235 struct sctp_sock *sp = sctp_sk(sk);
7237 params.assoc_value = sp->ep->strreset_enable;
7238 } else {
7239 retval = -EINVAL;
7240 goto out;
7243 if (put_user(len, optlen))
7244 goto out;
7246 if (copy_to_user(optval, &params, len))
7247 goto out;
7249 retval = 0;
7251 out:
7252 return retval;
7255 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7256 char __user *optval,
7257 int __user *optlen)
7259 struct sctp_assoc_value params;
7260 struct sctp_association *asoc;
7261 int retval = -EFAULT;
7263 if (len < sizeof(params)) {
7264 retval = -EINVAL;
7265 goto out;
7268 len = sizeof(params);
7269 if (copy_from_user(&params, optval, len))
7270 goto out;
7272 asoc = sctp_id2assoc(sk, params.assoc_id);
7273 if (!asoc) {
7274 retval = -EINVAL;
7275 goto out;
7278 params.assoc_value = sctp_sched_get_sched(asoc);
7280 if (put_user(len, optlen))
7281 goto out;
7283 if (copy_to_user(optval, &params, len))
7284 goto out;
7286 retval = 0;
7288 out:
7289 return retval;
7292 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7293 char __user *optval,
7294 int __user *optlen)
7296 struct sctp_stream_value params;
7297 struct sctp_association *asoc;
7298 int retval = -EFAULT;
7300 if (len < sizeof(params)) {
7301 retval = -EINVAL;
7302 goto out;
7305 len = sizeof(params);
7306 if (copy_from_user(&params, optval, len))
7307 goto out;
7309 asoc = sctp_id2assoc(sk, params.assoc_id);
7310 if (!asoc) {
7311 retval = -EINVAL;
7312 goto out;
7315 retval = sctp_sched_get_value(asoc, params.stream_id,
7316 &params.stream_value);
7317 if (retval)
7318 goto out;
7320 if (put_user(len, optlen)) {
7321 retval = -EFAULT;
7322 goto out;
7325 if (copy_to_user(optval, &params, len)) {
7326 retval = -EFAULT;
7327 goto out;
7330 out:
7331 return retval;
7334 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7335 char __user *optval,
7336 int __user *optlen)
7338 struct sctp_assoc_value params;
7339 struct sctp_association *asoc;
7340 int retval = -EFAULT;
7342 if (len < sizeof(params)) {
7343 retval = -EINVAL;
7344 goto out;
7347 len = sizeof(params);
7348 if (copy_from_user(&params, optval, len))
7349 goto out;
7351 asoc = sctp_id2assoc(sk, params.assoc_id);
7352 if (asoc) {
7353 params.assoc_value = asoc->intl_enable;
7354 } else if (!params.assoc_id) {
7355 struct sctp_sock *sp = sctp_sk(sk);
7357 params.assoc_value = sp->strm_interleave;
7358 } else {
7359 retval = -EINVAL;
7360 goto out;
7363 if (put_user(len, optlen))
7364 goto out;
7366 if (copy_to_user(optval, &params, len))
7367 goto out;
7369 retval = 0;
7371 out:
7372 return retval;
7375 static int sctp_getsockopt_reuse_port(struct sock *sk, int len,
7376 char __user *optval,
7377 int __user *optlen)
7379 int val;
7381 if (len < sizeof(int))
7382 return -EINVAL;
7384 len = sizeof(int);
7385 val = sctp_sk(sk)->reuse;
7386 if (put_user(len, optlen))
7387 return -EFAULT;
7389 if (copy_to_user(optval, &val, len))
7390 return -EFAULT;
7392 return 0;
7395 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7396 char __user *optval, int __user *optlen)
7398 int retval = 0;
7399 int len;
7401 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7403 /* I can hardly begin to describe how wrong this is. This is
7404 * so broken as to be worse than useless. The API draft
7405 * REALLY is NOT helpful here... I am not convinced that the
7406 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7407 * are at all well-founded.
7409 if (level != SOL_SCTP) {
7410 struct sctp_af *af = sctp_sk(sk)->pf->af;
7412 retval = af->getsockopt(sk, level, optname, optval, optlen);
7413 return retval;
7416 if (get_user(len, optlen))
7417 return -EFAULT;
7419 if (len < 0)
7420 return -EINVAL;
7422 lock_sock(sk);
7424 switch (optname) {
7425 case SCTP_STATUS:
7426 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7427 break;
7428 case SCTP_DISABLE_FRAGMENTS:
7429 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7430 optlen);
7431 break;
7432 case SCTP_EVENTS:
7433 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7434 break;
7435 case SCTP_AUTOCLOSE:
7436 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7437 break;
7438 case SCTP_SOCKOPT_PEELOFF:
7439 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7440 break;
7441 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7442 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7443 break;
7444 case SCTP_PEER_ADDR_PARAMS:
7445 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7446 optlen);
7447 break;
7448 case SCTP_DELAYED_SACK:
7449 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7450 optlen);
7451 break;
7452 case SCTP_INITMSG:
7453 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7454 break;
7455 case SCTP_GET_PEER_ADDRS:
7456 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7457 optlen);
7458 break;
7459 case SCTP_GET_LOCAL_ADDRS:
7460 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7461 optlen);
7462 break;
7463 case SCTP_SOCKOPT_CONNECTX3:
7464 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7465 break;
7466 case SCTP_DEFAULT_SEND_PARAM:
7467 retval = sctp_getsockopt_default_send_param(sk, len,
7468 optval, optlen);
7469 break;
7470 case SCTP_DEFAULT_SNDINFO:
7471 retval = sctp_getsockopt_default_sndinfo(sk, len,
7472 optval, optlen);
7473 break;
7474 case SCTP_PRIMARY_ADDR:
7475 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7476 break;
7477 case SCTP_NODELAY:
7478 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7479 break;
7480 case SCTP_RTOINFO:
7481 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7482 break;
7483 case SCTP_ASSOCINFO:
7484 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7485 break;
7486 case SCTP_I_WANT_MAPPED_V4_ADDR:
7487 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7488 break;
7489 case SCTP_MAXSEG:
7490 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7491 break;
7492 case SCTP_GET_PEER_ADDR_INFO:
7493 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7494 optlen);
7495 break;
7496 case SCTP_ADAPTATION_LAYER:
7497 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7498 optlen);
7499 break;
7500 case SCTP_CONTEXT:
7501 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7502 break;
7503 case SCTP_FRAGMENT_INTERLEAVE:
7504 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7505 optlen);
7506 break;
7507 case SCTP_PARTIAL_DELIVERY_POINT:
7508 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7509 optlen);
7510 break;
7511 case SCTP_MAX_BURST:
7512 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7513 break;
7514 case SCTP_AUTH_KEY:
7515 case SCTP_AUTH_CHUNK:
7516 case SCTP_AUTH_DELETE_KEY:
7517 case SCTP_AUTH_DEACTIVATE_KEY:
7518 retval = -EOPNOTSUPP;
7519 break;
7520 case SCTP_HMAC_IDENT:
7521 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7522 break;
7523 case SCTP_AUTH_ACTIVE_KEY:
7524 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7525 break;
7526 case SCTP_PEER_AUTH_CHUNKS:
7527 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7528 optlen);
7529 break;
7530 case SCTP_LOCAL_AUTH_CHUNKS:
7531 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7532 optlen);
7533 break;
7534 case SCTP_GET_ASSOC_NUMBER:
7535 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7536 break;
7537 case SCTP_GET_ASSOC_ID_LIST:
7538 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7539 break;
7540 case SCTP_AUTO_ASCONF:
7541 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7542 break;
7543 case SCTP_PEER_ADDR_THLDS:
7544 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7545 break;
7546 case SCTP_GET_ASSOC_STATS:
7547 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7548 break;
7549 case SCTP_RECVRCVINFO:
7550 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7551 break;
7552 case SCTP_RECVNXTINFO:
7553 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7554 break;
7555 case SCTP_PR_SUPPORTED:
7556 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7557 break;
7558 case SCTP_DEFAULT_PRINFO:
7559 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7560 optlen);
7561 break;
7562 case SCTP_PR_ASSOC_STATUS:
7563 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7564 optlen);
7565 break;
7566 case SCTP_PR_STREAM_STATUS:
7567 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7568 optlen);
7569 break;
7570 case SCTP_RECONFIG_SUPPORTED:
7571 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7572 optlen);
7573 break;
7574 case SCTP_ENABLE_STREAM_RESET:
7575 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7576 optlen);
7577 break;
7578 case SCTP_STREAM_SCHEDULER:
7579 retval = sctp_getsockopt_scheduler(sk, len, optval,
7580 optlen);
7581 break;
7582 case SCTP_STREAM_SCHEDULER_VALUE:
7583 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7584 optlen);
7585 break;
7586 case SCTP_INTERLEAVING_SUPPORTED:
7587 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7588 optlen);
7589 break;
7590 case SCTP_REUSE_PORT:
7591 retval = sctp_getsockopt_reuse_port(sk, len, optval, optlen);
7592 break;
7593 default:
7594 retval = -ENOPROTOOPT;
7595 break;
7598 release_sock(sk);
7599 return retval;
7602 static int sctp_hash(struct sock *sk)
7604 /* STUB */
7605 return 0;
7608 static void sctp_unhash(struct sock *sk)
7610 /* STUB */
7613 /* Check if port is acceptable. Possibly find first available port.
7615 * The port hash table (contained in the 'global' SCTP protocol storage
7616 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7617 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7618 * list (the list number is the port number hashed out, so as you
7619 * would expect from a hash function, all the ports in a given list have
7620 * such a number that hashes out to the same list number; you were
7621 * expecting that, right?); so each list has a set of ports, with a
7622 * link to the socket (struct sock) that uses it, the port number and
7623 * a fastreuse flag (FIXME: NPI ipg).
7625 static struct sctp_bind_bucket *sctp_bucket_create(
7626 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7628 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7630 bool reuse = (sk->sk_reuse || sctp_sk(sk)->reuse);
7631 struct sctp_bind_hashbucket *head; /* hash list */
7632 struct sctp_bind_bucket *pp;
7633 unsigned short snum;
7634 int ret;
7636 snum = ntohs(addr->v4.sin_port);
7638 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7640 local_bh_disable();
7642 if (snum == 0) {
7643 /* Search for an available port. */
7644 int low, high, remaining, index;
7645 unsigned int rover;
7646 struct net *net = sock_net(sk);
7648 inet_get_local_port_range(net, &low, &high);
7649 remaining = (high - low) + 1;
7650 rover = prandom_u32() % remaining + low;
7652 do {
7653 rover++;
7654 if ((rover < low) || (rover > high))
7655 rover = low;
7656 if (inet_is_local_reserved_port(net, rover))
7657 continue;
7658 index = sctp_phashfn(sock_net(sk), rover);
7659 head = &sctp_port_hashtable[index];
7660 spin_lock(&head->lock);
7661 sctp_for_each_hentry(pp, &head->chain)
7662 if ((pp->port == rover) &&
7663 net_eq(sock_net(sk), pp->net))
7664 goto next;
7665 break;
7666 next:
7667 spin_unlock(&head->lock);
7668 } while (--remaining > 0);
7670 /* Exhausted local port range during search? */
7671 ret = 1;
7672 if (remaining <= 0)
7673 goto fail;
7675 /* OK, here is the one we will use. HEAD (the port
7676 * hash table list entry) is non-NULL and we hold it's
7677 * mutex.
7679 snum = rover;
7680 } else {
7681 /* We are given an specific port number; we verify
7682 * that it is not being used. If it is used, we will
7683 * exahust the search in the hash list corresponding
7684 * to the port number (snum) - we detect that with the
7685 * port iterator, pp being NULL.
7687 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7688 spin_lock(&head->lock);
7689 sctp_for_each_hentry(pp, &head->chain) {
7690 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7691 goto pp_found;
7694 pp = NULL;
7695 goto pp_not_found;
7696 pp_found:
7697 if (!hlist_empty(&pp->owner)) {
7698 /* We had a port hash table hit - there is an
7699 * available port (pp != NULL) and it is being
7700 * used by other socket (pp->owner not empty); that other
7701 * socket is going to be sk2.
7703 struct sock *sk2;
7705 pr_debug("%s: found a possible match\n", __func__);
7707 if (pp->fastreuse && reuse && sk->sk_state != SCTP_SS_LISTENING)
7708 goto success;
7710 /* Run through the list of sockets bound to the port
7711 * (pp->port) [via the pointers bind_next and
7712 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7713 * we get the endpoint they describe and run through
7714 * the endpoint's list of IP (v4 or v6) addresses,
7715 * comparing each of the addresses with the address of
7716 * the socket sk. If we find a match, then that means
7717 * that this port/socket (sk) combination are already
7718 * in an endpoint.
7720 sk_for_each_bound(sk2, &pp->owner) {
7721 struct sctp_endpoint *ep2;
7722 ep2 = sctp_sk(sk2)->ep;
7724 if (sk == sk2 ||
7725 (reuse && (sk2->sk_reuse || sctp_sk(sk2)->reuse) &&
7726 sk2->sk_state != SCTP_SS_LISTENING))
7727 continue;
7729 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7730 sctp_sk(sk2), sctp_sk(sk))) {
7731 ret = (long)sk2;
7732 goto fail_unlock;
7736 pr_debug("%s: found a match\n", __func__);
7738 pp_not_found:
7739 /* If there was a hash table miss, create a new port. */
7740 ret = 1;
7741 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7742 goto fail_unlock;
7744 /* In either case (hit or miss), make sure fastreuse is 1 only
7745 * if sk->sk_reuse is too (that is, if the caller requested
7746 * SO_REUSEADDR on this socket -sk-).
7748 if (hlist_empty(&pp->owner)) {
7749 if (reuse && sk->sk_state != SCTP_SS_LISTENING)
7750 pp->fastreuse = 1;
7751 else
7752 pp->fastreuse = 0;
7753 } else if (pp->fastreuse &&
7754 (!reuse || sk->sk_state == SCTP_SS_LISTENING))
7755 pp->fastreuse = 0;
7757 /* We are set, so fill up all the data in the hash table
7758 * entry, tie the socket list information with the rest of the
7759 * sockets FIXME: Blurry, NPI (ipg).
7761 success:
7762 if (!sctp_sk(sk)->bind_hash) {
7763 inet_sk(sk)->inet_num = snum;
7764 sk_add_bind_node(sk, &pp->owner);
7765 sctp_sk(sk)->bind_hash = pp;
7767 ret = 0;
7769 fail_unlock:
7770 spin_unlock(&head->lock);
7772 fail:
7773 local_bh_enable();
7774 return ret;
7777 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7778 * port is requested.
7780 static int sctp_get_port(struct sock *sk, unsigned short snum)
7782 union sctp_addr addr;
7783 struct sctp_af *af = sctp_sk(sk)->pf->af;
7785 /* Set up a dummy address struct from the sk. */
7786 af->from_sk(&addr, sk);
7787 addr.v4.sin_port = htons(snum);
7789 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7790 return !!sctp_get_port_local(sk, &addr);
7794 * Move a socket to LISTENING state.
7796 static int sctp_listen_start(struct sock *sk, int backlog)
7798 struct sctp_sock *sp = sctp_sk(sk);
7799 struct sctp_endpoint *ep = sp->ep;
7800 struct crypto_shash *tfm = NULL;
7801 char alg[32];
7803 /* Allocate HMAC for generating cookie. */
7804 if (!sp->hmac && sp->sctp_hmac_alg) {
7805 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7806 tfm = crypto_alloc_shash(alg, 0, 0);
7807 if (IS_ERR(tfm)) {
7808 net_info_ratelimited("failed to load transform for %s: %ld\n",
7809 sp->sctp_hmac_alg, PTR_ERR(tfm));
7810 return -ENOSYS;
7812 sctp_sk(sk)->hmac = tfm;
7816 * If a bind() or sctp_bindx() is not called prior to a listen()
7817 * call that allows new associations to be accepted, the system
7818 * picks an ephemeral port and will choose an address set equivalent
7819 * to binding with a wildcard address.
7821 * This is not currently spelled out in the SCTP sockets
7822 * extensions draft, but follows the practice as seen in TCP
7823 * sockets.
7826 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7827 if (!ep->base.bind_addr.port) {
7828 if (sctp_autobind(sk))
7829 return -EAGAIN;
7830 } else {
7831 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7832 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7833 return -EADDRINUSE;
7837 sk->sk_max_ack_backlog = backlog;
7838 sctp_hash_endpoint(ep);
7839 return 0;
7843 * 4.1.3 / 5.1.3 listen()
7845 * By default, new associations are not accepted for UDP style sockets.
7846 * An application uses listen() to mark a socket as being able to
7847 * accept new associations.
7849 * On TCP style sockets, applications use listen() to ready the SCTP
7850 * endpoint for accepting inbound associations.
7852 * On both types of endpoints a backlog of '0' disables listening.
7854 * Move a socket to LISTENING state.
7856 int sctp_inet_listen(struct socket *sock, int backlog)
7858 struct sock *sk = sock->sk;
7859 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7860 int err = -EINVAL;
7862 if (unlikely(backlog < 0))
7863 return err;
7865 lock_sock(sk);
7867 /* Peeled-off sockets are not allowed to listen(). */
7868 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7869 goto out;
7871 if (sock->state != SS_UNCONNECTED)
7872 goto out;
7874 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7875 goto out;
7877 /* If backlog is zero, disable listening. */
7878 if (!backlog) {
7879 if (sctp_sstate(sk, CLOSED))
7880 goto out;
7882 err = 0;
7883 sctp_unhash_endpoint(ep);
7884 sk->sk_state = SCTP_SS_CLOSED;
7885 if (sk->sk_reuse || sctp_sk(sk)->reuse)
7886 sctp_sk(sk)->bind_hash->fastreuse = 1;
7887 goto out;
7890 /* If we are already listening, just update the backlog */
7891 if (sctp_sstate(sk, LISTENING))
7892 sk->sk_max_ack_backlog = backlog;
7893 else {
7894 err = sctp_listen_start(sk, backlog);
7895 if (err)
7896 goto out;
7899 err = 0;
7900 out:
7901 release_sock(sk);
7902 return err;
7906 * This function is done by modeling the current datagram_poll() and the
7907 * tcp_poll(). Note that, based on these implementations, we don't
7908 * lock the socket in this function, even though it seems that,
7909 * ideally, locking or some other mechanisms can be used to ensure
7910 * the integrity of the counters (sndbuf and wmem_alloc) used
7911 * in this place. We assume that we don't need locks either until proven
7912 * otherwise.
7914 * Another thing to note is that we include the Async I/O support
7915 * here, again, by modeling the current TCP/UDP code. We don't have
7916 * a good way to test with it yet.
7918 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7920 struct sock *sk = sock->sk;
7921 struct sctp_sock *sp = sctp_sk(sk);
7922 __poll_t mask;
7924 poll_wait(file, sk_sleep(sk), wait);
7926 sock_rps_record_flow(sk);
7928 /* A TCP-style listening socket becomes readable when the accept queue
7929 * is not empty.
7931 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7932 return (!list_empty(&sp->ep->asocs)) ?
7933 (EPOLLIN | EPOLLRDNORM) : 0;
7935 mask = 0;
7937 /* Is there any exceptional events? */
7938 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7939 mask |= EPOLLERR |
7940 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7941 if (sk->sk_shutdown & RCV_SHUTDOWN)
7942 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7943 if (sk->sk_shutdown == SHUTDOWN_MASK)
7944 mask |= EPOLLHUP;
7946 /* Is it readable? Reconsider this code with TCP-style support. */
7947 if (!skb_queue_empty(&sk->sk_receive_queue))
7948 mask |= EPOLLIN | EPOLLRDNORM;
7950 /* The association is either gone or not ready. */
7951 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7952 return mask;
7954 /* Is it writable? */
7955 if (sctp_writeable(sk)) {
7956 mask |= EPOLLOUT | EPOLLWRNORM;
7957 } else {
7958 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7960 * Since the socket is not locked, the buffer
7961 * might be made available after the writeable check and
7962 * before the bit is set. This could cause a lost I/O
7963 * signal. tcp_poll() has a race breaker for this race
7964 * condition. Based on their implementation, we put
7965 * in the following code to cover it as well.
7967 if (sctp_writeable(sk))
7968 mask |= EPOLLOUT | EPOLLWRNORM;
7970 return mask;
7973 /********************************************************************
7974 * 2nd Level Abstractions
7975 ********************************************************************/
7977 static struct sctp_bind_bucket *sctp_bucket_create(
7978 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7980 struct sctp_bind_bucket *pp;
7982 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7983 if (pp) {
7984 SCTP_DBG_OBJCNT_INC(bind_bucket);
7985 pp->port = snum;
7986 pp->fastreuse = 0;
7987 INIT_HLIST_HEAD(&pp->owner);
7988 pp->net = net;
7989 hlist_add_head(&pp->node, &head->chain);
7991 return pp;
7994 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7995 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
7997 if (pp && hlist_empty(&pp->owner)) {
7998 __hlist_del(&pp->node);
7999 kmem_cache_free(sctp_bucket_cachep, pp);
8000 SCTP_DBG_OBJCNT_DEC(bind_bucket);
8004 /* Release this socket's reference to a local port. */
8005 static inline void __sctp_put_port(struct sock *sk)
8007 struct sctp_bind_hashbucket *head =
8008 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
8009 inet_sk(sk)->inet_num)];
8010 struct sctp_bind_bucket *pp;
8012 spin_lock(&head->lock);
8013 pp = sctp_sk(sk)->bind_hash;
8014 __sk_del_bind_node(sk);
8015 sctp_sk(sk)->bind_hash = NULL;
8016 inet_sk(sk)->inet_num = 0;
8017 sctp_bucket_destroy(pp);
8018 spin_unlock(&head->lock);
8021 void sctp_put_port(struct sock *sk)
8023 local_bh_disable();
8024 __sctp_put_port(sk);
8025 local_bh_enable();
8029 * The system picks an ephemeral port and choose an address set equivalent
8030 * to binding with a wildcard address.
8031 * One of those addresses will be the primary address for the association.
8032 * This automatically enables the multihoming capability of SCTP.
8034 static int sctp_autobind(struct sock *sk)
8036 union sctp_addr autoaddr;
8037 struct sctp_af *af;
8038 __be16 port;
8040 /* Initialize a local sockaddr structure to INADDR_ANY. */
8041 af = sctp_sk(sk)->pf->af;
8043 port = htons(inet_sk(sk)->inet_num);
8044 af->inaddr_any(&autoaddr, port);
8046 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
8049 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
8051 * From RFC 2292
8052 * 4.2 The cmsghdr Structure *
8054 * When ancillary data is sent or received, any number of ancillary data
8055 * objects can be specified by the msg_control and msg_controllen members of
8056 * the msghdr structure, because each object is preceded by
8057 * a cmsghdr structure defining the object's length (the cmsg_len member).
8058 * Historically Berkeley-derived implementations have passed only one object
8059 * at a time, but this API allows multiple objects to be
8060 * passed in a single call to sendmsg() or recvmsg(). The following example
8061 * shows two ancillary data objects in a control buffer.
8063 * |<--------------------------- msg_controllen -------------------------->|
8064 * | |
8066 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
8068 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
8069 * | | |
8071 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
8073 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
8074 * | | | | |
8076 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8077 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
8079 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
8081 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8085 * msg_control
8086 * points here
8088 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
8090 struct msghdr *my_msg = (struct msghdr *)msg;
8091 struct cmsghdr *cmsg;
8093 for_each_cmsghdr(cmsg, my_msg) {
8094 if (!CMSG_OK(my_msg, cmsg))
8095 return -EINVAL;
8097 /* Should we parse this header or ignore? */
8098 if (cmsg->cmsg_level != IPPROTO_SCTP)
8099 continue;
8101 /* Strictly check lengths following example in SCM code. */
8102 switch (cmsg->cmsg_type) {
8103 case SCTP_INIT:
8104 /* SCTP Socket API Extension
8105 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
8107 * This cmsghdr structure provides information for
8108 * initializing new SCTP associations with sendmsg().
8109 * The SCTP_INITMSG socket option uses this same data
8110 * structure. This structure is not used for
8111 * recvmsg().
8113 * cmsg_level cmsg_type cmsg_data[]
8114 * ------------ ------------ ----------------------
8115 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
8117 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
8118 return -EINVAL;
8120 cmsgs->init = CMSG_DATA(cmsg);
8121 break;
8123 case SCTP_SNDRCV:
8124 /* SCTP Socket API Extension
8125 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
8127 * This cmsghdr structure specifies SCTP options for
8128 * sendmsg() and describes SCTP header information
8129 * about a received message through recvmsg().
8131 * cmsg_level cmsg_type cmsg_data[]
8132 * ------------ ------------ ----------------------
8133 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
8135 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
8136 return -EINVAL;
8138 cmsgs->srinfo = CMSG_DATA(cmsg);
8140 if (cmsgs->srinfo->sinfo_flags &
8141 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8142 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8143 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8144 return -EINVAL;
8145 break;
8147 case SCTP_SNDINFO:
8148 /* SCTP Socket API Extension
8149 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
8151 * This cmsghdr structure specifies SCTP options for
8152 * sendmsg(). This structure and SCTP_RCVINFO replaces
8153 * SCTP_SNDRCV which has been deprecated.
8155 * cmsg_level cmsg_type cmsg_data[]
8156 * ------------ ------------ ---------------------
8157 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
8159 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
8160 return -EINVAL;
8162 cmsgs->sinfo = CMSG_DATA(cmsg);
8164 if (cmsgs->sinfo->snd_flags &
8165 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8166 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8167 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8168 return -EINVAL;
8169 break;
8170 case SCTP_PRINFO:
8171 /* SCTP Socket API Extension
8172 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
8174 * This cmsghdr structure specifies SCTP options for sendmsg().
8176 * cmsg_level cmsg_type cmsg_data[]
8177 * ------------ ------------ ---------------------
8178 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
8180 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
8181 return -EINVAL;
8183 cmsgs->prinfo = CMSG_DATA(cmsg);
8184 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
8185 return -EINVAL;
8187 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
8188 cmsgs->prinfo->pr_value = 0;
8189 break;
8190 case SCTP_AUTHINFO:
8191 /* SCTP Socket API Extension
8192 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
8194 * This cmsghdr structure specifies SCTP options for sendmsg().
8196 * cmsg_level cmsg_type cmsg_data[]
8197 * ------------ ------------ ---------------------
8198 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
8200 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
8201 return -EINVAL;
8203 cmsgs->authinfo = CMSG_DATA(cmsg);
8204 break;
8205 case SCTP_DSTADDRV4:
8206 case SCTP_DSTADDRV6:
8207 /* SCTP Socket API Extension
8208 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
8210 * This cmsghdr structure specifies SCTP options for sendmsg().
8212 * cmsg_level cmsg_type cmsg_data[]
8213 * ------------ ------------ ---------------------
8214 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8215 * ------------ ------------ ---------------------
8216 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8218 cmsgs->addrs_msg = my_msg;
8219 break;
8220 default:
8221 return -EINVAL;
8225 return 0;
8229 * Wait for a packet..
8230 * Note: This function is the same function as in core/datagram.c
8231 * with a few modifications to make lksctp work.
8233 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8235 int error;
8236 DEFINE_WAIT(wait);
8238 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8240 /* Socket errors? */
8241 error = sock_error(sk);
8242 if (error)
8243 goto out;
8245 if (!skb_queue_empty(&sk->sk_receive_queue))
8246 goto ready;
8248 /* Socket shut down? */
8249 if (sk->sk_shutdown & RCV_SHUTDOWN)
8250 goto out;
8252 /* Sequenced packets can come disconnected. If so we report the
8253 * problem.
8255 error = -ENOTCONN;
8257 /* Is there a good reason to think that we may receive some data? */
8258 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8259 goto out;
8261 /* Handle signals. */
8262 if (signal_pending(current))
8263 goto interrupted;
8265 /* Let another process have a go. Since we are going to sleep
8266 * anyway. Note: This may cause odd behaviors if the message
8267 * does not fit in the user's buffer, but this seems to be the
8268 * only way to honor MSG_DONTWAIT realistically.
8270 release_sock(sk);
8271 *timeo_p = schedule_timeout(*timeo_p);
8272 lock_sock(sk);
8274 ready:
8275 finish_wait(sk_sleep(sk), &wait);
8276 return 0;
8278 interrupted:
8279 error = sock_intr_errno(*timeo_p);
8281 out:
8282 finish_wait(sk_sleep(sk), &wait);
8283 *err = error;
8284 return error;
8287 /* Receive a datagram.
8288 * Note: This is pretty much the same routine as in core/datagram.c
8289 * with a few changes to make lksctp work.
8291 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8292 int noblock, int *err)
8294 int error;
8295 struct sk_buff *skb;
8296 long timeo;
8298 timeo = sock_rcvtimeo(sk, noblock);
8300 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8301 MAX_SCHEDULE_TIMEOUT);
8303 do {
8304 /* Again only user level code calls this function,
8305 * so nothing interrupt level
8306 * will suddenly eat the receive_queue.
8308 * Look at current nfs client by the way...
8309 * However, this function was correct in any case. 8)
8311 if (flags & MSG_PEEK) {
8312 skb = skb_peek(&sk->sk_receive_queue);
8313 if (skb)
8314 refcount_inc(&skb->users);
8315 } else {
8316 skb = __skb_dequeue(&sk->sk_receive_queue);
8319 if (skb)
8320 return skb;
8322 /* Caller is allowed not to check sk->sk_err before calling. */
8323 error = sock_error(sk);
8324 if (error)
8325 goto no_packet;
8327 if (sk->sk_shutdown & RCV_SHUTDOWN)
8328 break;
8330 if (sk_can_busy_loop(sk)) {
8331 sk_busy_loop(sk, noblock);
8333 if (!skb_queue_empty(&sk->sk_receive_queue))
8334 continue;
8337 /* User doesn't want to wait. */
8338 error = -EAGAIN;
8339 if (!timeo)
8340 goto no_packet;
8341 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8343 return NULL;
8345 no_packet:
8346 *err = error;
8347 return NULL;
8350 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8351 static void __sctp_write_space(struct sctp_association *asoc)
8353 struct sock *sk = asoc->base.sk;
8355 if (sctp_wspace(asoc) <= 0)
8356 return;
8358 if (waitqueue_active(&asoc->wait))
8359 wake_up_interruptible(&asoc->wait);
8361 if (sctp_writeable(sk)) {
8362 struct socket_wq *wq;
8364 rcu_read_lock();
8365 wq = rcu_dereference(sk->sk_wq);
8366 if (wq) {
8367 if (waitqueue_active(&wq->wait))
8368 wake_up_interruptible(&wq->wait);
8370 /* Note that we try to include the Async I/O support
8371 * here by modeling from the current TCP/UDP code.
8372 * We have not tested with it yet.
8374 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8375 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8377 rcu_read_unlock();
8381 static void sctp_wake_up_waiters(struct sock *sk,
8382 struct sctp_association *asoc)
8384 struct sctp_association *tmp = asoc;
8386 /* We do accounting for the sndbuf space per association,
8387 * so we only need to wake our own association.
8389 if (asoc->ep->sndbuf_policy)
8390 return __sctp_write_space(asoc);
8392 /* If association goes down and is just flushing its
8393 * outq, then just normally notify others.
8395 if (asoc->base.dead)
8396 return sctp_write_space(sk);
8398 /* Accounting for the sndbuf space is per socket, so we
8399 * need to wake up others, try to be fair and in case of
8400 * other associations, let them have a go first instead
8401 * of just doing a sctp_write_space() call.
8403 * Note that we reach sctp_wake_up_waiters() only when
8404 * associations free up queued chunks, thus we are under
8405 * lock and the list of associations on a socket is
8406 * guaranteed not to change.
8408 for (tmp = list_next_entry(tmp, asocs); 1;
8409 tmp = list_next_entry(tmp, asocs)) {
8410 /* Manually skip the head element. */
8411 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8412 continue;
8413 /* Wake up association. */
8414 __sctp_write_space(tmp);
8415 /* We've reached the end. */
8416 if (tmp == asoc)
8417 break;
8421 /* Do accounting for the sndbuf space.
8422 * Decrement the used sndbuf space of the corresponding association by the
8423 * data size which was just transmitted(freed).
8425 static void sctp_wfree(struct sk_buff *skb)
8427 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8428 struct sctp_association *asoc = chunk->asoc;
8429 struct sock *sk = asoc->base.sk;
8431 sk_mem_uncharge(sk, skb->truesize);
8432 sk->sk_wmem_queued -= skb->truesize + sizeof(struct sctp_chunk);
8433 asoc->sndbuf_used -= skb->truesize + sizeof(struct sctp_chunk);
8434 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk),
8435 &sk->sk_wmem_alloc));
8437 if (chunk->shkey) {
8438 struct sctp_shared_key *shkey = chunk->shkey;
8440 /* refcnt == 2 and !list_empty mean after this release, it's
8441 * not being used anywhere, and it's time to notify userland
8442 * that this shkey can be freed if it's been deactivated.
8444 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8445 refcount_read(&shkey->refcnt) == 2) {
8446 struct sctp_ulpevent *ev;
8448 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8449 SCTP_AUTH_FREE_KEY,
8450 GFP_KERNEL);
8451 if (ev)
8452 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8454 sctp_auth_shkey_release(chunk->shkey);
8457 sock_wfree(skb);
8458 sctp_wake_up_waiters(sk, asoc);
8460 sctp_association_put(asoc);
8463 /* Do accounting for the receive space on the socket.
8464 * Accounting for the association is done in ulpevent.c
8465 * We set this as a destructor for the cloned data skbs so that
8466 * accounting is done at the correct time.
8468 void sctp_sock_rfree(struct sk_buff *skb)
8470 struct sock *sk = skb->sk;
8471 struct sctp_ulpevent *event = sctp_skb2event(skb);
8473 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8476 * Mimic the behavior of sock_rfree
8478 sk_mem_uncharge(sk, event->rmem_len);
8482 /* Helper function to wait for space in the sndbuf. */
8483 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8484 size_t msg_len)
8486 struct sock *sk = asoc->base.sk;
8487 long current_timeo = *timeo_p;
8488 DEFINE_WAIT(wait);
8489 int err = 0;
8491 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8492 *timeo_p, msg_len);
8494 /* Increment the association's refcnt. */
8495 sctp_association_hold(asoc);
8497 /* Wait on the association specific sndbuf space. */
8498 for (;;) {
8499 prepare_to_wait_exclusive(&asoc->wait, &wait,
8500 TASK_INTERRUPTIBLE);
8501 if (asoc->base.dead)
8502 goto do_dead;
8503 if (!*timeo_p)
8504 goto do_nonblock;
8505 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8506 goto do_error;
8507 if (signal_pending(current))
8508 goto do_interrupted;
8509 if ((int)msg_len <= sctp_wspace(asoc))
8510 break;
8512 /* Let another process have a go. Since we are going
8513 * to sleep anyway.
8515 release_sock(sk);
8516 current_timeo = schedule_timeout(current_timeo);
8517 lock_sock(sk);
8518 if (sk != asoc->base.sk)
8519 goto do_error;
8521 *timeo_p = current_timeo;
8524 out:
8525 finish_wait(&asoc->wait, &wait);
8527 /* Release the association's refcnt. */
8528 sctp_association_put(asoc);
8530 return err;
8532 do_dead:
8533 err = -ESRCH;
8534 goto out;
8536 do_error:
8537 err = -EPIPE;
8538 goto out;
8540 do_interrupted:
8541 err = sock_intr_errno(*timeo_p);
8542 goto out;
8544 do_nonblock:
8545 err = -EAGAIN;
8546 goto out;
8549 void sctp_data_ready(struct sock *sk)
8551 struct socket_wq *wq;
8553 rcu_read_lock();
8554 wq = rcu_dereference(sk->sk_wq);
8555 if (skwq_has_sleeper(wq))
8556 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8557 EPOLLRDNORM | EPOLLRDBAND);
8558 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8559 rcu_read_unlock();
8562 /* If socket sndbuf has changed, wake up all per association waiters. */
8563 void sctp_write_space(struct sock *sk)
8565 struct sctp_association *asoc;
8567 /* Wake up the tasks in each wait queue. */
8568 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8569 __sctp_write_space(asoc);
8573 /* Is there any sndbuf space available on the socket?
8575 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8576 * associations on the same socket. For a UDP-style socket with
8577 * multiple associations, it is possible for it to be "unwriteable"
8578 * prematurely. I assume that this is acceptable because
8579 * a premature "unwriteable" is better than an accidental "writeable" which
8580 * would cause an unwanted block under certain circumstances. For the 1-1
8581 * UDP-style sockets or TCP-style sockets, this code should work.
8582 * - Daisy
8584 static bool sctp_writeable(struct sock *sk)
8586 return sk->sk_sndbuf > sk->sk_wmem_queued;
8589 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8590 * returns immediately with EINPROGRESS.
8592 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8594 struct sock *sk = asoc->base.sk;
8595 int err = 0;
8596 long current_timeo = *timeo_p;
8597 DEFINE_WAIT(wait);
8599 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8601 /* Increment the association's refcnt. */
8602 sctp_association_hold(asoc);
8604 for (;;) {
8605 prepare_to_wait_exclusive(&asoc->wait, &wait,
8606 TASK_INTERRUPTIBLE);
8607 if (!*timeo_p)
8608 goto do_nonblock;
8609 if (sk->sk_shutdown & RCV_SHUTDOWN)
8610 break;
8611 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8612 asoc->base.dead)
8613 goto do_error;
8614 if (signal_pending(current))
8615 goto do_interrupted;
8617 if (sctp_state(asoc, ESTABLISHED))
8618 break;
8620 /* Let another process have a go. Since we are going
8621 * to sleep anyway.
8623 release_sock(sk);
8624 current_timeo = schedule_timeout(current_timeo);
8625 lock_sock(sk);
8627 *timeo_p = current_timeo;
8630 out:
8631 finish_wait(&asoc->wait, &wait);
8633 /* Release the association's refcnt. */
8634 sctp_association_put(asoc);
8636 return err;
8638 do_error:
8639 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8640 err = -ETIMEDOUT;
8641 else
8642 err = -ECONNREFUSED;
8643 goto out;
8645 do_interrupted:
8646 err = sock_intr_errno(*timeo_p);
8647 goto out;
8649 do_nonblock:
8650 err = -EINPROGRESS;
8651 goto out;
8654 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8656 struct sctp_endpoint *ep;
8657 int err = 0;
8658 DEFINE_WAIT(wait);
8660 ep = sctp_sk(sk)->ep;
8663 for (;;) {
8664 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8665 TASK_INTERRUPTIBLE);
8667 if (list_empty(&ep->asocs)) {
8668 release_sock(sk);
8669 timeo = schedule_timeout(timeo);
8670 lock_sock(sk);
8673 err = -EINVAL;
8674 if (!sctp_sstate(sk, LISTENING))
8675 break;
8677 err = 0;
8678 if (!list_empty(&ep->asocs))
8679 break;
8681 err = sock_intr_errno(timeo);
8682 if (signal_pending(current))
8683 break;
8685 err = -EAGAIN;
8686 if (!timeo)
8687 break;
8690 finish_wait(sk_sleep(sk), &wait);
8692 return err;
8695 static void sctp_wait_for_close(struct sock *sk, long timeout)
8697 DEFINE_WAIT(wait);
8699 do {
8700 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8701 if (list_empty(&sctp_sk(sk)->ep->asocs))
8702 break;
8703 release_sock(sk);
8704 timeout = schedule_timeout(timeout);
8705 lock_sock(sk);
8706 } while (!signal_pending(current) && timeout);
8708 finish_wait(sk_sleep(sk), &wait);
8711 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8713 struct sk_buff *frag;
8715 if (!skb->data_len)
8716 goto done;
8718 /* Don't forget the fragments. */
8719 skb_walk_frags(skb, frag)
8720 sctp_skb_set_owner_r_frag(frag, sk);
8722 done:
8723 sctp_skb_set_owner_r(skb, sk);
8726 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8727 struct sctp_association *asoc)
8729 struct inet_sock *inet = inet_sk(sk);
8730 struct inet_sock *newinet;
8731 struct sctp_sock *sp = sctp_sk(sk);
8732 struct sctp_endpoint *ep = sp->ep;
8734 newsk->sk_type = sk->sk_type;
8735 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8736 newsk->sk_flags = sk->sk_flags;
8737 newsk->sk_tsflags = sk->sk_tsflags;
8738 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8739 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8740 newsk->sk_reuse = sk->sk_reuse;
8741 sctp_sk(newsk)->reuse = sp->reuse;
8743 newsk->sk_shutdown = sk->sk_shutdown;
8744 newsk->sk_destruct = sctp_destruct_sock;
8745 newsk->sk_family = sk->sk_family;
8746 newsk->sk_protocol = IPPROTO_SCTP;
8747 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8748 newsk->sk_sndbuf = sk->sk_sndbuf;
8749 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8750 newsk->sk_lingertime = sk->sk_lingertime;
8751 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8752 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8753 newsk->sk_rxhash = sk->sk_rxhash;
8755 newinet = inet_sk(newsk);
8757 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8758 * getsockname() and getpeername()
8760 newinet->inet_sport = inet->inet_sport;
8761 newinet->inet_saddr = inet->inet_saddr;
8762 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8763 newinet->inet_dport = htons(asoc->peer.port);
8764 newinet->pmtudisc = inet->pmtudisc;
8765 newinet->inet_id = asoc->next_tsn ^ jiffies;
8767 newinet->uc_ttl = inet->uc_ttl;
8768 newinet->mc_loop = 1;
8769 newinet->mc_ttl = 1;
8770 newinet->mc_index = 0;
8771 newinet->mc_list = NULL;
8773 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8774 net_enable_timestamp();
8776 /* Set newsk security attributes from orginal sk and connection
8777 * security attribute from ep.
8779 security_sctp_sk_clone(ep, sk, newsk);
8782 static inline void sctp_copy_descendant(struct sock *sk_to,
8783 const struct sock *sk_from)
8785 int ancestor_size = sizeof(struct inet_sock) +
8786 sizeof(struct sctp_sock) -
8787 offsetof(struct sctp_sock, auto_asconf_list);
8789 if (sk_from->sk_family == PF_INET6)
8790 ancestor_size += sizeof(struct ipv6_pinfo);
8792 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8795 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8796 * and its messages to the newsk.
8798 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8799 struct sctp_association *assoc,
8800 enum sctp_socket_type type)
8802 struct sctp_sock *oldsp = sctp_sk(oldsk);
8803 struct sctp_sock *newsp = sctp_sk(newsk);
8804 struct sctp_bind_bucket *pp; /* hash list port iterator */
8805 struct sctp_endpoint *newep = newsp->ep;
8806 struct sk_buff *skb, *tmp;
8807 struct sctp_ulpevent *event;
8808 struct sctp_bind_hashbucket *head;
8810 /* Migrate socket buffer sizes and all the socket level options to the
8811 * new socket.
8813 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8814 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8815 /* Brute force copy old sctp opt. */
8816 sctp_copy_descendant(newsk, oldsk);
8818 /* Restore the ep value that was overwritten with the above structure
8819 * copy.
8821 newsp->ep = newep;
8822 newsp->hmac = NULL;
8824 /* Hook this new socket in to the bind_hash list. */
8825 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8826 inet_sk(oldsk)->inet_num)];
8827 spin_lock_bh(&head->lock);
8828 pp = sctp_sk(oldsk)->bind_hash;
8829 sk_add_bind_node(newsk, &pp->owner);
8830 sctp_sk(newsk)->bind_hash = pp;
8831 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8832 spin_unlock_bh(&head->lock);
8834 /* Copy the bind_addr list from the original endpoint to the new
8835 * endpoint so that we can handle restarts properly
8837 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8838 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8840 /* Move any messages in the old socket's receive queue that are for the
8841 * peeled off association to the new socket's receive queue.
8843 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8844 event = sctp_skb2event(skb);
8845 if (event->asoc == assoc) {
8846 __skb_unlink(skb, &oldsk->sk_receive_queue);
8847 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8848 sctp_skb_set_owner_r_frag(skb, newsk);
8852 /* Clean up any messages pending delivery due to partial
8853 * delivery. Three cases:
8854 * 1) No partial deliver; no work.
8855 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8856 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8858 skb_queue_head_init(&newsp->pd_lobby);
8859 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8861 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8862 struct sk_buff_head *queue;
8864 /* Decide which queue to move pd_lobby skbs to. */
8865 if (assoc->ulpq.pd_mode) {
8866 queue = &newsp->pd_lobby;
8867 } else
8868 queue = &newsk->sk_receive_queue;
8870 /* Walk through the pd_lobby, looking for skbs that
8871 * need moved to the new socket.
8873 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8874 event = sctp_skb2event(skb);
8875 if (event->asoc == assoc) {
8876 __skb_unlink(skb, &oldsp->pd_lobby);
8877 __skb_queue_tail(queue, skb);
8878 sctp_skb_set_owner_r_frag(skb, newsk);
8882 /* Clear up any skbs waiting for the partial
8883 * delivery to finish.
8885 if (assoc->ulpq.pd_mode)
8886 sctp_clear_pd(oldsk, NULL);
8890 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8892 /* Set the type of socket to indicate that it is peeled off from the
8893 * original UDP-style socket or created with the accept() call on a
8894 * TCP-style socket..
8896 newsp->type = type;
8898 /* Mark the new socket "in-use" by the user so that any packets
8899 * that may arrive on the association after we've moved it are
8900 * queued to the backlog. This prevents a potential race between
8901 * backlog processing on the old socket and new-packet processing
8902 * on the new socket.
8904 * The caller has just allocated newsk so we can guarantee that other
8905 * paths won't try to lock it and then oldsk.
8907 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8908 sctp_for_each_tx_datachunk(assoc, sctp_clear_owner_w);
8909 sctp_assoc_migrate(assoc, newsk);
8910 sctp_for_each_tx_datachunk(assoc, sctp_set_owner_w);
8912 /* If the association on the newsk is already closed before accept()
8913 * is called, set RCV_SHUTDOWN flag.
8915 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8916 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8917 newsk->sk_shutdown |= RCV_SHUTDOWN;
8918 } else {
8919 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8922 release_sock(newsk);
8926 /* This proto struct describes the ULP interface for SCTP. */
8927 struct proto sctp_prot = {
8928 .name = "SCTP",
8929 .owner = THIS_MODULE,
8930 .close = sctp_close,
8931 .disconnect = sctp_disconnect,
8932 .accept = sctp_accept,
8933 .ioctl = sctp_ioctl,
8934 .init = sctp_init_sock,
8935 .destroy = sctp_destroy_sock,
8936 .shutdown = sctp_shutdown,
8937 .setsockopt = sctp_setsockopt,
8938 .getsockopt = sctp_getsockopt,
8939 .sendmsg = sctp_sendmsg,
8940 .recvmsg = sctp_recvmsg,
8941 .bind = sctp_bind,
8942 .backlog_rcv = sctp_backlog_rcv,
8943 .hash = sctp_hash,
8944 .unhash = sctp_unhash,
8945 .get_port = sctp_get_port,
8946 .obj_size = sizeof(struct sctp_sock),
8947 .useroffset = offsetof(struct sctp_sock, subscribe),
8948 .usersize = offsetof(struct sctp_sock, initmsg) -
8949 offsetof(struct sctp_sock, subscribe) +
8950 sizeof_field(struct sctp_sock, initmsg),
8951 .sysctl_mem = sysctl_sctp_mem,
8952 .sysctl_rmem = sysctl_sctp_rmem,
8953 .sysctl_wmem = sysctl_sctp_wmem,
8954 .memory_pressure = &sctp_memory_pressure,
8955 .enter_memory_pressure = sctp_enter_memory_pressure,
8956 .memory_allocated = &sctp_memory_allocated,
8957 .sockets_allocated = &sctp_sockets_allocated,
8960 #if IS_ENABLED(CONFIG_IPV6)
8962 #include <net/transp_v6.h>
8963 static void sctp_v6_destroy_sock(struct sock *sk)
8965 sctp_destroy_sock(sk);
8966 inet6_destroy_sock(sk);
8969 struct proto sctpv6_prot = {
8970 .name = "SCTPv6",
8971 .owner = THIS_MODULE,
8972 .close = sctp_close,
8973 .disconnect = sctp_disconnect,
8974 .accept = sctp_accept,
8975 .ioctl = sctp_ioctl,
8976 .init = sctp_init_sock,
8977 .destroy = sctp_v6_destroy_sock,
8978 .shutdown = sctp_shutdown,
8979 .setsockopt = sctp_setsockopt,
8980 .getsockopt = sctp_getsockopt,
8981 .sendmsg = sctp_sendmsg,
8982 .recvmsg = sctp_recvmsg,
8983 .bind = sctp_bind,
8984 .backlog_rcv = sctp_backlog_rcv,
8985 .hash = sctp_hash,
8986 .unhash = sctp_unhash,
8987 .get_port = sctp_get_port,
8988 .obj_size = sizeof(struct sctp6_sock),
8989 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
8990 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
8991 offsetof(struct sctp6_sock, sctp.subscribe) +
8992 sizeof_field(struct sctp6_sock, sctp.initmsg),
8993 .sysctl_mem = sysctl_sctp_mem,
8994 .sysctl_rmem = sysctl_sctp_rmem,
8995 .sysctl_wmem = sysctl_sctp_wmem,
8996 .memory_pressure = &sctp_memory_pressure,
8997 .enter_memory_pressure = sctp_enter_memory_pressure,
8998 .memory_allocated = &sctp_memory_allocated,
8999 .sockets_allocated = &sctp_sockets_allocated,
9001 #endif /* IS_ENABLED(CONFIG_IPV6) */