xhci: prevent bus suspend if a roothub port detected a over-current condition
[linux/fpc-iii.git] / net / sctp / socket.c
blobc93be3ba5df2989ed41e62a904119a983c70c1c3
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 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
155 sizeof(struct sk_buff) +
156 sizeof(struct sctp_chunk);
158 refcount_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
159 sk->sk_wmem_queued += chunk->skb->truesize;
160 sk_mem_charge(sk, chunk->skb->truesize);
163 static void sctp_clear_owner_w(struct sctp_chunk *chunk)
165 skb_orphan(chunk->skb);
168 #define traverse_and_process() \
169 do { \
170 msg = chunk->msg; \
171 if (msg == prev_msg) \
172 continue; \
173 list_for_each_entry(c, &msg->chunks, frag_list) { \
174 if ((clear && asoc->base.sk == c->skb->sk) || \
175 (!clear && asoc->base.sk != c->skb->sk)) \
176 cb(c); \
178 prev_msg = msg; \
179 } while (0)
181 static void sctp_for_each_tx_datachunk(struct sctp_association *asoc,
182 bool clear,
183 void (*cb)(struct sctp_chunk *))
186 struct sctp_datamsg *msg, *prev_msg = NULL;
187 struct sctp_outq *q = &asoc->outqueue;
188 struct sctp_chunk *chunk, *c;
189 struct sctp_transport *t;
191 list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
192 list_for_each_entry(chunk, &t->transmitted, transmitted_list)
193 traverse_and_process();
195 list_for_each_entry(chunk, &q->retransmit, transmitted_list)
196 traverse_and_process();
198 list_for_each_entry(chunk, &q->sacked, transmitted_list)
199 traverse_and_process();
201 list_for_each_entry(chunk, &q->abandoned, transmitted_list)
202 traverse_and_process();
204 list_for_each_entry(chunk, &q->out_chunk_list, list)
205 traverse_and_process();
208 static void sctp_for_each_rx_skb(struct sctp_association *asoc, struct sock *sk,
209 void (*cb)(struct sk_buff *, struct sock *))
212 struct sk_buff *skb, *tmp;
214 sctp_skb_for_each(skb, &asoc->ulpq.lobby, tmp)
215 cb(skb, sk);
217 sctp_skb_for_each(skb, &asoc->ulpq.reasm, tmp)
218 cb(skb, sk);
220 sctp_skb_for_each(skb, &asoc->ulpq.reasm_uo, tmp)
221 cb(skb, sk);
224 /* Verify that this is a valid address. */
225 static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
226 int len)
228 struct sctp_af *af;
230 /* Verify basic sockaddr. */
231 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
232 if (!af)
233 return -EINVAL;
235 /* Is this a valid SCTP address? */
236 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
237 return -EINVAL;
239 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
240 return -EINVAL;
242 return 0;
245 /* Look up the association by its id. If this is not a UDP-style
246 * socket, the ID field is always ignored.
248 struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
250 struct sctp_association *asoc = NULL;
252 /* If this is not a UDP-style socket, assoc id should be ignored. */
253 if (!sctp_style(sk, UDP)) {
254 /* Return NULL if the socket state is not ESTABLISHED. It
255 * could be a TCP-style listening socket or a socket which
256 * hasn't yet called connect() to establish an association.
258 if (!sctp_sstate(sk, ESTABLISHED) && !sctp_sstate(sk, CLOSING))
259 return NULL;
261 /* Get the first and the only association from the list. */
262 if (!list_empty(&sctp_sk(sk)->ep->asocs))
263 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
264 struct sctp_association, asocs);
265 return asoc;
268 /* Otherwise this is a UDP-style socket. */
269 if (!id || (id == (sctp_assoc_t)-1))
270 return NULL;
272 spin_lock_bh(&sctp_assocs_id_lock);
273 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
274 if (asoc && (asoc->base.sk != sk || asoc->base.dead))
275 asoc = NULL;
276 spin_unlock_bh(&sctp_assocs_id_lock);
278 return asoc;
281 /* Look up the transport from an address and an assoc id. If both address and
282 * id are specified, the associations matching the address and the id should be
283 * the same.
285 static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
286 struct sockaddr_storage *addr,
287 sctp_assoc_t id)
289 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
290 struct sctp_af *af = sctp_get_af_specific(addr->ss_family);
291 union sctp_addr *laddr = (union sctp_addr *)addr;
292 struct sctp_transport *transport;
294 if (!af || sctp_verify_addr(sk, laddr, af->sockaddr_len))
295 return NULL;
297 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
298 laddr,
299 &transport);
301 if (!addr_asoc)
302 return NULL;
304 id_asoc = sctp_id2assoc(sk, id);
305 if (id_asoc && (id_asoc != addr_asoc))
306 return NULL;
308 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
309 (union sctp_addr *)addr);
311 return transport;
314 /* API 3.1.2 bind() - UDP Style Syntax
315 * The syntax of bind() is,
317 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
319 * sd - the socket descriptor returned by socket().
320 * addr - the address structure (struct sockaddr_in or struct
321 * sockaddr_in6 [RFC 2553]),
322 * addr_len - the size of the address structure.
324 static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
326 int retval = 0;
328 lock_sock(sk);
330 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
331 addr, addr_len);
333 /* Disallow binding twice. */
334 if (!sctp_sk(sk)->ep->base.bind_addr.port)
335 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
336 addr_len);
337 else
338 retval = -EINVAL;
340 release_sock(sk);
342 return retval;
345 static long sctp_get_port_local(struct sock *, union sctp_addr *);
347 /* Verify this is a valid sockaddr. */
348 static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
349 union sctp_addr *addr, int len)
351 struct sctp_af *af;
353 /* Check minimum size. */
354 if (len < sizeof (struct sockaddr))
355 return NULL;
357 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
358 return NULL;
360 if (addr->sa.sa_family == AF_INET6) {
361 if (len < SIN6_LEN_RFC2133)
362 return NULL;
363 /* V4 mapped address are really of AF_INET family */
364 if (ipv6_addr_v4mapped(&addr->v6.sin6_addr) &&
365 !opt->pf->af_supported(AF_INET, opt))
366 return NULL;
369 /* If we get this far, af is valid. */
370 af = sctp_get_af_specific(addr->sa.sa_family);
372 if (len < af->sockaddr_len)
373 return NULL;
375 return af;
378 /* Bind a local address either to an endpoint or to an association. */
379 static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
381 struct net *net = sock_net(sk);
382 struct sctp_sock *sp = sctp_sk(sk);
383 struct sctp_endpoint *ep = sp->ep;
384 struct sctp_bind_addr *bp = &ep->base.bind_addr;
385 struct sctp_af *af;
386 unsigned short snum;
387 int ret = 0;
389 /* Common sockaddr verification. */
390 af = sctp_sockaddr_af(sp, addr, len);
391 if (!af) {
392 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
393 __func__, sk, addr, len);
394 return -EINVAL;
397 snum = ntohs(addr->v4.sin_port);
399 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
400 __func__, sk, &addr->sa, bp->port, snum, len);
402 /* PF specific bind() address verification. */
403 if (!sp->pf->bind_verify(sp, addr))
404 return -EADDRNOTAVAIL;
406 /* We must either be unbound, or bind to the same port.
407 * It's OK to allow 0 ports if we are already bound.
408 * We'll just inhert an already bound port in this case
410 if (bp->port) {
411 if (!snum)
412 snum = bp->port;
413 else if (snum != bp->port) {
414 pr_debug("%s: new port %d doesn't match existing port "
415 "%d\n", __func__, snum, bp->port);
416 return -EINVAL;
420 if (snum && snum < inet_prot_sock(net) &&
421 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
422 return -EACCES;
424 /* See if the address matches any of the addresses we may have
425 * already bound before checking against other endpoints.
427 if (sctp_bind_addr_match(bp, addr, sp))
428 return -EINVAL;
430 /* Make sure we are allowed to bind here.
431 * The function sctp_get_port_local() does duplicate address
432 * detection.
434 addr->v4.sin_port = htons(snum);
435 if ((ret = sctp_get_port_local(sk, addr))) {
436 return -EADDRINUSE;
439 /* Refresh ephemeral port. */
440 if (!bp->port)
441 bp->port = inet_sk(sk)->inet_num;
443 /* Add the address to the bind address list.
444 * Use GFP_ATOMIC since BHs will be disabled.
446 ret = sctp_add_bind_addr(bp, addr, af->sockaddr_len,
447 SCTP_ADDR_SRC, GFP_ATOMIC);
449 /* Copy back into socket for getsockname() use. */
450 if (!ret) {
451 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
452 sp->pf->to_sk_saddr(addr, sk);
455 return ret;
458 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
460 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
461 * at any one time. If a sender, after sending an ASCONF chunk, decides
462 * it needs to transfer another ASCONF Chunk, it MUST wait until the
463 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
464 * subsequent ASCONF. Note this restriction binds each side, so at any
465 * time two ASCONF may be in-transit on any given association (one sent
466 * from each endpoint).
468 static int sctp_send_asconf(struct sctp_association *asoc,
469 struct sctp_chunk *chunk)
471 struct net *net = sock_net(asoc->base.sk);
472 int retval = 0;
474 /* If there is an outstanding ASCONF chunk, queue it for later
475 * transmission.
477 if (asoc->addip_last_asconf) {
478 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
479 goto out;
482 /* Hold the chunk until an ASCONF_ACK is received. */
483 sctp_chunk_hold(chunk);
484 retval = sctp_primitive_ASCONF(net, asoc, chunk);
485 if (retval)
486 sctp_chunk_free(chunk);
487 else
488 asoc->addip_last_asconf = chunk;
490 out:
491 return retval;
494 /* Add a list of addresses as bind addresses to local endpoint or
495 * association.
497 * Basically run through each address specified in the addrs/addrcnt
498 * array/length pair, determine if it is IPv6 or IPv4 and call
499 * sctp_do_bind() on it.
501 * If any of them fails, then the operation will be reversed and the
502 * ones that were added will be removed.
504 * Only sctp_setsockopt_bindx() is supposed to call this function.
506 static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
508 int cnt;
509 int retval = 0;
510 void *addr_buf;
511 struct sockaddr *sa_addr;
512 struct sctp_af *af;
514 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
515 addrs, addrcnt);
517 addr_buf = addrs;
518 for (cnt = 0; cnt < addrcnt; cnt++) {
519 /* The list may contain either IPv4 or IPv6 address;
520 * determine the address length for walking thru the list.
522 sa_addr = addr_buf;
523 af = sctp_get_af_specific(sa_addr->sa_family);
524 if (!af) {
525 retval = -EINVAL;
526 goto err_bindx_add;
529 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
530 af->sockaddr_len);
532 addr_buf += af->sockaddr_len;
534 err_bindx_add:
535 if (retval < 0) {
536 /* Failed. Cleanup the ones that have been added */
537 if (cnt > 0)
538 sctp_bindx_rem(sk, addrs, cnt);
539 return retval;
543 return retval;
546 /* Send an ASCONF chunk with Add IP address parameters to all the peers of the
547 * associations that are part of the endpoint indicating that a list of local
548 * addresses are added to the endpoint.
550 * If any of the addresses is already in the bind address list of the
551 * association, we do not send the chunk for that association. But it will not
552 * affect other associations.
554 * Only sctp_setsockopt_bindx() is supposed to call this function.
556 static int sctp_send_asconf_add_ip(struct sock *sk,
557 struct sockaddr *addrs,
558 int addrcnt)
560 struct net *net = sock_net(sk);
561 struct sctp_sock *sp;
562 struct sctp_endpoint *ep;
563 struct sctp_association *asoc;
564 struct sctp_bind_addr *bp;
565 struct sctp_chunk *chunk;
566 struct sctp_sockaddr_entry *laddr;
567 union sctp_addr *addr;
568 union sctp_addr saveaddr;
569 void *addr_buf;
570 struct sctp_af *af;
571 struct list_head *p;
572 int i;
573 int retval = 0;
575 if (!net->sctp.addip_enable)
576 return retval;
578 sp = sctp_sk(sk);
579 ep = sp->ep;
581 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
582 __func__, sk, addrs, addrcnt);
584 list_for_each_entry(asoc, &ep->asocs, asocs) {
585 if (!asoc->peer.asconf_capable)
586 continue;
588 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
589 continue;
591 if (!sctp_state(asoc, ESTABLISHED))
592 continue;
594 /* Check if any address in the packed array of addresses is
595 * in the bind address list of the association. If so,
596 * do not send the asconf chunk to its peer, but continue with
597 * other associations.
599 addr_buf = addrs;
600 for (i = 0; i < addrcnt; i++) {
601 addr = addr_buf;
602 af = sctp_get_af_specific(addr->v4.sin_family);
603 if (!af) {
604 retval = -EINVAL;
605 goto out;
608 if (sctp_assoc_lookup_laddr(asoc, addr))
609 break;
611 addr_buf += af->sockaddr_len;
613 if (i < addrcnt)
614 continue;
616 /* Use the first valid address in bind addr list of
617 * association as Address Parameter of ASCONF CHUNK.
619 bp = &asoc->base.bind_addr;
620 p = bp->address_list.next;
621 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
622 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
623 addrcnt, SCTP_PARAM_ADD_IP);
624 if (!chunk) {
625 retval = -ENOMEM;
626 goto out;
629 /* Add the new addresses to the bind address list with
630 * use_as_src set to 0.
632 addr_buf = addrs;
633 for (i = 0; i < addrcnt; i++) {
634 addr = addr_buf;
635 af = sctp_get_af_specific(addr->v4.sin_family);
636 memcpy(&saveaddr, addr, af->sockaddr_len);
637 retval = sctp_add_bind_addr(bp, &saveaddr,
638 sizeof(saveaddr),
639 SCTP_ADDR_NEW, GFP_ATOMIC);
640 addr_buf += af->sockaddr_len;
642 if (asoc->src_out_of_asoc_ok) {
643 struct sctp_transport *trans;
645 list_for_each_entry(trans,
646 &asoc->peer.transport_addr_list, transports) {
647 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
648 2*asoc->pathmtu, 4380));
649 trans->ssthresh = asoc->peer.i.a_rwnd;
650 trans->rto = asoc->rto_initial;
651 sctp_max_rto(asoc, trans);
652 trans->rtt = trans->srtt = trans->rttvar = 0;
653 /* Clear the source and route cache */
654 sctp_transport_route(trans, NULL,
655 sctp_sk(asoc->base.sk));
658 retval = sctp_send_asconf(asoc, chunk);
661 out:
662 return retval;
665 /* Remove a list of addresses from bind addresses list. Do not remove the
666 * last address.
668 * Basically run through each address specified in the addrs/addrcnt
669 * array/length pair, determine if it is IPv6 or IPv4 and call
670 * sctp_del_bind() on it.
672 * If any of them fails, then the operation will be reversed and the
673 * ones that were removed will be added back.
675 * At least one address has to be left; if only one address is
676 * available, the operation will return -EBUSY.
678 * Only sctp_setsockopt_bindx() is supposed to call this function.
680 static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
682 struct sctp_sock *sp = sctp_sk(sk);
683 struct sctp_endpoint *ep = sp->ep;
684 int cnt;
685 struct sctp_bind_addr *bp = &ep->base.bind_addr;
686 int retval = 0;
687 void *addr_buf;
688 union sctp_addr *sa_addr;
689 struct sctp_af *af;
691 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
692 __func__, sk, addrs, addrcnt);
694 addr_buf = addrs;
695 for (cnt = 0; cnt < addrcnt; cnt++) {
696 /* If the bind address list is empty or if there is only one
697 * bind address, there is nothing more to be removed (we need
698 * at least one address here).
700 if (list_empty(&bp->address_list) ||
701 (sctp_list_single_entry(&bp->address_list))) {
702 retval = -EBUSY;
703 goto err_bindx_rem;
706 sa_addr = addr_buf;
707 af = sctp_get_af_specific(sa_addr->sa.sa_family);
708 if (!af) {
709 retval = -EINVAL;
710 goto err_bindx_rem;
713 if (!af->addr_valid(sa_addr, sp, NULL)) {
714 retval = -EADDRNOTAVAIL;
715 goto err_bindx_rem;
718 if (sa_addr->v4.sin_port &&
719 sa_addr->v4.sin_port != htons(bp->port)) {
720 retval = -EINVAL;
721 goto err_bindx_rem;
724 if (!sa_addr->v4.sin_port)
725 sa_addr->v4.sin_port = htons(bp->port);
727 /* FIXME - There is probably a need to check if sk->sk_saddr and
728 * sk->sk_rcv_addr are currently set to one of the addresses to
729 * be removed. This is something which needs to be looked into
730 * when we are fixing the outstanding issues with multi-homing
731 * socket routing and failover schemes. Refer to comments in
732 * sctp_do_bind(). -daisy
734 retval = sctp_del_bind_addr(bp, sa_addr);
736 addr_buf += af->sockaddr_len;
737 err_bindx_rem:
738 if (retval < 0) {
739 /* Failed. Add the ones that has been removed back */
740 if (cnt > 0)
741 sctp_bindx_add(sk, addrs, cnt);
742 return retval;
746 return retval;
749 /* Send an ASCONF chunk with Delete IP address parameters to all the peers of
750 * the associations that are part of the endpoint indicating that a list of
751 * local addresses are removed from the endpoint.
753 * If any of the addresses is already in the bind address list of the
754 * association, we do not send the chunk for that association. But it will not
755 * affect other associations.
757 * Only sctp_setsockopt_bindx() is supposed to call this function.
759 static int sctp_send_asconf_del_ip(struct sock *sk,
760 struct sockaddr *addrs,
761 int addrcnt)
763 struct net *net = sock_net(sk);
764 struct sctp_sock *sp;
765 struct sctp_endpoint *ep;
766 struct sctp_association *asoc;
767 struct sctp_transport *transport;
768 struct sctp_bind_addr *bp;
769 struct sctp_chunk *chunk;
770 union sctp_addr *laddr;
771 void *addr_buf;
772 struct sctp_af *af;
773 struct sctp_sockaddr_entry *saddr;
774 int i;
775 int retval = 0;
776 int stored = 0;
778 chunk = NULL;
779 if (!net->sctp.addip_enable)
780 return retval;
782 sp = sctp_sk(sk);
783 ep = sp->ep;
785 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
786 __func__, sk, addrs, addrcnt);
788 list_for_each_entry(asoc, &ep->asocs, asocs) {
790 if (!asoc->peer.asconf_capable)
791 continue;
793 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
794 continue;
796 if (!sctp_state(asoc, ESTABLISHED))
797 continue;
799 /* Check if any address in the packed array of addresses is
800 * not present in the bind address list of the association.
801 * If so, do not send the asconf chunk to its peer, but
802 * continue with other associations.
804 addr_buf = addrs;
805 for (i = 0; i < addrcnt; i++) {
806 laddr = addr_buf;
807 af = sctp_get_af_specific(laddr->v4.sin_family);
808 if (!af) {
809 retval = -EINVAL;
810 goto out;
813 if (!sctp_assoc_lookup_laddr(asoc, laddr))
814 break;
816 addr_buf += af->sockaddr_len;
818 if (i < addrcnt)
819 continue;
821 /* Find one address in the association's bind address list
822 * that is not in the packed array of addresses. This is to
823 * make sure that we do not delete all the addresses in the
824 * association.
826 bp = &asoc->base.bind_addr;
827 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
828 addrcnt, sp);
829 if ((laddr == NULL) && (addrcnt == 1)) {
830 if (asoc->asconf_addr_del_pending)
831 continue;
832 asoc->asconf_addr_del_pending =
833 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
834 if (asoc->asconf_addr_del_pending == NULL) {
835 retval = -ENOMEM;
836 goto out;
838 asoc->asconf_addr_del_pending->sa.sa_family =
839 addrs->sa_family;
840 asoc->asconf_addr_del_pending->v4.sin_port =
841 htons(bp->port);
842 if (addrs->sa_family == AF_INET) {
843 struct sockaddr_in *sin;
845 sin = (struct sockaddr_in *)addrs;
846 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
847 } else if (addrs->sa_family == AF_INET6) {
848 struct sockaddr_in6 *sin6;
850 sin6 = (struct sockaddr_in6 *)addrs;
851 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
854 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
855 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
856 asoc->asconf_addr_del_pending);
858 asoc->src_out_of_asoc_ok = 1;
859 stored = 1;
860 goto skip_mkasconf;
863 if (laddr == NULL)
864 return -EINVAL;
866 /* We do not need RCU protection throughout this loop
867 * because this is done under a socket lock from the
868 * setsockopt call.
870 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
871 SCTP_PARAM_DEL_IP);
872 if (!chunk) {
873 retval = -ENOMEM;
874 goto out;
877 skip_mkasconf:
878 /* Reset use_as_src flag for the addresses in the bind address
879 * list that are to be deleted.
881 addr_buf = addrs;
882 for (i = 0; i < addrcnt; i++) {
883 laddr = addr_buf;
884 af = sctp_get_af_specific(laddr->v4.sin_family);
885 list_for_each_entry(saddr, &bp->address_list, list) {
886 if (sctp_cmp_addr_exact(&saddr->a, laddr))
887 saddr->state = SCTP_ADDR_DEL;
889 addr_buf += af->sockaddr_len;
892 /* Update the route and saddr entries for all the transports
893 * as some of the addresses in the bind address list are
894 * about to be deleted and cannot be used as source addresses.
896 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
897 transports) {
898 sctp_transport_route(transport, NULL,
899 sctp_sk(asoc->base.sk));
902 if (stored)
903 /* We don't need to transmit ASCONF */
904 continue;
905 retval = sctp_send_asconf(asoc, chunk);
907 out:
908 return retval;
911 /* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
912 int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
914 struct sock *sk = sctp_opt2sk(sp);
915 union sctp_addr *addr;
916 struct sctp_af *af;
918 /* It is safe to write port space in caller. */
919 addr = &addrw->a;
920 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
921 af = sctp_get_af_specific(addr->sa.sa_family);
922 if (!af)
923 return -EINVAL;
924 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
925 return -EINVAL;
927 if (addrw->state == SCTP_ADDR_NEW)
928 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
929 else
930 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
933 /* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
935 * API 8.1
936 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
937 * int flags);
939 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
940 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
941 * or IPv6 addresses.
943 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
944 * Section 3.1.2 for this usage.
946 * addrs is a pointer to an array of one or more socket addresses. Each
947 * address is contained in its appropriate structure (i.e. struct
948 * sockaddr_in or struct sockaddr_in6) the family of the address type
949 * must be used to distinguish the address length (note that this
950 * representation is termed a "packed array" of addresses). The caller
951 * specifies the number of addresses in the array with addrcnt.
953 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
954 * -1, and sets errno to the appropriate error code.
956 * For SCTP, the port given in each socket address must be the same, or
957 * sctp_bindx() will fail, setting errno to EINVAL.
959 * The flags parameter is formed from the bitwise OR of zero or more of
960 * the following currently defined flags:
962 * SCTP_BINDX_ADD_ADDR
964 * SCTP_BINDX_REM_ADDR
966 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
967 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
968 * addresses from the association. The two flags are mutually exclusive;
969 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
970 * not remove all addresses from an association; sctp_bindx() will
971 * reject such an attempt with EINVAL.
973 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
974 * additional addresses with an endpoint after calling bind(). Or use
975 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
976 * socket is associated with so that no new association accepted will be
977 * associated with those addresses. If the endpoint supports dynamic
978 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
979 * endpoint to send the appropriate message to the peer to change the
980 * peers address lists.
982 * Adding and removing addresses from a connected association is
983 * optional functionality. Implementations that do not support this
984 * functionality should return EOPNOTSUPP.
986 * Basically do nothing but copying the addresses from user to kernel
987 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
988 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
989 * from userspace.
991 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
992 * it.
994 * sk The sk of the socket
995 * addrs The pointer to the addresses in user land
996 * addrssize Size of the addrs buffer
997 * op Operation to perform (add or remove, see the flags of
998 * sctp_bindx)
1000 * Returns 0 if ok, <0 errno code on error.
1002 static int sctp_setsockopt_bindx(struct sock *sk,
1003 struct sockaddr __user *addrs,
1004 int addrs_size, int op)
1006 struct sockaddr *kaddrs;
1007 int err;
1008 int addrcnt = 0;
1009 int walk_size = 0;
1010 struct sockaddr *sa_addr;
1011 void *addr_buf;
1012 struct sctp_af *af;
1014 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
1015 __func__, sk, addrs, addrs_size, op);
1017 if (unlikely(addrs_size <= 0))
1018 return -EINVAL;
1020 kaddrs = memdup_user(addrs, addrs_size);
1021 if (unlikely(IS_ERR(kaddrs)))
1022 return PTR_ERR(kaddrs);
1024 /* Walk through the addrs buffer and count the number of addresses. */
1025 addr_buf = kaddrs;
1026 while (walk_size < addrs_size) {
1027 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1028 kfree(kaddrs);
1029 return -EINVAL;
1032 sa_addr = addr_buf;
1033 af = sctp_get_af_specific(sa_addr->sa_family);
1035 /* If the address family is not supported or if this address
1036 * causes the address buffer to overflow return EINVAL.
1038 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1039 kfree(kaddrs);
1040 return -EINVAL;
1042 addrcnt++;
1043 addr_buf += af->sockaddr_len;
1044 walk_size += af->sockaddr_len;
1047 /* Do the work. */
1048 switch (op) {
1049 case SCTP_BINDX_ADD_ADDR:
1050 /* Allow security module to validate bindx addresses. */
1051 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_BINDX_ADD,
1052 (struct sockaddr *)kaddrs,
1053 addrs_size);
1054 if (err)
1055 goto out;
1056 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1057 if (err)
1058 goto out;
1059 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1060 break;
1062 case SCTP_BINDX_REM_ADDR:
1063 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1064 if (err)
1065 goto out;
1066 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1067 break;
1069 default:
1070 err = -EINVAL;
1071 break;
1074 out:
1075 kfree(kaddrs);
1077 return err;
1080 /* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1082 * Common routine for handling connect() and sctp_connectx().
1083 * Connect will come in with just a single address.
1085 static int __sctp_connect(struct sock *sk,
1086 struct sockaddr *kaddrs,
1087 int addrs_size, int flags,
1088 sctp_assoc_t *assoc_id)
1090 struct net *net = sock_net(sk);
1091 struct sctp_sock *sp;
1092 struct sctp_endpoint *ep;
1093 struct sctp_association *asoc = NULL;
1094 struct sctp_association *asoc2;
1095 struct sctp_transport *transport;
1096 union sctp_addr to;
1097 enum sctp_scope scope;
1098 long timeo;
1099 int err = 0;
1100 int addrcnt = 0;
1101 int walk_size = 0;
1102 union sctp_addr *sa_addr = NULL;
1103 void *addr_buf;
1104 unsigned short port;
1106 sp = sctp_sk(sk);
1107 ep = sp->ep;
1109 /* connect() cannot be done on a socket that is already in ESTABLISHED
1110 * state - UDP-style peeled off socket or a TCP-style socket that
1111 * is already connected.
1112 * It cannot be done even on a TCP-style listening socket.
1114 if (sctp_sstate(sk, ESTABLISHED) || sctp_sstate(sk, CLOSING) ||
1115 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1116 err = -EISCONN;
1117 goto out_free;
1120 /* Walk through the addrs buffer and count the number of addresses. */
1121 addr_buf = kaddrs;
1122 while (walk_size < addrs_size) {
1123 struct sctp_af *af;
1125 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1126 err = -EINVAL;
1127 goto out_free;
1130 sa_addr = addr_buf;
1131 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1133 /* If the address family is not supported or if this address
1134 * causes the address buffer to overflow return EINVAL.
1136 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1137 err = -EINVAL;
1138 goto out_free;
1141 port = ntohs(sa_addr->v4.sin_port);
1143 /* Save current address so we can work with it */
1144 memcpy(&to, sa_addr, af->sockaddr_len);
1146 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1147 if (err)
1148 goto out_free;
1150 /* Make sure the destination port is correctly set
1151 * in all addresses.
1153 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1154 err = -EINVAL;
1155 goto out_free;
1158 /* Check if there already is a matching association on the
1159 * endpoint (other than the one created here).
1161 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1162 if (asoc2 && asoc2 != asoc) {
1163 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1164 err = -EISCONN;
1165 else
1166 err = -EALREADY;
1167 goto out_free;
1170 /* If we could not find a matching association on the endpoint,
1171 * make sure that there is no peeled-off association matching
1172 * the peer address even on another socket.
1174 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1175 err = -EADDRNOTAVAIL;
1176 goto out_free;
1179 if (!asoc) {
1180 /* If a bind() or sctp_bindx() is not called prior to
1181 * an sctp_connectx() call, the system picks an
1182 * ephemeral port and will choose an address set
1183 * equivalent to binding with a wildcard address.
1185 if (!ep->base.bind_addr.port) {
1186 if (sctp_autobind(sk)) {
1187 err = -EAGAIN;
1188 goto out_free;
1190 } else {
1192 * If an unprivileged user inherits a 1-many
1193 * style socket with open associations on a
1194 * privileged port, it MAY be permitted to
1195 * accept new associations, but it SHOULD NOT
1196 * be permitted to open new associations.
1198 if (ep->base.bind_addr.port <
1199 inet_prot_sock(net) &&
1200 !ns_capable(net->user_ns,
1201 CAP_NET_BIND_SERVICE)) {
1202 err = -EACCES;
1203 goto out_free;
1207 scope = sctp_scope(&to);
1208 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1209 if (!asoc) {
1210 err = -ENOMEM;
1211 goto out_free;
1214 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1215 GFP_KERNEL);
1216 if (err < 0) {
1217 goto out_free;
1222 /* Prime the peer's transport structures. */
1223 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1224 SCTP_UNKNOWN);
1225 if (!transport) {
1226 err = -ENOMEM;
1227 goto out_free;
1230 addrcnt++;
1231 addr_buf += af->sockaddr_len;
1232 walk_size += af->sockaddr_len;
1235 /* In case the user of sctp_connectx() wants an association
1236 * id back, assign one now.
1238 if (assoc_id) {
1239 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1240 if (err < 0)
1241 goto out_free;
1244 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1245 if (err < 0) {
1246 goto out_free;
1249 /* Initialize sk's dport and daddr for getpeername() */
1250 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1251 sp->pf->to_sk_daddr(sa_addr, sk);
1252 sk->sk_err = 0;
1254 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1256 if (assoc_id)
1257 *assoc_id = asoc->assoc_id;
1259 err = sctp_wait_for_connect(asoc, &timeo);
1260 /* Note: the asoc may be freed after the return of
1261 * sctp_wait_for_connect.
1264 /* Don't free association on exit. */
1265 asoc = NULL;
1267 out_free:
1268 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1269 __func__, asoc, kaddrs, err);
1271 if (asoc) {
1272 /* sctp_primitive_ASSOCIATE may have added this association
1273 * To the hash table, try to unhash it, just in case, its a noop
1274 * if it wasn't hashed so we're safe
1276 sctp_association_free(asoc);
1278 return err;
1281 /* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1283 * API 8.9
1284 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1285 * sctp_assoc_t *asoc);
1287 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1288 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1289 * or IPv6 addresses.
1291 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1292 * Section 3.1.2 for this usage.
1294 * addrs is a pointer to an array of one or more socket addresses. Each
1295 * address is contained in its appropriate structure (i.e. struct
1296 * sockaddr_in or struct sockaddr_in6) the family of the address type
1297 * must be used to distengish the address length (note that this
1298 * representation is termed a "packed array" of addresses). The caller
1299 * specifies the number of addresses in the array with addrcnt.
1301 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1302 * the association id of the new association. On failure, sctp_connectx()
1303 * returns -1, and sets errno to the appropriate error code. The assoc_id
1304 * is not touched by the kernel.
1306 * For SCTP, the port given in each socket address must be the same, or
1307 * sctp_connectx() will fail, setting errno to EINVAL.
1309 * An application can use sctp_connectx to initiate an association with
1310 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1311 * allows a caller to specify multiple addresses at which a peer can be
1312 * reached. The way the SCTP stack uses the list of addresses to set up
1313 * the association is implementation dependent. This function only
1314 * specifies that the stack will try to make use of all the addresses in
1315 * the list when needed.
1317 * Note that the list of addresses passed in is only used for setting up
1318 * the association. It does not necessarily equal the set of addresses
1319 * the peer uses for the resulting association. If the caller wants to
1320 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1321 * retrieve them after the association has been set up.
1323 * Basically do nothing but copying the addresses from user to kernel
1324 * land and invoking either sctp_connectx(). This is used for tunneling
1325 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1327 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1328 * it.
1330 * sk The sk of the socket
1331 * addrs The pointer to the addresses in user land
1332 * addrssize Size of the addrs buffer
1334 * Returns >=0 if ok, <0 errno code on error.
1336 static int __sctp_setsockopt_connectx(struct sock *sk,
1337 struct sockaddr __user *addrs,
1338 int addrs_size,
1339 sctp_assoc_t *assoc_id)
1341 struct sockaddr *kaddrs;
1342 int err = 0, flags = 0;
1344 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1345 __func__, sk, addrs, addrs_size);
1347 if (unlikely(addrs_size <= 0))
1348 return -EINVAL;
1350 kaddrs = memdup_user(addrs, addrs_size);
1351 if (unlikely(IS_ERR(kaddrs)))
1352 return PTR_ERR(kaddrs);
1354 /* Allow security module to validate connectx addresses. */
1355 err = security_sctp_bind_connect(sk, SCTP_SOCKOPT_CONNECTX,
1356 (struct sockaddr *)kaddrs,
1357 addrs_size);
1358 if (err)
1359 goto out_free;
1361 /* in-kernel sockets don't generally have a file allocated to them
1362 * if all they do is call sock_create_kern().
1364 if (sk->sk_socket->file)
1365 flags = sk->sk_socket->file->f_flags;
1367 err = __sctp_connect(sk, kaddrs, addrs_size, flags, assoc_id);
1369 out_free:
1370 kfree(kaddrs);
1372 return err;
1376 * This is an older interface. It's kept for backward compatibility
1377 * to the option that doesn't provide association id.
1379 static int sctp_setsockopt_connectx_old(struct sock *sk,
1380 struct sockaddr __user *addrs,
1381 int addrs_size)
1383 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1387 * New interface for the API. The since the API is done with a socket
1388 * option, to make it simple we feed back the association id is as a return
1389 * indication to the call. Error is always negative and association id is
1390 * always positive.
1392 static int sctp_setsockopt_connectx(struct sock *sk,
1393 struct sockaddr __user *addrs,
1394 int addrs_size)
1396 sctp_assoc_t assoc_id = 0;
1397 int err = 0;
1399 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1401 if (err)
1402 return err;
1403 else
1404 return assoc_id;
1408 * New (hopefully final) interface for the API.
1409 * We use the sctp_getaddrs_old structure so that use-space library
1410 * can avoid any unnecessary allocations. The only different part
1411 * is that we store the actual length of the address buffer into the
1412 * addrs_num structure member. That way we can re-use the existing
1413 * code.
1415 #ifdef CONFIG_COMPAT
1416 struct compat_sctp_getaddrs_old {
1417 sctp_assoc_t assoc_id;
1418 s32 addr_num;
1419 compat_uptr_t addrs; /* struct sockaddr * */
1421 #endif
1423 static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1424 char __user *optval,
1425 int __user *optlen)
1427 struct sctp_getaddrs_old param;
1428 sctp_assoc_t assoc_id = 0;
1429 int err = 0;
1431 #ifdef CONFIG_COMPAT
1432 if (in_compat_syscall()) {
1433 struct compat_sctp_getaddrs_old param32;
1435 if (len < sizeof(param32))
1436 return -EINVAL;
1437 if (copy_from_user(&param32, optval, sizeof(param32)))
1438 return -EFAULT;
1440 param.assoc_id = param32.assoc_id;
1441 param.addr_num = param32.addr_num;
1442 param.addrs = compat_ptr(param32.addrs);
1443 } else
1444 #endif
1446 if (len < sizeof(param))
1447 return -EINVAL;
1448 if (copy_from_user(&param, optval, sizeof(param)))
1449 return -EFAULT;
1452 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1453 param.addrs, param.addr_num,
1454 &assoc_id);
1455 if (err == 0 || err == -EINPROGRESS) {
1456 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1457 return -EFAULT;
1458 if (put_user(sizeof(assoc_id), optlen))
1459 return -EFAULT;
1462 return err;
1465 /* API 3.1.4 close() - UDP Style Syntax
1466 * Applications use close() to perform graceful shutdown (as described in
1467 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1468 * by a UDP-style socket.
1470 * The syntax is
1472 * ret = close(int sd);
1474 * sd - the socket descriptor of the associations to be closed.
1476 * To gracefully shutdown a specific association represented by the
1477 * UDP-style socket, an application should use the sendmsg() call,
1478 * passing no user data, but including the appropriate flag in the
1479 * ancillary data (see Section xxxx).
1481 * If sd in the close() call is a branched-off socket representing only
1482 * one association, the shutdown is performed on that association only.
1484 * 4.1.6 close() - TCP Style Syntax
1486 * Applications use close() to gracefully close down an association.
1488 * The syntax is:
1490 * int close(int sd);
1492 * sd - the socket descriptor of the association to be closed.
1494 * After an application calls close() on a socket descriptor, no further
1495 * socket operations will succeed on that descriptor.
1497 * API 7.1.4 SO_LINGER
1499 * An application using the TCP-style socket can use this option to
1500 * perform the SCTP ABORT primitive. The linger option structure is:
1502 * struct linger {
1503 * int l_onoff; // option on/off
1504 * int l_linger; // linger time
1505 * };
1507 * To enable the option, set l_onoff to 1. If the l_linger value is set
1508 * to 0, calling close() is the same as the ABORT primitive. If the
1509 * value is set to a negative value, the setsockopt() call will return
1510 * an error. If the value is set to a positive value linger_time, the
1511 * close() can be blocked for at most linger_time ms. If the graceful
1512 * shutdown phase does not finish during this period, close() will
1513 * return but the graceful shutdown phase continues in the system.
1515 static void sctp_close(struct sock *sk, long timeout)
1517 struct net *net = sock_net(sk);
1518 struct sctp_endpoint *ep;
1519 struct sctp_association *asoc;
1520 struct list_head *pos, *temp;
1521 unsigned int data_was_unread;
1523 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1525 lock_sock_nested(sk, SINGLE_DEPTH_NESTING);
1526 sk->sk_shutdown = SHUTDOWN_MASK;
1527 inet_sk_set_state(sk, SCTP_SS_CLOSING);
1529 ep = sctp_sk(sk)->ep;
1531 /* Clean up any skbs sitting on the receive queue. */
1532 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1533 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1535 /* Walk all associations on an endpoint. */
1536 list_for_each_safe(pos, temp, &ep->asocs) {
1537 asoc = list_entry(pos, struct sctp_association, asocs);
1539 if (sctp_style(sk, TCP)) {
1540 /* A closed association can still be in the list if
1541 * it belongs to a TCP-style listening socket that is
1542 * not yet accepted. If so, free it. If not, send an
1543 * ABORT or SHUTDOWN based on the linger options.
1545 if (sctp_state(asoc, CLOSED)) {
1546 sctp_association_free(asoc);
1547 continue;
1551 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1552 !skb_queue_empty(&asoc->ulpq.reasm) ||
1553 !skb_queue_empty(&asoc->ulpq.reasm_uo) ||
1554 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1555 struct sctp_chunk *chunk;
1557 chunk = sctp_make_abort_user(asoc, NULL, 0);
1558 sctp_primitive_ABORT(net, asoc, chunk);
1559 } else
1560 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1563 /* On a TCP-style socket, block for at most linger_time if set. */
1564 if (sctp_style(sk, TCP) && timeout)
1565 sctp_wait_for_close(sk, timeout);
1567 /* This will run the backlog queue. */
1568 release_sock(sk);
1570 /* Supposedly, no process has access to the socket, but
1571 * the net layers still may.
1572 * Also, sctp_destroy_sock() needs to be called with addr_wq_lock
1573 * held and that should be grabbed before socket lock.
1575 spin_lock_bh(&net->sctp.addr_wq_lock);
1576 bh_lock_sock_nested(sk);
1578 /* Hold the sock, since sk_common_release() will put sock_put()
1579 * and we have just a little more cleanup.
1581 sock_hold(sk);
1582 sk_common_release(sk);
1584 bh_unlock_sock(sk);
1585 spin_unlock_bh(&net->sctp.addr_wq_lock);
1587 sock_put(sk);
1589 SCTP_DBG_OBJCNT_DEC(sock);
1592 /* Handle EPIPE error. */
1593 static int sctp_error(struct sock *sk, int flags, int err)
1595 if (err == -EPIPE)
1596 err = sock_error(sk) ? : -EPIPE;
1597 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1598 send_sig(SIGPIPE, current, 0);
1599 return err;
1602 /* API 3.1.3 sendmsg() - UDP Style Syntax
1604 * An application uses sendmsg() and recvmsg() calls to transmit data to
1605 * and receive data from its peer.
1607 * ssize_t sendmsg(int socket, const struct msghdr *message,
1608 * int flags);
1610 * socket - the socket descriptor of the endpoint.
1611 * message - pointer to the msghdr structure which contains a single
1612 * user message and possibly some ancillary data.
1614 * See Section 5 for complete description of the data
1615 * structures.
1617 * flags - flags sent or received with the user message, see Section
1618 * 5 for complete description of the flags.
1620 * Note: This function could use a rewrite especially when explicit
1621 * connect support comes in.
1623 /* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1625 static int sctp_msghdr_parse(const struct msghdr *msg,
1626 struct sctp_cmsgs *cmsgs);
1628 static int sctp_sendmsg_parse(struct sock *sk, struct sctp_cmsgs *cmsgs,
1629 struct sctp_sndrcvinfo *srinfo,
1630 const struct msghdr *msg, size_t msg_len)
1632 __u16 sflags;
1633 int err;
1635 if (sctp_sstate(sk, LISTENING) && sctp_style(sk, TCP))
1636 return -EPIPE;
1638 if (msg_len > sk->sk_sndbuf)
1639 return -EMSGSIZE;
1641 memset(cmsgs, 0, sizeof(*cmsgs));
1642 err = sctp_msghdr_parse(msg, cmsgs);
1643 if (err) {
1644 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1645 return err;
1648 memset(srinfo, 0, sizeof(*srinfo));
1649 if (cmsgs->srinfo) {
1650 srinfo->sinfo_stream = cmsgs->srinfo->sinfo_stream;
1651 srinfo->sinfo_flags = cmsgs->srinfo->sinfo_flags;
1652 srinfo->sinfo_ppid = cmsgs->srinfo->sinfo_ppid;
1653 srinfo->sinfo_context = cmsgs->srinfo->sinfo_context;
1654 srinfo->sinfo_assoc_id = cmsgs->srinfo->sinfo_assoc_id;
1655 srinfo->sinfo_timetolive = cmsgs->srinfo->sinfo_timetolive;
1658 if (cmsgs->sinfo) {
1659 srinfo->sinfo_stream = cmsgs->sinfo->snd_sid;
1660 srinfo->sinfo_flags = cmsgs->sinfo->snd_flags;
1661 srinfo->sinfo_ppid = cmsgs->sinfo->snd_ppid;
1662 srinfo->sinfo_context = cmsgs->sinfo->snd_context;
1663 srinfo->sinfo_assoc_id = cmsgs->sinfo->snd_assoc_id;
1666 if (cmsgs->prinfo) {
1667 srinfo->sinfo_timetolive = cmsgs->prinfo->pr_value;
1668 SCTP_PR_SET_POLICY(srinfo->sinfo_flags,
1669 cmsgs->prinfo->pr_policy);
1672 sflags = srinfo->sinfo_flags;
1673 if (!sflags && msg_len)
1674 return 0;
1676 if (sctp_style(sk, TCP) && (sflags & (SCTP_EOF | SCTP_ABORT)))
1677 return -EINVAL;
1679 if (((sflags & SCTP_EOF) && msg_len > 0) ||
1680 (!(sflags & (SCTP_EOF | SCTP_ABORT)) && msg_len == 0))
1681 return -EINVAL;
1683 if ((sflags & SCTP_ADDR_OVER) && !msg->msg_name)
1684 return -EINVAL;
1686 return 0;
1689 static int sctp_sendmsg_new_asoc(struct sock *sk, __u16 sflags,
1690 struct sctp_cmsgs *cmsgs,
1691 union sctp_addr *daddr,
1692 struct sctp_transport **tp)
1694 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
1695 struct net *net = sock_net(sk);
1696 struct sctp_association *asoc;
1697 enum sctp_scope scope;
1698 struct cmsghdr *cmsg;
1699 __be32 flowinfo = 0;
1700 struct sctp_af *af;
1701 int err;
1703 *tp = NULL;
1705 if (sflags & (SCTP_EOF | SCTP_ABORT))
1706 return -EINVAL;
1708 if (sctp_style(sk, TCP) && (sctp_sstate(sk, ESTABLISHED) ||
1709 sctp_sstate(sk, CLOSING)))
1710 return -EADDRNOTAVAIL;
1712 if (sctp_endpoint_is_peeled_off(ep, daddr))
1713 return -EADDRNOTAVAIL;
1715 if (!ep->base.bind_addr.port) {
1716 if (sctp_autobind(sk))
1717 return -EAGAIN;
1718 } else {
1719 if (ep->base.bind_addr.port < inet_prot_sock(net) &&
1720 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
1721 return -EACCES;
1724 scope = sctp_scope(daddr);
1726 /* Label connection socket for first association 1-to-many
1727 * style for client sequence socket()->sendmsg(). This
1728 * needs to be done before sctp_assoc_add_peer() as that will
1729 * set up the initial packet that needs to account for any
1730 * security ip options (CIPSO/CALIPSO) added to the packet.
1732 af = sctp_get_af_specific(daddr->sa.sa_family);
1733 if (!af)
1734 return -EINVAL;
1735 err = security_sctp_bind_connect(sk, SCTP_SENDMSG_CONNECT,
1736 (struct sockaddr *)daddr,
1737 af->sockaddr_len);
1738 if (err < 0)
1739 return err;
1741 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1742 if (!asoc)
1743 return -ENOMEM;
1745 if (sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL) < 0) {
1746 err = -ENOMEM;
1747 goto free;
1750 if (cmsgs->init) {
1751 struct sctp_initmsg *init = cmsgs->init;
1753 if (init->sinit_num_ostreams) {
1754 __u16 outcnt = init->sinit_num_ostreams;
1756 asoc->c.sinit_num_ostreams = outcnt;
1757 /* outcnt has been changed, need to re-init stream */
1758 err = sctp_stream_init(&asoc->stream, outcnt, 0,
1759 GFP_KERNEL);
1760 if (err)
1761 goto free;
1764 if (init->sinit_max_instreams)
1765 asoc->c.sinit_max_instreams = init->sinit_max_instreams;
1767 if (init->sinit_max_attempts)
1768 asoc->max_init_attempts = init->sinit_max_attempts;
1770 if (init->sinit_max_init_timeo)
1771 asoc->max_init_timeo =
1772 msecs_to_jiffies(init->sinit_max_init_timeo);
1775 *tp = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL, SCTP_UNKNOWN);
1776 if (!*tp) {
1777 err = -ENOMEM;
1778 goto free;
1781 if (!cmsgs->addrs_msg)
1782 return 0;
1784 if (daddr->sa.sa_family == AF_INET6)
1785 flowinfo = daddr->v6.sin6_flowinfo;
1787 /* sendv addr list parse */
1788 for_each_cmsghdr(cmsg, cmsgs->addrs_msg) {
1789 struct sctp_transport *transport;
1790 struct sctp_association *old;
1791 union sctp_addr _daddr;
1792 int dlen;
1794 if (cmsg->cmsg_level != IPPROTO_SCTP ||
1795 (cmsg->cmsg_type != SCTP_DSTADDRV4 &&
1796 cmsg->cmsg_type != SCTP_DSTADDRV6))
1797 continue;
1799 daddr = &_daddr;
1800 memset(daddr, 0, sizeof(*daddr));
1801 dlen = cmsg->cmsg_len - sizeof(struct cmsghdr);
1802 if (cmsg->cmsg_type == SCTP_DSTADDRV4) {
1803 if (dlen < sizeof(struct in_addr)) {
1804 err = -EINVAL;
1805 goto free;
1808 dlen = sizeof(struct in_addr);
1809 daddr->v4.sin_family = AF_INET;
1810 daddr->v4.sin_port = htons(asoc->peer.port);
1811 memcpy(&daddr->v4.sin_addr, CMSG_DATA(cmsg), dlen);
1812 } else {
1813 if (dlen < sizeof(struct in6_addr)) {
1814 err = -EINVAL;
1815 goto free;
1818 dlen = sizeof(struct in6_addr);
1819 daddr->v6.sin6_flowinfo = flowinfo;
1820 daddr->v6.sin6_family = AF_INET6;
1821 daddr->v6.sin6_port = htons(asoc->peer.port);
1822 memcpy(&daddr->v6.sin6_addr, CMSG_DATA(cmsg), dlen);
1824 err = sctp_verify_addr(sk, daddr, sizeof(*daddr));
1825 if (err)
1826 goto free;
1828 old = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
1829 if (old && old != asoc) {
1830 if (old->state >= SCTP_STATE_ESTABLISHED)
1831 err = -EISCONN;
1832 else
1833 err = -EALREADY;
1834 goto free;
1837 if (sctp_endpoint_is_peeled_off(ep, daddr)) {
1838 err = -EADDRNOTAVAIL;
1839 goto free;
1842 transport = sctp_assoc_add_peer(asoc, daddr, GFP_KERNEL,
1843 SCTP_UNKNOWN);
1844 if (!transport) {
1845 err = -ENOMEM;
1846 goto free;
1850 return 0;
1852 free:
1853 sctp_association_free(asoc);
1854 return err;
1857 static int sctp_sendmsg_check_sflags(struct sctp_association *asoc,
1858 __u16 sflags, struct msghdr *msg,
1859 size_t msg_len)
1861 struct sock *sk = asoc->base.sk;
1862 struct net *net = sock_net(sk);
1864 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP))
1865 return -EPIPE;
1867 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP) &&
1868 !sctp_state(asoc, ESTABLISHED))
1869 return 0;
1871 if (sflags & SCTP_EOF) {
1872 pr_debug("%s: shutting down association:%p\n", __func__, asoc);
1873 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1875 return 0;
1878 if (sflags & SCTP_ABORT) {
1879 struct sctp_chunk *chunk;
1881 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1882 if (!chunk)
1883 return -ENOMEM;
1885 pr_debug("%s: aborting association:%p\n", __func__, asoc);
1886 sctp_primitive_ABORT(net, asoc, chunk);
1887 iov_iter_revert(&msg->msg_iter, msg_len);
1889 return 0;
1892 return 1;
1895 static int sctp_sendmsg_to_asoc(struct sctp_association *asoc,
1896 struct msghdr *msg, size_t msg_len,
1897 struct sctp_transport *transport,
1898 struct sctp_sndrcvinfo *sinfo)
1900 struct sock *sk = asoc->base.sk;
1901 struct sctp_sock *sp = sctp_sk(sk);
1902 struct net *net = sock_net(sk);
1903 struct sctp_datamsg *datamsg;
1904 bool wait_connect = false;
1905 struct sctp_chunk *chunk;
1906 long timeo;
1907 int err;
1909 if (sinfo->sinfo_stream >= asoc->stream.outcnt) {
1910 err = -EINVAL;
1911 goto err;
1914 if (unlikely(!SCTP_SO(&asoc->stream, sinfo->sinfo_stream)->ext)) {
1915 err = sctp_stream_init_ext(&asoc->stream, sinfo->sinfo_stream);
1916 if (err)
1917 goto err;
1920 if (sp->disable_fragments && msg_len > asoc->frag_point) {
1921 err = -EMSGSIZE;
1922 goto err;
1925 if (asoc->pmtu_pending) {
1926 if (sp->param_flags & SPP_PMTUD_ENABLE)
1927 sctp_assoc_sync_pmtu(asoc);
1928 asoc->pmtu_pending = 0;
1931 if (sctp_wspace(asoc) < (int)msg_len)
1932 sctp_prsctp_prune(asoc, sinfo, msg_len - sctp_wspace(asoc));
1934 if (sctp_wspace(asoc) <= 0) {
1935 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1936 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1937 if (err)
1938 goto err;
1941 if (sctp_state(asoc, CLOSED)) {
1942 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1943 if (err)
1944 goto err;
1946 if (sp->strm_interleave) {
1947 timeo = sock_sndtimeo(sk, 0);
1948 err = sctp_wait_for_connect(asoc, &timeo);
1949 if (err) {
1950 err = -ESRCH;
1951 goto err;
1953 } else {
1954 wait_connect = true;
1957 pr_debug("%s: we associated primitively\n", __func__);
1960 datamsg = sctp_datamsg_from_user(asoc, sinfo, &msg->msg_iter);
1961 if (IS_ERR(datamsg)) {
1962 err = PTR_ERR(datamsg);
1963 goto err;
1966 asoc->force_delay = !!(msg->msg_flags & MSG_MORE);
1968 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1969 sctp_chunk_hold(chunk);
1970 sctp_set_owner_w(chunk);
1971 chunk->transport = transport;
1974 err = sctp_primitive_SEND(net, asoc, datamsg);
1975 if (err) {
1976 sctp_datamsg_free(datamsg);
1977 goto err;
1980 pr_debug("%s: we sent primitively\n", __func__);
1982 sctp_datamsg_put(datamsg);
1984 if (unlikely(wait_connect)) {
1985 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1986 sctp_wait_for_connect(asoc, &timeo);
1989 err = msg_len;
1991 err:
1992 return err;
1995 static union sctp_addr *sctp_sendmsg_get_daddr(struct sock *sk,
1996 const struct msghdr *msg,
1997 struct sctp_cmsgs *cmsgs)
1999 union sctp_addr *daddr = NULL;
2000 int err;
2002 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
2003 int len = msg->msg_namelen;
2005 if (len > sizeof(*daddr))
2006 len = sizeof(*daddr);
2008 daddr = (union sctp_addr *)msg->msg_name;
2010 err = sctp_verify_addr(sk, daddr, len);
2011 if (err)
2012 return ERR_PTR(err);
2015 return daddr;
2018 static void sctp_sendmsg_update_sinfo(struct sctp_association *asoc,
2019 struct sctp_sndrcvinfo *sinfo,
2020 struct sctp_cmsgs *cmsgs)
2022 if (!cmsgs->srinfo && !cmsgs->sinfo) {
2023 sinfo->sinfo_stream = asoc->default_stream;
2024 sinfo->sinfo_ppid = asoc->default_ppid;
2025 sinfo->sinfo_context = asoc->default_context;
2026 sinfo->sinfo_assoc_id = sctp_assoc2id(asoc);
2028 if (!cmsgs->prinfo)
2029 sinfo->sinfo_flags = asoc->default_flags;
2032 if (!cmsgs->srinfo && !cmsgs->prinfo)
2033 sinfo->sinfo_timetolive = asoc->default_timetolive;
2035 if (cmsgs->authinfo) {
2036 /* Reuse sinfo_tsn to indicate that authinfo was set and
2037 * sinfo_ssn to save the keyid on tx path.
2039 sinfo->sinfo_tsn = 1;
2040 sinfo->sinfo_ssn = cmsgs->authinfo->auth_keynumber;
2044 static int sctp_sendmsg(struct sock *sk, struct msghdr *msg, size_t msg_len)
2046 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
2047 struct sctp_transport *transport = NULL;
2048 struct sctp_sndrcvinfo _sinfo, *sinfo;
2049 struct sctp_association *asoc, *tmp;
2050 struct sctp_cmsgs cmsgs;
2051 union sctp_addr *daddr;
2052 bool new = false;
2053 __u16 sflags;
2054 int err;
2056 /* Parse and get snd_info */
2057 err = sctp_sendmsg_parse(sk, &cmsgs, &_sinfo, msg, msg_len);
2058 if (err)
2059 goto out;
2061 sinfo = &_sinfo;
2062 sflags = sinfo->sinfo_flags;
2064 /* Get daddr from msg */
2065 daddr = sctp_sendmsg_get_daddr(sk, msg, &cmsgs);
2066 if (IS_ERR(daddr)) {
2067 err = PTR_ERR(daddr);
2068 goto out;
2071 lock_sock(sk);
2073 /* SCTP_SENDALL process */
2074 if ((sflags & SCTP_SENDALL) && sctp_style(sk, UDP)) {
2075 list_for_each_entry_safe(asoc, tmp, &ep->asocs, asocs) {
2076 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2077 msg_len);
2078 if (err == 0)
2079 continue;
2080 if (err < 0)
2081 goto out_unlock;
2083 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2085 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len,
2086 NULL, sinfo);
2087 if (err < 0)
2088 goto out_unlock;
2090 iov_iter_revert(&msg->msg_iter, err);
2093 goto out_unlock;
2096 /* Get and check or create asoc */
2097 if (daddr) {
2098 asoc = sctp_endpoint_lookup_assoc(ep, daddr, &transport);
2099 if (asoc) {
2100 err = sctp_sendmsg_check_sflags(asoc, sflags, msg,
2101 msg_len);
2102 if (err <= 0)
2103 goto out_unlock;
2104 } else {
2105 err = sctp_sendmsg_new_asoc(sk, sflags, &cmsgs, daddr,
2106 &transport);
2107 if (err)
2108 goto out_unlock;
2110 asoc = transport->asoc;
2111 new = true;
2114 if (!sctp_style(sk, TCP) && !(sflags & SCTP_ADDR_OVER))
2115 transport = NULL;
2116 } else {
2117 asoc = sctp_id2assoc(sk, sinfo->sinfo_assoc_id);
2118 if (!asoc) {
2119 err = -EPIPE;
2120 goto out_unlock;
2123 err = sctp_sendmsg_check_sflags(asoc, sflags, msg, msg_len);
2124 if (err <= 0)
2125 goto out_unlock;
2128 /* Update snd_info with the asoc */
2129 sctp_sendmsg_update_sinfo(asoc, sinfo, &cmsgs);
2131 /* Send msg to the asoc */
2132 err = sctp_sendmsg_to_asoc(asoc, msg, msg_len, transport, sinfo);
2133 if (err < 0 && err != -ESRCH && new)
2134 sctp_association_free(asoc);
2136 out_unlock:
2137 release_sock(sk);
2138 out:
2139 return sctp_error(sk, msg->msg_flags, err);
2142 /* This is an extended version of skb_pull() that removes the data from the
2143 * start of a skb even when data is spread across the list of skb's in the
2144 * frag_list. len specifies the total amount of data that needs to be removed.
2145 * when 'len' bytes could be removed from the skb, it returns 0.
2146 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2147 * could not be removed.
2149 static int sctp_skb_pull(struct sk_buff *skb, int len)
2151 struct sk_buff *list;
2152 int skb_len = skb_headlen(skb);
2153 int rlen;
2155 if (len <= skb_len) {
2156 __skb_pull(skb, len);
2157 return 0;
2159 len -= skb_len;
2160 __skb_pull(skb, skb_len);
2162 skb_walk_frags(skb, list) {
2163 rlen = sctp_skb_pull(list, len);
2164 skb->len -= (len-rlen);
2165 skb->data_len -= (len-rlen);
2167 if (!rlen)
2168 return 0;
2170 len = rlen;
2173 return len;
2176 /* API 3.1.3 recvmsg() - UDP Style Syntax
2178 * ssize_t recvmsg(int socket, struct msghdr *message,
2179 * int flags);
2181 * socket - the socket descriptor of the endpoint.
2182 * message - pointer to the msghdr structure which contains a single
2183 * user message and possibly some ancillary data.
2185 * See Section 5 for complete description of the data
2186 * structures.
2188 * flags - flags sent or received with the user message, see Section
2189 * 5 for complete description of the flags.
2191 static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
2192 int noblock, int flags, int *addr_len)
2194 struct sctp_ulpevent *event = NULL;
2195 struct sctp_sock *sp = sctp_sk(sk);
2196 struct sk_buff *skb, *head_skb;
2197 int copied;
2198 int err = 0;
2199 int skb_len;
2201 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2202 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2203 addr_len);
2205 lock_sock(sk);
2207 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
2208 !sctp_sstate(sk, CLOSING) && !sctp_sstate(sk, CLOSED)) {
2209 err = -ENOTCONN;
2210 goto out;
2213 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2214 if (!skb)
2215 goto out;
2217 /* Get the total length of the skb including any skb's in the
2218 * frag_list.
2220 skb_len = skb->len;
2222 copied = skb_len;
2223 if (copied > len)
2224 copied = len;
2226 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2228 event = sctp_skb2event(skb);
2230 if (err)
2231 goto out_free;
2233 if (event->chunk && event->chunk->head_skb)
2234 head_skb = event->chunk->head_skb;
2235 else
2236 head_skb = skb;
2237 sock_recv_ts_and_drops(msg, sk, head_skb);
2238 if (sctp_ulpevent_is_notification(event)) {
2239 msg->msg_flags |= MSG_NOTIFICATION;
2240 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2241 } else {
2242 sp->pf->skb_msgname(head_skb, msg->msg_name, addr_len);
2245 /* Check if we allow SCTP_NXTINFO. */
2246 if (sp->recvnxtinfo)
2247 sctp_ulpevent_read_nxtinfo(event, msg, sk);
2248 /* Check if we allow SCTP_RCVINFO. */
2249 if (sp->recvrcvinfo)
2250 sctp_ulpevent_read_rcvinfo(event, msg);
2251 /* Check if we allow SCTP_SNDRCVINFO. */
2252 if (sp->subscribe.sctp_data_io_event)
2253 sctp_ulpevent_read_sndrcvinfo(event, msg);
2255 err = copied;
2257 /* If skb's length exceeds the user's buffer, update the skb and
2258 * push it back to the receive_queue so that the next call to
2259 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2261 if (skb_len > copied) {
2262 msg->msg_flags &= ~MSG_EOR;
2263 if (flags & MSG_PEEK)
2264 goto out_free;
2265 sctp_skb_pull(skb, copied);
2266 skb_queue_head(&sk->sk_receive_queue, skb);
2268 /* When only partial message is copied to the user, increase
2269 * rwnd by that amount. If all the data in the skb is read,
2270 * rwnd is updated when the event is freed.
2272 if (!sctp_ulpevent_is_notification(event))
2273 sctp_assoc_rwnd_increase(event->asoc, copied);
2274 goto out;
2275 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2276 (event->msg_flags & MSG_EOR))
2277 msg->msg_flags |= MSG_EOR;
2278 else
2279 msg->msg_flags &= ~MSG_EOR;
2281 out_free:
2282 if (flags & MSG_PEEK) {
2283 /* Release the skb reference acquired after peeking the skb in
2284 * sctp_skb_recv_datagram().
2286 kfree_skb(skb);
2287 } else {
2288 /* Free the event which includes releasing the reference to
2289 * the owner of the skb, freeing the skb and updating the
2290 * rwnd.
2292 sctp_ulpevent_free(event);
2294 out:
2295 release_sock(sk);
2296 return err;
2299 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2301 * This option is a on/off flag. If enabled no SCTP message
2302 * fragmentation will be performed. Instead if a message being sent
2303 * exceeds the current PMTU size, the message will NOT be sent and
2304 * instead a error will be indicated to the user.
2306 static int sctp_setsockopt_disable_fragments(struct sock *sk,
2307 char __user *optval,
2308 unsigned int optlen)
2310 int val;
2312 if (optlen < sizeof(int))
2313 return -EINVAL;
2315 if (get_user(val, (int __user *)optval))
2316 return -EFAULT;
2318 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2320 return 0;
2323 static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2324 unsigned int optlen)
2326 struct sctp_association *asoc;
2327 struct sctp_ulpevent *event;
2329 if (optlen > sizeof(struct sctp_event_subscribe))
2330 return -EINVAL;
2331 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2332 return -EFAULT;
2334 /* At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2335 * if there is no data to be sent or retransmit, the stack will
2336 * immediately send up this notification.
2338 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2339 &sctp_sk(sk)->subscribe)) {
2340 asoc = sctp_id2assoc(sk, 0);
2342 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2343 event = sctp_ulpevent_make_sender_dry_event(asoc,
2344 GFP_USER | __GFP_NOWARN);
2345 if (!event)
2346 return -ENOMEM;
2348 asoc->stream.si->enqueue_event(&asoc->ulpq, event);
2352 return 0;
2355 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2357 * This socket option is applicable to the UDP-style socket only. When
2358 * set it will cause associations that are idle for more than the
2359 * specified number of seconds to automatically close. An association
2360 * being idle is defined an association that has NOT sent or received
2361 * user data. The special value of '0' indicates that no automatic
2362 * close of any associations should be performed. The option expects an
2363 * integer defining the number of seconds of idle time before an
2364 * association is closed.
2366 static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2367 unsigned int optlen)
2369 struct sctp_sock *sp = sctp_sk(sk);
2370 struct net *net = sock_net(sk);
2372 /* Applicable to UDP-style socket only */
2373 if (sctp_style(sk, TCP))
2374 return -EOPNOTSUPP;
2375 if (optlen != sizeof(int))
2376 return -EINVAL;
2377 if (copy_from_user(&sp->autoclose, optval, optlen))
2378 return -EFAULT;
2380 if (sp->autoclose > net->sctp.max_autoclose)
2381 sp->autoclose = net->sctp.max_autoclose;
2383 return 0;
2386 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2388 * Applications can enable or disable heartbeats for any peer address of
2389 * an association, modify an address's heartbeat interval, force a
2390 * heartbeat to be sent immediately, and adjust the address's maximum
2391 * number of retransmissions sent before an address is considered
2392 * unreachable. The following structure is used to access and modify an
2393 * address's parameters:
2395 * struct sctp_paddrparams {
2396 * sctp_assoc_t spp_assoc_id;
2397 * struct sockaddr_storage spp_address;
2398 * uint32_t spp_hbinterval;
2399 * uint16_t spp_pathmaxrxt;
2400 * uint32_t spp_pathmtu;
2401 * uint32_t spp_sackdelay;
2402 * uint32_t spp_flags;
2403 * uint32_t spp_ipv6_flowlabel;
2404 * uint8_t spp_dscp;
2405 * };
2407 * spp_assoc_id - (one-to-many style socket) This is filled in the
2408 * application, and identifies the association for
2409 * this query.
2410 * spp_address - This specifies which address is of interest.
2411 * spp_hbinterval - This contains the value of the heartbeat interval,
2412 * in milliseconds. If a value of zero
2413 * is present in this field then no changes are to
2414 * be made to this parameter.
2415 * spp_pathmaxrxt - This contains the maximum number of
2416 * retransmissions before this address shall be
2417 * considered unreachable. If a value of zero
2418 * is present in this field then no changes are to
2419 * be made to this parameter.
2420 * spp_pathmtu - When Path MTU discovery is disabled the value
2421 * specified here will be the "fixed" path mtu.
2422 * Note that if the spp_address field is empty
2423 * then all associations on this address will
2424 * have this fixed path mtu set upon them.
2426 * spp_sackdelay - When delayed sack is enabled, this value specifies
2427 * the number of milliseconds that sacks will be delayed
2428 * for. This value will apply to all addresses of an
2429 * association if the spp_address field is empty. Note
2430 * also, that if delayed sack is enabled and this
2431 * value is set to 0, no change is made to the last
2432 * recorded delayed sack timer value.
2434 * spp_flags - These flags are used to control various features
2435 * on an association. The flag field may contain
2436 * zero or more of the following options.
2438 * SPP_HB_ENABLE - Enable heartbeats on the
2439 * specified address. Note that if the address
2440 * field is empty all addresses for the association
2441 * have heartbeats enabled upon them.
2443 * SPP_HB_DISABLE - Disable heartbeats on the
2444 * speicifed address. Note that if the address
2445 * field is empty all addresses for the association
2446 * will have their heartbeats disabled. Note also
2447 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2448 * mutually exclusive, only one of these two should
2449 * be specified. Enabling both fields will have
2450 * undetermined results.
2452 * SPP_HB_DEMAND - Request a user initiated heartbeat
2453 * to be made immediately.
2455 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2456 * heartbeat delayis to be set to the value of 0
2457 * milliseconds.
2459 * SPP_PMTUD_ENABLE - This field will enable PMTU
2460 * discovery upon the specified address. Note that
2461 * if the address feild is empty then all addresses
2462 * on the association are effected.
2464 * SPP_PMTUD_DISABLE - This field will disable PMTU
2465 * discovery upon the specified address. Note that
2466 * if the address feild is empty then all addresses
2467 * on the association are effected. Not also that
2468 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2469 * exclusive. Enabling both will have undetermined
2470 * results.
2472 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2473 * on delayed sack. The time specified in spp_sackdelay
2474 * is used to specify the sack delay for this address. Note
2475 * that if spp_address is empty then all addresses will
2476 * enable delayed sack and take on the sack delay
2477 * value specified in spp_sackdelay.
2478 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2479 * off delayed sack. If the spp_address field is blank then
2480 * delayed sack is disabled for the entire association. Note
2481 * also that this field is mutually exclusive to
2482 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2483 * results.
2485 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
2486 * setting of the IPV6 flow label value. The value is
2487 * contained in the spp_ipv6_flowlabel field.
2488 * Upon retrieval, this flag will be set to indicate that
2489 * the spp_ipv6_flowlabel field has a valid value returned.
2490 * If a specific destination address is set (in the
2491 * spp_address field), then the value returned is that of
2492 * the address. If just an association is specified (and
2493 * no address), then the association's default flow label
2494 * is returned. If neither an association nor a destination
2495 * is specified, then the socket's default flow label is
2496 * returned. For non-IPv6 sockets, this flag will be left
2497 * cleared.
2499 * SPP_DSCP: Setting this flag enables the setting of the
2500 * Differentiated Services Code Point (DSCP) value
2501 * associated with either the association or a specific
2502 * address. The value is obtained in the spp_dscp field.
2503 * Upon retrieval, this flag will be set to indicate that
2504 * the spp_dscp field has a valid value returned. If a
2505 * specific destination address is set when called (in the
2506 * spp_address field), then that specific destination
2507 * address's DSCP value is returned. If just an association
2508 * is specified, then the association's default DSCP is
2509 * returned. If neither an association nor a destination is
2510 * specified, then the socket's default DSCP is returned.
2512 * spp_ipv6_flowlabel
2513 * - This field is used in conjunction with the
2514 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
2515 * The 20 least significant bits are used for the flow
2516 * label. This setting has precedence over any IPv6-layer
2517 * setting.
2519 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
2520 * and contains the DSCP. The 6 most significant bits are
2521 * used for the DSCP. This setting has precedence over any
2522 * IPv4- or IPv6- layer setting.
2524 static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2525 struct sctp_transport *trans,
2526 struct sctp_association *asoc,
2527 struct sctp_sock *sp,
2528 int hb_change,
2529 int pmtud_change,
2530 int sackdelay_change)
2532 int error;
2534 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2535 struct net *net = sock_net(trans->asoc->base.sk);
2537 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2538 if (error)
2539 return error;
2542 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2543 * this field is ignored. Note also that a value of zero indicates
2544 * the current setting should be left unchanged.
2546 if (params->spp_flags & SPP_HB_ENABLE) {
2548 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2549 * set. This lets us use 0 value when this flag
2550 * is set.
2552 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2553 params->spp_hbinterval = 0;
2555 if (params->spp_hbinterval ||
2556 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2557 if (trans) {
2558 trans->hbinterval =
2559 msecs_to_jiffies(params->spp_hbinterval);
2560 } else if (asoc) {
2561 asoc->hbinterval =
2562 msecs_to_jiffies(params->spp_hbinterval);
2563 } else {
2564 sp->hbinterval = params->spp_hbinterval;
2569 if (hb_change) {
2570 if (trans) {
2571 trans->param_flags =
2572 (trans->param_flags & ~SPP_HB) | hb_change;
2573 } else if (asoc) {
2574 asoc->param_flags =
2575 (asoc->param_flags & ~SPP_HB) | hb_change;
2576 } else {
2577 sp->param_flags =
2578 (sp->param_flags & ~SPP_HB) | hb_change;
2582 /* When Path MTU discovery is disabled the value specified here will
2583 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2584 * include the flag SPP_PMTUD_DISABLE for this field to have any
2585 * effect).
2587 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2588 if (trans) {
2589 trans->pathmtu = params->spp_pathmtu;
2590 sctp_assoc_sync_pmtu(asoc);
2591 } else if (asoc) {
2592 sctp_assoc_set_pmtu(asoc, params->spp_pathmtu);
2593 } else {
2594 sp->pathmtu = params->spp_pathmtu;
2598 if (pmtud_change) {
2599 if (trans) {
2600 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2601 (params->spp_flags & SPP_PMTUD_ENABLE);
2602 trans->param_flags =
2603 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2604 if (update) {
2605 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2606 sctp_assoc_sync_pmtu(asoc);
2608 } else if (asoc) {
2609 asoc->param_flags =
2610 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2611 } else {
2612 sp->param_flags =
2613 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2617 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2618 * value of this field is ignored. Note also that a value of zero
2619 * indicates the current setting should be left unchanged.
2621 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2622 if (trans) {
2623 trans->sackdelay =
2624 msecs_to_jiffies(params->spp_sackdelay);
2625 } else if (asoc) {
2626 asoc->sackdelay =
2627 msecs_to_jiffies(params->spp_sackdelay);
2628 } else {
2629 sp->sackdelay = params->spp_sackdelay;
2633 if (sackdelay_change) {
2634 if (trans) {
2635 trans->param_flags =
2636 (trans->param_flags & ~SPP_SACKDELAY) |
2637 sackdelay_change;
2638 } else if (asoc) {
2639 asoc->param_flags =
2640 (asoc->param_flags & ~SPP_SACKDELAY) |
2641 sackdelay_change;
2642 } else {
2643 sp->param_flags =
2644 (sp->param_flags & ~SPP_SACKDELAY) |
2645 sackdelay_change;
2649 /* Note that a value of zero indicates the current setting should be
2650 left unchanged.
2652 if (params->spp_pathmaxrxt) {
2653 if (trans) {
2654 trans->pathmaxrxt = params->spp_pathmaxrxt;
2655 } else if (asoc) {
2656 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2657 } else {
2658 sp->pathmaxrxt = params->spp_pathmaxrxt;
2662 if (params->spp_flags & SPP_IPV6_FLOWLABEL) {
2663 if (trans) {
2664 if (trans->ipaddr.sa.sa_family == AF_INET6) {
2665 trans->flowlabel = params->spp_ipv6_flowlabel &
2666 SCTP_FLOWLABEL_VAL_MASK;
2667 trans->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2669 } else if (asoc) {
2670 struct sctp_transport *t;
2672 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2673 transports) {
2674 if (t->ipaddr.sa.sa_family != AF_INET6)
2675 continue;
2676 t->flowlabel = params->spp_ipv6_flowlabel &
2677 SCTP_FLOWLABEL_VAL_MASK;
2678 t->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2680 asoc->flowlabel = params->spp_ipv6_flowlabel &
2681 SCTP_FLOWLABEL_VAL_MASK;
2682 asoc->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2683 } else if (sctp_opt2sk(sp)->sk_family == AF_INET6) {
2684 sp->flowlabel = params->spp_ipv6_flowlabel &
2685 SCTP_FLOWLABEL_VAL_MASK;
2686 sp->flowlabel |= SCTP_FLOWLABEL_SET_MASK;
2690 if (params->spp_flags & SPP_DSCP) {
2691 if (trans) {
2692 trans->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2693 trans->dscp |= SCTP_DSCP_SET_MASK;
2694 } else if (asoc) {
2695 struct sctp_transport *t;
2697 list_for_each_entry(t, &asoc->peer.transport_addr_list,
2698 transports) {
2699 t->dscp = params->spp_dscp &
2700 SCTP_DSCP_VAL_MASK;
2701 t->dscp |= SCTP_DSCP_SET_MASK;
2703 asoc->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2704 asoc->dscp |= SCTP_DSCP_SET_MASK;
2705 } else {
2706 sp->dscp = params->spp_dscp & SCTP_DSCP_VAL_MASK;
2707 sp->dscp |= SCTP_DSCP_SET_MASK;
2711 return 0;
2714 static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2715 char __user *optval,
2716 unsigned int optlen)
2718 struct sctp_paddrparams params;
2719 struct sctp_transport *trans = NULL;
2720 struct sctp_association *asoc = NULL;
2721 struct sctp_sock *sp = sctp_sk(sk);
2722 int error;
2723 int hb_change, pmtud_change, sackdelay_change;
2725 if (optlen == sizeof(params)) {
2726 if (copy_from_user(&params, optval, optlen))
2727 return -EFAULT;
2728 } else if (optlen == ALIGN(offsetof(struct sctp_paddrparams,
2729 spp_ipv6_flowlabel), 4)) {
2730 if (copy_from_user(&params, optval, optlen))
2731 return -EFAULT;
2732 if (params.spp_flags & (SPP_DSCP | SPP_IPV6_FLOWLABEL))
2733 return -EINVAL;
2734 } else {
2735 return -EINVAL;
2738 /* Validate flags and value parameters. */
2739 hb_change = params.spp_flags & SPP_HB;
2740 pmtud_change = params.spp_flags & SPP_PMTUD;
2741 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2743 if (hb_change == SPP_HB ||
2744 pmtud_change == SPP_PMTUD ||
2745 sackdelay_change == SPP_SACKDELAY ||
2746 params.spp_sackdelay > 500 ||
2747 (params.spp_pathmtu &&
2748 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2749 return -EINVAL;
2751 /* If an address other than INADDR_ANY is specified, and
2752 * no transport is found, then the request is invalid.
2754 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
2755 trans = sctp_addr_id2transport(sk, &params.spp_address,
2756 params.spp_assoc_id);
2757 if (!trans)
2758 return -EINVAL;
2761 /* Get association, if assoc_id != 0 and the socket is a one
2762 * to many style socket, and an association was not found, then
2763 * the id was invalid.
2765 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2766 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2767 return -EINVAL;
2769 /* Heartbeat demand can only be sent on a transport or
2770 * association, but not a socket.
2772 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2773 return -EINVAL;
2775 /* Process parameters. */
2776 error = sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2777 hb_change, pmtud_change,
2778 sackdelay_change);
2780 if (error)
2781 return error;
2783 /* If changes are for association, also apply parameters to each
2784 * transport.
2786 if (!trans && asoc) {
2787 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2788 transports) {
2789 sctp_apply_peer_addr_params(&params, trans, asoc, sp,
2790 hb_change, pmtud_change,
2791 sackdelay_change);
2795 return 0;
2798 static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2800 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2803 static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2805 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2809 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2811 * This option will effect the way delayed acks are performed. This
2812 * option allows you to get or set the delayed ack time, in
2813 * milliseconds. It also allows changing the delayed ack frequency.
2814 * Changing the frequency to 1 disables the delayed sack algorithm. If
2815 * the assoc_id is 0, then this sets or gets the endpoints default
2816 * values. If the assoc_id field is non-zero, then the set or get
2817 * effects the specified association for the one to many model (the
2818 * assoc_id field is ignored by the one to one model). Note that if
2819 * sack_delay or sack_freq are 0 when setting this option, then the
2820 * current values will remain unchanged.
2822 * struct sctp_sack_info {
2823 * sctp_assoc_t sack_assoc_id;
2824 * uint32_t sack_delay;
2825 * uint32_t sack_freq;
2826 * };
2828 * sack_assoc_id - This parameter, indicates which association the user
2829 * is performing an action upon. Note that if this field's value is
2830 * zero then the endpoints default value is changed (effecting future
2831 * associations only).
2833 * sack_delay - This parameter contains the number of milliseconds that
2834 * the user is requesting the delayed ACK timer be set to. Note that
2835 * this value is defined in the standard to be between 200 and 500
2836 * milliseconds.
2838 * sack_freq - This parameter contains the number of packets that must
2839 * be received before a sack is sent without waiting for the delay
2840 * timer to expire. The default value for this is 2, setting this
2841 * value to 1 will disable the delayed sack algorithm.
2844 static int sctp_setsockopt_delayed_ack(struct sock *sk,
2845 char __user *optval, unsigned int optlen)
2847 struct sctp_sack_info params;
2848 struct sctp_transport *trans = NULL;
2849 struct sctp_association *asoc = NULL;
2850 struct sctp_sock *sp = sctp_sk(sk);
2852 if (optlen == sizeof(struct sctp_sack_info)) {
2853 if (copy_from_user(&params, optval, optlen))
2854 return -EFAULT;
2856 if (params.sack_delay == 0 && params.sack_freq == 0)
2857 return 0;
2858 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2859 pr_warn_ratelimited(DEPRECATED
2860 "%s (pid %d) "
2861 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2862 "Use struct sctp_sack_info instead\n",
2863 current->comm, task_pid_nr(current));
2864 if (copy_from_user(&params, optval, optlen))
2865 return -EFAULT;
2867 if (params.sack_delay == 0)
2868 params.sack_freq = 1;
2869 else
2870 params.sack_freq = 0;
2871 } else
2872 return -EINVAL;
2874 /* Validate value parameter. */
2875 if (params.sack_delay > 500)
2876 return -EINVAL;
2878 /* Get association, if sack_assoc_id != 0 and the socket is a one
2879 * to many style socket, and an association was not found, then
2880 * the id was invalid.
2882 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2883 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2884 return -EINVAL;
2886 if (params.sack_delay) {
2887 if (asoc) {
2888 asoc->sackdelay =
2889 msecs_to_jiffies(params.sack_delay);
2890 asoc->param_flags =
2891 sctp_spp_sackdelay_enable(asoc->param_flags);
2892 } else {
2893 sp->sackdelay = params.sack_delay;
2894 sp->param_flags =
2895 sctp_spp_sackdelay_enable(sp->param_flags);
2899 if (params.sack_freq == 1) {
2900 if (asoc) {
2901 asoc->param_flags =
2902 sctp_spp_sackdelay_disable(asoc->param_flags);
2903 } else {
2904 sp->param_flags =
2905 sctp_spp_sackdelay_disable(sp->param_flags);
2907 } else if (params.sack_freq > 1) {
2908 if (asoc) {
2909 asoc->sackfreq = params.sack_freq;
2910 asoc->param_flags =
2911 sctp_spp_sackdelay_enable(asoc->param_flags);
2912 } else {
2913 sp->sackfreq = params.sack_freq;
2914 sp->param_flags =
2915 sctp_spp_sackdelay_enable(sp->param_flags);
2919 /* If change is for association, also apply to each transport. */
2920 if (asoc) {
2921 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2922 transports) {
2923 if (params.sack_delay) {
2924 trans->sackdelay =
2925 msecs_to_jiffies(params.sack_delay);
2926 trans->param_flags =
2927 sctp_spp_sackdelay_enable(trans->param_flags);
2929 if (params.sack_freq == 1) {
2930 trans->param_flags =
2931 sctp_spp_sackdelay_disable(trans->param_flags);
2932 } else if (params.sack_freq > 1) {
2933 trans->sackfreq = params.sack_freq;
2934 trans->param_flags =
2935 sctp_spp_sackdelay_enable(trans->param_flags);
2940 return 0;
2943 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2945 * Applications can specify protocol parameters for the default association
2946 * initialization. The option name argument to setsockopt() and getsockopt()
2947 * is SCTP_INITMSG.
2949 * Setting initialization parameters is effective only on an unconnected
2950 * socket (for UDP-style sockets only future associations are effected
2951 * by the change). With TCP-style sockets, this option is inherited by
2952 * sockets derived from a listener socket.
2954 static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2956 struct sctp_initmsg sinit;
2957 struct sctp_sock *sp = sctp_sk(sk);
2959 if (optlen != sizeof(struct sctp_initmsg))
2960 return -EINVAL;
2961 if (copy_from_user(&sinit, optval, optlen))
2962 return -EFAULT;
2964 if (sinit.sinit_num_ostreams)
2965 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2966 if (sinit.sinit_max_instreams)
2967 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2968 if (sinit.sinit_max_attempts)
2969 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2970 if (sinit.sinit_max_init_timeo)
2971 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2973 return 0;
2977 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2979 * Applications that wish to use the sendto() system call may wish to
2980 * specify a default set of parameters that would normally be supplied
2981 * through the inclusion of ancillary data. This socket option allows
2982 * such an application to set the default sctp_sndrcvinfo structure.
2983 * The application that wishes to use this socket option simply passes
2984 * in to this call the sctp_sndrcvinfo structure defined in Section
2985 * 5.2.2) The input parameters accepted by this call include
2986 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2987 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2988 * to this call if the caller is using the UDP model.
2990 static int sctp_setsockopt_default_send_param(struct sock *sk,
2991 char __user *optval,
2992 unsigned int optlen)
2994 struct sctp_sock *sp = sctp_sk(sk);
2995 struct sctp_association *asoc;
2996 struct sctp_sndrcvinfo info;
2998 if (optlen != sizeof(info))
2999 return -EINVAL;
3000 if (copy_from_user(&info, optval, optlen))
3001 return -EFAULT;
3002 if (info.sinfo_flags &
3003 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3004 SCTP_ABORT | SCTP_EOF))
3005 return -EINVAL;
3007 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
3008 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
3009 return -EINVAL;
3010 if (asoc) {
3011 asoc->default_stream = info.sinfo_stream;
3012 asoc->default_flags = info.sinfo_flags;
3013 asoc->default_ppid = info.sinfo_ppid;
3014 asoc->default_context = info.sinfo_context;
3015 asoc->default_timetolive = info.sinfo_timetolive;
3016 } else {
3017 sp->default_stream = info.sinfo_stream;
3018 sp->default_flags = info.sinfo_flags;
3019 sp->default_ppid = info.sinfo_ppid;
3020 sp->default_context = info.sinfo_context;
3021 sp->default_timetolive = info.sinfo_timetolive;
3024 return 0;
3027 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
3028 * (SCTP_DEFAULT_SNDINFO)
3030 static int sctp_setsockopt_default_sndinfo(struct sock *sk,
3031 char __user *optval,
3032 unsigned int optlen)
3034 struct sctp_sock *sp = sctp_sk(sk);
3035 struct sctp_association *asoc;
3036 struct sctp_sndinfo info;
3038 if (optlen != sizeof(info))
3039 return -EINVAL;
3040 if (copy_from_user(&info, optval, optlen))
3041 return -EFAULT;
3042 if (info.snd_flags &
3043 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
3044 SCTP_ABORT | SCTP_EOF))
3045 return -EINVAL;
3047 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
3048 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
3049 return -EINVAL;
3050 if (asoc) {
3051 asoc->default_stream = info.snd_sid;
3052 asoc->default_flags = info.snd_flags;
3053 asoc->default_ppid = info.snd_ppid;
3054 asoc->default_context = info.snd_context;
3055 } else {
3056 sp->default_stream = info.snd_sid;
3057 sp->default_flags = info.snd_flags;
3058 sp->default_ppid = info.snd_ppid;
3059 sp->default_context = info.snd_context;
3062 return 0;
3065 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
3067 * Requests that the local SCTP stack use the enclosed peer address as
3068 * the association primary. The enclosed address must be one of the
3069 * association peer's addresses.
3071 static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
3072 unsigned int optlen)
3074 struct sctp_prim prim;
3075 struct sctp_transport *trans;
3076 struct sctp_af *af;
3077 int err;
3079 if (optlen != sizeof(struct sctp_prim))
3080 return -EINVAL;
3082 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
3083 return -EFAULT;
3085 /* Allow security module to validate address but need address len. */
3086 af = sctp_get_af_specific(prim.ssp_addr.ss_family);
3087 if (!af)
3088 return -EINVAL;
3090 err = security_sctp_bind_connect(sk, SCTP_PRIMARY_ADDR,
3091 (struct sockaddr *)&prim.ssp_addr,
3092 af->sockaddr_len);
3093 if (err)
3094 return err;
3096 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
3097 if (!trans)
3098 return -EINVAL;
3100 sctp_assoc_set_primary(trans->asoc, trans);
3102 return 0;
3106 * 7.1.5 SCTP_NODELAY
3108 * Turn on/off any Nagle-like algorithm. This means that packets are
3109 * generally sent as soon as possible and no unnecessary delays are
3110 * introduced, at the cost of more packets in the network. Expects an
3111 * integer boolean flag.
3113 static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
3114 unsigned int optlen)
3116 int val;
3118 if (optlen < sizeof(int))
3119 return -EINVAL;
3120 if (get_user(val, (int __user *)optval))
3121 return -EFAULT;
3123 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
3124 return 0;
3129 * 7.1.1 SCTP_RTOINFO
3131 * The protocol parameters used to initialize and bound retransmission
3132 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
3133 * and modify these parameters.
3134 * All parameters are time values, in milliseconds. A value of 0, when
3135 * modifying the parameters, indicates that the current value should not
3136 * be changed.
3139 static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
3141 struct sctp_rtoinfo rtoinfo;
3142 struct sctp_association *asoc;
3143 unsigned long rto_min, rto_max;
3144 struct sctp_sock *sp = sctp_sk(sk);
3146 if (optlen != sizeof (struct sctp_rtoinfo))
3147 return -EINVAL;
3149 if (copy_from_user(&rtoinfo, optval, optlen))
3150 return -EFAULT;
3152 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
3154 /* Set the values to the specific association */
3155 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
3156 return -EINVAL;
3158 rto_max = rtoinfo.srto_max;
3159 rto_min = rtoinfo.srto_min;
3161 if (rto_max)
3162 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
3163 else
3164 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
3166 if (rto_min)
3167 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
3168 else
3169 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
3171 if (rto_min > rto_max)
3172 return -EINVAL;
3174 if (asoc) {
3175 if (rtoinfo.srto_initial != 0)
3176 asoc->rto_initial =
3177 msecs_to_jiffies(rtoinfo.srto_initial);
3178 asoc->rto_max = rto_max;
3179 asoc->rto_min = rto_min;
3180 } else {
3181 /* If there is no association or the association-id = 0
3182 * set the values to the endpoint.
3184 if (rtoinfo.srto_initial != 0)
3185 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
3186 sp->rtoinfo.srto_max = rto_max;
3187 sp->rtoinfo.srto_min = rto_min;
3190 return 0;
3195 * 7.1.2 SCTP_ASSOCINFO
3197 * This option is used to tune the maximum retransmission attempts
3198 * of the association.
3199 * Returns an error if the new association retransmission value is
3200 * greater than the sum of the retransmission value of the peer.
3201 * See [SCTP] for more information.
3204 static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
3207 struct sctp_assocparams assocparams;
3208 struct sctp_association *asoc;
3210 if (optlen != sizeof(struct sctp_assocparams))
3211 return -EINVAL;
3212 if (copy_from_user(&assocparams, optval, optlen))
3213 return -EFAULT;
3215 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
3217 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
3218 return -EINVAL;
3220 /* Set the values to the specific association */
3221 if (asoc) {
3222 if (assocparams.sasoc_asocmaxrxt != 0) {
3223 __u32 path_sum = 0;
3224 int paths = 0;
3225 struct sctp_transport *peer_addr;
3227 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
3228 transports) {
3229 path_sum += peer_addr->pathmaxrxt;
3230 paths++;
3233 /* Only validate asocmaxrxt if we have more than
3234 * one path/transport. We do this because path
3235 * retransmissions are only counted when we have more
3236 * then one path.
3238 if (paths > 1 &&
3239 assocparams.sasoc_asocmaxrxt > path_sum)
3240 return -EINVAL;
3242 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
3245 if (assocparams.sasoc_cookie_life != 0)
3246 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
3247 } else {
3248 /* Set the values to the endpoint */
3249 struct sctp_sock *sp = sctp_sk(sk);
3251 if (assocparams.sasoc_asocmaxrxt != 0)
3252 sp->assocparams.sasoc_asocmaxrxt =
3253 assocparams.sasoc_asocmaxrxt;
3254 if (assocparams.sasoc_cookie_life != 0)
3255 sp->assocparams.sasoc_cookie_life =
3256 assocparams.sasoc_cookie_life;
3258 return 0;
3262 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
3264 * This socket option is a boolean flag which turns on or off mapped V4
3265 * addresses. If this option is turned on and the socket is type
3266 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
3267 * If this option is turned off, then no mapping will be done of V4
3268 * addresses and a user will receive both PF_INET6 and PF_INET type
3269 * addresses on the socket.
3271 static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
3273 int val;
3274 struct sctp_sock *sp = sctp_sk(sk);
3276 if (optlen < sizeof(int))
3277 return -EINVAL;
3278 if (get_user(val, (int __user *)optval))
3279 return -EFAULT;
3280 if (val)
3281 sp->v4mapped = 1;
3282 else
3283 sp->v4mapped = 0;
3285 return 0;
3289 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
3290 * This option will get or set the maximum size to put in any outgoing
3291 * SCTP DATA chunk. If a message is larger than this size it will be
3292 * fragmented by SCTP into the specified size. Note that the underlying
3293 * SCTP implementation may fragment into smaller sized chunks when the
3294 * PMTU of the underlying association is smaller than the value set by
3295 * the user. The default value for this option is '0' which indicates
3296 * the user is NOT limiting fragmentation and only the PMTU will effect
3297 * SCTP's choice of DATA chunk size. Note also that values set larger
3298 * than the maximum size of an IP datagram will effectively let SCTP
3299 * control fragmentation (i.e. the same as setting this option to 0).
3301 * The following structure is used to access and modify this parameter:
3303 * struct sctp_assoc_value {
3304 * sctp_assoc_t assoc_id;
3305 * uint32_t assoc_value;
3306 * };
3308 * assoc_id: This parameter is ignored for one-to-one style sockets.
3309 * For one-to-many style sockets this parameter indicates which
3310 * association the user is performing an action upon. Note that if
3311 * this field's value is zero then the endpoints default value is
3312 * changed (effecting future associations only).
3313 * assoc_value: This parameter specifies the maximum size in bytes.
3315 static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3317 struct sctp_sock *sp = sctp_sk(sk);
3318 struct sctp_assoc_value params;
3319 struct sctp_association *asoc;
3320 int val;
3322 if (optlen == sizeof(int)) {
3323 pr_warn_ratelimited(DEPRECATED
3324 "%s (pid %d) "
3325 "Use of int in maxseg socket option.\n"
3326 "Use struct sctp_assoc_value instead\n",
3327 current->comm, task_pid_nr(current));
3328 if (copy_from_user(&val, optval, optlen))
3329 return -EFAULT;
3330 params.assoc_id = 0;
3331 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3332 if (copy_from_user(&params, optval, optlen))
3333 return -EFAULT;
3334 val = params.assoc_value;
3335 } else {
3336 return -EINVAL;
3339 asoc = sctp_id2assoc(sk, params.assoc_id);
3341 if (val) {
3342 int min_len, max_len;
3343 __u16 datasize = asoc ? sctp_datachk_len(&asoc->stream) :
3344 sizeof(struct sctp_data_chunk);
3346 min_len = sctp_min_frag_point(sp, datasize);
3347 max_len = SCTP_MAX_CHUNK_LEN - datasize;
3349 if (val < min_len || val > max_len)
3350 return -EINVAL;
3353 if (asoc) {
3354 asoc->user_frag = val;
3355 sctp_assoc_update_frag_point(asoc);
3356 } else {
3357 if (params.assoc_id && sctp_style(sk, UDP))
3358 return -EINVAL;
3359 sp->user_frag = val;
3362 return 0;
3367 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3369 * Requests that the peer mark the enclosed address as the association
3370 * primary. The enclosed address must be one of the association's
3371 * locally bound addresses. The following structure is used to make a
3372 * set primary request:
3374 static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3375 unsigned int optlen)
3377 struct net *net = sock_net(sk);
3378 struct sctp_sock *sp;
3379 struct sctp_association *asoc = NULL;
3380 struct sctp_setpeerprim prim;
3381 struct sctp_chunk *chunk;
3382 struct sctp_af *af;
3383 int err;
3385 sp = sctp_sk(sk);
3387 if (!net->sctp.addip_enable)
3388 return -EPERM;
3390 if (optlen != sizeof(struct sctp_setpeerprim))
3391 return -EINVAL;
3393 if (copy_from_user(&prim, optval, optlen))
3394 return -EFAULT;
3396 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3397 if (!asoc)
3398 return -EINVAL;
3400 if (!asoc->peer.asconf_capable)
3401 return -EPERM;
3403 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3404 return -EPERM;
3406 if (!sctp_state(asoc, ESTABLISHED))
3407 return -ENOTCONN;
3409 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3410 if (!af)
3411 return -EINVAL;
3413 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3414 return -EADDRNOTAVAIL;
3416 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3417 return -EADDRNOTAVAIL;
3419 /* Allow security module to validate address. */
3420 err = security_sctp_bind_connect(sk, SCTP_SET_PEER_PRIMARY_ADDR,
3421 (struct sockaddr *)&prim.sspp_addr,
3422 af->sockaddr_len);
3423 if (err)
3424 return err;
3426 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3427 chunk = sctp_make_asconf_set_prim(asoc,
3428 (union sctp_addr *)&prim.sspp_addr);
3429 if (!chunk)
3430 return -ENOMEM;
3432 err = sctp_send_asconf(asoc, chunk);
3434 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3436 return err;
3439 static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3440 unsigned int optlen)
3442 struct sctp_setadaptation adaptation;
3444 if (optlen != sizeof(struct sctp_setadaptation))
3445 return -EINVAL;
3446 if (copy_from_user(&adaptation, optval, optlen))
3447 return -EFAULT;
3449 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3451 return 0;
3455 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3457 * The context field in the sctp_sndrcvinfo structure is normally only
3458 * used when a failed message is retrieved holding the value that was
3459 * sent down on the actual send call. This option allows the setting of
3460 * a default context on an association basis that will be received on
3461 * reading messages from the peer. This is especially helpful in the
3462 * one-2-many model for an application to keep some reference to an
3463 * internal state machine that is processing messages on the
3464 * association. Note that the setting of this value only effects
3465 * received messages from the peer and does not effect the value that is
3466 * saved with outbound messages.
3468 static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3469 unsigned int optlen)
3471 struct sctp_assoc_value params;
3472 struct sctp_sock *sp;
3473 struct sctp_association *asoc;
3475 if (optlen != sizeof(struct sctp_assoc_value))
3476 return -EINVAL;
3477 if (copy_from_user(&params, optval, optlen))
3478 return -EFAULT;
3480 sp = sctp_sk(sk);
3482 if (params.assoc_id != 0) {
3483 asoc = sctp_id2assoc(sk, params.assoc_id);
3484 if (!asoc)
3485 return -EINVAL;
3486 asoc->default_rcv_context = params.assoc_value;
3487 } else {
3488 sp->default_rcv_context = params.assoc_value;
3491 return 0;
3495 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3497 * This options will at a minimum specify if the implementation is doing
3498 * fragmented interleave. Fragmented interleave, for a one to many
3499 * socket, is when subsequent calls to receive a message may return
3500 * parts of messages from different associations. Some implementations
3501 * may allow you to turn this value on or off. If so, when turned off,
3502 * no fragment interleave will occur (which will cause a head of line
3503 * blocking amongst multiple associations sharing the same one to many
3504 * socket). When this option is turned on, then each receive call may
3505 * come from a different association (thus the user must receive data
3506 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3507 * association each receive belongs to.
3509 * This option takes a boolean value. A non-zero value indicates that
3510 * fragmented interleave is on. A value of zero indicates that
3511 * fragmented interleave is off.
3513 * Note that it is important that an implementation that allows this
3514 * option to be turned on, have it off by default. Otherwise an unaware
3515 * application using the one to many model may become confused and act
3516 * incorrectly.
3518 static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3519 char __user *optval,
3520 unsigned int optlen)
3522 int val;
3524 if (optlen != sizeof(int))
3525 return -EINVAL;
3526 if (get_user(val, (int __user *)optval))
3527 return -EFAULT;
3529 sctp_sk(sk)->frag_interleave = !!val;
3531 if (!sctp_sk(sk)->frag_interleave)
3532 sctp_sk(sk)->strm_interleave = 0;
3534 return 0;
3538 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3539 * (SCTP_PARTIAL_DELIVERY_POINT)
3541 * This option will set or get the SCTP partial delivery point. This
3542 * point is the size of a message where the partial delivery API will be
3543 * invoked to help free up rwnd space for the peer. Setting this to a
3544 * lower value will cause partial deliveries to happen more often. The
3545 * calls argument is an integer that sets or gets the partial delivery
3546 * point. Note also that the call will fail if the user attempts to set
3547 * this value larger than the socket receive buffer size.
3549 * Note that any single message having a length smaller than or equal to
3550 * the SCTP partial delivery point will be delivered in one single read
3551 * call as long as the user provided buffer is large enough to hold the
3552 * message.
3554 static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3555 char __user *optval,
3556 unsigned int optlen)
3558 u32 val;
3560 if (optlen != sizeof(u32))
3561 return -EINVAL;
3562 if (get_user(val, (int __user *)optval))
3563 return -EFAULT;
3565 /* Note: We double the receive buffer from what the user sets
3566 * it to be, also initial rwnd is based on rcvbuf/2.
3568 if (val > (sk->sk_rcvbuf >> 1))
3569 return -EINVAL;
3571 sctp_sk(sk)->pd_point = val;
3573 return 0; /* is this the right error code? */
3577 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3579 * This option will allow a user to change the maximum burst of packets
3580 * that can be emitted by this association. Note that the default value
3581 * is 4, and some implementations may restrict this setting so that it
3582 * can only be lowered.
3584 * NOTE: This text doesn't seem right. Do this on a socket basis with
3585 * future associations inheriting the socket value.
3587 static int sctp_setsockopt_maxburst(struct sock *sk,
3588 char __user *optval,
3589 unsigned int optlen)
3591 struct sctp_assoc_value params;
3592 struct sctp_sock *sp;
3593 struct sctp_association *asoc;
3594 int val;
3595 int assoc_id = 0;
3597 if (optlen == sizeof(int)) {
3598 pr_warn_ratelimited(DEPRECATED
3599 "%s (pid %d) "
3600 "Use of int in max_burst socket option deprecated.\n"
3601 "Use struct sctp_assoc_value instead\n",
3602 current->comm, task_pid_nr(current));
3603 if (copy_from_user(&val, optval, optlen))
3604 return -EFAULT;
3605 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3606 if (copy_from_user(&params, optval, optlen))
3607 return -EFAULT;
3608 val = params.assoc_value;
3609 assoc_id = params.assoc_id;
3610 } else
3611 return -EINVAL;
3613 sp = sctp_sk(sk);
3615 if (assoc_id != 0) {
3616 asoc = sctp_id2assoc(sk, assoc_id);
3617 if (!asoc)
3618 return -EINVAL;
3619 asoc->max_burst = val;
3620 } else
3621 sp->max_burst = val;
3623 return 0;
3627 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3629 * This set option adds a chunk type that the user is requesting to be
3630 * received only in an authenticated way. Changes to the list of chunks
3631 * will only effect future associations on the socket.
3633 static int sctp_setsockopt_auth_chunk(struct sock *sk,
3634 char __user *optval,
3635 unsigned int optlen)
3637 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3638 struct sctp_authchunk val;
3640 if (!ep->auth_enable)
3641 return -EACCES;
3643 if (optlen != sizeof(struct sctp_authchunk))
3644 return -EINVAL;
3645 if (copy_from_user(&val, optval, optlen))
3646 return -EFAULT;
3648 switch (val.sauth_chunk) {
3649 case SCTP_CID_INIT:
3650 case SCTP_CID_INIT_ACK:
3651 case SCTP_CID_SHUTDOWN_COMPLETE:
3652 case SCTP_CID_AUTH:
3653 return -EINVAL;
3656 /* add this chunk id to the endpoint */
3657 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3661 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3663 * This option gets or sets the list of HMAC algorithms that the local
3664 * endpoint requires the peer to use.
3666 static int sctp_setsockopt_hmac_ident(struct sock *sk,
3667 char __user *optval,
3668 unsigned int optlen)
3670 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3671 struct sctp_hmacalgo *hmacs;
3672 u32 idents;
3673 int err;
3675 if (!ep->auth_enable)
3676 return -EACCES;
3678 if (optlen < sizeof(struct sctp_hmacalgo))
3679 return -EINVAL;
3680 optlen = min_t(unsigned int, optlen, sizeof(struct sctp_hmacalgo) +
3681 SCTP_AUTH_NUM_HMACS * sizeof(u16));
3683 hmacs = memdup_user(optval, optlen);
3684 if (IS_ERR(hmacs))
3685 return PTR_ERR(hmacs);
3687 idents = hmacs->shmac_num_idents;
3688 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3689 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3690 err = -EINVAL;
3691 goto out;
3694 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3695 out:
3696 kfree(hmacs);
3697 return err;
3701 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3703 * This option will set a shared secret key which is used to build an
3704 * association shared key.
3706 static int sctp_setsockopt_auth_key(struct sock *sk,
3707 char __user *optval,
3708 unsigned int optlen)
3710 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3711 struct sctp_authkey *authkey;
3712 struct sctp_association *asoc;
3713 int ret;
3715 if (!ep->auth_enable)
3716 return -EACCES;
3718 if (optlen <= sizeof(struct sctp_authkey))
3719 return -EINVAL;
3720 /* authkey->sca_keylength is u16, so optlen can't be bigger than
3721 * this.
3723 optlen = min_t(unsigned int, optlen, USHRT_MAX +
3724 sizeof(struct sctp_authkey));
3726 authkey = memdup_user(optval, optlen);
3727 if (IS_ERR(authkey))
3728 return PTR_ERR(authkey);
3730 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3731 ret = -EINVAL;
3732 goto out;
3735 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3736 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3737 ret = -EINVAL;
3738 goto out;
3741 ret = sctp_auth_set_key(ep, asoc, authkey);
3742 out:
3743 kzfree(authkey);
3744 return ret;
3748 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3750 * This option will get or set the active shared key to be used to build
3751 * the association shared key.
3753 static int sctp_setsockopt_active_key(struct sock *sk,
3754 char __user *optval,
3755 unsigned int optlen)
3757 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3758 struct sctp_authkeyid val;
3759 struct sctp_association *asoc;
3761 if (!ep->auth_enable)
3762 return -EACCES;
3764 if (optlen != sizeof(struct sctp_authkeyid))
3765 return -EINVAL;
3766 if (copy_from_user(&val, optval, optlen))
3767 return -EFAULT;
3769 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3770 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3771 return -EINVAL;
3773 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3777 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3779 * This set option will delete a shared secret key from use.
3781 static int sctp_setsockopt_del_key(struct sock *sk,
3782 char __user *optval,
3783 unsigned int optlen)
3785 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3786 struct sctp_authkeyid val;
3787 struct sctp_association *asoc;
3789 if (!ep->auth_enable)
3790 return -EACCES;
3792 if (optlen != sizeof(struct sctp_authkeyid))
3793 return -EINVAL;
3794 if (copy_from_user(&val, optval, optlen))
3795 return -EFAULT;
3797 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3798 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3799 return -EINVAL;
3801 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3806 * 8.3.4 Deactivate a Shared Key (SCTP_AUTH_DEACTIVATE_KEY)
3808 * This set option will deactivate a shared secret key.
3810 static int sctp_setsockopt_deactivate_key(struct sock *sk, char __user *optval,
3811 unsigned int optlen)
3813 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3814 struct sctp_authkeyid val;
3815 struct sctp_association *asoc;
3817 if (!ep->auth_enable)
3818 return -EACCES;
3820 if (optlen != sizeof(struct sctp_authkeyid))
3821 return -EINVAL;
3822 if (copy_from_user(&val, optval, optlen))
3823 return -EFAULT;
3825 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3826 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3827 return -EINVAL;
3829 return sctp_auth_deact_key_id(ep, asoc, val.scact_keynumber);
3833 * 8.1.23 SCTP_AUTO_ASCONF
3835 * This option will enable or disable the use of the automatic generation of
3836 * ASCONF chunks to add and delete addresses to an existing association. Note
3837 * that this option has two caveats namely: a) it only affects sockets that
3838 * are bound to all addresses available to the SCTP stack, and b) the system
3839 * administrator may have an overriding control that turns the ASCONF feature
3840 * off no matter what setting the socket option may have.
3841 * This option expects an integer boolean flag, where a non-zero value turns on
3842 * the option, and a zero value turns off the option.
3843 * Note. In this implementation, socket operation overrides default parameter
3844 * being set by sysctl as well as FreeBSD implementation
3846 static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3847 unsigned int optlen)
3849 int val;
3850 struct sctp_sock *sp = sctp_sk(sk);
3852 if (optlen < sizeof(int))
3853 return -EINVAL;
3854 if (get_user(val, (int __user *)optval))
3855 return -EFAULT;
3856 if (!sctp_is_ep_boundall(sk) && val)
3857 return -EINVAL;
3858 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3859 return 0;
3861 spin_lock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3862 if (val == 0 && sp->do_auto_asconf) {
3863 list_del(&sp->auto_asconf_list);
3864 sp->do_auto_asconf = 0;
3865 } else if (val && !sp->do_auto_asconf) {
3866 list_add_tail(&sp->auto_asconf_list,
3867 &sock_net(sk)->sctp.auto_asconf_splist);
3868 sp->do_auto_asconf = 1;
3870 spin_unlock_bh(&sock_net(sk)->sctp.addr_wq_lock);
3871 return 0;
3875 * SCTP_PEER_ADDR_THLDS
3877 * This option allows us to alter the partially failed threshold for one or all
3878 * transports in an association. See Section 6.1 of:
3879 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3881 static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3882 char __user *optval,
3883 unsigned int optlen)
3885 struct sctp_paddrthlds val;
3886 struct sctp_transport *trans;
3887 struct sctp_association *asoc;
3889 if (optlen < sizeof(struct sctp_paddrthlds))
3890 return -EINVAL;
3891 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3892 sizeof(struct sctp_paddrthlds)))
3893 return -EFAULT;
3896 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3897 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3898 if (!asoc)
3899 return -ENOENT;
3900 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3901 transports) {
3902 if (val.spt_pathmaxrxt)
3903 trans->pathmaxrxt = val.spt_pathmaxrxt;
3904 trans->pf_retrans = val.spt_pathpfthld;
3907 if (val.spt_pathmaxrxt)
3908 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3909 asoc->pf_retrans = val.spt_pathpfthld;
3910 } else {
3911 trans = sctp_addr_id2transport(sk, &val.spt_address,
3912 val.spt_assoc_id);
3913 if (!trans)
3914 return -ENOENT;
3916 if (val.spt_pathmaxrxt)
3917 trans->pathmaxrxt = val.spt_pathmaxrxt;
3918 trans->pf_retrans = val.spt_pathpfthld;
3921 return 0;
3924 static int sctp_setsockopt_recvrcvinfo(struct sock *sk,
3925 char __user *optval,
3926 unsigned int optlen)
3928 int val;
3930 if (optlen < sizeof(int))
3931 return -EINVAL;
3932 if (get_user(val, (int __user *) optval))
3933 return -EFAULT;
3935 sctp_sk(sk)->recvrcvinfo = (val == 0) ? 0 : 1;
3937 return 0;
3940 static int sctp_setsockopt_recvnxtinfo(struct sock *sk,
3941 char __user *optval,
3942 unsigned int optlen)
3944 int val;
3946 if (optlen < sizeof(int))
3947 return -EINVAL;
3948 if (get_user(val, (int __user *) optval))
3949 return -EFAULT;
3951 sctp_sk(sk)->recvnxtinfo = (val == 0) ? 0 : 1;
3953 return 0;
3956 static int sctp_setsockopt_pr_supported(struct sock *sk,
3957 char __user *optval,
3958 unsigned int optlen)
3960 struct sctp_assoc_value params;
3962 if (optlen != sizeof(params))
3963 return -EINVAL;
3965 if (copy_from_user(&params, optval, optlen))
3966 return -EFAULT;
3968 sctp_sk(sk)->ep->prsctp_enable = !!params.assoc_value;
3970 return 0;
3973 static int sctp_setsockopt_default_prinfo(struct sock *sk,
3974 char __user *optval,
3975 unsigned int optlen)
3977 struct sctp_default_prinfo info;
3978 struct sctp_association *asoc;
3979 int retval = -EINVAL;
3981 if (optlen != sizeof(info))
3982 goto out;
3984 if (copy_from_user(&info, optval, sizeof(info))) {
3985 retval = -EFAULT;
3986 goto out;
3989 if (info.pr_policy & ~SCTP_PR_SCTP_MASK)
3990 goto out;
3992 if (info.pr_policy == SCTP_PR_SCTP_NONE)
3993 info.pr_value = 0;
3995 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
3996 if (asoc) {
3997 SCTP_PR_SET_POLICY(asoc->default_flags, info.pr_policy);
3998 asoc->default_timetolive = info.pr_value;
3999 } else if (!info.pr_assoc_id) {
4000 struct sctp_sock *sp = sctp_sk(sk);
4002 SCTP_PR_SET_POLICY(sp->default_flags, info.pr_policy);
4003 sp->default_timetolive = info.pr_value;
4004 } else {
4005 goto out;
4008 retval = 0;
4010 out:
4011 return retval;
4014 static int sctp_setsockopt_reconfig_supported(struct sock *sk,
4015 char __user *optval,
4016 unsigned int optlen)
4018 struct sctp_assoc_value params;
4019 struct sctp_association *asoc;
4020 int retval = -EINVAL;
4022 if (optlen != sizeof(params))
4023 goto out;
4025 if (copy_from_user(&params, optval, optlen)) {
4026 retval = -EFAULT;
4027 goto out;
4030 asoc = sctp_id2assoc(sk, params.assoc_id);
4031 if (asoc) {
4032 asoc->reconf_enable = !!params.assoc_value;
4033 } else if (!params.assoc_id) {
4034 struct sctp_sock *sp = sctp_sk(sk);
4036 sp->ep->reconf_enable = !!params.assoc_value;
4037 } else {
4038 goto out;
4041 retval = 0;
4043 out:
4044 return retval;
4047 static int sctp_setsockopt_enable_strreset(struct sock *sk,
4048 char __user *optval,
4049 unsigned int optlen)
4051 struct sctp_assoc_value params;
4052 struct sctp_association *asoc;
4053 int retval = -EINVAL;
4055 if (optlen != sizeof(params))
4056 goto out;
4058 if (copy_from_user(&params, optval, optlen)) {
4059 retval = -EFAULT;
4060 goto out;
4063 if (params.assoc_value & (~SCTP_ENABLE_STRRESET_MASK))
4064 goto out;
4066 asoc = sctp_id2assoc(sk, params.assoc_id);
4067 if (asoc) {
4068 asoc->strreset_enable = params.assoc_value;
4069 } else if (!params.assoc_id) {
4070 struct sctp_sock *sp = sctp_sk(sk);
4072 sp->ep->strreset_enable = params.assoc_value;
4073 } else {
4074 goto out;
4077 retval = 0;
4079 out:
4080 return retval;
4083 static int sctp_setsockopt_reset_streams(struct sock *sk,
4084 char __user *optval,
4085 unsigned int optlen)
4087 struct sctp_reset_streams *params;
4088 struct sctp_association *asoc;
4089 int retval = -EINVAL;
4091 if (optlen < sizeof(*params))
4092 return -EINVAL;
4093 /* srs_number_streams is u16, so optlen can't be bigger than this. */
4094 optlen = min_t(unsigned int, optlen, USHRT_MAX +
4095 sizeof(__u16) * sizeof(*params));
4097 params = memdup_user(optval, optlen);
4098 if (IS_ERR(params))
4099 return PTR_ERR(params);
4101 if (params->srs_number_streams * sizeof(__u16) >
4102 optlen - sizeof(*params))
4103 goto out;
4105 asoc = sctp_id2assoc(sk, params->srs_assoc_id);
4106 if (!asoc)
4107 goto out;
4109 retval = sctp_send_reset_streams(asoc, params);
4111 out:
4112 kfree(params);
4113 return retval;
4116 static int sctp_setsockopt_reset_assoc(struct sock *sk,
4117 char __user *optval,
4118 unsigned int optlen)
4120 struct sctp_association *asoc;
4121 sctp_assoc_t associd;
4122 int retval = -EINVAL;
4124 if (optlen != sizeof(associd))
4125 goto out;
4127 if (copy_from_user(&associd, optval, optlen)) {
4128 retval = -EFAULT;
4129 goto out;
4132 asoc = sctp_id2assoc(sk, associd);
4133 if (!asoc)
4134 goto out;
4136 retval = sctp_send_reset_assoc(asoc);
4138 out:
4139 return retval;
4142 static int sctp_setsockopt_add_streams(struct sock *sk,
4143 char __user *optval,
4144 unsigned int optlen)
4146 struct sctp_association *asoc;
4147 struct sctp_add_streams params;
4148 int retval = -EINVAL;
4150 if (optlen != sizeof(params))
4151 goto out;
4153 if (copy_from_user(&params, optval, optlen)) {
4154 retval = -EFAULT;
4155 goto out;
4158 asoc = sctp_id2assoc(sk, params.sas_assoc_id);
4159 if (!asoc)
4160 goto out;
4162 retval = sctp_send_add_streams(asoc, &params);
4164 out:
4165 return retval;
4168 static int sctp_setsockopt_scheduler(struct sock *sk,
4169 char __user *optval,
4170 unsigned int optlen)
4172 struct sctp_association *asoc;
4173 struct sctp_assoc_value params;
4174 int retval = -EINVAL;
4176 if (optlen < sizeof(params))
4177 goto out;
4179 optlen = sizeof(params);
4180 if (copy_from_user(&params, optval, optlen)) {
4181 retval = -EFAULT;
4182 goto out;
4185 if (params.assoc_value > SCTP_SS_MAX)
4186 goto out;
4188 asoc = sctp_id2assoc(sk, params.assoc_id);
4189 if (!asoc)
4190 goto out;
4192 retval = sctp_sched_set_sched(asoc, params.assoc_value);
4194 out:
4195 return retval;
4198 static int sctp_setsockopt_scheduler_value(struct sock *sk,
4199 char __user *optval,
4200 unsigned int optlen)
4202 struct sctp_association *asoc;
4203 struct sctp_stream_value params;
4204 int retval = -EINVAL;
4206 if (optlen < sizeof(params))
4207 goto out;
4209 optlen = sizeof(params);
4210 if (copy_from_user(&params, optval, optlen)) {
4211 retval = -EFAULT;
4212 goto out;
4215 asoc = sctp_id2assoc(sk, params.assoc_id);
4216 if (!asoc)
4217 goto out;
4219 retval = sctp_sched_set_value(asoc, params.stream_id,
4220 params.stream_value, GFP_KERNEL);
4222 out:
4223 return retval;
4226 static int sctp_setsockopt_interleaving_supported(struct sock *sk,
4227 char __user *optval,
4228 unsigned int optlen)
4230 struct sctp_sock *sp = sctp_sk(sk);
4231 struct net *net = sock_net(sk);
4232 struct sctp_assoc_value params;
4233 int retval = -EINVAL;
4235 if (optlen < sizeof(params))
4236 goto out;
4238 optlen = sizeof(params);
4239 if (copy_from_user(&params, optval, optlen)) {
4240 retval = -EFAULT;
4241 goto out;
4244 if (params.assoc_id)
4245 goto out;
4247 if (!net->sctp.intl_enable || !sp->frag_interleave) {
4248 retval = -EPERM;
4249 goto out;
4252 sp->strm_interleave = !!params.assoc_value;
4254 retval = 0;
4256 out:
4257 return retval;
4260 static int sctp_setsockopt_reuse_port(struct sock *sk, char __user *optval,
4261 unsigned int optlen)
4263 int val;
4265 if (!sctp_style(sk, TCP))
4266 return -EOPNOTSUPP;
4268 if (sctp_sk(sk)->ep->base.bind_addr.port)
4269 return -EFAULT;
4271 if (optlen < sizeof(int))
4272 return -EINVAL;
4274 if (get_user(val, (int __user *)optval))
4275 return -EFAULT;
4277 sctp_sk(sk)->reuse = !!val;
4279 return 0;
4282 /* API 6.2 setsockopt(), getsockopt()
4284 * Applications use setsockopt() and getsockopt() to set or retrieve
4285 * socket options. Socket options are used to change the default
4286 * behavior of sockets calls. They are described in Section 7.
4288 * The syntax is:
4290 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
4291 * int __user *optlen);
4292 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
4293 * int optlen);
4295 * sd - the socket descript.
4296 * level - set to IPPROTO_SCTP for all SCTP options.
4297 * optname - the option name.
4298 * optval - the buffer to store the value of the option.
4299 * optlen - the size of the buffer.
4301 static int sctp_setsockopt(struct sock *sk, int level, int optname,
4302 char __user *optval, unsigned int optlen)
4304 int retval = 0;
4306 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
4308 /* I can hardly begin to describe how wrong this is. This is
4309 * so broken as to be worse than useless. The API draft
4310 * REALLY is NOT helpful here... I am not convinced that the
4311 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
4312 * are at all well-founded.
4314 if (level != SOL_SCTP) {
4315 struct sctp_af *af = sctp_sk(sk)->pf->af;
4316 retval = af->setsockopt(sk, level, optname, optval, optlen);
4317 goto out_nounlock;
4320 lock_sock(sk);
4322 switch (optname) {
4323 case SCTP_SOCKOPT_BINDX_ADD:
4324 /* 'optlen' is the size of the addresses buffer. */
4325 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4326 optlen, SCTP_BINDX_ADD_ADDR);
4327 break;
4329 case SCTP_SOCKOPT_BINDX_REM:
4330 /* 'optlen' is the size of the addresses buffer. */
4331 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
4332 optlen, SCTP_BINDX_REM_ADDR);
4333 break;
4335 case SCTP_SOCKOPT_CONNECTX_OLD:
4336 /* 'optlen' is the size of the addresses buffer. */
4337 retval = sctp_setsockopt_connectx_old(sk,
4338 (struct sockaddr __user *)optval,
4339 optlen);
4340 break;
4342 case SCTP_SOCKOPT_CONNECTX:
4343 /* 'optlen' is the size of the addresses buffer. */
4344 retval = sctp_setsockopt_connectx(sk,
4345 (struct sockaddr __user *)optval,
4346 optlen);
4347 break;
4349 case SCTP_DISABLE_FRAGMENTS:
4350 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
4351 break;
4353 case SCTP_EVENTS:
4354 retval = sctp_setsockopt_events(sk, optval, optlen);
4355 break;
4357 case SCTP_AUTOCLOSE:
4358 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
4359 break;
4361 case SCTP_PEER_ADDR_PARAMS:
4362 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
4363 break;
4365 case SCTP_DELAYED_SACK:
4366 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
4367 break;
4368 case SCTP_PARTIAL_DELIVERY_POINT:
4369 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
4370 break;
4372 case SCTP_INITMSG:
4373 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
4374 break;
4375 case SCTP_DEFAULT_SEND_PARAM:
4376 retval = sctp_setsockopt_default_send_param(sk, optval,
4377 optlen);
4378 break;
4379 case SCTP_DEFAULT_SNDINFO:
4380 retval = sctp_setsockopt_default_sndinfo(sk, optval, optlen);
4381 break;
4382 case SCTP_PRIMARY_ADDR:
4383 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
4384 break;
4385 case SCTP_SET_PEER_PRIMARY_ADDR:
4386 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
4387 break;
4388 case SCTP_NODELAY:
4389 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
4390 break;
4391 case SCTP_RTOINFO:
4392 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
4393 break;
4394 case SCTP_ASSOCINFO:
4395 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
4396 break;
4397 case SCTP_I_WANT_MAPPED_V4_ADDR:
4398 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
4399 break;
4400 case SCTP_MAXSEG:
4401 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
4402 break;
4403 case SCTP_ADAPTATION_LAYER:
4404 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
4405 break;
4406 case SCTP_CONTEXT:
4407 retval = sctp_setsockopt_context(sk, optval, optlen);
4408 break;
4409 case SCTP_FRAGMENT_INTERLEAVE:
4410 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
4411 break;
4412 case SCTP_MAX_BURST:
4413 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
4414 break;
4415 case SCTP_AUTH_CHUNK:
4416 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
4417 break;
4418 case SCTP_HMAC_IDENT:
4419 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
4420 break;
4421 case SCTP_AUTH_KEY:
4422 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
4423 break;
4424 case SCTP_AUTH_ACTIVE_KEY:
4425 retval = sctp_setsockopt_active_key(sk, optval, optlen);
4426 break;
4427 case SCTP_AUTH_DELETE_KEY:
4428 retval = sctp_setsockopt_del_key(sk, optval, optlen);
4429 break;
4430 case SCTP_AUTH_DEACTIVATE_KEY:
4431 retval = sctp_setsockopt_deactivate_key(sk, optval, optlen);
4432 break;
4433 case SCTP_AUTO_ASCONF:
4434 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
4435 break;
4436 case SCTP_PEER_ADDR_THLDS:
4437 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
4438 break;
4439 case SCTP_RECVRCVINFO:
4440 retval = sctp_setsockopt_recvrcvinfo(sk, optval, optlen);
4441 break;
4442 case SCTP_RECVNXTINFO:
4443 retval = sctp_setsockopt_recvnxtinfo(sk, optval, optlen);
4444 break;
4445 case SCTP_PR_SUPPORTED:
4446 retval = sctp_setsockopt_pr_supported(sk, optval, optlen);
4447 break;
4448 case SCTP_DEFAULT_PRINFO:
4449 retval = sctp_setsockopt_default_prinfo(sk, optval, optlen);
4450 break;
4451 case SCTP_RECONFIG_SUPPORTED:
4452 retval = sctp_setsockopt_reconfig_supported(sk, optval, optlen);
4453 break;
4454 case SCTP_ENABLE_STREAM_RESET:
4455 retval = sctp_setsockopt_enable_strreset(sk, optval, optlen);
4456 break;
4457 case SCTP_RESET_STREAMS:
4458 retval = sctp_setsockopt_reset_streams(sk, optval, optlen);
4459 break;
4460 case SCTP_RESET_ASSOC:
4461 retval = sctp_setsockopt_reset_assoc(sk, optval, optlen);
4462 break;
4463 case SCTP_ADD_STREAMS:
4464 retval = sctp_setsockopt_add_streams(sk, optval, optlen);
4465 break;
4466 case SCTP_STREAM_SCHEDULER:
4467 retval = sctp_setsockopt_scheduler(sk, optval, optlen);
4468 break;
4469 case SCTP_STREAM_SCHEDULER_VALUE:
4470 retval = sctp_setsockopt_scheduler_value(sk, optval, optlen);
4471 break;
4472 case SCTP_INTERLEAVING_SUPPORTED:
4473 retval = sctp_setsockopt_interleaving_supported(sk, optval,
4474 optlen);
4475 break;
4476 case SCTP_REUSE_PORT:
4477 retval = sctp_setsockopt_reuse_port(sk, optval, optlen);
4478 break;
4479 default:
4480 retval = -ENOPROTOOPT;
4481 break;
4484 release_sock(sk);
4486 out_nounlock:
4487 return retval;
4490 /* API 3.1.6 connect() - UDP Style Syntax
4492 * An application may use the connect() call in the UDP model to initiate an
4493 * association without sending data.
4495 * The syntax is:
4497 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
4499 * sd: the socket descriptor to have a new association added to.
4501 * nam: the address structure (either struct sockaddr_in or struct
4502 * sockaddr_in6 defined in RFC2553 [7]).
4504 * len: the size of the address.
4506 static int sctp_connect(struct sock *sk, struct sockaddr *addr,
4507 int addr_len, int flags)
4509 struct sctp_af *af;
4510 int err = -EINVAL;
4512 lock_sock(sk);
4514 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
4515 addr, addr_len);
4517 /* Validate addr_len before calling common connect/connectx routine. */
4518 af = sctp_get_af_specific(addr->sa_family);
4519 if (af && addr_len >= af->sockaddr_len)
4520 err = __sctp_connect(sk, addr, af->sockaddr_len, flags, NULL);
4522 release_sock(sk);
4523 return err;
4526 int sctp_inet_connect(struct socket *sock, struct sockaddr *uaddr,
4527 int addr_len, int flags)
4529 if (addr_len < sizeof(uaddr->sa_family))
4530 return -EINVAL;
4532 if (uaddr->sa_family == AF_UNSPEC)
4533 return -EOPNOTSUPP;
4535 return sctp_connect(sock->sk, uaddr, addr_len, flags);
4538 /* FIXME: Write comments. */
4539 static int sctp_disconnect(struct sock *sk, int flags)
4541 return -EOPNOTSUPP; /* STUB */
4544 /* 4.1.4 accept() - TCP Style Syntax
4546 * Applications use accept() call to remove an established SCTP
4547 * association from the accept queue of the endpoint. A new socket
4548 * descriptor will be returned from accept() to represent the newly
4549 * formed association.
4551 static struct sock *sctp_accept(struct sock *sk, int flags, int *err, bool kern)
4553 struct sctp_sock *sp;
4554 struct sctp_endpoint *ep;
4555 struct sock *newsk = NULL;
4556 struct sctp_association *asoc;
4557 long timeo;
4558 int error = 0;
4560 lock_sock(sk);
4562 sp = sctp_sk(sk);
4563 ep = sp->ep;
4565 if (!sctp_style(sk, TCP)) {
4566 error = -EOPNOTSUPP;
4567 goto out;
4570 if (!sctp_sstate(sk, LISTENING)) {
4571 error = -EINVAL;
4572 goto out;
4575 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
4577 error = sctp_wait_for_accept(sk, timeo);
4578 if (error)
4579 goto out;
4581 /* We treat the list of associations on the endpoint as the accept
4582 * queue and pick the first association on the list.
4584 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
4586 newsk = sp->pf->create_accept_sk(sk, asoc, kern);
4587 if (!newsk) {
4588 error = -ENOMEM;
4589 goto out;
4592 /* Populate the fields of the newsk from the oldsk and migrate the
4593 * asoc to the newsk.
4595 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
4597 out:
4598 release_sock(sk);
4599 *err = error;
4600 return newsk;
4603 /* The SCTP ioctl handler. */
4604 static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
4606 int rc = -ENOTCONN;
4608 lock_sock(sk);
4611 * SEQPACKET-style sockets in LISTENING state are valid, for
4612 * SCTP, so only discard TCP-style sockets in LISTENING state.
4614 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
4615 goto out;
4617 switch (cmd) {
4618 case SIOCINQ: {
4619 struct sk_buff *skb;
4620 unsigned int amount = 0;
4622 skb = skb_peek(&sk->sk_receive_queue);
4623 if (skb != NULL) {
4625 * We will only return the amount of this packet since
4626 * that is all that will be read.
4628 amount = skb->len;
4630 rc = put_user(amount, (int __user *)arg);
4631 break;
4633 default:
4634 rc = -ENOIOCTLCMD;
4635 break;
4637 out:
4638 release_sock(sk);
4639 return rc;
4642 /* This is the function which gets called during socket creation to
4643 * initialized the SCTP-specific portion of the sock.
4644 * The sock structure should already be zero-filled memory.
4646 static int sctp_init_sock(struct sock *sk)
4648 struct net *net = sock_net(sk);
4649 struct sctp_sock *sp;
4651 pr_debug("%s: sk:%p\n", __func__, sk);
4653 sp = sctp_sk(sk);
4655 /* Initialize the SCTP per socket area. */
4656 switch (sk->sk_type) {
4657 case SOCK_SEQPACKET:
4658 sp->type = SCTP_SOCKET_UDP;
4659 break;
4660 case SOCK_STREAM:
4661 sp->type = SCTP_SOCKET_TCP;
4662 break;
4663 default:
4664 return -ESOCKTNOSUPPORT;
4667 sk->sk_gso_type = SKB_GSO_SCTP;
4669 /* Initialize default send parameters. These parameters can be
4670 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
4672 sp->default_stream = 0;
4673 sp->default_ppid = 0;
4674 sp->default_flags = 0;
4675 sp->default_context = 0;
4676 sp->default_timetolive = 0;
4678 sp->default_rcv_context = 0;
4679 sp->max_burst = net->sctp.max_burst;
4681 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
4683 /* Initialize default setup parameters. These parameters
4684 * can be modified with the SCTP_INITMSG socket option or
4685 * overridden by the SCTP_INIT CMSG.
4687 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
4688 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
4689 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
4690 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
4692 /* Initialize default RTO related parameters. These parameters can
4693 * be modified for with the SCTP_RTOINFO socket option.
4695 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
4696 sp->rtoinfo.srto_max = net->sctp.rto_max;
4697 sp->rtoinfo.srto_min = net->sctp.rto_min;
4699 /* Initialize default association related parameters. These parameters
4700 * can be modified with the SCTP_ASSOCINFO socket option.
4702 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
4703 sp->assocparams.sasoc_number_peer_destinations = 0;
4704 sp->assocparams.sasoc_peer_rwnd = 0;
4705 sp->assocparams.sasoc_local_rwnd = 0;
4706 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
4708 /* Initialize default event subscriptions. By default, all the
4709 * options are off.
4711 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
4713 /* Default Peer Address Parameters. These defaults can
4714 * be modified via SCTP_PEER_ADDR_PARAMS
4716 sp->hbinterval = net->sctp.hb_interval;
4717 sp->pathmaxrxt = net->sctp.max_retrans_path;
4718 sp->pathmtu = 0; /* allow default discovery */
4719 sp->sackdelay = net->sctp.sack_timeout;
4720 sp->sackfreq = 2;
4721 sp->param_flags = SPP_HB_ENABLE |
4722 SPP_PMTUD_ENABLE |
4723 SPP_SACKDELAY_ENABLE;
4725 /* If enabled no SCTP message fragmentation will be performed.
4726 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
4728 sp->disable_fragments = 0;
4730 /* Enable Nagle algorithm by default. */
4731 sp->nodelay = 0;
4733 sp->recvrcvinfo = 0;
4734 sp->recvnxtinfo = 0;
4736 /* Enable by default. */
4737 sp->v4mapped = 1;
4739 /* Auto-close idle associations after the configured
4740 * number of seconds. A value of 0 disables this
4741 * feature. Configure through the SCTP_AUTOCLOSE socket option,
4742 * for UDP-style sockets only.
4744 sp->autoclose = 0;
4746 /* User specified fragmentation limit. */
4747 sp->user_frag = 0;
4749 sp->adaptation_ind = 0;
4751 sp->pf = sctp_get_pf_specific(sk->sk_family);
4753 /* Control variables for partial data delivery. */
4754 atomic_set(&sp->pd_mode, 0);
4755 skb_queue_head_init(&sp->pd_lobby);
4756 sp->frag_interleave = 0;
4758 /* Create a per socket endpoint structure. Even if we
4759 * change the data structure relationships, this may still
4760 * be useful for storing pre-connect address information.
4762 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4763 if (!sp->ep)
4764 return -ENOMEM;
4766 sp->hmac = NULL;
4768 sk->sk_destruct = sctp_destruct_sock;
4770 SCTP_DBG_OBJCNT_INC(sock);
4772 local_bh_disable();
4773 sk_sockets_allocated_inc(sk);
4774 sock_prot_inuse_add(net, sk->sk_prot, 1);
4776 /* Nothing can fail after this block, otherwise
4777 * sctp_destroy_sock() will be called without addr_wq_lock held
4779 if (net->sctp.default_auto_asconf) {
4780 spin_lock(&sock_net(sk)->sctp.addr_wq_lock);
4781 list_add_tail(&sp->auto_asconf_list,
4782 &net->sctp.auto_asconf_splist);
4783 sp->do_auto_asconf = 1;
4784 spin_unlock(&sock_net(sk)->sctp.addr_wq_lock);
4785 } else {
4786 sp->do_auto_asconf = 0;
4789 local_bh_enable();
4791 return 0;
4794 /* Cleanup any SCTP per socket resources. Must be called with
4795 * sock_net(sk)->sctp.addr_wq_lock held if sp->do_auto_asconf is true
4797 static void sctp_destroy_sock(struct sock *sk)
4799 struct sctp_sock *sp;
4801 pr_debug("%s: sk:%p\n", __func__, sk);
4803 /* Release our hold on the endpoint. */
4804 sp = sctp_sk(sk);
4805 /* This could happen during socket init, thus we bail out
4806 * early, since the rest of the below is not setup either.
4808 if (sp->ep == NULL)
4809 return;
4811 if (sp->do_auto_asconf) {
4812 sp->do_auto_asconf = 0;
4813 list_del(&sp->auto_asconf_list);
4815 sctp_endpoint_free(sp->ep);
4816 local_bh_disable();
4817 sk_sockets_allocated_dec(sk);
4818 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4819 local_bh_enable();
4822 /* Triggered when there are no references on the socket anymore */
4823 static void sctp_destruct_sock(struct sock *sk)
4825 struct sctp_sock *sp = sctp_sk(sk);
4827 /* Free up the HMAC transform. */
4828 crypto_free_shash(sp->hmac);
4830 inet_sock_destruct(sk);
4833 /* API 4.1.7 shutdown() - TCP Style Syntax
4834 * int shutdown(int socket, int how);
4836 * sd - the socket descriptor of the association to be closed.
4837 * how - Specifies the type of shutdown. The values are
4838 * as follows:
4839 * SHUT_RD
4840 * Disables further receive operations. No SCTP
4841 * protocol action is taken.
4842 * SHUT_WR
4843 * Disables further send operations, and initiates
4844 * the SCTP shutdown sequence.
4845 * SHUT_RDWR
4846 * Disables further send and receive operations
4847 * and initiates the SCTP shutdown sequence.
4849 static void sctp_shutdown(struct sock *sk, int how)
4851 struct net *net = sock_net(sk);
4852 struct sctp_endpoint *ep;
4854 if (!sctp_style(sk, TCP))
4855 return;
4857 ep = sctp_sk(sk)->ep;
4858 if (how & SEND_SHUTDOWN && !list_empty(&ep->asocs)) {
4859 struct sctp_association *asoc;
4861 inet_sk_set_state(sk, SCTP_SS_CLOSING);
4862 asoc = list_entry(ep->asocs.next,
4863 struct sctp_association, asocs);
4864 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4868 int sctp_get_sctp_info(struct sock *sk, struct sctp_association *asoc,
4869 struct sctp_info *info)
4871 struct sctp_transport *prim;
4872 struct list_head *pos;
4873 int mask;
4875 memset(info, 0, sizeof(*info));
4876 if (!asoc) {
4877 struct sctp_sock *sp = sctp_sk(sk);
4879 info->sctpi_s_autoclose = sp->autoclose;
4880 info->sctpi_s_adaptation_ind = sp->adaptation_ind;
4881 info->sctpi_s_pd_point = sp->pd_point;
4882 info->sctpi_s_nodelay = sp->nodelay;
4883 info->sctpi_s_disable_fragments = sp->disable_fragments;
4884 info->sctpi_s_v4mapped = sp->v4mapped;
4885 info->sctpi_s_frag_interleave = sp->frag_interleave;
4886 info->sctpi_s_type = sp->type;
4888 return 0;
4891 info->sctpi_tag = asoc->c.my_vtag;
4892 info->sctpi_state = asoc->state;
4893 info->sctpi_rwnd = asoc->a_rwnd;
4894 info->sctpi_unackdata = asoc->unack_data;
4895 info->sctpi_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4896 info->sctpi_instrms = asoc->stream.incnt;
4897 info->sctpi_outstrms = asoc->stream.outcnt;
4898 list_for_each(pos, &asoc->base.inqueue.in_chunk_list)
4899 info->sctpi_inqueue++;
4900 list_for_each(pos, &asoc->outqueue.out_chunk_list)
4901 info->sctpi_outqueue++;
4902 info->sctpi_overall_error = asoc->overall_error_count;
4903 info->sctpi_max_burst = asoc->max_burst;
4904 info->sctpi_maxseg = asoc->frag_point;
4905 info->sctpi_peer_rwnd = asoc->peer.rwnd;
4906 info->sctpi_peer_tag = asoc->c.peer_vtag;
4908 mask = asoc->peer.ecn_capable << 1;
4909 mask = (mask | asoc->peer.ipv4_address) << 1;
4910 mask = (mask | asoc->peer.ipv6_address) << 1;
4911 mask = (mask | asoc->peer.hostname_address) << 1;
4912 mask = (mask | asoc->peer.asconf_capable) << 1;
4913 mask = (mask | asoc->peer.prsctp_capable) << 1;
4914 mask = (mask | asoc->peer.auth_capable);
4915 info->sctpi_peer_capable = mask;
4916 mask = asoc->peer.sack_needed << 1;
4917 mask = (mask | asoc->peer.sack_generation) << 1;
4918 mask = (mask | asoc->peer.zero_window_announced);
4919 info->sctpi_peer_sack = mask;
4921 info->sctpi_isacks = asoc->stats.isacks;
4922 info->sctpi_osacks = asoc->stats.osacks;
4923 info->sctpi_opackets = asoc->stats.opackets;
4924 info->sctpi_ipackets = asoc->stats.ipackets;
4925 info->sctpi_rtxchunks = asoc->stats.rtxchunks;
4926 info->sctpi_outofseqtsns = asoc->stats.outofseqtsns;
4927 info->sctpi_idupchunks = asoc->stats.idupchunks;
4928 info->sctpi_gapcnt = asoc->stats.gapcnt;
4929 info->sctpi_ouodchunks = asoc->stats.ouodchunks;
4930 info->sctpi_iuodchunks = asoc->stats.iuodchunks;
4931 info->sctpi_oodchunks = asoc->stats.oodchunks;
4932 info->sctpi_iodchunks = asoc->stats.iodchunks;
4933 info->sctpi_octrlchunks = asoc->stats.octrlchunks;
4934 info->sctpi_ictrlchunks = asoc->stats.ictrlchunks;
4936 prim = asoc->peer.primary_path;
4937 memcpy(&info->sctpi_p_address, &prim->ipaddr, sizeof(prim->ipaddr));
4938 info->sctpi_p_state = prim->state;
4939 info->sctpi_p_cwnd = prim->cwnd;
4940 info->sctpi_p_srtt = prim->srtt;
4941 info->sctpi_p_rto = jiffies_to_msecs(prim->rto);
4942 info->sctpi_p_hbinterval = prim->hbinterval;
4943 info->sctpi_p_pathmaxrxt = prim->pathmaxrxt;
4944 info->sctpi_p_sackdelay = jiffies_to_msecs(prim->sackdelay);
4945 info->sctpi_p_ssthresh = prim->ssthresh;
4946 info->sctpi_p_partial_bytes_acked = prim->partial_bytes_acked;
4947 info->sctpi_p_flight_size = prim->flight_size;
4948 info->sctpi_p_error = prim->error_count;
4950 return 0;
4952 EXPORT_SYMBOL_GPL(sctp_get_sctp_info);
4954 /* use callback to avoid exporting the core structure */
4955 void sctp_transport_walk_start(struct rhashtable_iter *iter)
4957 rhltable_walk_enter(&sctp_transport_hashtable, iter);
4959 rhashtable_walk_start(iter);
4962 void sctp_transport_walk_stop(struct rhashtable_iter *iter)
4964 rhashtable_walk_stop(iter);
4965 rhashtable_walk_exit(iter);
4968 struct sctp_transport *sctp_transport_get_next(struct net *net,
4969 struct rhashtable_iter *iter)
4971 struct sctp_transport *t;
4973 t = rhashtable_walk_next(iter);
4974 for (; t; t = rhashtable_walk_next(iter)) {
4975 if (IS_ERR(t)) {
4976 if (PTR_ERR(t) == -EAGAIN)
4977 continue;
4978 break;
4981 if (!sctp_transport_hold(t))
4982 continue;
4984 if (net_eq(sock_net(t->asoc->base.sk), net) &&
4985 t->asoc->peer.primary_path == t)
4986 break;
4988 sctp_transport_put(t);
4991 return t;
4994 struct sctp_transport *sctp_transport_get_idx(struct net *net,
4995 struct rhashtable_iter *iter,
4996 int pos)
4998 struct sctp_transport *t;
5000 if (!pos)
5001 return SEQ_START_TOKEN;
5003 while ((t = sctp_transport_get_next(net, iter)) && !IS_ERR(t)) {
5004 if (!--pos)
5005 break;
5006 sctp_transport_put(t);
5009 return t;
5012 int sctp_for_each_endpoint(int (*cb)(struct sctp_endpoint *, void *),
5013 void *p) {
5014 int err = 0;
5015 int hash = 0;
5016 struct sctp_ep_common *epb;
5017 struct sctp_hashbucket *head;
5019 for (head = sctp_ep_hashtable; hash < sctp_ep_hashsize;
5020 hash++, head++) {
5021 read_lock_bh(&head->lock);
5022 sctp_for_each_hentry(epb, &head->chain) {
5023 err = cb(sctp_ep(epb), p);
5024 if (err)
5025 break;
5027 read_unlock_bh(&head->lock);
5030 return err;
5032 EXPORT_SYMBOL_GPL(sctp_for_each_endpoint);
5034 int sctp_transport_lookup_process(int (*cb)(struct sctp_transport *, void *),
5035 struct net *net,
5036 const union sctp_addr *laddr,
5037 const union sctp_addr *paddr, void *p)
5039 struct sctp_transport *transport;
5040 int err;
5042 rcu_read_lock();
5043 transport = sctp_addrs_lookup_transport(net, laddr, paddr);
5044 rcu_read_unlock();
5045 if (!transport)
5046 return -ENOENT;
5048 err = cb(transport, p);
5049 sctp_transport_put(transport);
5051 return err;
5053 EXPORT_SYMBOL_GPL(sctp_transport_lookup_process);
5055 int sctp_for_each_transport(int (*cb)(struct sctp_transport *, void *),
5056 int (*cb_done)(struct sctp_transport *, void *),
5057 struct net *net, int *pos, void *p) {
5058 struct rhashtable_iter hti;
5059 struct sctp_transport *tsp;
5060 int ret;
5062 again:
5063 ret = 0;
5064 sctp_transport_walk_start(&hti);
5066 tsp = sctp_transport_get_idx(net, &hti, *pos + 1);
5067 for (; !IS_ERR_OR_NULL(tsp); tsp = sctp_transport_get_next(net, &hti)) {
5068 ret = cb(tsp, p);
5069 if (ret)
5070 break;
5071 (*pos)++;
5072 sctp_transport_put(tsp);
5074 sctp_transport_walk_stop(&hti);
5076 if (ret) {
5077 if (cb_done && !cb_done(tsp, p)) {
5078 (*pos)++;
5079 sctp_transport_put(tsp);
5080 goto again;
5082 sctp_transport_put(tsp);
5085 return ret;
5087 EXPORT_SYMBOL_GPL(sctp_for_each_transport);
5089 /* 7.2.1 Association Status (SCTP_STATUS)
5091 * Applications can retrieve current status information about an
5092 * association, including association state, peer receiver window size,
5093 * number of unacked data chunks, and number of data chunks pending
5094 * receipt. This information is read-only.
5096 static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
5097 char __user *optval,
5098 int __user *optlen)
5100 struct sctp_status status;
5101 struct sctp_association *asoc = NULL;
5102 struct sctp_transport *transport;
5103 sctp_assoc_t associd;
5104 int retval = 0;
5106 if (len < sizeof(status)) {
5107 retval = -EINVAL;
5108 goto out;
5111 len = sizeof(status);
5112 if (copy_from_user(&status, optval, len)) {
5113 retval = -EFAULT;
5114 goto out;
5117 associd = status.sstat_assoc_id;
5118 asoc = sctp_id2assoc(sk, associd);
5119 if (!asoc) {
5120 retval = -EINVAL;
5121 goto out;
5124 transport = asoc->peer.primary_path;
5126 status.sstat_assoc_id = sctp_assoc2id(asoc);
5127 status.sstat_state = sctp_assoc_to_state(asoc);
5128 status.sstat_rwnd = asoc->peer.rwnd;
5129 status.sstat_unackdata = asoc->unack_data;
5131 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
5132 status.sstat_instrms = asoc->stream.incnt;
5133 status.sstat_outstrms = asoc->stream.outcnt;
5134 status.sstat_fragmentation_point = asoc->frag_point;
5135 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5136 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
5137 transport->af_specific->sockaddr_len);
5138 /* Map ipv4 address into v4-mapped-on-v6 address. */
5139 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sctp_sk(sk),
5140 (union sctp_addr *)&status.sstat_primary.spinfo_address);
5141 status.sstat_primary.spinfo_state = transport->state;
5142 status.sstat_primary.spinfo_cwnd = transport->cwnd;
5143 status.sstat_primary.spinfo_srtt = transport->srtt;
5144 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
5145 status.sstat_primary.spinfo_mtu = transport->pathmtu;
5147 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
5148 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
5150 if (put_user(len, optlen)) {
5151 retval = -EFAULT;
5152 goto out;
5155 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
5156 __func__, len, status.sstat_state, status.sstat_rwnd,
5157 status.sstat_assoc_id);
5159 if (copy_to_user(optval, &status, len)) {
5160 retval = -EFAULT;
5161 goto out;
5164 out:
5165 return retval;
5169 /* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
5171 * Applications can retrieve information about a specific peer address
5172 * of an association, including its reachability state, congestion
5173 * window, and retransmission timer values. This information is
5174 * read-only.
5176 static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
5177 char __user *optval,
5178 int __user *optlen)
5180 struct sctp_paddrinfo pinfo;
5181 struct sctp_transport *transport;
5182 int retval = 0;
5184 if (len < sizeof(pinfo)) {
5185 retval = -EINVAL;
5186 goto out;
5189 len = sizeof(pinfo);
5190 if (copy_from_user(&pinfo, optval, len)) {
5191 retval = -EFAULT;
5192 goto out;
5195 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
5196 pinfo.spinfo_assoc_id);
5197 if (!transport)
5198 return -EINVAL;
5200 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
5201 pinfo.spinfo_state = transport->state;
5202 pinfo.spinfo_cwnd = transport->cwnd;
5203 pinfo.spinfo_srtt = transport->srtt;
5204 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
5205 pinfo.spinfo_mtu = transport->pathmtu;
5207 if (pinfo.spinfo_state == SCTP_UNKNOWN)
5208 pinfo.spinfo_state = SCTP_ACTIVE;
5210 if (put_user(len, optlen)) {
5211 retval = -EFAULT;
5212 goto out;
5215 if (copy_to_user(optval, &pinfo, len)) {
5216 retval = -EFAULT;
5217 goto out;
5220 out:
5221 return retval;
5224 /* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
5226 * This option is a on/off flag. If enabled no SCTP message
5227 * fragmentation will be performed. Instead if a message being sent
5228 * exceeds the current PMTU size, the message will NOT be sent and
5229 * instead a error will be indicated to the user.
5231 static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
5232 char __user *optval, int __user *optlen)
5234 int val;
5236 if (len < sizeof(int))
5237 return -EINVAL;
5239 len = sizeof(int);
5240 val = (sctp_sk(sk)->disable_fragments == 1);
5241 if (put_user(len, optlen))
5242 return -EFAULT;
5243 if (copy_to_user(optval, &val, len))
5244 return -EFAULT;
5245 return 0;
5248 /* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
5250 * This socket option is used to specify various notifications and
5251 * ancillary data the user wishes to receive.
5253 static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
5254 int __user *optlen)
5256 if (len == 0)
5257 return -EINVAL;
5258 if (len > sizeof(struct sctp_event_subscribe))
5259 len = sizeof(struct sctp_event_subscribe);
5260 if (put_user(len, optlen))
5261 return -EFAULT;
5262 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
5263 return -EFAULT;
5264 return 0;
5267 /* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
5269 * This socket option is applicable to the UDP-style socket only. When
5270 * set it will cause associations that are idle for more than the
5271 * specified number of seconds to automatically close. An association
5272 * being idle is defined an association that has NOT sent or received
5273 * user data. The special value of '0' indicates that no automatic
5274 * close of any associations should be performed. The option expects an
5275 * integer defining the number of seconds of idle time before an
5276 * association is closed.
5278 static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
5280 /* Applicable to UDP-style socket only */
5281 if (sctp_style(sk, TCP))
5282 return -EOPNOTSUPP;
5283 if (len < sizeof(int))
5284 return -EINVAL;
5285 len = sizeof(int);
5286 if (put_user(len, optlen))
5287 return -EFAULT;
5288 if (put_user(sctp_sk(sk)->autoclose, (int __user *)optval))
5289 return -EFAULT;
5290 return 0;
5293 /* Helper routine to branch off an association to a new socket. */
5294 int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
5296 struct sctp_association *asoc = sctp_id2assoc(sk, id);
5297 struct sctp_sock *sp = sctp_sk(sk);
5298 struct socket *sock;
5299 int err = 0;
5301 /* Do not peel off from one netns to another one. */
5302 if (!net_eq(current->nsproxy->net_ns, sock_net(sk)))
5303 return -EINVAL;
5305 if (!asoc)
5306 return -EINVAL;
5308 /* An association cannot be branched off from an already peeled-off
5309 * socket, nor is this supported for tcp style sockets.
5311 if (!sctp_style(sk, UDP))
5312 return -EINVAL;
5314 /* Create a new socket. */
5315 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
5316 if (err < 0)
5317 return err;
5319 sctp_copy_sock(sock->sk, sk, asoc);
5321 /* Make peeled-off sockets more like 1-1 accepted sockets.
5322 * Set the daddr and initialize id to something more random and also
5323 * copy over any ip options.
5325 sp->pf->to_sk_daddr(&asoc->peer.primary_addr, sk);
5326 sp->pf->copy_ip_options(sk, sock->sk);
5328 /* Populate the fields of the newsk from the oldsk and migrate the
5329 * asoc to the newsk.
5331 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
5333 *sockp = sock;
5335 return err;
5337 EXPORT_SYMBOL(sctp_do_peeloff);
5339 static int sctp_getsockopt_peeloff_common(struct sock *sk, sctp_peeloff_arg_t *peeloff,
5340 struct file **newfile, unsigned flags)
5342 struct socket *newsock;
5343 int retval;
5345 retval = sctp_do_peeloff(sk, peeloff->associd, &newsock);
5346 if (retval < 0)
5347 goto out;
5349 /* Map the socket to an unused fd that can be returned to the user. */
5350 retval = get_unused_fd_flags(flags & SOCK_CLOEXEC);
5351 if (retval < 0) {
5352 sock_release(newsock);
5353 goto out;
5356 *newfile = sock_alloc_file(newsock, 0, NULL);
5357 if (IS_ERR(*newfile)) {
5358 put_unused_fd(retval);
5359 retval = PTR_ERR(*newfile);
5360 *newfile = NULL;
5361 return retval;
5364 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
5365 retval);
5367 peeloff->sd = retval;
5369 if (flags & SOCK_NONBLOCK)
5370 (*newfile)->f_flags |= O_NONBLOCK;
5371 out:
5372 return retval;
5375 static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
5377 sctp_peeloff_arg_t peeloff;
5378 struct file *newfile = NULL;
5379 int retval = 0;
5381 if (len < sizeof(sctp_peeloff_arg_t))
5382 return -EINVAL;
5383 len = sizeof(sctp_peeloff_arg_t);
5384 if (copy_from_user(&peeloff, optval, len))
5385 return -EFAULT;
5387 retval = sctp_getsockopt_peeloff_common(sk, &peeloff, &newfile, 0);
5388 if (retval < 0)
5389 goto out;
5391 /* Return the fd mapped to the new socket. */
5392 if (put_user(len, optlen)) {
5393 fput(newfile);
5394 put_unused_fd(retval);
5395 return -EFAULT;
5398 if (copy_to_user(optval, &peeloff, len)) {
5399 fput(newfile);
5400 put_unused_fd(retval);
5401 return -EFAULT;
5403 fd_install(retval, newfile);
5404 out:
5405 return retval;
5408 static int sctp_getsockopt_peeloff_flags(struct sock *sk, int len,
5409 char __user *optval, int __user *optlen)
5411 sctp_peeloff_flags_arg_t peeloff;
5412 struct file *newfile = NULL;
5413 int retval = 0;
5415 if (len < sizeof(sctp_peeloff_flags_arg_t))
5416 return -EINVAL;
5417 len = sizeof(sctp_peeloff_flags_arg_t);
5418 if (copy_from_user(&peeloff, optval, len))
5419 return -EFAULT;
5421 retval = sctp_getsockopt_peeloff_common(sk, &peeloff.p_arg,
5422 &newfile, peeloff.flags);
5423 if (retval < 0)
5424 goto out;
5426 /* Return the fd mapped to the new socket. */
5427 if (put_user(len, optlen)) {
5428 fput(newfile);
5429 put_unused_fd(retval);
5430 return -EFAULT;
5433 if (copy_to_user(optval, &peeloff, len)) {
5434 fput(newfile);
5435 put_unused_fd(retval);
5436 return -EFAULT;
5438 fd_install(retval, newfile);
5439 out:
5440 return retval;
5443 /* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
5445 * Applications can enable or disable heartbeats for any peer address of
5446 * an association, modify an address's heartbeat interval, force a
5447 * heartbeat to be sent immediately, and adjust the address's maximum
5448 * number of retransmissions sent before an address is considered
5449 * unreachable. The following structure is used to access and modify an
5450 * address's parameters:
5452 * struct sctp_paddrparams {
5453 * sctp_assoc_t spp_assoc_id;
5454 * struct sockaddr_storage spp_address;
5455 * uint32_t spp_hbinterval;
5456 * uint16_t spp_pathmaxrxt;
5457 * uint32_t spp_pathmtu;
5458 * uint32_t spp_sackdelay;
5459 * uint32_t spp_flags;
5460 * };
5462 * spp_assoc_id - (one-to-many style socket) This is filled in the
5463 * application, and identifies the association for
5464 * this query.
5465 * spp_address - This specifies which address is of interest.
5466 * spp_hbinterval - This contains the value of the heartbeat interval,
5467 * in milliseconds. If a value of zero
5468 * is present in this field then no changes are to
5469 * be made to this parameter.
5470 * spp_pathmaxrxt - This contains the maximum number of
5471 * retransmissions before this address shall be
5472 * considered unreachable. If a value of zero
5473 * is present in this field then no changes are to
5474 * be made to this parameter.
5475 * spp_pathmtu - When Path MTU discovery is disabled the value
5476 * specified here will be the "fixed" path mtu.
5477 * Note that if the spp_address field is empty
5478 * then all associations on this address will
5479 * have this fixed path mtu set upon them.
5481 * spp_sackdelay - When delayed sack is enabled, this value specifies
5482 * the number of milliseconds that sacks will be delayed
5483 * for. This value will apply to all addresses of an
5484 * association if the spp_address field is empty. Note
5485 * also, that if delayed sack is enabled and this
5486 * value is set to 0, no change is made to the last
5487 * recorded delayed sack timer value.
5489 * spp_flags - These flags are used to control various features
5490 * on an association. The flag field may contain
5491 * zero or more of the following options.
5493 * SPP_HB_ENABLE - Enable heartbeats on the
5494 * specified address. Note that if the address
5495 * field is empty all addresses for the association
5496 * have heartbeats enabled upon them.
5498 * SPP_HB_DISABLE - Disable heartbeats on the
5499 * speicifed address. Note that if the address
5500 * field is empty all addresses for the association
5501 * will have their heartbeats disabled. Note also
5502 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
5503 * mutually exclusive, only one of these two should
5504 * be specified. Enabling both fields will have
5505 * undetermined results.
5507 * SPP_HB_DEMAND - Request a user initiated heartbeat
5508 * to be made immediately.
5510 * SPP_PMTUD_ENABLE - This field will enable PMTU
5511 * discovery upon the specified address. Note that
5512 * if the address feild is empty then all addresses
5513 * on the association are effected.
5515 * SPP_PMTUD_DISABLE - This field will disable PMTU
5516 * discovery upon the specified address. Note that
5517 * if the address feild is empty then all addresses
5518 * on the association are effected. Not also that
5519 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
5520 * exclusive. Enabling both will have undetermined
5521 * results.
5523 * SPP_SACKDELAY_ENABLE - Setting this flag turns
5524 * on delayed sack. The time specified in spp_sackdelay
5525 * is used to specify the sack delay for this address. Note
5526 * that if spp_address is empty then all addresses will
5527 * enable delayed sack and take on the sack delay
5528 * value specified in spp_sackdelay.
5529 * SPP_SACKDELAY_DISABLE - Setting this flag turns
5530 * off delayed sack. If the spp_address field is blank then
5531 * delayed sack is disabled for the entire association. Note
5532 * also that this field is mutually exclusive to
5533 * SPP_SACKDELAY_ENABLE, setting both will have undefined
5534 * results.
5536 * SPP_IPV6_FLOWLABEL: Setting this flag enables the
5537 * setting of the IPV6 flow label value. The value is
5538 * contained in the spp_ipv6_flowlabel field.
5539 * Upon retrieval, this flag will be set to indicate that
5540 * the spp_ipv6_flowlabel field has a valid value returned.
5541 * If a specific destination address is set (in the
5542 * spp_address field), then the value returned is that of
5543 * the address. If just an association is specified (and
5544 * no address), then the association's default flow label
5545 * is returned. If neither an association nor a destination
5546 * is specified, then the socket's default flow label is
5547 * returned. For non-IPv6 sockets, this flag will be left
5548 * cleared.
5550 * SPP_DSCP: Setting this flag enables the setting of the
5551 * Differentiated Services Code Point (DSCP) value
5552 * associated with either the association or a specific
5553 * address. The value is obtained in the spp_dscp field.
5554 * Upon retrieval, this flag will be set to indicate that
5555 * the spp_dscp field has a valid value returned. If a
5556 * specific destination address is set when called (in the
5557 * spp_address field), then that specific destination
5558 * address's DSCP value is returned. If just an association
5559 * is specified, then the association's default DSCP is
5560 * returned. If neither an association nor a destination is
5561 * specified, then the socket's default DSCP is returned.
5563 * spp_ipv6_flowlabel
5564 * - This field is used in conjunction with the
5565 * SPP_IPV6_FLOWLABEL flag and contains the IPv6 flow label.
5566 * The 20 least significant bits are used for the flow
5567 * label. This setting has precedence over any IPv6-layer
5568 * setting.
5570 * spp_dscp - This field is used in conjunction with the SPP_DSCP flag
5571 * and contains the DSCP. The 6 most significant bits are
5572 * used for the DSCP. This setting has precedence over any
5573 * IPv4- or IPv6- layer setting.
5575 static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
5576 char __user *optval, int __user *optlen)
5578 struct sctp_paddrparams params;
5579 struct sctp_transport *trans = NULL;
5580 struct sctp_association *asoc = NULL;
5581 struct sctp_sock *sp = sctp_sk(sk);
5583 if (len >= sizeof(params))
5584 len = sizeof(params);
5585 else if (len >= ALIGN(offsetof(struct sctp_paddrparams,
5586 spp_ipv6_flowlabel), 4))
5587 len = ALIGN(offsetof(struct sctp_paddrparams,
5588 spp_ipv6_flowlabel), 4);
5589 else
5590 return -EINVAL;
5592 if (copy_from_user(&params, optval, len))
5593 return -EFAULT;
5595 /* If an address other than INADDR_ANY is specified, and
5596 * no transport is found, then the request is invalid.
5598 if (!sctp_is_any(sk, (union sctp_addr *)&params.spp_address)) {
5599 trans = sctp_addr_id2transport(sk, &params.spp_address,
5600 params.spp_assoc_id);
5601 if (!trans) {
5602 pr_debug("%s: failed no transport\n", __func__);
5603 return -EINVAL;
5607 /* Get association, if assoc_id != 0 and the socket is a one
5608 * to many style socket, and an association was not found, then
5609 * the id was invalid.
5611 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
5612 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
5613 pr_debug("%s: failed no association\n", __func__);
5614 return -EINVAL;
5617 if (trans) {
5618 /* Fetch transport values. */
5619 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
5620 params.spp_pathmtu = trans->pathmtu;
5621 params.spp_pathmaxrxt = trans->pathmaxrxt;
5622 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
5624 /*draft-11 doesn't say what to return in spp_flags*/
5625 params.spp_flags = trans->param_flags;
5626 if (trans->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5627 params.spp_ipv6_flowlabel = trans->flowlabel &
5628 SCTP_FLOWLABEL_VAL_MASK;
5629 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5631 if (trans->dscp & SCTP_DSCP_SET_MASK) {
5632 params.spp_dscp = trans->dscp & SCTP_DSCP_VAL_MASK;
5633 params.spp_flags |= SPP_DSCP;
5635 } else if (asoc) {
5636 /* Fetch association values. */
5637 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
5638 params.spp_pathmtu = asoc->pathmtu;
5639 params.spp_pathmaxrxt = asoc->pathmaxrxt;
5640 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
5642 /*draft-11 doesn't say what to return in spp_flags*/
5643 params.spp_flags = asoc->param_flags;
5644 if (asoc->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5645 params.spp_ipv6_flowlabel = asoc->flowlabel &
5646 SCTP_FLOWLABEL_VAL_MASK;
5647 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5649 if (asoc->dscp & SCTP_DSCP_SET_MASK) {
5650 params.spp_dscp = asoc->dscp & SCTP_DSCP_VAL_MASK;
5651 params.spp_flags |= SPP_DSCP;
5653 } else {
5654 /* Fetch socket values. */
5655 params.spp_hbinterval = sp->hbinterval;
5656 params.spp_pathmtu = sp->pathmtu;
5657 params.spp_sackdelay = sp->sackdelay;
5658 params.spp_pathmaxrxt = sp->pathmaxrxt;
5660 /*draft-11 doesn't say what to return in spp_flags*/
5661 params.spp_flags = sp->param_flags;
5662 if (sp->flowlabel & SCTP_FLOWLABEL_SET_MASK) {
5663 params.spp_ipv6_flowlabel = sp->flowlabel &
5664 SCTP_FLOWLABEL_VAL_MASK;
5665 params.spp_flags |= SPP_IPV6_FLOWLABEL;
5667 if (sp->dscp & SCTP_DSCP_SET_MASK) {
5668 params.spp_dscp = sp->dscp & SCTP_DSCP_VAL_MASK;
5669 params.spp_flags |= SPP_DSCP;
5673 if (copy_to_user(optval, &params, len))
5674 return -EFAULT;
5676 if (put_user(len, optlen))
5677 return -EFAULT;
5679 return 0;
5683 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
5685 * This option will effect the way delayed acks are performed. This
5686 * option allows you to get or set the delayed ack time, in
5687 * milliseconds. It also allows changing the delayed ack frequency.
5688 * Changing the frequency to 1 disables the delayed sack algorithm. If
5689 * the assoc_id is 0, then this sets or gets the endpoints default
5690 * values. If the assoc_id field is non-zero, then the set or get
5691 * effects the specified association for the one to many model (the
5692 * assoc_id field is ignored by the one to one model). Note that if
5693 * sack_delay or sack_freq are 0 when setting this option, then the
5694 * current values will remain unchanged.
5696 * struct sctp_sack_info {
5697 * sctp_assoc_t sack_assoc_id;
5698 * uint32_t sack_delay;
5699 * uint32_t sack_freq;
5700 * };
5702 * sack_assoc_id - This parameter, indicates which association the user
5703 * is performing an action upon. Note that if this field's value is
5704 * zero then the endpoints default value is changed (effecting future
5705 * associations only).
5707 * sack_delay - This parameter contains the number of milliseconds that
5708 * the user is requesting the delayed ACK timer be set to. Note that
5709 * this value is defined in the standard to be between 200 and 500
5710 * milliseconds.
5712 * sack_freq - This parameter contains the number of packets that must
5713 * be received before a sack is sent without waiting for the delay
5714 * timer to expire. The default value for this is 2, setting this
5715 * value to 1 will disable the delayed sack algorithm.
5717 static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
5718 char __user *optval,
5719 int __user *optlen)
5721 struct sctp_sack_info params;
5722 struct sctp_association *asoc = NULL;
5723 struct sctp_sock *sp = sctp_sk(sk);
5725 if (len >= sizeof(struct sctp_sack_info)) {
5726 len = sizeof(struct sctp_sack_info);
5728 if (copy_from_user(&params, optval, len))
5729 return -EFAULT;
5730 } else if (len == sizeof(struct sctp_assoc_value)) {
5731 pr_warn_ratelimited(DEPRECATED
5732 "%s (pid %d) "
5733 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
5734 "Use struct sctp_sack_info instead\n",
5735 current->comm, task_pid_nr(current));
5736 if (copy_from_user(&params, optval, len))
5737 return -EFAULT;
5738 } else
5739 return -EINVAL;
5741 /* Get association, if sack_assoc_id != 0 and the socket is a one
5742 * to many style socket, and an association was not found, then
5743 * the id was invalid.
5745 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
5746 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
5747 return -EINVAL;
5749 if (asoc) {
5750 /* Fetch association values. */
5751 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
5752 params.sack_delay = jiffies_to_msecs(
5753 asoc->sackdelay);
5754 params.sack_freq = asoc->sackfreq;
5756 } else {
5757 params.sack_delay = 0;
5758 params.sack_freq = 1;
5760 } else {
5761 /* Fetch socket values. */
5762 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
5763 params.sack_delay = sp->sackdelay;
5764 params.sack_freq = sp->sackfreq;
5765 } else {
5766 params.sack_delay = 0;
5767 params.sack_freq = 1;
5771 if (copy_to_user(optval, &params, len))
5772 return -EFAULT;
5774 if (put_user(len, optlen))
5775 return -EFAULT;
5777 return 0;
5780 /* 7.1.3 Initialization Parameters (SCTP_INITMSG)
5782 * Applications can specify protocol parameters for the default association
5783 * initialization. The option name argument to setsockopt() and getsockopt()
5784 * is SCTP_INITMSG.
5786 * Setting initialization parameters is effective only on an unconnected
5787 * socket (for UDP-style sockets only future associations are effected
5788 * by the change). With TCP-style sockets, this option is inherited by
5789 * sockets derived from a listener socket.
5791 static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
5793 if (len < sizeof(struct sctp_initmsg))
5794 return -EINVAL;
5795 len = sizeof(struct sctp_initmsg);
5796 if (put_user(len, optlen))
5797 return -EFAULT;
5798 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
5799 return -EFAULT;
5800 return 0;
5804 static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
5805 char __user *optval, int __user *optlen)
5807 struct sctp_association *asoc;
5808 int cnt = 0;
5809 struct sctp_getaddrs getaddrs;
5810 struct sctp_transport *from;
5811 void __user *to;
5812 union sctp_addr temp;
5813 struct sctp_sock *sp = sctp_sk(sk);
5814 int addrlen;
5815 size_t space_left;
5816 int bytes_copied;
5818 if (len < sizeof(struct sctp_getaddrs))
5819 return -EINVAL;
5821 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5822 return -EFAULT;
5824 /* For UDP-style sockets, id specifies the association to query. */
5825 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5826 if (!asoc)
5827 return -EINVAL;
5829 to = optval + offsetof(struct sctp_getaddrs, addrs);
5830 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5832 list_for_each_entry(from, &asoc->peer.transport_addr_list,
5833 transports) {
5834 memcpy(&temp, &from->ipaddr, sizeof(temp));
5835 addrlen = sctp_get_pf_specific(sk->sk_family)
5836 ->addr_to_user(sp, &temp);
5837 if (space_left < addrlen)
5838 return -ENOMEM;
5839 if (copy_to_user(to, &temp, addrlen))
5840 return -EFAULT;
5841 to += addrlen;
5842 cnt++;
5843 space_left -= addrlen;
5846 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
5847 return -EFAULT;
5848 bytes_copied = ((char __user *)to) - optval;
5849 if (put_user(bytes_copied, optlen))
5850 return -EFAULT;
5852 return 0;
5855 static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
5856 size_t space_left, int *bytes_copied)
5858 struct sctp_sockaddr_entry *addr;
5859 union sctp_addr temp;
5860 int cnt = 0;
5861 int addrlen;
5862 struct net *net = sock_net(sk);
5864 rcu_read_lock();
5865 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
5866 if (!addr->valid)
5867 continue;
5869 if ((PF_INET == sk->sk_family) &&
5870 (AF_INET6 == addr->a.sa.sa_family))
5871 continue;
5872 if ((PF_INET6 == sk->sk_family) &&
5873 inet_v6_ipv6only(sk) &&
5874 (AF_INET == addr->a.sa.sa_family))
5875 continue;
5876 memcpy(&temp, &addr->a, sizeof(temp));
5877 if (!temp.v4.sin_port)
5878 temp.v4.sin_port = htons(port);
5880 addrlen = sctp_get_pf_specific(sk->sk_family)
5881 ->addr_to_user(sctp_sk(sk), &temp);
5883 if (space_left < addrlen) {
5884 cnt = -ENOMEM;
5885 break;
5887 memcpy(to, &temp, addrlen);
5889 to += addrlen;
5890 cnt++;
5891 space_left -= addrlen;
5892 *bytes_copied += addrlen;
5894 rcu_read_unlock();
5896 return cnt;
5900 static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
5901 char __user *optval, int __user *optlen)
5903 struct sctp_bind_addr *bp;
5904 struct sctp_association *asoc;
5905 int cnt = 0;
5906 struct sctp_getaddrs getaddrs;
5907 struct sctp_sockaddr_entry *addr;
5908 void __user *to;
5909 union sctp_addr temp;
5910 struct sctp_sock *sp = sctp_sk(sk);
5911 int addrlen;
5912 int err = 0;
5913 size_t space_left;
5914 int bytes_copied = 0;
5915 void *addrs;
5916 void *buf;
5918 if (len < sizeof(struct sctp_getaddrs))
5919 return -EINVAL;
5921 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
5922 return -EFAULT;
5925 * For UDP-style sockets, id specifies the association to query.
5926 * If the id field is set to the value '0' then the locally bound
5927 * addresses are returned without regard to any particular
5928 * association.
5930 if (0 == getaddrs.assoc_id) {
5931 bp = &sctp_sk(sk)->ep->base.bind_addr;
5932 } else {
5933 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
5934 if (!asoc)
5935 return -EINVAL;
5936 bp = &asoc->base.bind_addr;
5939 to = optval + offsetof(struct sctp_getaddrs, addrs);
5940 space_left = len - offsetof(struct sctp_getaddrs, addrs);
5942 addrs = kmalloc(space_left, GFP_USER | __GFP_NOWARN);
5943 if (!addrs)
5944 return -ENOMEM;
5946 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
5947 * addresses from the global local address list.
5949 if (sctp_list_single_entry(&bp->address_list)) {
5950 addr = list_entry(bp->address_list.next,
5951 struct sctp_sockaddr_entry, list);
5952 if (sctp_is_any(sk, &addr->a)) {
5953 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
5954 space_left, &bytes_copied);
5955 if (cnt < 0) {
5956 err = cnt;
5957 goto out;
5959 goto copy_getaddrs;
5963 buf = addrs;
5964 /* Protection on the bound address list is not needed since
5965 * in the socket option context we hold a socket lock and
5966 * thus the bound address list can't change.
5968 list_for_each_entry(addr, &bp->address_list, list) {
5969 memcpy(&temp, &addr->a, sizeof(temp));
5970 addrlen = sctp_get_pf_specific(sk->sk_family)
5971 ->addr_to_user(sp, &temp);
5972 if (space_left < addrlen) {
5973 err = -ENOMEM; /*fixme: right error?*/
5974 goto out;
5976 memcpy(buf, &temp, addrlen);
5977 buf += addrlen;
5978 bytes_copied += addrlen;
5979 cnt++;
5980 space_left -= addrlen;
5983 copy_getaddrs:
5984 if (copy_to_user(to, addrs, bytes_copied)) {
5985 err = -EFAULT;
5986 goto out;
5988 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
5989 err = -EFAULT;
5990 goto out;
5992 /* XXX: We should have accounted for sizeof(struct sctp_getaddrs) too,
5993 * but we can't change it anymore.
5995 if (put_user(bytes_copied, optlen))
5996 err = -EFAULT;
5997 out:
5998 kfree(addrs);
5999 return err;
6002 /* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
6004 * Requests that the local SCTP stack use the enclosed peer address as
6005 * the association primary. The enclosed address must be one of the
6006 * association peer's addresses.
6008 static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
6009 char __user *optval, int __user *optlen)
6011 struct sctp_prim prim;
6012 struct sctp_association *asoc;
6013 struct sctp_sock *sp = sctp_sk(sk);
6015 if (len < sizeof(struct sctp_prim))
6016 return -EINVAL;
6018 len = sizeof(struct sctp_prim);
6020 if (copy_from_user(&prim, optval, len))
6021 return -EFAULT;
6023 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
6024 if (!asoc)
6025 return -EINVAL;
6027 if (!asoc->peer.primary_path)
6028 return -ENOTCONN;
6030 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
6031 asoc->peer.primary_path->af_specific->sockaddr_len);
6033 sctp_get_pf_specific(sk->sk_family)->addr_to_user(sp,
6034 (union sctp_addr *)&prim.ssp_addr);
6036 if (put_user(len, optlen))
6037 return -EFAULT;
6038 if (copy_to_user(optval, &prim, len))
6039 return -EFAULT;
6041 return 0;
6045 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
6047 * Requests that the local endpoint set the specified Adaptation Layer
6048 * Indication parameter for all future INIT and INIT-ACK exchanges.
6050 static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
6051 char __user *optval, int __user *optlen)
6053 struct sctp_setadaptation adaptation;
6055 if (len < sizeof(struct sctp_setadaptation))
6056 return -EINVAL;
6058 len = sizeof(struct sctp_setadaptation);
6060 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
6062 if (put_user(len, optlen))
6063 return -EFAULT;
6064 if (copy_to_user(optval, &adaptation, len))
6065 return -EFAULT;
6067 return 0;
6072 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
6074 * Applications that wish to use the sendto() system call may wish to
6075 * specify a default set of parameters that would normally be supplied
6076 * through the inclusion of ancillary data. This socket option allows
6077 * such an application to set the default sctp_sndrcvinfo structure.
6080 * The application that wishes to use this socket option simply passes
6081 * in to this call the sctp_sndrcvinfo structure defined in Section
6082 * 5.2.2) The input parameters accepted by this call include
6083 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
6084 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
6085 * to this call if the caller is using the UDP model.
6087 * For getsockopt, it get the default sctp_sndrcvinfo structure.
6089 static int sctp_getsockopt_default_send_param(struct sock *sk,
6090 int len, char __user *optval,
6091 int __user *optlen)
6093 struct sctp_sock *sp = sctp_sk(sk);
6094 struct sctp_association *asoc;
6095 struct sctp_sndrcvinfo info;
6097 if (len < sizeof(info))
6098 return -EINVAL;
6100 len = sizeof(info);
6102 if (copy_from_user(&info, optval, len))
6103 return -EFAULT;
6105 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
6106 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
6107 return -EINVAL;
6108 if (asoc) {
6109 info.sinfo_stream = asoc->default_stream;
6110 info.sinfo_flags = asoc->default_flags;
6111 info.sinfo_ppid = asoc->default_ppid;
6112 info.sinfo_context = asoc->default_context;
6113 info.sinfo_timetolive = asoc->default_timetolive;
6114 } else {
6115 info.sinfo_stream = sp->default_stream;
6116 info.sinfo_flags = sp->default_flags;
6117 info.sinfo_ppid = sp->default_ppid;
6118 info.sinfo_context = sp->default_context;
6119 info.sinfo_timetolive = sp->default_timetolive;
6122 if (put_user(len, optlen))
6123 return -EFAULT;
6124 if (copy_to_user(optval, &info, len))
6125 return -EFAULT;
6127 return 0;
6130 /* RFC6458, Section 8.1.31. Set/get Default Send Parameters
6131 * (SCTP_DEFAULT_SNDINFO)
6133 static int sctp_getsockopt_default_sndinfo(struct sock *sk, int len,
6134 char __user *optval,
6135 int __user *optlen)
6137 struct sctp_sock *sp = sctp_sk(sk);
6138 struct sctp_association *asoc;
6139 struct sctp_sndinfo info;
6141 if (len < sizeof(info))
6142 return -EINVAL;
6144 len = sizeof(info);
6146 if (copy_from_user(&info, optval, len))
6147 return -EFAULT;
6149 asoc = sctp_id2assoc(sk, info.snd_assoc_id);
6150 if (!asoc && info.snd_assoc_id && sctp_style(sk, UDP))
6151 return -EINVAL;
6152 if (asoc) {
6153 info.snd_sid = asoc->default_stream;
6154 info.snd_flags = asoc->default_flags;
6155 info.snd_ppid = asoc->default_ppid;
6156 info.snd_context = asoc->default_context;
6157 } else {
6158 info.snd_sid = sp->default_stream;
6159 info.snd_flags = sp->default_flags;
6160 info.snd_ppid = sp->default_ppid;
6161 info.snd_context = sp->default_context;
6164 if (put_user(len, optlen))
6165 return -EFAULT;
6166 if (copy_to_user(optval, &info, len))
6167 return -EFAULT;
6169 return 0;
6174 * 7.1.5 SCTP_NODELAY
6176 * Turn on/off any Nagle-like algorithm. This means that packets are
6177 * generally sent as soon as possible and no unnecessary delays are
6178 * introduced, at the cost of more packets in the network. Expects an
6179 * integer boolean flag.
6182 static int sctp_getsockopt_nodelay(struct sock *sk, int len,
6183 char __user *optval, int __user *optlen)
6185 int val;
6187 if (len < sizeof(int))
6188 return -EINVAL;
6190 len = sizeof(int);
6191 val = (sctp_sk(sk)->nodelay == 1);
6192 if (put_user(len, optlen))
6193 return -EFAULT;
6194 if (copy_to_user(optval, &val, len))
6195 return -EFAULT;
6196 return 0;
6201 * 7.1.1 SCTP_RTOINFO
6203 * The protocol parameters used to initialize and bound retransmission
6204 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
6205 * and modify these parameters.
6206 * All parameters are time values, in milliseconds. A value of 0, when
6207 * modifying the parameters, indicates that the current value should not
6208 * be changed.
6211 static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
6212 char __user *optval,
6213 int __user *optlen) {
6214 struct sctp_rtoinfo rtoinfo;
6215 struct sctp_association *asoc;
6217 if (len < sizeof (struct sctp_rtoinfo))
6218 return -EINVAL;
6220 len = sizeof(struct sctp_rtoinfo);
6222 if (copy_from_user(&rtoinfo, optval, len))
6223 return -EFAULT;
6225 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
6227 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
6228 return -EINVAL;
6230 /* Values corresponding to the specific association. */
6231 if (asoc) {
6232 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
6233 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
6234 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
6235 } else {
6236 /* Values corresponding to the endpoint. */
6237 struct sctp_sock *sp = sctp_sk(sk);
6239 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
6240 rtoinfo.srto_max = sp->rtoinfo.srto_max;
6241 rtoinfo.srto_min = sp->rtoinfo.srto_min;
6244 if (put_user(len, optlen))
6245 return -EFAULT;
6247 if (copy_to_user(optval, &rtoinfo, len))
6248 return -EFAULT;
6250 return 0;
6255 * 7.1.2 SCTP_ASSOCINFO
6257 * This option is used to tune the maximum retransmission attempts
6258 * of the association.
6259 * Returns an error if the new association retransmission value is
6260 * greater than the sum of the retransmission value of the peer.
6261 * See [SCTP] for more information.
6264 static int sctp_getsockopt_associnfo(struct sock *sk, int len,
6265 char __user *optval,
6266 int __user *optlen)
6269 struct sctp_assocparams assocparams;
6270 struct sctp_association *asoc;
6271 struct list_head *pos;
6272 int cnt = 0;
6274 if (len < sizeof (struct sctp_assocparams))
6275 return -EINVAL;
6277 len = sizeof(struct sctp_assocparams);
6279 if (copy_from_user(&assocparams, optval, len))
6280 return -EFAULT;
6282 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
6284 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
6285 return -EINVAL;
6287 /* Values correspoinding to the specific association */
6288 if (asoc) {
6289 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
6290 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
6291 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
6292 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
6294 list_for_each(pos, &asoc->peer.transport_addr_list) {
6295 cnt++;
6298 assocparams.sasoc_number_peer_destinations = cnt;
6299 } else {
6300 /* Values corresponding to the endpoint */
6301 struct sctp_sock *sp = sctp_sk(sk);
6303 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
6304 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
6305 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
6306 assocparams.sasoc_cookie_life =
6307 sp->assocparams.sasoc_cookie_life;
6308 assocparams.sasoc_number_peer_destinations =
6309 sp->assocparams.
6310 sasoc_number_peer_destinations;
6313 if (put_user(len, optlen))
6314 return -EFAULT;
6316 if (copy_to_user(optval, &assocparams, len))
6317 return -EFAULT;
6319 return 0;
6323 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
6325 * This socket option is a boolean flag which turns on or off mapped V4
6326 * addresses. If this option is turned on and the socket is type
6327 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
6328 * If this option is turned off, then no mapping will be done of V4
6329 * addresses and a user will receive both PF_INET6 and PF_INET type
6330 * addresses on the socket.
6332 static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
6333 char __user *optval, int __user *optlen)
6335 int val;
6336 struct sctp_sock *sp = sctp_sk(sk);
6338 if (len < sizeof(int))
6339 return -EINVAL;
6341 len = sizeof(int);
6342 val = sp->v4mapped;
6343 if (put_user(len, optlen))
6344 return -EFAULT;
6345 if (copy_to_user(optval, &val, len))
6346 return -EFAULT;
6348 return 0;
6352 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
6353 * (chapter and verse is quoted at sctp_setsockopt_context())
6355 static int sctp_getsockopt_context(struct sock *sk, int len,
6356 char __user *optval, int __user *optlen)
6358 struct sctp_assoc_value params;
6359 struct sctp_sock *sp;
6360 struct sctp_association *asoc;
6362 if (len < sizeof(struct sctp_assoc_value))
6363 return -EINVAL;
6365 len = sizeof(struct sctp_assoc_value);
6367 if (copy_from_user(&params, optval, len))
6368 return -EFAULT;
6370 sp = sctp_sk(sk);
6372 if (params.assoc_id != 0) {
6373 asoc = sctp_id2assoc(sk, params.assoc_id);
6374 if (!asoc)
6375 return -EINVAL;
6376 params.assoc_value = asoc->default_rcv_context;
6377 } else {
6378 params.assoc_value = sp->default_rcv_context;
6381 if (put_user(len, optlen))
6382 return -EFAULT;
6383 if (copy_to_user(optval, &params, len))
6384 return -EFAULT;
6386 return 0;
6390 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
6391 * This option will get or set the maximum size to put in any outgoing
6392 * SCTP DATA chunk. If a message is larger than this size it will be
6393 * fragmented by SCTP into the specified size. Note that the underlying
6394 * SCTP implementation may fragment into smaller sized chunks when the
6395 * PMTU of the underlying association is smaller than the value set by
6396 * the user. The default value for this option is '0' which indicates
6397 * the user is NOT limiting fragmentation and only the PMTU will effect
6398 * SCTP's choice of DATA chunk size. Note also that values set larger
6399 * than the maximum size of an IP datagram will effectively let SCTP
6400 * control fragmentation (i.e. the same as setting this option to 0).
6402 * The following structure is used to access and modify this parameter:
6404 * struct sctp_assoc_value {
6405 * sctp_assoc_t assoc_id;
6406 * uint32_t assoc_value;
6407 * };
6409 * assoc_id: This parameter is ignored for one-to-one style sockets.
6410 * For one-to-many style sockets this parameter indicates which
6411 * association the user is performing an action upon. Note that if
6412 * this field's value is zero then the endpoints default value is
6413 * changed (effecting future associations only).
6414 * assoc_value: This parameter specifies the maximum size in bytes.
6416 static int sctp_getsockopt_maxseg(struct sock *sk, int len,
6417 char __user *optval, int __user *optlen)
6419 struct sctp_assoc_value params;
6420 struct sctp_association *asoc;
6422 if (len == sizeof(int)) {
6423 pr_warn_ratelimited(DEPRECATED
6424 "%s (pid %d) "
6425 "Use of int in maxseg socket option.\n"
6426 "Use struct sctp_assoc_value instead\n",
6427 current->comm, task_pid_nr(current));
6428 params.assoc_id = 0;
6429 } else if (len >= sizeof(struct sctp_assoc_value)) {
6430 len = sizeof(struct sctp_assoc_value);
6431 if (copy_from_user(&params, optval, len))
6432 return -EFAULT;
6433 } else
6434 return -EINVAL;
6436 asoc = sctp_id2assoc(sk, params.assoc_id);
6437 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
6438 return -EINVAL;
6440 if (asoc)
6441 params.assoc_value = asoc->frag_point;
6442 else
6443 params.assoc_value = sctp_sk(sk)->user_frag;
6445 if (put_user(len, optlen))
6446 return -EFAULT;
6447 if (len == sizeof(int)) {
6448 if (copy_to_user(optval, &params.assoc_value, len))
6449 return -EFAULT;
6450 } else {
6451 if (copy_to_user(optval, &params, len))
6452 return -EFAULT;
6455 return 0;
6459 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
6460 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
6462 static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
6463 char __user *optval, int __user *optlen)
6465 int val;
6467 if (len < sizeof(int))
6468 return -EINVAL;
6470 len = sizeof(int);
6472 val = sctp_sk(sk)->frag_interleave;
6473 if (put_user(len, optlen))
6474 return -EFAULT;
6475 if (copy_to_user(optval, &val, len))
6476 return -EFAULT;
6478 return 0;
6482 * 7.1.25. Set or Get the sctp partial delivery point
6483 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
6485 static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
6486 char __user *optval,
6487 int __user *optlen)
6489 u32 val;
6491 if (len < sizeof(u32))
6492 return -EINVAL;
6494 len = sizeof(u32);
6496 val = sctp_sk(sk)->pd_point;
6497 if (put_user(len, optlen))
6498 return -EFAULT;
6499 if (copy_to_user(optval, &val, len))
6500 return -EFAULT;
6502 return 0;
6506 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
6507 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
6509 static int sctp_getsockopt_maxburst(struct sock *sk, int len,
6510 char __user *optval,
6511 int __user *optlen)
6513 struct sctp_assoc_value params;
6514 struct sctp_sock *sp;
6515 struct sctp_association *asoc;
6517 if (len == sizeof(int)) {
6518 pr_warn_ratelimited(DEPRECATED
6519 "%s (pid %d) "
6520 "Use of int in max_burst socket option.\n"
6521 "Use struct sctp_assoc_value instead\n",
6522 current->comm, task_pid_nr(current));
6523 params.assoc_id = 0;
6524 } else if (len >= sizeof(struct sctp_assoc_value)) {
6525 len = sizeof(struct sctp_assoc_value);
6526 if (copy_from_user(&params, optval, len))
6527 return -EFAULT;
6528 } else
6529 return -EINVAL;
6531 sp = sctp_sk(sk);
6533 if (params.assoc_id != 0) {
6534 asoc = sctp_id2assoc(sk, params.assoc_id);
6535 if (!asoc)
6536 return -EINVAL;
6537 params.assoc_value = asoc->max_burst;
6538 } else
6539 params.assoc_value = sp->max_burst;
6541 if (len == sizeof(int)) {
6542 if (copy_to_user(optval, &params.assoc_value, len))
6543 return -EFAULT;
6544 } else {
6545 if (copy_to_user(optval, &params, len))
6546 return -EFAULT;
6549 return 0;
6553 static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
6554 char __user *optval, int __user *optlen)
6556 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6557 struct sctp_hmacalgo __user *p = (void __user *)optval;
6558 struct sctp_hmac_algo_param *hmacs;
6559 __u16 data_len = 0;
6560 u32 num_idents;
6561 int i;
6563 if (!ep->auth_enable)
6564 return -EACCES;
6566 hmacs = ep->auth_hmacs_list;
6567 data_len = ntohs(hmacs->param_hdr.length) -
6568 sizeof(struct sctp_paramhdr);
6570 if (len < sizeof(struct sctp_hmacalgo) + data_len)
6571 return -EINVAL;
6573 len = sizeof(struct sctp_hmacalgo) + data_len;
6574 num_idents = data_len / sizeof(u16);
6576 if (put_user(len, optlen))
6577 return -EFAULT;
6578 if (put_user(num_idents, &p->shmac_num_idents))
6579 return -EFAULT;
6580 for (i = 0; i < num_idents; i++) {
6581 __u16 hmacid = ntohs(hmacs->hmac_ids[i]);
6583 if (copy_to_user(&p->shmac_idents[i], &hmacid, sizeof(__u16)))
6584 return -EFAULT;
6586 return 0;
6589 static int sctp_getsockopt_active_key(struct sock *sk, int len,
6590 char __user *optval, int __user *optlen)
6592 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6593 struct sctp_authkeyid val;
6594 struct sctp_association *asoc;
6596 if (!ep->auth_enable)
6597 return -EACCES;
6599 if (len < sizeof(struct sctp_authkeyid))
6600 return -EINVAL;
6602 len = sizeof(struct sctp_authkeyid);
6603 if (copy_from_user(&val, optval, len))
6604 return -EFAULT;
6606 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
6607 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
6608 return -EINVAL;
6610 if (asoc)
6611 val.scact_keynumber = asoc->active_key_id;
6612 else
6613 val.scact_keynumber = ep->active_key_id;
6615 if (put_user(len, optlen))
6616 return -EFAULT;
6617 if (copy_to_user(optval, &val, len))
6618 return -EFAULT;
6620 return 0;
6623 static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
6624 char __user *optval, int __user *optlen)
6626 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6627 struct sctp_authchunks __user *p = (void __user *)optval;
6628 struct sctp_authchunks val;
6629 struct sctp_association *asoc;
6630 struct sctp_chunks_param *ch;
6631 u32 num_chunks = 0;
6632 char __user *to;
6634 if (!ep->auth_enable)
6635 return -EACCES;
6637 if (len < sizeof(struct sctp_authchunks))
6638 return -EINVAL;
6640 if (copy_from_user(&val, optval, sizeof(val)))
6641 return -EFAULT;
6643 to = p->gauth_chunks;
6644 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6645 if (!asoc)
6646 return -EINVAL;
6648 ch = asoc->peer.peer_chunks;
6649 if (!ch)
6650 goto num;
6652 /* See if the user provided enough room for all the data */
6653 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6654 if (len < num_chunks)
6655 return -EINVAL;
6657 if (copy_to_user(to, ch->chunks, num_chunks))
6658 return -EFAULT;
6659 num:
6660 len = sizeof(struct sctp_authchunks) + num_chunks;
6661 if (put_user(len, optlen))
6662 return -EFAULT;
6663 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6664 return -EFAULT;
6665 return 0;
6668 static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
6669 char __user *optval, int __user *optlen)
6671 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6672 struct sctp_authchunks __user *p = (void __user *)optval;
6673 struct sctp_authchunks val;
6674 struct sctp_association *asoc;
6675 struct sctp_chunks_param *ch;
6676 u32 num_chunks = 0;
6677 char __user *to;
6679 if (!ep->auth_enable)
6680 return -EACCES;
6682 if (len < sizeof(struct sctp_authchunks))
6683 return -EINVAL;
6685 if (copy_from_user(&val, optval, sizeof(val)))
6686 return -EFAULT;
6688 to = p->gauth_chunks;
6689 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
6690 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
6691 return -EINVAL;
6693 if (asoc)
6694 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
6695 else
6696 ch = ep->auth_chunk_list;
6698 if (!ch)
6699 goto num;
6701 num_chunks = ntohs(ch->param_hdr.length) - sizeof(struct sctp_paramhdr);
6702 if (len < sizeof(struct sctp_authchunks) + num_chunks)
6703 return -EINVAL;
6705 if (copy_to_user(to, ch->chunks, num_chunks))
6706 return -EFAULT;
6707 num:
6708 len = sizeof(struct sctp_authchunks) + num_chunks;
6709 if (put_user(len, optlen))
6710 return -EFAULT;
6711 if (put_user(num_chunks, &p->gauth_number_of_chunks))
6712 return -EFAULT;
6714 return 0;
6718 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
6719 * This option gets the current number of associations that are attached
6720 * to a one-to-many style socket. The option value is an uint32_t.
6722 static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
6723 char __user *optval, int __user *optlen)
6725 struct sctp_sock *sp = sctp_sk(sk);
6726 struct sctp_association *asoc;
6727 u32 val = 0;
6729 if (sctp_style(sk, TCP))
6730 return -EOPNOTSUPP;
6732 if (len < sizeof(u32))
6733 return -EINVAL;
6735 len = sizeof(u32);
6737 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6738 val++;
6741 if (put_user(len, optlen))
6742 return -EFAULT;
6743 if (copy_to_user(optval, &val, len))
6744 return -EFAULT;
6746 return 0;
6750 * 8.1.23 SCTP_AUTO_ASCONF
6751 * See the corresponding setsockopt entry as description
6753 static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
6754 char __user *optval, int __user *optlen)
6756 int val = 0;
6758 if (len < sizeof(int))
6759 return -EINVAL;
6761 len = sizeof(int);
6762 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
6763 val = 1;
6764 if (put_user(len, optlen))
6765 return -EFAULT;
6766 if (copy_to_user(optval, &val, len))
6767 return -EFAULT;
6768 return 0;
6772 * 8.2.6. Get the Current Identifiers of Associations
6773 * (SCTP_GET_ASSOC_ID_LIST)
6775 * This option gets the current list of SCTP association identifiers of
6776 * the SCTP associations handled by a one-to-many style socket.
6778 static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
6779 char __user *optval, int __user *optlen)
6781 struct sctp_sock *sp = sctp_sk(sk);
6782 struct sctp_association *asoc;
6783 struct sctp_assoc_ids *ids;
6784 u32 num = 0;
6786 if (sctp_style(sk, TCP))
6787 return -EOPNOTSUPP;
6789 if (len < sizeof(struct sctp_assoc_ids))
6790 return -EINVAL;
6792 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6793 num++;
6796 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
6797 return -EINVAL;
6799 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
6801 ids = kmalloc(len, GFP_USER | __GFP_NOWARN);
6802 if (unlikely(!ids))
6803 return -ENOMEM;
6805 ids->gaids_number_of_ids = num;
6806 num = 0;
6807 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
6808 ids->gaids_assoc_id[num++] = asoc->assoc_id;
6811 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
6812 kfree(ids);
6813 return -EFAULT;
6816 kfree(ids);
6817 return 0;
6821 * SCTP_PEER_ADDR_THLDS
6823 * This option allows us to fetch the partially failed threshold for one or all
6824 * transports in an association. See Section 6.1 of:
6825 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
6827 static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
6828 char __user *optval,
6829 int len,
6830 int __user *optlen)
6832 struct sctp_paddrthlds val;
6833 struct sctp_transport *trans;
6834 struct sctp_association *asoc;
6836 if (len < sizeof(struct sctp_paddrthlds))
6837 return -EINVAL;
6838 len = sizeof(struct sctp_paddrthlds);
6839 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
6840 return -EFAULT;
6842 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
6843 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
6844 if (!asoc)
6845 return -ENOENT;
6847 val.spt_pathpfthld = asoc->pf_retrans;
6848 val.spt_pathmaxrxt = asoc->pathmaxrxt;
6849 } else {
6850 trans = sctp_addr_id2transport(sk, &val.spt_address,
6851 val.spt_assoc_id);
6852 if (!trans)
6853 return -ENOENT;
6855 val.spt_pathmaxrxt = trans->pathmaxrxt;
6856 val.spt_pathpfthld = trans->pf_retrans;
6859 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
6860 return -EFAULT;
6862 return 0;
6866 * SCTP_GET_ASSOC_STATS
6868 * This option retrieves local per endpoint statistics. It is modeled
6869 * after OpenSolaris' implementation
6871 static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
6872 char __user *optval,
6873 int __user *optlen)
6875 struct sctp_assoc_stats sas;
6876 struct sctp_association *asoc = NULL;
6878 /* User must provide at least the assoc id */
6879 if (len < sizeof(sctp_assoc_t))
6880 return -EINVAL;
6882 /* Allow the struct to grow and fill in as much as possible */
6883 len = min_t(size_t, len, sizeof(sas));
6885 if (copy_from_user(&sas, optval, len))
6886 return -EFAULT;
6888 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
6889 if (!asoc)
6890 return -EINVAL;
6892 sas.sas_rtxchunks = asoc->stats.rtxchunks;
6893 sas.sas_gapcnt = asoc->stats.gapcnt;
6894 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
6895 sas.sas_osacks = asoc->stats.osacks;
6896 sas.sas_isacks = asoc->stats.isacks;
6897 sas.sas_octrlchunks = asoc->stats.octrlchunks;
6898 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
6899 sas.sas_oodchunks = asoc->stats.oodchunks;
6900 sas.sas_iodchunks = asoc->stats.iodchunks;
6901 sas.sas_ouodchunks = asoc->stats.ouodchunks;
6902 sas.sas_iuodchunks = asoc->stats.iuodchunks;
6903 sas.sas_idupchunks = asoc->stats.idupchunks;
6904 sas.sas_opackets = asoc->stats.opackets;
6905 sas.sas_ipackets = asoc->stats.ipackets;
6907 /* New high max rto observed, will return 0 if not a single
6908 * RTO update took place. obs_rto_ipaddr will be bogus
6909 * in such a case
6911 sas.sas_maxrto = asoc->stats.max_obs_rto;
6912 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
6913 sizeof(struct sockaddr_storage));
6915 /* Mark beginning of a new observation period */
6916 asoc->stats.max_obs_rto = asoc->rto_min;
6918 if (put_user(len, optlen))
6919 return -EFAULT;
6921 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
6923 if (copy_to_user(optval, &sas, len))
6924 return -EFAULT;
6926 return 0;
6929 static int sctp_getsockopt_recvrcvinfo(struct sock *sk, int len,
6930 char __user *optval,
6931 int __user *optlen)
6933 int val = 0;
6935 if (len < sizeof(int))
6936 return -EINVAL;
6938 len = sizeof(int);
6939 if (sctp_sk(sk)->recvrcvinfo)
6940 val = 1;
6941 if (put_user(len, optlen))
6942 return -EFAULT;
6943 if (copy_to_user(optval, &val, len))
6944 return -EFAULT;
6946 return 0;
6949 static int sctp_getsockopt_recvnxtinfo(struct sock *sk, int len,
6950 char __user *optval,
6951 int __user *optlen)
6953 int val = 0;
6955 if (len < sizeof(int))
6956 return -EINVAL;
6958 len = sizeof(int);
6959 if (sctp_sk(sk)->recvnxtinfo)
6960 val = 1;
6961 if (put_user(len, optlen))
6962 return -EFAULT;
6963 if (copy_to_user(optval, &val, len))
6964 return -EFAULT;
6966 return 0;
6969 static int sctp_getsockopt_pr_supported(struct sock *sk, int len,
6970 char __user *optval,
6971 int __user *optlen)
6973 struct sctp_assoc_value params;
6974 struct sctp_association *asoc;
6975 int retval = -EFAULT;
6977 if (len < sizeof(params)) {
6978 retval = -EINVAL;
6979 goto out;
6982 len = sizeof(params);
6983 if (copy_from_user(&params, optval, len))
6984 goto out;
6986 asoc = sctp_id2assoc(sk, params.assoc_id);
6987 if (asoc) {
6988 params.assoc_value = asoc->prsctp_enable;
6989 } else if (!params.assoc_id) {
6990 struct sctp_sock *sp = sctp_sk(sk);
6992 params.assoc_value = sp->ep->prsctp_enable;
6993 } else {
6994 retval = -EINVAL;
6995 goto out;
6998 if (put_user(len, optlen))
6999 goto out;
7001 if (copy_to_user(optval, &params, len))
7002 goto out;
7004 retval = 0;
7006 out:
7007 return retval;
7010 static int sctp_getsockopt_default_prinfo(struct sock *sk, int len,
7011 char __user *optval,
7012 int __user *optlen)
7014 struct sctp_default_prinfo info;
7015 struct sctp_association *asoc;
7016 int retval = -EFAULT;
7018 if (len < sizeof(info)) {
7019 retval = -EINVAL;
7020 goto out;
7023 len = sizeof(info);
7024 if (copy_from_user(&info, optval, len))
7025 goto out;
7027 asoc = sctp_id2assoc(sk, info.pr_assoc_id);
7028 if (asoc) {
7029 info.pr_policy = SCTP_PR_POLICY(asoc->default_flags);
7030 info.pr_value = asoc->default_timetolive;
7031 } else if (!info.pr_assoc_id) {
7032 struct sctp_sock *sp = sctp_sk(sk);
7034 info.pr_policy = SCTP_PR_POLICY(sp->default_flags);
7035 info.pr_value = sp->default_timetolive;
7036 } else {
7037 retval = -EINVAL;
7038 goto out;
7041 if (put_user(len, optlen))
7042 goto out;
7044 if (copy_to_user(optval, &info, len))
7045 goto out;
7047 retval = 0;
7049 out:
7050 return retval;
7053 static int sctp_getsockopt_pr_assocstatus(struct sock *sk, int len,
7054 char __user *optval,
7055 int __user *optlen)
7057 struct sctp_prstatus params;
7058 struct sctp_association *asoc;
7059 int policy;
7060 int retval = -EINVAL;
7062 if (len < sizeof(params))
7063 goto out;
7065 len = sizeof(params);
7066 if (copy_from_user(&params, optval, len)) {
7067 retval = -EFAULT;
7068 goto out;
7071 policy = params.sprstat_policy;
7072 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7073 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7074 goto out;
7076 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7077 if (!asoc)
7078 goto out;
7080 if (policy == SCTP_PR_SCTP_ALL) {
7081 params.sprstat_abandoned_unsent = 0;
7082 params.sprstat_abandoned_sent = 0;
7083 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7084 params.sprstat_abandoned_unsent +=
7085 asoc->abandoned_unsent[policy];
7086 params.sprstat_abandoned_sent +=
7087 asoc->abandoned_sent[policy];
7089 } else {
7090 params.sprstat_abandoned_unsent =
7091 asoc->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7092 params.sprstat_abandoned_sent =
7093 asoc->abandoned_sent[__SCTP_PR_INDEX(policy)];
7096 if (put_user(len, optlen)) {
7097 retval = -EFAULT;
7098 goto out;
7101 if (copy_to_user(optval, &params, len)) {
7102 retval = -EFAULT;
7103 goto out;
7106 retval = 0;
7108 out:
7109 return retval;
7112 static int sctp_getsockopt_pr_streamstatus(struct sock *sk, int len,
7113 char __user *optval,
7114 int __user *optlen)
7116 struct sctp_stream_out_ext *streamoute;
7117 struct sctp_association *asoc;
7118 struct sctp_prstatus params;
7119 int retval = -EINVAL;
7120 int policy;
7122 if (len < sizeof(params))
7123 goto out;
7125 len = sizeof(params);
7126 if (copy_from_user(&params, optval, len)) {
7127 retval = -EFAULT;
7128 goto out;
7131 policy = params.sprstat_policy;
7132 if (!policy || (policy & ~(SCTP_PR_SCTP_MASK | SCTP_PR_SCTP_ALL)) ||
7133 ((policy & SCTP_PR_SCTP_ALL) && (policy & SCTP_PR_SCTP_MASK)))
7134 goto out;
7136 asoc = sctp_id2assoc(sk, params.sprstat_assoc_id);
7137 if (!asoc || params.sprstat_sid >= asoc->stream.outcnt)
7138 goto out;
7140 streamoute = SCTP_SO(&asoc->stream, params.sprstat_sid)->ext;
7141 if (!streamoute) {
7142 /* Not allocated yet, means all stats are 0 */
7143 params.sprstat_abandoned_unsent = 0;
7144 params.sprstat_abandoned_sent = 0;
7145 retval = 0;
7146 goto out;
7149 if (policy == SCTP_PR_SCTP_ALL) {
7150 params.sprstat_abandoned_unsent = 0;
7151 params.sprstat_abandoned_sent = 0;
7152 for (policy = 0; policy <= SCTP_PR_INDEX(MAX); policy++) {
7153 params.sprstat_abandoned_unsent +=
7154 streamoute->abandoned_unsent[policy];
7155 params.sprstat_abandoned_sent +=
7156 streamoute->abandoned_sent[policy];
7158 } else {
7159 params.sprstat_abandoned_unsent =
7160 streamoute->abandoned_unsent[__SCTP_PR_INDEX(policy)];
7161 params.sprstat_abandoned_sent =
7162 streamoute->abandoned_sent[__SCTP_PR_INDEX(policy)];
7165 if (put_user(len, optlen) || copy_to_user(optval, &params, len)) {
7166 retval = -EFAULT;
7167 goto out;
7170 retval = 0;
7172 out:
7173 return retval;
7176 static int sctp_getsockopt_reconfig_supported(struct sock *sk, int len,
7177 char __user *optval,
7178 int __user *optlen)
7180 struct sctp_assoc_value params;
7181 struct sctp_association *asoc;
7182 int retval = -EFAULT;
7184 if (len < sizeof(params)) {
7185 retval = -EINVAL;
7186 goto out;
7189 len = sizeof(params);
7190 if (copy_from_user(&params, optval, len))
7191 goto out;
7193 asoc = sctp_id2assoc(sk, params.assoc_id);
7194 if (asoc) {
7195 params.assoc_value = asoc->reconf_enable;
7196 } else if (!params.assoc_id) {
7197 struct sctp_sock *sp = sctp_sk(sk);
7199 params.assoc_value = sp->ep->reconf_enable;
7200 } else {
7201 retval = -EINVAL;
7202 goto out;
7205 if (put_user(len, optlen))
7206 goto out;
7208 if (copy_to_user(optval, &params, len))
7209 goto out;
7211 retval = 0;
7213 out:
7214 return retval;
7217 static int sctp_getsockopt_enable_strreset(struct sock *sk, int len,
7218 char __user *optval,
7219 int __user *optlen)
7221 struct sctp_assoc_value params;
7222 struct sctp_association *asoc;
7223 int retval = -EFAULT;
7225 if (len < sizeof(params)) {
7226 retval = -EINVAL;
7227 goto out;
7230 len = sizeof(params);
7231 if (copy_from_user(&params, optval, len))
7232 goto out;
7234 asoc = sctp_id2assoc(sk, params.assoc_id);
7235 if (asoc) {
7236 params.assoc_value = asoc->strreset_enable;
7237 } else if (!params.assoc_id) {
7238 struct sctp_sock *sp = sctp_sk(sk);
7240 params.assoc_value = sp->ep->strreset_enable;
7241 } else {
7242 retval = -EINVAL;
7243 goto out;
7246 if (put_user(len, optlen))
7247 goto out;
7249 if (copy_to_user(optval, &params, len))
7250 goto out;
7252 retval = 0;
7254 out:
7255 return retval;
7258 static int sctp_getsockopt_scheduler(struct sock *sk, int len,
7259 char __user *optval,
7260 int __user *optlen)
7262 struct sctp_assoc_value params;
7263 struct sctp_association *asoc;
7264 int retval = -EFAULT;
7266 if (len < sizeof(params)) {
7267 retval = -EINVAL;
7268 goto out;
7271 len = sizeof(params);
7272 if (copy_from_user(&params, optval, len))
7273 goto out;
7275 asoc = sctp_id2assoc(sk, params.assoc_id);
7276 if (!asoc) {
7277 retval = -EINVAL;
7278 goto out;
7281 params.assoc_value = sctp_sched_get_sched(asoc);
7283 if (put_user(len, optlen))
7284 goto out;
7286 if (copy_to_user(optval, &params, len))
7287 goto out;
7289 retval = 0;
7291 out:
7292 return retval;
7295 static int sctp_getsockopt_scheduler_value(struct sock *sk, int len,
7296 char __user *optval,
7297 int __user *optlen)
7299 struct sctp_stream_value params;
7300 struct sctp_association *asoc;
7301 int retval = -EFAULT;
7303 if (len < sizeof(params)) {
7304 retval = -EINVAL;
7305 goto out;
7308 len = sizeof(params);
7309 if (copy_from_user(&params, optval, len))
7310 goto out;
7312 asoc = sctp_id2assoc(sk, params.assoc_id);
7313 if (!asoc) {
7314 retval = -EINVAL;
7315 goto out;
7318 retval = sctp_sched_get_value(asoc, params.stream_id,
7319 &params.stream_value);
7320 if (retval)
7321 goto out;
7323 if (put_user(len, optlen)) {
7324 retval = -EFAULT;
7325 goto out;
7328 if (copy_to_user(optval, &params, len)) {
7329 retval = -EFAULT;
7330 goto out;
7333 out:
7334 return retval;
7337 static int sctp_getsockopt_interleaving_supported(struct sock *sk, int len,
7338 char __user *optval,
7339 int __user *optlen)
7341 struct sctp_assoc_value params;
7342 struct sctp_association *asoc;
7343 int retval = -EFAULT;
7345 if (len < sizeof(params)) {
7346 retval = -EINVAL;
7347 goto out;
7350 len = sizeof(params);
7351 if (copy_from_user(&params, optval, len))
7352 goto out;
7354 asoc = sctp_id2assoc(sk, params.assoc_id);
7355 if (asoc) {
7356 params.assoc_value = asoc->intl_enable;
7357 } else if (!params.assoc_id) {
7358 struct sctp_sock *sp = sctp_sk(sk);
7360 params.assoc_value = sp->strm_interleave;
7361 } else {
7362 retval = -EINVAL;
7363 goto out;
7366 if (put_user(len, optlen))
7367 goto out;
7369 if (copy_to_user(optval, &params, len))
7370 goto out;
7372 retval = 0;
7374 out:
7375 return retval;
7378 static int sctp_getsockopt_reuse_port(struct sock *sk, int len,
7379 char __user *optval,
7380 int __user *optlen)
7382 int val;
7384 if (len < sizeof(int))
7385 return -EINVAL;
7387 len = sizeof(int);
7388 val = sctp_sk(sk)->reuse;
7389 if (put_user(len, optlen))
7390 return -EFAULT;
7392 if (copy_to_user(optval, &val, len))
7393 return -EFAULT;
7395 return 0;
7398 static int sctp_getsockopt(struct sock *sk, int level, int optname,
7399 char __user *optval, int __user *optlen)
7401 int retval = 0;
7402 int len;
7404 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
7406 /* I can hardly begin to describe how wrong this is. This is
7407 * so broken as to be worse than useless. The API draft
7408 * REALLY is NOT helpful here... I am not convinced that the
7409 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
7410 * are at all well-founded.
7412 if (level != SOL_SCTP) {
7413 struct sctp_af *af = sctp_sk(sk)->pf->af;
7415 retval = af->getsockopt(sk, level, optname, optval, optlen);
7416 return retval;
7419 if (get_user(len, optlen))
7420 return -EFAULT;
7422 if (len < 0)
7423 return -EINVAL;
7425 lock_sock(sk);
7427 switch (optname) {
7428 case SCTP_STATUS:
7429 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
7430 break;
7431 case SCTP_DISABLE_FRAGMENTS:
7432 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
7433 optlen);
7434 break;
7435 case SCTP_EVENTS:
7436 retval = sctp_getsockopt_events(sk, len, optval, optlen);
7437 break;
7438 case SCTP_AUTOCLOSE:
7439 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
7440 break;
7441 case SCTP_SOCKOPT_PEELOFF:
7442 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
7443 break;
7444 case SCTP_SOCKOPT_PEELOFF_FLAGS:
7445 retval = sctp_getsockopt_peeloff_flags(sk, len, optval, optlen);
7446 break;
7447 case SCTP_PEER_ADDR_PARAMS:
7448 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
7449 optlen);
7450 break;
7451 case SCTP_DELAYED_SACK:
7452 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
7453 optlen);
7454 break;
7455 case SCTP_INITMSG:
7456 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
7457 break;
7458 case SCTP_GET_PEER_ADDRS:
7459 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
7460 optlen);
7461 break;
7462 case SCTP_GET_LOCAL_ADDRS:
7463 retval = sctp_getsockopt_local_addrs(sk, len, optval,
7464 optlen);
7465 break;
7466 case SCTP_SOCKOPT_CONNECTX3:
7467 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
7468 break;
7469 case SCTP_DEFAULT_SEND_PARAM:
7470 retval = sctp_getsockopt_default_send_param(sk, len,
7471 optval, optlen);
7472 break;
7473 case SCTP_DEFAULT_SNDINFO:
7474 retval = sctp_getsockopt_default_sndinfo(sk, len,
7475 optval, optlen);
7476 break;
7477 case SCTP_PRIMARY_ADDR:
7478 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
7479 break;
7480 case SCTP_NODELAY:
7481 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
7482 break;
7483 case SCTP_RTOINFO:
7484 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
7485 break;
7486 case SCTP_ASSOCINFO:
7487 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
7488 break;
7489 case SCTP_I_WANT_MAPPED_V4_ADDR:
7490 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
7491 break;
7492 case SCTP_MAXSEG:
7493 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
7494 break;
7495 case SCTP_GET_PEER_ADDR_INFO:
7496 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
7497 optlen);
7498 break;
7499 case SCTP_ADAPTATION_LAYER:
7500 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
7501 optlen);
7502 break;
7503 case SCTP_CONTEXT:
7504 retval = sctp_getsockopt_context(sk, len, optval, optlen);
7505 break;
7506 case SCTP_FRAGMENT_INTERLEAVE:
7507 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
7508 optlen);
7509 break;
7510 case SCTP_PARTIAL_DELIVERY_POINT:
7511 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
7512 optlen);
7513 break;
7514 case SCTP_MAX_BURST:
7515 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
7516 break;
7517 case SCTP_AUTH_KEY:
7518 case SCTP_AUTH_CHUNK:
7519 case SCTP_AUTH_DELETE_KEY:
7520 case SCTP_AUTH_DEACTIVATE_KEY:
7521 retval = -EOPNOTSUPP;
7522 break;
7523 case SCTP_HMAC_IDENT:
7524 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
7525 break;
7526 case SCTP_AUTH_ACTIVE_KEY:
7527 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
7528 break;
7529 case SCTP_PEER_AUTH_CHUNKS:
7530 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
7531 optlen);
7532 break;
7533 case SCTP_LOCAL_AUTH_CHUNKS:
7534 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
7535 optlen);
7536 break;
7537 case SCTP_GET_ASSOC_NUMBER:
7538 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
7539 break;
7540 case SCTP_GET_ASSOC_ID_LIST:
7541 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
7542 break;
7543 case SCTP_AUTO_ASCONF:
7544 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
7545 break;
7546 case SCTP_PEER_ADDR_THLDS:
7547 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
7548 break;
7549 case SCTP_GET_ASSOC_STATS:
7550 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
7551 break;
7552 case SCTP_RECVRCVINFO:
7553 retval = sctp_getsockopt_recvrcvinfo(sk, len, optval, optlen);
7554 break;
7555 case SCTP_RECVNXTINFO:
7556 retval = sctp_getsockopt_recvnxtinfo(sk, len, optval, optlen);
7557 break;
7558 case SCTP_PR_SUPPORTED:
7559 retval = sctp_getsockopt_pr_supported(sk, len, optval, optlen);
7560 break;
7561 case SCTP_DEFAULT_PRINFO:
7562 retval = sctp_getsockopt_default_prinfo(sk, len, optval,
7563 optlen);
7564 break;
7565 case SCTP_PR_ASSOC_STATUS:
7566 retval = sctp_getsockopt_pr_assocstatus(sk, len, optval,
7567 optlen);
7568 break;
7569 case SCTP_PR_STREAM_STATUS:
7570 retval = sctp_getsockopt_pr_streamstatus(sk, len, optval,
7571 optlen);
7572 break;
7573 case SCTP_RECONFIG_SUPPORTED:
7574 retval = sctp_getsockopt_reconfig_supported(sk, len, optval,
7575 optlen);
7576 break;
7577 case SCTP_ENABLE_STREAM_RESET:
7578 retval = sctp_getsockopt_enable_strreset(sk, len, optval,
7579 optlen);
7580 break;
7581 case SCTP_STREAM_SCHEDULER:
7582 retval = sctp_getsockopt_scheduler(sk, len, optval,
7583 optlen);
7584 break;
7585 case SCTP_STREAM_SCHEDULER_VALUE:
7586 retval = sctp_getsockopt_scheduler_value(sk, len, optval,
7587 optlen);
7588 break;
7589 case SCTP_INTERLEAVING_SUPPORTED:
7590 retval = sctp_getsockopt_interleaving_supported(sk, len, optval,
7591 optlen);
7592 break;
7593 case SCTP_REUSE_PORT:
7594 retval = sctp_getsockopt_reuse_port(sk, len, optval, optlen);
7595 break;
7596 default:
7597 retval = -ENOPROTOOPT;
7598 break;
7601 release_sock(sk);
7602 return retval;
7605 static int sctp_hash(struct sock *sk)
7607 /* STUB */
7608 return 0;
7611 static void sctp_unhash(struct sock *sk)
7613 /* STUB */
7616 /* Check if port is acceptable. Possibly find first available port.
7618 * The port hash table (contained in the 'global' SCTP protocol storage
7619 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
7620 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
7621 * list (the list number is the port number hashed out, so as you
7622 * would expect from a hash function, all the ports in a given list have
7623 * such a number that hashes out to the same list number; you were
7624 * expecting that, right?); so each list has a set of ports, with a
7625 * link to the socket (struct sock) that uses it, the port number and
7626 * a fastreuse flag (FIXME: NPI ipg).
7628 static struct sctp_bind_bucket *sctp_bucket_create(
7629 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
7631 static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
7633 bool reuse = (sk->sk_reuse || sctp_sk(sk)->reuse);
7634 struct sctp_bind_hashbucket *head; /* hash list */
7635 struct sctp_bind_bucket *pp;
7636 unsigned short snum;
7637 int ret;
7639 snum = ntohs(addr->v4.sin_port);
7641 pr_debug("%s: begins, snum:%d\n", __func__, snum);
7643 local_bh_disable();
7645 if (snum == 0) {
7646 /* Search for an available port. */
7647 int low, high, remaining, index;
7648 unsigned int rover;
7649 struct net *net = sock_net(sk);
7651 inet_get_local_port_range(net, &low, &high);
7652 remaining = (high - low) + 1;
7653 rover = prandom_u32() % remaining + low;
7655 do {
7656 rover++;
7657 if ((rover < low) || (rover > high))
7658 rover = low;
7659 if (inet_is_local_reserved_port(net, rover))
7660 continue;
7661 index = sctp_phashfn(sock_net(sk), rover);
7662 head = &sctp_port_hashtable[index];
7663 spin_lock(&head->lock);
7664 sctp_for_each_hentry(pp, &head->chain)
7665 if ((pp->port == rover) &&
7666 net_eq(sock_net(sk), pp->net))
7667 goto next;
7668 break;
7669 next:
7670 spin_unlock(&head->lock);
7671 } while (--remaining > 0);
7673 /* Exhausted local port range during search? */
7674 ret = 1;
7675 if (remaining <= 0)
7676 goto fail;
7678 /* OK, here is the one we will use. HEAD (the port
7679 * hash table list entry) is non-NULL and we hold it's
7680 * mutex.
7682 snum = rover;
7683 } else {
7684 /* We are given an specific port number; we verify
7685 * that it is not being used. If it is used, we will
7686 * exahust the search in the hash list corresponding
7687 * to the port number (snum) - we detect that with the
7688 * port iterator, pp being NULL.
7690 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
7691 spin_lock(&head->lock);
7692 sctp_for_each_hentry(pp, &head->chain) {
7693 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
7694 goto pp_found;
7697 pp = NULL;
7698 goto pp_not_found;
7699 pp_found:
7700 if (!hlist_empty(&pp->owner)) {
7701 /* We had a port hash table hit - there is an
7702 * available port (pp != NULL) and it is being
7703 * used by other socket (pp->owner not empty); that other
7704 * socket is going to be sk2.
7706 struct sock *sk2;
7708 pr_debug("%s: found a possible match\n", __func__);
7710 if (pp->fastreuse && reuse && sk->sk_state != SCTP_SS_LISTENING)
7711 goto success;
7713 /* Run through the list of sockets bound to the port
7714 * (pp->port) [via the pointers bind_next and
7715 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
7716 * we get the endpoint they describe and run through
7717 * the endpoint's list of IP (v4 or v6) addresses,
7718 * comparing each of the addresses with the address of
7719 * the socket sk. If we find a match, then that means
7720 * that this port/socket (sk) combination are already
7721 * in an endpoint.
7723 sk_for_each_bound(sk2, &pp->owner) {
7724 struct sctp_endpoint *ep2;
7725 ep2 = sctp_sk(sk2)->ep;
7727 if (sk == sk2 ||
7728 (reuse && (sk2->sk_reuse || sctp_sk(sk2)->reuse) &&
7729 sk2->sk_state != SCTP_SS_LISTENING))
7730 continue;
7732 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
7733 sctp_sk(sk2), sctp_sk(sk))) {
7734 ret = (long)sk2;
7735 goto fail_unlock;
7739 pr_debug("%s: found a match\n", __func__);
7741 pp_not_found:
7742 /* If there was a hash table miss, create a new port. */
7743 ret = 1;
7744 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
7745 goto fail_unlock;
7747 /* In either case (hit or miss), make sure fastreuse is 1 only
7748 * if sk->sk_reuse is too (that is, if the caller requested
7749 * SO_REUSEADDR on this socket -sk-).
7751 if (hlist_empty(&pp->owner)) {
7752 if (reuse && sk->sk_state != SCTP_SS_LISTENING)
7753 pp->fastreuse = 1;
7754 else
7755 pp->fastreuse = 0;
7756 } else if (pp->fastreuse &&
7757 (!reuse || sk->sk_state == SCTP_SS_LISTENING))
7758 pp->fastreuse = 0;
7760 /* We are set, so fill up all the data in the hash table
7761 * entry, tie the socket list information with the rest of the
7762 * sockets FIXME: Blurry, NPI (ipg).
7764 success:
7765 if (!sctp_sk(sk)->bind_hash) {
7766 inet_sk(sk)->inet_num = snum;
7767 sk_add_bind_node(sk, &pp->owner);
7768 sctp_sk(sk)->bind_hash = pp;
7770 ret = 0;
7772 fail_unlock:
7773 spin_unlock(&head->lock);
7775 fail:
7776 local_bh_enable();
7777 return ret;
7780 /* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
7781 * port is requested.
7783 static int sctp_get_port(struct sock *sk, unsigned short snum)
7785 union sctp_addr addr;
7786 struct sctp_af *af = sctp_sk(sk)->pf->af;
7788 /* Set up a dummy address struct from the sk. */
7789 af->from_sk(&addr, sk);
7790 addr.v4.sin_port = htons(snum);
7792 /* Note: sk->sk_num gets filled in if ephemeral port request. */
7793 return !!sctp_get_port_local(sk, &addr);
7797 * Move a socket to LISTENING state.
7799 static int sctp_listen_start(struct sock *sk, int backlog)
7801 struct sctp_sock *sp = sctp_sk(sk);
7802 struct sctp_endpoint *ep = sp->ep;
7803 struct crypto_shash *tfm = NULL;
7804 char alg[32];
7806 /* Allocate HMAC for generating cookie. */
7807 if (!sp->hmac && sp->sctp_hmac_alg) {
7808 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
7809 tfm = crypto_alloc_shash(alg, 0, 0);
7810 if (IS_ERR(tfm)) {
7811 net_info_ratelimited("failed to load transform for %s: %ld\n",
7812 sp->sctp_hmac_alg, PTR_ERR(tfm));
7813 return -ENOSYS;
7815 sctp_sk(sk)->hmac = tfm;
7819 * If a bind() or sctp_bindx() is not called prior to a listen()
7820 * call that allows new associations to be accepted, the system
7821 * picks an ephemeral port and will choose an address set equivalent
7822 * to binding with a wildcard address.
7824 * This is not currently spelled out in the SCTP sockets
7825 * extensions draft, but follows the practice as seen in TCP
7826 * sockets.
7829 inet_sk_set_state(sk, SCTP_SS_LISTENING);
7830 if (!ep->base.bind_addr.port) {
7831 if (sctp_autobind(sk))
7832 return -EAGAIN;
7833 } else {
7834 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
7835 inet_sk_set_state(sk, SCTP_SS_CLOSED);
7836 return -EADDRINUSE;
7840 sk->sk_max_ack_backlog = backlog;
7841 sctp_hash_endpoint(ep);
7842 return 0;
7846 * 4.1.3 / 5.1.3 listen()
7848 * By default, new associations are not accepted for UDP style sockets.
7849 * An application uses listen() to mark a socket as being able to
7850 * accept new associations.
7852 * On TCP style sockets, applications use listen() to ready the SCTP
7853 * endpoint for accepting inbound associations.
7855 * On both types of endpoints a backlog of '0' disables listening.
7857 * Move a socket to LISTENING state.
7859 int sctp_inet_listen(struct socket *sock, int backlog)
7861 struct sock *sk = sock->sk;
7862 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
7863 int err = -EINVAL;
7865 if (unlikely(backlog < 0))
7866 return err;
7868 lock_sock(sk);
7870 /* Peeled-off sockets are not allowed to listen(). */
7871 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
7872 goto out;
7874 if (sock->state != SS_UNCONNECTED)
7875 goto out;
7877 if (!sctp_sstate(sk, LISTENING) && !sctp_sstate(sk, CLOSED))
7878 goto out;
7880 /* If backlog is zero, disable listening. */
7881 if (!backlog) {
7882 if (sctp_sstate(sk, CLOSED))
7883 goto out;
7885 err = 0;
7886 sctp_unhash_endpoint(ep);
7887 sk->sk_state = SCTP_SS_CLOSED;
7888 if (sk->sk_reuse || sctp_sk(sk)->reuse)
7889 sctp_sk(sk)->bind_hash->fastreuse = 1;
7890 goto out;
7893 /* If we are already listening, just update the backlog */
7894 if (sctp_sstate(sk, LISTENING))
7895 sk->sk_max_ack_backlog = backlog;
7896 else {
7897 err = sctp_listen_start(sk, backlog);
7898 if (err)
7899 goto out;
7902 err = 0;
7903 out:
7904 release_sock(sk);
7905 return err;
7909 * This function is done by modeling the current datagram_poll() and the
7910 * tcp_poll(). Note that, based on these implementations, we don't
7911 * lock the socket in this function, even though it seems that,
7912 * ideally, locking or some other mechanisms can be used to ensure
7913 * the integrity of the counters (sndbuf and wmem_alloc) used
7914 * in this place. We assume that we don't need locks either until proven
7915 * otherwise.
7917 * Another thing to note is that we include the Async I/O support
7918 * here, again, by modeling the current TCP/UDP code. We don't have
7919 * a good way to test with it yet.
7921 __poll_t sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
7923 struct sock *sk = sock->sk;
7924 struct sctp_sock *sp = sctp_sk(sk);
7925 __poll_t mask;
7927 poll_wait(file, sk_sleep(sk), wait);
7929 sock_rps_record_flow(sk);
7931 /* A TCP-style listening socket becomes readable when the accept queue
7932 * is not empty.
7934 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
7935 return (!list_empty(&sp->ep->asocs)) ?
7936 (EPOLLIN | EPOLLRDNORM) : 0;
7938 mask = 0;
7940 /* Is there any exceptional events? */
7941 if (sk->sk_err || !skb_queue_empty_lockless(&sk->sk_error_queue))
7942 mask |= EPOLLERR |
7943 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? EPOLLPRI : 0);
7944 if (sk->sk_shutdown & RCV_SHUTDOWN)
7945 mask |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
7946 if (sk->sk_shutdown == SHUTDOWN_MASK)
7947 mask |= EPOLLHUP;
7949 /* Is it readable? Reconsider this code with TCP-style support. */
7950 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
7951 mask |= EPOLLIN | EPOLLRDNORM;
7953 /* The association is either gone or not ready. */
7954 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
7955 return mask;
7957 /* Is it writable? */
7958 if (sctp_writeable(sk)) {
7959 mask |= EPOLLOUT | EPOLLWRNORM;
7960 } else {
7961 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
7963 * Since the socket is not locked, the buffer
7964 * might be made available after the writeable check and
7965 * before the bit is set. This could cause a lost I/O
7966 * signal. tcp_poll() has a race breaker for this race
7967 * condition. Based on their implementation, we put
7968 * in the following code to cover it as well.
7970 if (sctp_writeable(sk))
7971 mask |= EPOLLOUT | EPOLLWRNORM;
7973 return mask;
7976 /********************************************************************
7977 * 2nd Level Abstractions
7978 ********************************************************************/
7980 static struct sctp_bind_bucket *sctp_bucket_create(
7981 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
7983 struct sctp_bind_bucket *pp;
7985 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
7986 if (pp) {
7987 SCTP_DBG_OBJCNT_INC(bind_bucket);
7988 pp->port = snum;
7989 pp->fastreuse = 0;
7990 INIT_HLIST_HEAD(&pp->owner);
7991 pp->net = net;
7992 hlist_add_head(&pp->node, &head->chain);
7994 return pp;
7997 /* Caller must hold hashbucket lock for this tb with local BH disabled */
7998 static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
8000 if (pp && hlist_empty(&pp->owner)) {
8001 __hlist_del(&pp->node);
8002 kmem_cache_free(sctp_bucket_cachep, pp);
8003 SCTP_DBG_OBJCNT_DEC(bind_bucket);
8007 /* Release this socket's reference to a local port. */
8008 static inline void __sctp_put_port(struct sock *sk)
8010 struct sctp_bind_hashbucket *head =
8011 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
8012 inet_sk(sk)->inet_num)];
8013 struct sctp_bind_bucket *pp;
8015 spin_lock(&head->lock);
8016 pp = sctp_sk(sk)->bind_hash;
8017 __sk_del_bind_node(sk);
8018 sctp_sk(sk)->bind_hash = NULL;
8019 inet_sk(sk)->inet_num = 0;
8020 sctp_bucket_destroy(pp);
8021 spin_unlock(&head->lock);
8024 void sctp_put_port(struct sock *sk)
8026 local_bh_disable();
8027 __sctp_put_port(sk);
8028 local_bh_enable();
8032 * The system picks an ephemeral port and choose an address set equivalent
8033 * to binding with a wildcard address.
8034 * One of those addresses will be the primary address for the association.
8035 * This automatically enables the multihoming capability of SCTP.
8037 static int sctp_autobind(struct sock *sk)
8039 union sctp_addr autoaddr;
8040 struct sctp_af *af;
8041 __be16 port;
8043 /* Initialize a local sockaddr structure to INADDR_ANY. */
8044 af = sctp_sk(sk)->pf->af;
8046 port = htons(inet_sk(sk)->inet_num);
8047 af->inaddr_any(&autoaddr, port);
8049 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
8052 /* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
8054 * From RFC 2292
8055 * 4.2 The cmsghdr Structure *
8057 * When ancillary data is sent or received, any number of ancillary data
8058 * objects can be specified by the msg_control and msg_controllen members of
8059 * the msghdr structure, because each object is preceded by
8060 * a cmsghdr structure defining the object's length (the cmsg_len member).
8061 * Historically Berkeley-derived implementations have passed only one object
8062 * at a time, but this API allows multiple objects to be
8063 * passed in a single call to sendmsg() or recvmsg(). The following example
8064 * shows two ancillary data objects in a control buffer.
8066 * |<--------------------------- msg_controllen -------------------------->|
8067 * | |
8069 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
8071 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
8072 * | | |
8074 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
8076 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
8077 * | | | | |
8079 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8080 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
8082 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
8084 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
8088 * msg_control
8089 * points here
8091 static int sctp_msghdr_parse(const struct msghdr *msg, struct sctp_cmsgs *cmsgs)
8093 struct msghdr *my_msg = (struct msghdr *)msg;
8094 struct cmsghdr *cmsg;
8096 for_each_cmsghdr(cmsg, my_msg) {
8097 if (!CMSG_OK(my_msg, cmsg))
8098 return -EINVAL;
8100 /* Should we parse this header or ignore? */
8101 if (cmsg->cmsg_level != IPPROTO_SCTP)
8102 continue;
8104 /* Strictly check lengths following example in SCM code. */
8105 switch (cmsg->cmsg_type) {
8106 case SCTP_INIT:
8107 /* SCTP Socket API Extension
8108 * 5.3.1 SCTP Initiation Structure (SCTP_INIT)
8110 * This cmsghdr structure provides information for
8111 * initializing new SCTP associations with sendmsg().
8112 * The SCTP_INITMSG socket option uses this same data
8113 * structure. This structure is not used for
8114 * recvmsg().
8116 * cmsg_level cmsg_type cmsg_data[]
8117 * ------------ ------------ ----------------------
8118 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
8120 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_initmsg)))
8121 return -EINVAL;
8123 cmsgs->init = CMSG_DATA(cmsg);
8124 break;
8126 case SCTP_SNDRCV:
8127 /* SCTP Socket API Extension
8128 * 5.3.2 SCTP Header Information Structure(SCTP_SNDRCV)
8130 * This cmsghdr structure specifies SCTP options for
8131 * sendmsg() and describes SCTP header information
8132 * about a received message through recvmsg().
8134 * cmsg_level cmsg_type cmsg_data[]
8135 * ------------ ------------ ----------------------
8136 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
8138 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
8139 return -EINVAL;
8141 cmsgs->srinfo = CMSG_DATA(cmsg);
8143 if (cmsgs->srinfo->sinfo_flags &
8144 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8145 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8146 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8147 return -EINVAL;
8148 break;
8150 case SCTP_SNDINFO:
8151 /* SCTP Socket API Extension
8152 * 5.3.4 SCTP Send Information Structure (SCTP_SNDINFO)
8154 * This cmsghdr structure specifies SCTP options for
8155 * sendmsg(). This structure and SCTP_RCVINFO replaces
8156 * SCTP_SNDRCV which has been deprecated.
8158 * cmsg_level cmsg_type cmsg_data[]
8159 * ------------ ------------ ---------------------
8160 * IPPROTO_SCTP SCTP_SNDINFO struct sctp_sndinfo
8162 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_sndinfo)))
8163 return -EINVAL;
8165 cmsgs->sinfo = CMSG_DATA(cmsg);
8167 if (cmsgs->sinfo->snd_flags &
8168 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
8169 SCTP_SACK_IMMEDIATELY | SCTP_SENDALL |
8170 SCTP_PR_SCTP_MASK | SCTP_ABORT | SCTP_EOF))
8171 return -EINVAL;
8172 break;
8173 case SCTP_PRINFO:
8174 /* SCTP Socket API Extension
8175 * 5.3.7 SCTP PR-SCTP Information Structure (SCTP_PRINFO)
8177 * This cmsghdr structure specifies SCTP options for sendmsg().
8179 * cmsg_level cmsg_type cmsg_data[]
8180 * ------------ ------------ ---------------------
8181 * IPPROTO_SCTP SCTP_PRINFO struct sctp_prinfo
8183 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_prinfo)))
8184 return -EINVAL;
8186 cmsgs->prinfo = CMSG_DATA(cmsg);
8187 if (cmsgs->prinfo->pr_policy & ~SCTP_PR_SCTP_MASK)
8188 return -EINVAL;
8190 if (cmsgs->prinfo->pr_policy == SCTP_PR_SCTP_NONE)
8191 cmsgs->prinfo->pr_value = 0;
8192 break;
8193 case SCTP_AUTHINFO:
8194 /* SCTP Socket API Extension
8195 * 5.3.8 SCTP AUTH Information Structure (SCTP_AUTHINFO)
8197 * This cmsghdr structure specifies SCTP options for sendmsg().
8199 * cmsg_level cmsg_type cmsg_data[]
8200 * ------------ ------------ ---------------------
8201 * IPPROTO_SCTP SCTP_AUTHINFO struct sctp_authinfo
8203 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct sctp_authinfo)))
8204 return -EINVAL;
8206 cmsgs->authinfo = CMSG_DATA(cmsg);
8207 break;
8208 case SCTP_DSTADDRV4:
8209 case SCTP_DSTADDRV6:
8210 /* SCTP Socket API Extension
8211 * 5.3.9/10 SCTP Destination IPv4/6 Address Structure (SCTP_DSTADDRV4/6)
8213 * This cmsghdr structure specifies SCTP options for sendmsg().
8215 * cmsg_level cmsg_type cmsg_data[]
8216 * ------------ ------------ ---------------------
8217 * IPPROTO_SCTP SCTP_DSTADDRV4 struct in_addr
8218 * ------------ ------------ ---------------------
8219 * IPPROTO_SCTP SCTP_DSTADDRV6 struct in6_addr
8221 cmsgs->addrs_msg = my_msg;
8222 break;
8223 default:
8224 return -EINVAL;
8228 return 0;
8232 * Wait for a packet..
8233 * Note: This function is the same function as in core/datagram.c
8234 * with a few modifications to make lksctp work.
8236 static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
8238 int error;
8239 DEFINE_WAIT(wait);
8241 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8243 /* Socket errors? */
8244 error = sock_error(sk);
8245 if (error)
8246 goto out;
8248 if (!skb_queue_empty(&sk->sk_receive_queue))
8249 goto ready;
8251 /* Socket shut down? */
8252 if (sk->sk_shutdown & RCV_SHUTDOWN)
8253 goto out;
8255 /* Sequenced packets can come disconnected. If so we report the
8256 * problem.
8258 error = -ENOTCONN;
8260 /* Is there a good reason to think that we may receive some data? */
8261 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
8262 goto out;
8264 /* Handle signals. */
8265 if (signal_pending(current))
8266 goto interrupted;
8268 /* Let another process have a go. Since we are going to sleep
8269 * anyway. Note: This may cause odd behaviors if the message
8270 * does not fit in the user's buffer, but this seems to be the
8271 * only way to honor MSG_DONTWAIT realistically.
8273 release_sock(sk);
8274 *timeo_p = schedule_timeout(*timeo_p);
8275 lock_sock(sk);
8277 ready:
8278 finish_wait(sk_sleep(sk), &wait);
8279 return 0;
8281 interrupted:
8282 error = sock_intr_errno(*timeo_p);
8284 out:
8285 finish_wait(sk_sleep(sk), &wait);
8286 *err = error;
8287 return error;
8290 /* Receive a datagram.
8291 * Note: This is pretty much the same routine as in core/datagram.c
8292 * with a few changes to make lksctp work.
8294 struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
8295 int noblock, int *err)
8297 int error;
8298 struct sk_buff *skb;
8299 long timeo;
8301 timeo = sock_rcvtimeo(sk, noblock);
8303 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
8304 MAX_SCHEDULE_TIMEOUT);
8306 do {
8307 /* Again only user level code calls this function,
8308 * so nothing interrupt level
8309 * will suddenly eat the receive_queue.
8311 * Look at current nfs client by the way...
8312 * However, this function was correct in any case. 8)
8314 if (flags & MSG_PEEK) {
8315 skb = skb_peek(&sk->sk_receive_queue);
8316 if (skb)
8317 refcount_inc(&skb->users);
8318 } else {
8319 skb = __skb_dequeue(&sk->sk_receive_queue);
8322 if (skb)
8323 return skb;
8325 /* Caller is allowed not to check sk->sk_err before calling. */
8326 error = sock_error(sk);
8327 if (error)
8328 goto no_packet;
8330 if (sk->sk_shutdown & RCV_SHUTDOWN)
8331 break;
8333 if (sk_can_busy_loop(sk)) {
8334 sk_busy_loop(sk, noblock);
8336 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
8337 continue;
8340 /* User doesn't want to wait. */
8341 error = -EAGAIN;
8342 if (!timeo)
8343 goto no_packet;
8344 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
8346 return NULL;
8348 no_packet:
8349 *err = error;
8350 return NULL;
8353 /* If sndbuf has changed, wake up per association sndbuf waiters. */
8354 static void __sctp_write_space(struct sctp_association *asoc)
8356 struct sock *sk = asoc->base.sk;
8358 if (sctp_wspace(asoc) <= 0)
8359 return;
8361 if (waitqueue_active(&asoc->wait))
8362 wake_up_interruptible(&asoc->wait);
8364 if (sctp_writeable(sk)) {
8365 struct socket_wq *wq;
8367 rcu_read_lock();
8368 wq = rcu_dereference(sk->sk_wq);
8369 if (wq) {
8370 if (waitqueue_active(&wq->wait))
8371 wake_up_interruptible(&wq->wait);
8373 /* Note that we try to include the Async I/O support
8374 * here by modeling from the current TCP/UDP code.
8375 * We have not tested with it yet.
8377 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
8378 sock_wake_async(wq, SOCK_WAKE_SPACE, POLL_OUT);
8380 rcu_read_unlock();
8384 static void sctp_wake_up_waiters(struct sock *sk,
8385 struct sctp_association *asoc)
8387 struct sctp_association *tmp = asoc;
8389 /* We do accounting for the sndbuf space per association,
8390 * so we only need to wake our own association.
8392 if (asoc->ep->sndbuf_policy)
8393 return __sctp_write_space(asoc);
8395 /* If association goes down and is just flushing its
8396 * outq, then just normally notify others.
8398 if (asoc->base.dead)
8399 return sctp_write_space(sk);
8401 /* Accounting for the sndbuf space is per socket, so we
8402 * need to wake up others, try to be fair and in case of
8403 * other associations, let them have a go first instead
8404 * of just doing a sctp_write_space() call.
8406 * Note that we reach sctp_wake_up_waiters() only when
8407 * associations free up queued chunks, thus we are under
8408 * lock and the list of associations on a socket is
8409 * guaranteed not to change.
8411 for (tmp = list_next_entry(tmp, asocs); 1;
8412 tmp = list_next_entry(tmp, asocs)) {
8413 /* Manually skip the head element. */
8414 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
8415 continue;
8416 /* Wake up association. */
8417 __sctp_write_space(tmp);
8418 /* We've reached the end. */
8419 if (tmp == asoc)
8420 break;
8424 /* Do accounting for the sndbuf space.
8425 * Decrement the used sndbuf space of the corresponding association by the
8426 * data size which was just transmitted(freed).
8428 static void sctp_wfree(struct sk_buff *skb)
8430 struct sctp_chunk *chunk = skb_shinfo(skb)->destructor_arg;
8431 struct sctp_association *asoc = chunk->asoc;
8432 struct sock *sk = asoc->base.sk;
8434 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
8435 sizeof(struct sk_buff) +
8436 sizeof(struct sctp_chunk);
8438 WARN_ON(refcount_sub_and_test(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc));
8441 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
8443 sk->sk_wmem_queued -= skb->truesize;
8444 sk_mem_uncharge(sk, skb->truesize);
8446 if (chunk->shkey) {
8447 struct sctp_shared_key *shkey = chunk->shkey;
8449 /* refcnt == 2 and !list_empty mean after this release, it's
8450 * not being used anywhere, and it's time to notify userland
8451 * that this shkey can be freed if it's been deactivated.
8453 if (shkey->deactivated && !list_empty(&shkey->key_list) &&
8454 refcount_read(&shkey->refcnt) == 2) {
8455 struct sctp_ulpevent *ev;
8457 ev = sctp_ulpevent_make_authkey(asoc, shkey->key_id,
8458 SCTP_AUTH_FREE_KEY,
8459 GFP_KERNEL);
8460 if (ev)
8461 asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
8463 sctp_auth_shkey_release(chunk->shkey);
8466 sock_wfree(skb);
8467 sctp_wake_up_waiters(sk, asoc);
8469 sctp_association_put(asoc);
8472 /* Do accounting for the receive space on the socket.
8473 * Accounting for the association is done in ulpevent.c
8474 * We set this as a destructor for the cloned data skbs so that
8475 * accounting is done at the correct time.
8477 void sctp_sock_rfree(struct sk_buff *skb)
8479 struct sock *sk = skb->sk;
8480 struct sctp_ulpevent *event = sctp_skb2event(skb);
8482 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
8485 * Mimic the behavior of sock_rfree
8487 sk_mem_uncharge(sk, event->rmem_len);
8491 /* Helper function to wait for space in the sndbuf. */
8492 static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
8493 size_t msg_len)
8495 struct sock *sk = asoc->base.sk;
8496 long current_timeo = *timeo_p;
8497 DEFINE_WAIT(wait);
8498 int err = 0;
8500 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
8501 *timeo_p, msg_len);
8503 /* Increment the association's refcnt. */
8504 sctp_association_hold(asoc);
8506 /* Wait on the association specific sndbuf space. */
8507 for (;;) {
8508 prepare_to_wait_exclusive(&asoc->wait, &wait,
8509 TASK_INTERRUPTIBLE);
8510 if (asoc->base.dead)
8511 goto do_dead;
8512 if (!*timeo_p)
8513 goto do_nonblock;
8514 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING)
8515 goto do_error;
8516 if (signal_pending(current))
8517 goto do_interrupted;
8518 if ((int)msg_len <= sctp_wspace(asoc))
8519 break;
8521 /* Let another process have a go. Since we are going
8522 * to sleep anyway.
8524 release_sock(sk);
8525 current_timeo = schedule_timeout(current_timeo);
8526 lock_sock(sk);
8527 if (sk != asoc->base.sk)
8528 goto do_error;
8530 *timeo_p = current_timeo;
8533 out:
8534 finish_wait(&asoc->wait, &wait);
8536 /* Release the association's refcnt. */
8537 sctp_association_put(asoc);
8539 return err;
8541 do_dead:
8542 err = -ESRCH;
8543 goto out;
8545 do_error:
8546 err = -EPIPE;
8547 goto out;
8549 do_interrupted:
8550 err = sock_intr_errno(*timeo_p);
8551 goto out;
8553 do_nonblock:
8554 err = -EAGAIN;
8555 goto out;
8558 void sctp_data_ready(struct sock *sk)
8560 struct socket_wq *wq;
8562 rcu_read_lock();
8563 wq = rcu_dereference(sk->sk_wq);
8564 if (skwq_has_sleeper(wq))
8565 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
8566 EPOLLRDNORM | EPOLLRDBAND);
8567 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
8568 rcu_read_unlock();
8571 /* If socket sndbuf has changed, wake up all per association waiters. */
8572 void sctp_write_space(struct sock *sk)
8574 struct sctp_association *asoc;
8576 /* Wake up the tasks in each wait queue. */
8577 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
8578 __sctp_write_space(asoc);
8582 /* Is there any sndbuf space available on the socket?
8584 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
8585 * associations on the same socket. For a UDP-style socket with
8586 * multiple associations, it is possible for it to be "unwriteable"
8587 * prematurely. I assume that this is acceptable because
8588 * a premature "unwriteable" is better than an accidental "writeable" which
8589 * would cause an unwanted block under certain circumstances. For the 1-1
8590 * UDP-style sockets or TCP-style sockets, this code should work.
8591 * - Daisy
8593 static bool sctp_writeable(struct sock *sk)
8595 return sk->sk_sndbuf > sk->sk_wmem_queued;
8598 /* Wait for an association to go into ESTABLISHED state. If timeout is 0,
8599 * returns immediately with EINPROGRESS.
8601 static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
8603 struct sock *sk = asoc->base.sk;
8604 int err = 0;
8605 long current_timeo = *timeo_p;
8606 DEFINE_WAIT(wait);
8608 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
8610 /* Increment the association's refcnt. */
8611 sctp_association_hold(asoc);
8613 for (;;) {
8614 prepare_to_wait_exclusive(&asoc->wait, &wait,
8615 TASK_INTERRUPTIBLE);
8616 if (!*timeo_p)
8617 goto do_nonblock;
8618 if (sk->sk_shutdown & RCV_SHUTDOWN)
8619 break;
8620 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
8621 asoc->base.dead)
8622 goto do_error;
8623 if (signal_pending(current))
8624 goto do_interrupted;
8626 if (sctp_state(asoc, ESTABLISHED))
8627 break;
8629 /* Let another process have a go. Since we are going
8630 * to sleep anyway.
8632 release_sock(sk);
8633 current_timeo = schedule_timeout(current_timeo);
8634 lock_sock(sk);
8636 *timeo_p = current_timeo;
8639 out:
8640 finish_wait(&asoc->wait, &wait);
8642 /* Release the association's refcnt. */
8643 sctp_association_put(asoc);
8645 return err;
8647 do_error:
8648 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
8649 err = -ETIMEDOUT;
8650 else
8651 err = -ECONNREFUSED;
8652 goto out;
8654 do_interrupted:
8655 err = sock_intr_errno(*timeo_p);
8656 goto out;
8658 do_nonblock:
8659 err = -EINPROGRESS;
8660 goto out;
8663 static int sctp_wait_for_accept(struct sock *sk, long timeo)
8665 struct sctp_endpoint *ep;
8666 int err = 0;
8667 DEFINE_WAIT(wait);
8669 ep = sctp_sk(sk)->ep;
8672 for (;;) {
8673 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
8674 TASK_INTERRUPTIBLE);
8676 if (list_empty(&ep->asocs)) {
8677 release_sock(sk);
8678 timeo = schedule_timeout(timeo);
8679 lock_sock(sk);
8682 err = -EINVAL;
8683 if (!sctp_sstate(sk, LISTENING))
8684 break;
8686 err = 0;
8687 if (!list_empty(&ep->asocs))
8688 break;
8690 err = sock_intr_errno(timeo);
8691 if (signal_pending(current))
8692 break;
8694 err = -EAGAIN;
8695 if (!timeo)
8696 break;
8699 finish_wait(sk_sleep(sk), &wait);
8701 return err;
8704 static void sctp_wait_for_close(struct sock *sk, long timeout)
8706 DEFINE_WAIT(wait);
8708 do {
8709 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
8710 if (list_empty(&sctp_sk(sk)->ep->asocs))
8711 break;
8712 release_sock(sk);
8713 timeout = schedule_timeout(timeout);
8714 lock_sock(sk);
8715 } while (!signal_pending(current) && timeout);
8717 finish_wait(sk_sleep(sk), &wait);
8720 static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
8722 struct sk_buff *frag;
8724 if (!skb->data_len)
8725 goto done;
8727 /* Don't forget the fragments. */
8728 skb_walk_frags(skb, frag)
8729 sctp_skb_set_owner_r_frag(frag, sk);
8731 done:
8732 sctp_skb_set_owner_r(skb, sk);
8735 void sctp_copy_sock(struct sock *newsk, struct sock *sk,
8736 struct sctp_association *asoc)
8738 struct inet_sock *inet = inet_sk(sk);
8739 struct inet_sock *newinet;
8740 struct sctp_sock *sp = sctp_sk(sk);
8741 struct sctp_endpoint *ep = sp->ep;
8743 newsk->sk_type = sk->sk_type;
8744 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
8745 newsk->sk_flags = sk->sk_flags;
8746 newsk->sk_tsflags = sk->sk_tsflags;
8747 newsk->sk_no_check_tx = sk->sk_no_check_tx;
8748 newsk->sk_no_check_rx = sk->sk_no_check_rx;
8749 newsk->sk_reuse = sk->sk_reuse;
8750 sctp_sk(newsk)->reuse = sp->reuse;
8752 newsk->sk_shutdown = sk->sk_shutdown;
8753 newsk->sk_destruct = sctp_destruct_sock;
8754 newsk->sk_family = sk->sk_family;
8755 newsk->sk_protocol = IPPROTO_SCTP;
8756 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
8757 newsk->sk_sndbuf = sk->sk_sndbuf;
8758 newsk->sk_rcvbuf = sk->sk_rcvbuf;
8759 newsk->sk_lingertime = sk->sk_lingertime;
8760 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
8761 newsk->sk_sndtimeo = sk->sk_sndtimeo;
8762 newsk->sk_rxhash = sk->sk_rxhash;
8764 newinet = inet_sk(newsk);
8766 /* Initialize sk's sport, dport, rcv_saddr and daddr for
8767 * getsockname() and getpeername()
8769 newinet->inet_sport = inet->inet_sport;
8770 newinet->inet_saddr = inet->inet_saddr;
8771 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
8772 newinet->inet_dport = htons(asoc->peer.port);
8773 newinet->pmtudisc = inet->pmtudisc;
8774 newinet->inet_id = prandom_u32();
8776 newinet->uc_ttl = inet->uc_ttl;
8777 newinet->mc_loop = 1;
8778 newinet->mc_ttl = 1;
8779 newinet->mc_index = 0;
8780 newinet->mc_list = NULL;
8782 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
8783 net_enable_timestamp();
8785 /* Set newsk security attributes from orginal sk and connection
8786 * security attribute from ep.
8788 security_sctp_sk_clone(ep, sk, newsk);
8791 static inline void sctp_copy_descendant(struct sock *sk_to,
8792 const struct sock *sk_from)
8794 int ancestor_size = sizeof(struct inet_sock) +
8795 sizeof(struct sctp_sock) -
8796 offsetof(struct sctp_sock, auto_asconf_list);
8798 if (sk_from->sk_family == PF_INET6)
8799 ancestor_size += sizeof(struct ipv6_pinfo);
8801 __inet_sk_copy_descendant(sk_to, sk_from, ancestor_size);
8804 /* Populate the fields of the newsk from the oldsk and migrate the assoc
8805 * and its messages to the newsk.
8807 static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
8808 struct sctp_association *assoc,
8809 enum sctp_socket_type type)
8811 struct sctp_sock *oldsp = sctp_sk(oldsk);
8812 struct sctp_sock *newsp = sctp_sk(newsk);
8813 struct sctp_bind_bucket *pp; /* hash list port iterator */
8814 struct sctp_endpoint *newep = newsp->ep;
8815 struct sk_buff *skb, *tmp;
8816 struct sctp_ulpevent *event;
8817 struct sctp_bind_hashbucket *head;
8819 /* Migrate socket buffer sizes and all the socket level options to the
8820 * new socket.
8822 newsk->sk_sndbuf = oldsk->sk_sndbuf;
8823 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
8824 /* Brute force copy old sctp opt. */
8825 sctp_copy_descendant(newsk, oldsk);
8827 /* Restore the ep value that was overwritten with the above structure
8828 * copy.
8830 newsp->ep = newep;
8831 newsp->hmac = NULL;
8833 /* Hook this new socket in to the bind_hash list. */
8834 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
8835 inet_sk(oldsk)->inet_num)];
8836 spin_lock_bh(&head->lock);
8837 pp = sctp_sk(oldsk)->bind_hash;
8838 sk_add_bind_node(newsk, &pp->owner);
8839 sctp_sk(newsk)->bind_hash = pp;
8840 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
8841 spin_unlock_bh(&head->lock);
8843 /* Copy the bind_addr list from the original endpoint to the new
8844 * endpoint so that we can handle restarts properly
8846 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
8847 &oldsp->ep->base.bind_addr, GFP_KERNEL);
8849 /* Move any messages in the old socket's receive queue that are for the
8850 * peeled off association to the new socket's receive queue.
8852 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
8853 event = sctp_skb2event(skb);
8854 if (event->asoc == assoc) {
8855 __skb_unlink(skb, &oldsk->sk_receive_queue);
8856 __skb_queue_tail(&newsk->sk_receive_queue, skb);
8857 sctp_skb_set_owner_r_frag(skb, newsk);
8861 /* Clean up any messages pending delivery due to partial
8862 * delivery. Three cases:
8863 * 1) No partial deliver; no work.
8864 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
8865 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
8867 skb_queue_head_init(&newsp->pd_lobby);
8868 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
8870 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
8871 struct sk_buff_head *queue;
8873 /* Decide which queue to move pd_lobby skbs to. */
8874 if (assoc->ulpq.pd_mode) {
8875 queue = &newsp->pd_lobby;
8876 } else
8877 queue = &newsk->sk_receive_queue;
8879 /* Walk through the pd_lobby, looking for skbs that
8880 * need moved to the new socket.
8882 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
8883 event = sctp_skb2event(skb);
8884 if (event->asoc == assoc) {
8885 __skb_unlink(skb, &oldsp->pd_lobby);
8886 __skb_queue_tail(queue, skb);
8887 sctp_skb_set_owner_r_frag(skb, newsk);
8891 /* Clear up any skbs waiting for the partial
8892 * delivery to finish.
8894 if (assoc->ulpq.pd_mode)
8895 sctp_clear_pd(oldsk, NULL);
8899 sctp_for_each_rx_skb(assoc, newsk, sctp_skb_set_owner_r_frag);
8901 /* Set the type of socket to indicate that it is peeled off from the
8902 * original UDP-style socket or created with the accept() call on a
8903 * TCP-style socket..
8905 newsp->type = type;
8907 /* Mark the new socket "in-use" by the user so that any packets
8908 * that may arrive on the association after we've moved it are
8909 * queued to the backlog. This prevents a potential race between
8910 * backlog processing on the old socket and new-packet processing
8911 * on the new socket.
8913 * The caller has just allocated newsk so we can guarantee that other
8914 * paths won't try to lock it and then oldsk.
8916 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
8917 sctp_for_each_tx_datachunk(assoc, true, sctp_clear_owner_w);
8918 sctp_assoc_migrate(assoc, newsk);
8919 sctp_for_each_tx_datachunk(assoc, false, sctp_set_owner_w);
8921 /* If the association on the newsk is already closed before accept()
8922 * is called, set RCV_SHUTDOWN flag.
8924 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP)) {
8925 inet_sk_set_state(newsk, SCTP_SS_CLOSED);
8926 newsk->sk_shutdown |= RCV_SHUTDOWN;
8927 } else {
8928 inet_sk_set_state(newsk, SCTP_SS_ESTABLISHED);
8931 release_sock(newsk);
8935 /* This proto struct describes the ULP interface for SCTP. */
8936 struct proto sctp_prot = {
8937 .name = "SCTP",
8938 .owner = THIS_MODULE,
8939 .close = sctp_close,
8940 .disconnect = sctp_disconnect,
8941 .accept = sctp_accept,
8942 .ioctl = sctp_ioctl,
8943 .init = sctp_init_sock,
8944 .destroy = sctp_destroy_sock,
8945 .shutdown = sctp_shutdown,
8946 .setsockopt = sctp_setsockopt,
8947 .getsockopt = sctp_getsockopt,
8948 .sendmsg = sctp_sendmsg,
8949 .recvmsg = sctp_recvmsg,
8950 .bind = sctp_bind,
8951 .backlog_rcv = sctp_backlog_rcv,
8952 .hash = sctp_hash,
8953 .unhash = sctp_unhash,
8954 .no_autobind = true,
8955 .obj_size = sizeof(struct sctp_sock),
8956 .useroffset = offsetof(struct sctp_sock, subscribe),
8957 .usersize = offsetof(struct sctp_sock, initmsg) -
8958 offsetof(struct sctp_sock, subscribe) +
8959 sizeof_field(struct sctp_sock, initmsg),
8960 .sysctl_mem = sysctl_sctp_mem,
8961 .sysctl_rmem = sysctl_sctp_rmem,
8962 .sysctl_wmem = sysctl_sctp_wmem,
8963 .memory_pressure = &sctp_memory_pressure,
8964 .enter_memory_pressure = sctp_enter_memory_pressure,
8965 .memory_allocated = &sctp_memory_allocated,
8966 .sockets_allocated = &sctp_sockets_allocated,
8969 #if IS_ENABLED(CONFIG_IPV6)
8971 #include <net/transp_v6.h>
8972 static void sctp_v6_destroy_sock(struct sock *sk)
8974 sctp_destroy_sock(sk);
8975 inet6_destroy_sock(sk);
8978 struct proto sctpv6_prot = {
8979 .name = "SCTPv6",
8980 .owner = THIS_MODULE,
8981 .close = sctp_close,
8982 .disconnect = sctp_disconnect,
8983 .accept = sctp_accept,
8984 .ioctl = sctp_ioctl,
8985 .init = sctp_init_sock,
8986 .destroy = sctp_v6_destroy_sock,
8987 .shutdown = sctp_shutdown,
8988 .setsockopt = sctp_setsockopt,
8989 .getsockopt = sctp_getsockopt,
8990 .sendmsg = sctp_sendmsg,
8991 .recvmsg = sctp_recvmsg,
8992 .bind = sctp_bind,
8993 .backlog_rcv = sctp_backlog_rcv,
8994 .hash = sctp_hash,
8995 .unhash = sctp_unhash,
8996 .no_autobind = true,
8997 .obj_size = sizeof(struct sctp6_sock),
8998 .useroffset = offsetof(struct sctp6_sock, sctp.subscribe),
8999 .usersize = offsetof(struct sctp6_sock, sctp.initmsg) -
9000 offsetof(struct sctp6_sock, sctp.subscribe) +
9001 sizeof_field(struct sctp6_sock, sctp.initmsg),
9002 .sysctl_mem = sysctl_sctp_mem,
9003 .sysctl_rmem = sysctl_sctp_rmem,
9004 .sysctl_wmem = sysctl_sctp_wmem,
9005 .memory_pressure = &sctp_memory_pressure,
9006 .enter_memory_pressure = sctp_enter_memory_pressure,
9007 .memory_allocated = &sctp_memory_allocated,
9008 .sockets_allocated = &sctp_sockets_allocated,
9010 #endif /* IS_ENABLED(CONFIG_IPV6) */