Loading...
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
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
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.
17 *
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.
23 *
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.
29 *
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/>.
33 *
34 * Please send any bug reports or fixes you make to the
35 * email address(es):
36 * lksctp developers <linux-sctp@vger.kernel.org>
37 *
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>
51 */
52
53#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
54
55#include <linux/types.h>
56#include <linux/kernel.h>
57#include <linux/wait.h>
58#include <linux/time.h>
59#include <linux/ip.h>
60#include <linux/capability.h>
61#include <linux/fcntl.h>
62#include <linux/poll.h>
63#include <linux/init.h>
64#include <linux/crypto.h>
65#include <linux/slab.h>
66#include <linux/file.h>
67#include <linux/compat.h>
68
69#include <net/ip.h>
70#include <net/icmp.h>
71#include <net/route.h>
72#include <net/ipv6.h>
73#include <net/inet_common.h>
74
75#include <linux/socket.h> /* for sa_family_t */
76#include <linux/export.h>
77#include <net/sock.h>
78#include <net/sctp/sctp.h>
79#include <net/sctp/sm.h>
80
81/* Forward declarations for internal helper functions. */
82static int sctp_writeable(struct sock *sk);
83static void sctp_wfree(struct sk_buff *skb);
84static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
85 size_t msg_len);
86static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p);
87static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
88static int sctp_wait_for_accept(struct sock *sk, long timeo);
89static void sctp_wait_for_close(struct sock *sk, long timeo);
90static void sctp_destruct_sock(struct sock *sk);
91static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
92 union sctp_addr *addr, int len);
93static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
94static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
95static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
96static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
97static int sctp_send_asconf(struct sctp_association *asoc,
98 struct sctp_chunk *chunk);
99static int sctp_do_bind(struct sock *, union sctp_addr *, int);
100static int sctp_autobind(struct sock *sk);
101static void sctp_sock_migrate(struct sock *, struct sock *,
102 struct sctp_association *, sctp_socket_type_t);
103
104extern struct kmem_cache *sctp_bucket_cachep;
105extern long sysctl_sctp_mem[3];
106extern int sysctl_sctp_rmem[3];
107extern int sysctl_sctp_wmem[3];
108
109static int sctp_memory_pressure;
110static atomic_long_t sctp_memory_allocated;
111struct percpu_counter sctp_sockets_allocated;
112
113static void sctp_enter_memory_pressure(struct sock *sk)
114{
115 sctp_memory_pressure = 1;
116}
117
118
119/* Get the sndbuf space available at the time on the association. */
120static inline int sctp_wspace(struct sctp_association *asoc)
121{
122 int amt;
123
124 if (asoc->ep->sndbuf_policy)
125 amt = asoc->sndbuf_used;
126 else
127 amt = sk_wmem_alloc_get(asoc->base.sk);
128
129 if (amt >= asoc->base.sk->sk_sndbuf) {
130 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
131 amt = 0;
132 else {
133 amt = sk_stream_wspace(asoc->base.sk);
134 if (amt < 0)
135 amt = 0;
136 }
137 } else {
138 amt = asoc->base.sk->sk_sndbuf - amt;
139 }
140 return amt;
141}
142
143/* Increment the used sndbuf space count of the corresponding association by
144 * the size of the outgoing data chunk.
145 * Also, set the skb destructor for sndbuf accounting later.
146 *
147 * Since it is always 1-1 between chunk and skb, and also a new skb is always
148 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
149 * destructor in the data chunk skb for the purpose of the sndbuf space
150 * tracking.
151 */
152static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
153{
154 struct sctp_association *asoc = chunk->asoc;
155 struct sock *sk = asoc->base.sk;
156
157 /* The sndbuf space is tracked per association. */
158 sctp_association_hold(asoc);
159
160 skb_set_owner_w(chunk->skb, sk);
161
162 chunk->skb->destructor = sctp_wfree;
163 /* Save the chunk pointer in skb for sctp_wfree to use later. */
164 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
165
166 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
167 sizeof(struct sk_buff) +
168 sizeof(struct sctp_chunk);
169
170 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
171 sk->sk_wmem_queued += chunk->skb->truesize;
172 sk_mem_charge(sk, chunk->skb->truesize);
173}
174
175/* Verify that this is a valid address. */
176static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
177 int len)
178{
179 struct sctp_af *af;
180
181 /* Verify basic sockaddr. */
182 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
183 if (!af)
184 return -EINVAL;
185
186 /* Is this a valid SCTP address? */
187 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
188 return -EINVAL;
189
190 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
191 return -EINVAL;
192
193 return 0;
194}
195
196/* Look up the association by its id. If this is not a UDP-style
197 * socket, the ID field is always ignored.
198 */
199struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
200{
201 struct sctp_association *asoc = NULL;
202
203 /* If this is not a UDP-style socket, assoc id should be ignored. */
204 if (!sctp_style(sk, UDP)) {
205 /* Return NULL if the socket state is not ESTABLISHED. It
206 * could be a TCP-style listening socket or a socket which
207 * hasn't yet called connect() to establish an association.
208 */
209 if (!sctp_sstate(sk, ESTABLISHED))
210 return NULL;
211
212 /* Get the first and the only association from the list. */
213 if (!list_empty(&sctp_sk(sk)->ep->asocs))
214 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
215 struct sctp_association, asocs);
216 return asoc;
217 }
218
219 /* Otherwise this is a UDP-style socket. */
220 if (!id || (id == (sctp_assoc_t)-1))
221 return NULL;
222
223 spin_lock_bh(&sctp_assocs_id_lock);
224 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
225 spin_unlock_bh(&sctp_assocs_id_lock);
226
227 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
228 return NULL;
229
230 return asoc;
231}
232
233/* Look up the transport from an address and an assoc id. If both address and
234 * id are specified, the associations matching the address and the id should be
235 * the same.
236 */
237static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
238 struct sockaddr_storage *addr,
239 sctp_assoc_t id)
240{
241 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
242 struct sctp_transport *transport;
243 union sctp_addr *laddr = (union sctp_addr *)addr;
244
245 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
246 laddr,
247 &transport);
248
249 if (!addr_asoc)
250 return NULL;
251
252 id_asoc = sctp_id2assoc(sk, id);
253 if (id_asoc && (id_asoc != addr_asoc))
254 return NULL;
255
256 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
257 (union sctp_addr *)addr);
258
259 return transport;
260}
261
262/* API 3.1.2 bind() - UDP Style Syntax
263 * The syntax of bind() is,
264 *
265 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
266 *
267 * sd - the socket descriptor returned by socket().
268 * addr - the address structure (struct sockaddr_in or struct
269 * sockaddr_in6 [RFC 2553]),
270 * addr_len - the size of the address structure.
271 */
272static int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
273{
274 int retval = 0;
275
276 lock_sock(sk);
277
278 pr_debug("%s: sk:%p, addr:%p, addr_len:%d\n", __func__, sk,
279 addr, addr_len);
280
281 /* Disallow binding twice. */
282 if (!sctp_sk(sk)->ep->base.bind_addr.port)
283 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
284 addr_len);
285 else
286 retval = -EINVAL;
287
288 release_sock(sk);
289
290 return retval;
291}
292
293static long sctp_get_port_local(struct sock *, union sctp_addr *);
294
295/* Verify this is a valid sockaddr. */
296static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
297 union sctp_addr *addr, int len)
298{
299 struct sctp_af *af;
300
301 /* Check minimum size. */
302 if (len < sizeof (struct sockaddr))
303 return NULL;
304
305 /* V4 mapped address are really of AF_INET family */
306 if (addr->sa.sa_family == AF_INET6 &&
307 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
308 if (!opt->pf->af_supported(AF_INET, opt))
309 return NULL;
310 } else {
311 /* Does this PF support this AF? */
312 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
313 return NULL;
314 }
315
316 /* If we get this far, af is valid. */
317 af = sctp_get_af_specific(addr->sa.sa_family);
318
319 if (len < af->sockaddr_len)
320 return NULL;
321
322 return af;
323}
324
325/* Bind a local address either to an endpoint or to an association. */
326static int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
327{
328 struct net *net = sock_net(sk);
329 struct sctp_sock *sp = sctp_sk(sk);
330 struct sctp_endpoint *ep = sp->ep;
331 struct sctp_bind_addr *bp = &ep->base.bind_addr;
332 struct sctp_af *af;
333 unsigned short snum;
334 int ret = 0;
335
336 /* Common sockaddr verification. */
337 af = sctp_sockaddr_af(sp, addr, len);
338 if (!af) {
339 pr_debug("%s: sk:%p, newaddr:%p, len:%d EINVAL\n",
340 __func__, sk, addr, len);
341 return -EINVAL;
342 }
343
344 snum = ntohs(addr->v4.sin_port);
345
346 pr_debug("%s: sk:%p, new addr:%pISc, port:%d, new port:%d, len:%d\n",
347 __func__, sk, &addr->sa, bp->port, snum, len);
348
349 /* PF specific bind() address verification. */
350 if (!sp->pf->bind_verify(sp, addr))
351 return -EADDRNOTAVAIL;
352
353 /* We must either be unbound, or bind to the same port.
354 * It's OK to allow 0 ports if we are already bound.
355 * We'll just inhert an already bound port in this case
356 */
357 if (bp->port) {
358 if (!snum)
359 snum = bp->port;
360 else if (snum != bp->port) {
361 pr_debug("%s: new port %d doesn't match existing port "
362 "%d\n", __func__, snum, bp->port);
363 return -EINVAL;
364 }
365 }
366
367 if (snum && snum < PROT_SOCK &&
368 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE))
369 return -EACCES;
370
371 /* See if the address matches any of the addresses we may have
372 * already bound before checking against other endpoints.
373 */
374 if (sctp_bind_addr_match(bp, addr, sp))
375 return -EINVAL;
376
377 /* Make sure we are allowed to bind here.
378 * The function sctp_get_port_local() does duplicate address
379 * detection.
380 */
381 addr->v4.sin_port = htons(snum);
382 if ((ret = sctp_get_port_local(sk, addr))) {
383 return -EADDRINUSE;
384 }
385
386 /* Refresh ephemeral port. */
387 if (!bp->port)
388 bp->port = inet_sk(sk)->inet_num;
389
390 /* Add the address to the bind address list.
391 * Use GFP_ATOMIC since BHs will be disabled.
392 */
393 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
394
395 /* Copy back into socket for getsockname() use. */
396 if (!ret) {
397 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
398 af->to_sk_saddr(addr, sk);
399 }
400
401 return ret;
402}
403
404 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
405 *
406 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
407 * at any one time. If a sender, after sending an ASCONF chunk, decides
408 * it needs to transfer another ASCONF Chunk, it MUST wait until the
409 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
410 * subsequent ASCONF. Note this restriction binds each side, so at any
411 * time two ASCONF may be in-transit on any given association (one sent
412 * from each endpoint).
413 */
414static int sctp_send_asconf(struct sctp_association *asoc,
415 struct sctp_chunk *chunk)
416{
417 struct net *net = sock_net(asoc->base.sk);
418 int retval = 0;
419
420 /* If there is an outstanding ASCONF chunk, queue it for later
421 * transmission.
422 */
423 if (asoc->addip_last_asconf) {
424 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
425 goto out;
426 }
427
428 /* Hold the chunk until an ASCONF_ACK is received. */
429 sctp_chunk_hold(chunk);
430 retval = sctp_primitive_ASCONF(net, asoc, chunk);
431 if (retval)
432 sctp_chunk_free(chunk);
433 else
434 asoc->addip_last_asconf = chunk;
435
436out:
437 return retval;
438}
439
440/* Add a list of addresses as bind addresses to local endpoint or
441 * association.
442 *
443 * Basically run through each address specified in the addrs/addrcnt
444 * array/length pair, determine if it is IPv6 or IPv4 and call
445 * sctp_do_bind() on it.
446 *
447 * If any of them fails, then the operation will be reversed and the
448 * ones that were added will be removed.
449 *
450 * Only sctp_setsockopt_bindx() is supposed to call this function.
451 */
452static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
453{
454 int cnt;
455 int retval = 0;
456 void *addr_buf;
457 struct sockaddr *sa_addr;
458 struct sctp_af *af;
459
460 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n", __func__, sk,
461 addrs, addrcnt);
462
463 addr_buf = addrs;
464 for (cnt = 0; cnt < addrcnt; cnt++) {
465 /* The list may contain either IPv4 or IPv6 address;
466 * determine the address length for walking thru the list.
467 */
468 sa_addr = addr_buf;
469 af = sctp_get_af_specific(sa_addr->sa_family);
470 if (!af) {
471 retval = -EINVAL;
472 goto err_bindx_add;
473 }
474
475 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
476 af->sockaddr_len);
477
478 addr_buf += af->sockaddr_len;
479
480err_bindx_add:
481 if (retval < 0) {
482 /* Failed. Cleanup the ones that have been added */
483 if (cnt > 0)
484 sctp_bindx_rem(sk, addrs, cnt);
485 return retval;
486 }
487 }
488
489 return retval;
490}
491
492/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
493 * associations that are part of the endpoint indicating that a list of local
494 * addresses are added to the endpoint.
495 *
496 * If any of the addresses is already in the bind address list of the
497 * association, we do not send the chunk for that association. But it will not
498 * affect other associations.
499 *
500 * Only sctp_setsockopt_bindx() is supposed to call this function.
501 */
502static int sctp_send_asconf_add_ip(struct sock *sk,
503 struct sockaddr *addrs,
504 int addrcnt)
505{
506 struct net *net = sock_net(sk);
507 struct sctp_sock *sp;
508 struct sctp_endpoint *ep;
509 struct sctp_association *asoc;
510 struct sctp_bind_addr *bp;
511 struct sctp_chunk *chunk;
512 struct sctp_sockaddr_entry *laddr;
513 union sctp_addr *addr;
514 union sctp_addr saveaddr;
515 void *addr_buf;
516 struct sctp_af *af;
517 struct list_head *p;
518 int i;
519 int retval = 0;
520
521 if (!net->sctp.addip_enable)
522 return retval;
523
524 sp = sctp_sk(sk);
525 ep = sp->ep;
526
527 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
528 __func__, sk, addrs, addrcnt);
529
530 list_for_each_entry(asoc, &ep->asocs, asocs) {
531 if (!asoc->peer.asconf_capable)
532 continue;
533
534 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
535 continue;
536
537 if (!sctp_state(asoc, ESTABLISHED))
538 continue;
539
540 /* Check if any address in the packed array of addresses is
541 * in the bind address list of the association. If so,
542 * do not send the asconf chunk to its peer, but continue with
543 * other associations.
544 */
545 addr_buf = addrs;
546 for (i = 0; i < addrcnt; i++) {
547 addr = addr_buf;
548 af = sctp_get_af_specific(addr->v4.sin_family);
549 if (!af) {
550 retval = -EINVAL;
551 goto out;
552 }
553
554 if (sctp_assoc_lookup_laddr(asoc, addr))
555 break;
556
557 addr_buf += af->sockaddr_len;
558 }
559 if (i < addrcnt)
560 continue;
561
562 /* Use the first valid address in bind addr list of
563 * association as Address Parameter of ASCONF CHUNK.
564 */
565 bp = &asoc->base.bind_addr;
566 p = bp->address_list.next;
567 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
568 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
569 addrcnt, SCTP_PARAM_ADD_IP);
570 if (!chunk) {
571 retval = -ENOMEM;
572 goto out;
573 }
574
575 /* Add the new addresses to the bind address list with
576 * use_as_src set to 0.
577 */
578 addr_buf = addrs;
579 for (i = 0; i < addrcnt; i++) {
580 addr = addr_buf;
581 af = sctp_get_af_specific(addr->v4.sin_family);
582 memcpy(&saveaddr, addr, af->sockaddr_len);
583 retval = sctp_add_bind_addr(bp, &saveaddr,
584 SCTP_ADDR_NEW, GFP_ATOMIC);
585 addr_buf += af->sockaddr_len;
586 }
587 if (asoc->src_out_of_asoc_ok) {
588 struct sctp_transport *trans;
589
590 list_for_each_entry(trans,
591 &asoc->peer.transport_addr_list, transports) {
592 /* Clear the source and route cache */
593 dst_release(trans->dst);
594 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
595 2*asoc->pathmtu, 4380));
596 trans->ssthresh = asoc->peer.i.a_rwnd;
597 trans->rto = asoc->rto_initial;
598 sctp_max_rto(asoc, trans);
599 trans->rtt = trans->srtt = trans->rttvar = 0;
600 sctp_transport_route(trans, NULL,
601 sctp_sk(asoc->base.sk));
602 }
603 }
604 retval = sctp_send_asconf(asoc, chunk);
605 }
606
607out:
608 return retval;
609}
610
611/* Remove a list of addresses from bind addresses list. Do not remove the
612 * last address.
613 *
614 * Basically run through each address specified in the addrs/addrcnt
615 * array/length pair, determine if it is IPv6 or IPv4 and call
616 * sctp_del_bind() on it.
617 *
618 * If any of them fails, then the operation will be reversed and the
619 * ones that were removed will be added back.
620 *
621 * At least one address has to be left; if only one address is
622 * available, the operation will return -EBUSY.
623 *
624 * Only sctp_setsockopt_bindx() is supposed to call this function.
625 */
626static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
627{
628 struct sctp_sock *sp = sctp_sk(sk);
629 struct sctp_endpoint *ep = sp->ep;
630 int cnt;
631 struct sctp_bind_addr *bp = &ep->base.bind_addr;
632 int retval = 0;
633 void *addr_buf;
634 union sctp_addr *sa_addr;
635 struct sctp_af *af;
636
637 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
638 __func__, sk, addrs, addrcnt);
639
640 addr_buf = addrs;
641 for (cnt = 0; cnt < addrcnt; cnt++) {
642 /* If the bind address list is empty or if there is only one
643 * bind address, there is nothing more to be removed (we need
644 * at least one address here).
645 */
646 if (list_empty(&bp->address_list) ||
647 (sctp_list_single_entry(&bp->address_list))) {
648 retval = -EBUSY;
649 goto err_bindx_rem;
650 }
651
652 sa_addr = addr_buf;
653 af = sctp_get_af_specific(sa_addr->sa.sa_family);
654 if (!af) {
655 retval = -EINVAL;
656 goto err_bindx_rem;
657 }
658
659 if (!af->addr_valid(sa_addr, sp, NULL)) {
660 retval = -EADDRNOTAVAIL;
661 goto err_bindx_rem;
662 }
663
664 if (sa_addr->v4.sin_port &&
665 sa_addr->v4.sin_port != htons(bp->port)) {
666 retval = -EINVAL;
667 goto err_bindx_rem;
668 }
669
670 if (!sa_addr->v4.sin_port)
671 sa_addr->v4.sin_port = htons(bp->port);
672
673 /* FIXME - There is probably a need to check if sk->sk_saddr and
674 * sk->sk_rcv_addr are currently set to one of the addresses to
675 * be removed. This is something which needs to be looked into
676 * when we are fixing the outstanding issues with multi-homing
677 * socket routing and failover schemes. Refer to comments in
678 * sctp_do_bind(). -daisy
679 */
680 retval = sctp_del_bind_addr(bp, sa_addr);
681
682 addr_buf += af->sockaddr_len;
683err_bindx_rem:
684 if (retval < 0) {
685 /* Failed. Add the ones that has been removed back */
686 if (cnt > 0)
687 sctp_bindx_add(sk, addrs, cnt);
688 return retval;
689 }
690 }
691
692 return retval;
693}
694
695/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
696 * the associations that are part of the endpoint indicating that a list of
697 * local addresses are removed from the endpoint.
698 *
699 * If any of the addresses is already in the bind address list of the
700 * association, we do not send the chunk for that association. But it will not
701 * affect other associations.
702 *
703 * Only sctp_setsockopt_bindx() is supposed to call this function.
704 */
705static int sctp_send_asconf_del_ip(struct sock *sk,
706 struct sockaddr *addrs,
707 int addrcnt)
708{
709 struct net *net = sock_net(sk);
710 struct sctp_sock *sp;
711 struct sctp_endpoint *ep;
712 struct sctp_association *asoc;
713 struct sctp_transport *transport;
714 struct sctp_bind_addr *bp;
715 struct sctp_chunk *chunk;
716 union sctp_addr *laddr;
717 void *addr_buf;
718 struct sctp_af *af;
719 struct sctp_sockaddr_entry *saddr;
720 int i;
721 int retval = 0;
722 int stored = 0;
723
724 chunk = NULL;
725 if (!net->sctp.addip_enable)
726 return retval;
727
728 sp = sctp_sk(sk);
729 ep = sp->ep;
730
731 pr_debug("%s: sk:%p, addrs:%p, addrcnt:%d\n",
732 __func__, sk, addrs, addrcnt);
733
734 list_for_each_entry(asoc, &ep->asocs, asocs) {
735
736 if (!asoc->peer.asconf_capable)
737 continue;
738
739 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
740 continue;
741
742 if (!sctp_state(asoc, ESTABLISHED))
743 continue;
744
745 /* Check if any address in the packed array of addresses is
746 * not present in the bind address list of the association.
747 * If so, do not send the asconf chunk to its peer, but
748 * continue with other associations.
749 */
750 addr_buf = addrs;
751 for (i = 0; i < addrcnt; i++) {
752 laddr = addr_buf;
753 af = sctp_get_af_specific(laddr->v4.sin_family);
754 if (!af) {
755 retval = -EINVAL;
756 goto out;
757 }
758
759 if (!sctp_assoc_lookup_laddr(asoc, laddr))
760 break;
761
762 addr_buf += af->sockaddr_len;
763 }
764 if (i < addrcnt)
765 continue;
766
767 /* Find one address in the association's bind address list
768 * that is not in the packed array of addresses. This is to
769 * make sure that we do not delete all the addresses in the
770 * association.
771 */
772 bp = &asoc->base.bind_addr;
773 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
774 addrcnt, sp);
775 if ((laddr == NULL) && (addrcnt == 1)) {
776 if (asoc->asconf_addr_del_pending)
777 continue;
778 asoc->asconf_addr_del_pending =
779 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
780 if (asoc->asconf_addr_del_pending == NULL) {
781 retval = -ENOMEM;
782 goto out;
783 }
784 asoc->asconf_addr_del_pending->sa.sa_family =
785 addrs->sa_family;
786 asoc->asconf_addr_del_pending->v4.sin_port =
787 htons(bp->port);
788 if (addrs->sa_family == AF_INET) {
789 struct sockaddr_in *sin;
790
791 sin = (struct sockaddr_in *)addrs;
792 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
793 } else if (addrs->sa_family == AF_INET6) {
794 struct sockaddr_in6 *sin6;
795
796 sin6 = (struct sockaddr_in6 *)addrs;
797 asoc->asconf_addr_del_pending->v6.sin6_addr = sin6->sin6_addr;
798 }
799
800 pr_debug("%s: keep the last address asoc:%p %pISc at %p\n",
801 __func__, asoc, &asoc->asconf_addr_del_pending->sa,
802 asoc->asconf_addr_del_pending);
803
804 asoc->src_out_of_asoc_ok = 1;
805 stored = 1;
806 goto skip_mkasconf;
807 }
808
809 if (laddr == NULL)
810 return -EINVAL;
811
812 /* We do not need RCU protection throughout this loop
813 * because this is done under a socket lock from the
814 * setsockopt call.
815 */
816 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
817 SCTP_PARAM_DEL_IP);
818 if (!chunk) {
819 retval = -ENOMEM;
820 goto out;
821 }
822
823skip_mkasconf:
824 /* Reset use_as_src flag for the addresses in the bind address
825 * list that are to be deleted.
826 */
827 addr_buf = addrs;
828 for (i = 0; i < addrcnt; i++) {
829 laddr = addr_buf;
830 af = sctp_get_af_specific(laddr->v4.sin_family);
831 list_for_each_entry(saddr, &bp->address_list, list) {
832 if (sctp_cmp_addr_exact(&saddr->a, laddr))
833 saddr->state = SCTP_ADDR_DEL;
834 }
835 addr_buf += af->sockaddr_len;
836 }
837
838 /* Update the route and saddr entries for all the transports
839 * as some of the addresses in the bind address list are
840 * about to be deleted and cannot be used as source addresses.
841 */
842 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
843 transports) {
844 dst_release(transport->dst);
845 sctp_transport_route(transport, NULL,
846 sctp_sk(asoc->base.sk));
847 }
848
849 if (stored)
850 /* We don't need to transmit ASCONF */
851 continue;
852 retval = sctp_send_asconf(asoc, chunk);
853 }
854out:
855 return retval;
856}
857
858/* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
859int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
860{
861 struct sock *sk = sctp_opt2sk(sp);
862 union sctp_addr *addr;
863 struct sctp_af *af;
864
865 /* It is safe to write port space in caller. */
866 addr = &addrw->a;
867 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
868 af = sctp_get_af_specific(addr->sa.sa_family);
869 if (!af)
870 return -EINVAL;
871 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
872 return -EINVAL;
873
874 if (addrw->state == SCTP_ADDR_NEW)
875 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
876 else
877 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
878}
879
880/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
881 *
882 * API 8.1
883 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
884 * int flags);
885 *
886 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
887 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
888 * or IPv6 addresses.
889 *
890 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
891 * Section 3.1.2 for this usage.
892 *
893 * addrs is a pointer to an array of one or more socket addresses. Each
894 * address is contained in its appropriate structure (i.e. struct
895 * sockaddr_in or struct sockaddr_in6) the family of the address type
896 * must be used to distinguish the address length (note that this
897 * representation is termed a "packed array" of addresses). The caller
898 * specifies the number of addresses in the array with addrcnt.
899 *
900 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
901 * -1, and sets errno to the appropriate error code.
902 *
903 * For SCTP, the port given in each socket address must be the same, or
904 * sctp_bindx() will fail, setting errno to EINVAL.
905 *
906 * The flags parameter is formed from the bitwise OR of zero or more of
907 * the following currently defined flags:
908 *
909 * SCTP_BINDX_ADD_ADDR
910 *
911 * SCTP_BINDX_REM_ADDR
912 *
913 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
914 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
915 * addresses from the association. The two flags are mutually exclusive;
916 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
917 * not remove all addresses from an association; sctp_bindx() will
918 * reject such an attempt with EINVAL.
919 *
920 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
921 * additional addresses with an endpoint after calling bind(). Or use
922 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
923 * socket is associated with so that no new association accepted will be
924 * associated with those addresses. If the endpoint supports dynamic
925 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
926 * endpoint to send the appropriate message to the peer to change the
927 * peers address lists.
928 *
929 * Adding and removing addresses from a connected association is
930 * optional functionality. Implementations that do not support this
931 * functionality should return EOPNOTSUPP.
932 *
933 * Basically do nothing but copying the addresses from user to kernel
934 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
935 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
936 * from userspace.
937 *
938 * We don't use copy_from_user() for optimization: we first do the
939 * sanity checks (buffer size -fast- and access check-healthy
940 * pointer); if all of those succeed, then we can alloc the memory
941 * (expensive operation) needed to copy the data to kernel. Then we do
942 * the copying without checking the user space area
943 * (__copy_from_user()).
944 *
945 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
946 * it.
947 *
948 * sk The sk of the socket
949 * addrs The pointer to the addresses in user land
950 * addrssize Size of the addrs buffer
951 * op Operation to perform (add or remove, see the flags of
952 * sctp_bindx)
953 *
954 * Returns 0 if ok, <0 errno code on error.
955 */
956static int sctp_setsockopt_bindx(struct sock *sk,
957 struct sockaddr __user *addrs,
958 int addrs_size, int op)
959{
960 struct sockaddr *kaddrs;
961 int err;
962 int addrcnt = 0;
963 int walk_size = 0;
964 struct sockaddr *sa_addr;
965 void *addr_buf;
966 struct sctp_af *af;
967
968 pr_debug("%s: sk:%p addrs:%p addrs_size:%d opt:%d\n",
969 __func__, sk, addrs, addrs_size, op);
970
971 if (unlikely(addrs_size <= 0))
972 return -EINVAL;
973
974 /* Check the user passed a healthy pointer. */
975 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
976 return -EFAULT;
977
978 /* Alloc space for the address array in kernel memory. */
979 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
980 if (unlikely(!kaddrs))
981 return -ENOMEM;
982
983 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
984 kfree(kaddrs);
985 return -EFAULT;
986 }
987
988 /* Walk through the addrs buffer and count the number of addresses. */
989 addr_buf = kaddrs;
990 while (walk_size < addrs_size) {
991 if (walk_size + sizeof(sa_family_t) > addrs_size) {
992 kfree(kaddrs);
993 return -EINVAL;
994 }
995
996 sa_addr = addr_buf;
997 af = sctp_get_af_specific(sa_addr->sa_family);
998
999 /* If the address family is not supported or if this address
1000 * causes the address buffer to overflow return EINVAL.
1001 */
1002 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1003 kfree(kaddrs);
1004 return -EINVAL;
1005 }
1006 addrcnt++;
1007 addr_buf += af->sockaddr_len;
1008 walk_size += af->sockaddr_len;
1009 }
1010
1011 /* Do the work. */
1012 switch (op) {
1013 case SCTP_BINDX_ADD_ADDR:
1014 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1015 if (err)
1016 goto out;
1017 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1018 break;
1019
1020 case SCTP_BINDX_REM_ADDR:
1021 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1022 if (err)
1023 goto out;
1024 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1025 break;
1026
1027 default:
1028 err = -EINVAL;
1029 break;
1030 }
1031
1032out:
1033 kfree(kaddrs);
1034
1035 return err;
1036}
1037
1038/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1039 *
1040 * Common routine for handling connect() and sctp_connectx().
1041 * Connect will come in with just a single address.
1042 */
1043static int __sctp_connect(struct sock *sk,
1044 struct sockaddr *kaddrs,
1045 int addrs_size,
1046 sctp_assoc_t *assoc_id)
1047{
1048 struct net *net = sock_net(sk);
1049 struct sctp_sock *sp;
1050 struct sctp_endpoint *ep;
1051 struct sctp_association *asoc = NULL;
1052 struct sctp_association *asoc2;
1053 struct sctp_transport *transport;
1054 union sctp_addr to;
1055 struct sctp_af *af;
1056 sctp_scope_t scope;
1057 long timeo;
1058 int err = 0;
1059 int addrcnt = 0;
1060 int walk_size = 0;
1061 union sctp_addr *sa_addr = NULL;
1062 void *addr_buf;
1063 unsigned short port;
1064 unsigned int f_flags = 0;
1065
1066 sp = sctp_sk(sk);
1067 ep = sp->ep;
1068
1069 /* connect() cannot be done on a socket that is already in ESTABLISHED
1070 * state - UDP-style peeled off socket or a TCP-style socket that
1071 * is already connected.
1072 * It cannot be done even on a TCP-style listening socket.
1073 */
1074 if (sctp_sstate(sk, ESTABLISHED) ||
1075 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1076 err = -EISCONN;
1077 goto out_free;
1078 }
1079
1080 /* Walk through the addrs buffer and count the number of addresses. */
1081 addr_buf = kaddrs;
1082 while (walk_size < addrs_size) {
1083 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1084 err = -EINVAL;
1085 goto out_free;
1086 }
1087
1088 sa_addr = addr_buf;
1089 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1090
1091 /* If the address family is not supported or if this address
1092 * causes the address buffer to overflow return EINVAL.
1093 */
1094 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1095 err = -EINVAL;
1096 goto out_free;
1097 }
1098
1099 port = ntohs(sa_addr->v4.sin_port);
1100
1101 /* Save current address so we can work with it */
1102 memcpy(&to, sa_addr, af->sockaddr_len);
1103
1104 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1105 if (err)
1106 goto out_free;
1107
1108 /* Make sure the destination port is correctly set
1109 * in all addresses.
1110 */
1111 if (asoc && asoc->peer.port && asoc->peer.port != port) {
1112 err = -EINVAL;
1113 goto out_free;
1114 }
1115
1116 /* Check if there already is a matching association on the
1117 * endpoint (other than the one created here).
1118 */
1119 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1120 if (asoc2 && asoc2 != asoc) {
1121 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1122 err = -EISCONN;
1123 else
1124 err = -EALREADY;
1125 goto out_free;
1126 }
1127
1128 /* If we could not find a matching association on the endpoint,
1129 * make sure that there is no peeled-off association matching
1130 * the peer address even on another socket.
1131 */
1132 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1133 err = -EADDRNOTAVAIL;
1134 goto out_free;
1135 }
1136
1137 if (!asoc) {
1138 /* If a bind() or sctp_bindx() is not called prior to
1139 * an sctp_connectx() call, the system picks an
1140 * ephemeral port and will choose an address set
1141 * equivalent to binding with a wildcard address.
1142 */
1143 if (!ep->base.bind_addr.port) {
1144 if (sctp_autobind(sk)) {
1145 err = -EAGAIN;
1146 goto out_free;
1147 }
1148 } else {
1149 /*
1150 * If an unprivileged user inherits a 1-many
1151 * style socket with open associations on a
1152 * privileged port, it MAY be permitted to
1153 * accept new associations, but it SHOULD NOT
1154 * be permitted to open new associations.
1155 */
1156 if (ep->base.bind_addr.port < PROT_SOCK &&
1157 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1158 err = -EACCES;
1159 goto out_free;
1160 }
1161 }
1162
1163 scope = sctp_scope(&to);
1164 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1165 if (!asoc) {
1166 err = -ENOMEM;
1167 goto out_free;
1168 }
1169
1170 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1171 GFP_KERNEL);
1172 if (err < 0) {
1173 goto out_free;
1174 }
1175
1176 }
1177
1178 /* Prime the peer's transport structures. */
1179 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1180 SCTP_UNKNOWN);
1181 if (!transport) {
1182 err = -ENOMEM;
1183 goto out_free;
1184 }
1185
1186 addrcnt++;
1187 addr_buf += af->sockaddr_len;
1188 walk_size += af->sockaddr_len;
1189 }
1190
1191 /* In case the user of sctp_connectx() wants an association
1192 * id back, assign one now.
1193 */
1194 if (assoc_id) {
1195 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1196 if (err < 0)
1197 goto out_free;
1198 }
1199
1200 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1201 if (err < 0) {
1202 goto out_free;
1203 }
1204
1205 /* Initialize sk's dport and daddr for getpeername() */
1206 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1207 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1208 af->to_sk_daddr(sa_addr, sk);
1209 sk->sk_err = 0;
1210
1211 /* in-kernel sockets don't generally have a file allocated to them
1212 * if all they do is call sock_create_kern().
1213 */
1214 if (sk->sk_socket->file)
1215 f_flags = sk->sk_socket->file->f_flags;
1216
1217 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1218
1219 err = sctp_wait_for_connect(asoc, &timeo);
1220 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1221 *assoc_id = asoc->assoc_id;
1222
1223 /* Don't free association on exit. */
1224 asoc = NULL;
1225
1226out_free:
1227 pr_debug("%s: took out_free path with asoc:%p kaddrs:%p err:%d\n",
1228 __func__, asoc, kaddrs, err);
1229
1230 if (asoc) {
1231 /* sctp_primitive_ASSOCIATE may have added this association
1232 * To the hash table, try to unhash it, just in case, its a noop
1233 * if it wasn't hashed so we're safe
1234 */
1235 sctp_unhash_established(asoc);
1236 sctp_association_free(asoc);
1237 }
1238 return err;
1239}
1240
1241/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1242 *
1243 * API 8.9
1244 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1245 * sctp_assoc_t *asoc);
1246 *
1247 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1248 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1249 * or IPv6 addresses.
1250 *
1251 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1252 * Section 3.1.2 for this usage.
1253 *
1254 * addrs is a pointer to an array of one or more socket addresses. Each
1255 * address is contained in its appropriate structure (i.e. struct
1256 * sockaddr_in or struct sockaddr_in6) the family of the address type
1257 * must be used to distengish the address length (note that this
1258 * representation is termed a "packed array" of addresses). The caller
1259 * specifies the number of addresses in the array with addrcnt.
1260 *
1261 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1262 * the association id of the new association. On failure, sctp_connectx()
1263 * returns -1, and sets errno to the appropriate error code. The assoc_id
1264 * is not touched by the kernel.
1265 *
1266 * For SCTP, the port given in each socket address must be the same, or
1267 * sctp_connectx() will fail, setting errno to EINVAL.
1268 *
1269 * An application can use sctp_connectx to initiate an association with
1270 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1271 * allows a caller to specify multiple addresses at which a peer can be
1272 * reached. The way the SCTP stack uses the list of addresses to set up
1273 * the association is implementation dependent. This function only
1274 * specifies that the stack will try to make use of all the addresses in
1275 * the list when needed.
1276 *
1277 * Note that the list of addresses passed in is only used for setting up
1278 * the association. It does not necessarily equal the set of addresses
1279 * the peer uses for the resulting association. If the caller wants to
1280 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1281 * retrieve them after the association has been set up.
1282 *
1283 * Basically do nothing but copying the addresses from user to kernel
1284 * land and invoking either sctp_connectx(). This is used for tunneling
1285 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1286 *
1287 * We don't use copy_from_user() for optimization: we first do the
1288 * sanity checks (buffer size -fast- and access check-healthy
1289 * pointer); if all of those succeed, then we can alloc the memory
1290 * (expensive operation) needed to copy the data to kernel. Then we do
1291 * the copying without checking the user space area
1292 * (__copy_from_user()).
1293 *
1294 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1295 * it.
1296 *
1297 * sk The sk of the socket
1298 * addrs The pointer to the addresses in user land
1299 * addrssize Size of the addrs buffer
1300 *
1301 * Returns >=0 if ok, <0 errno code on error.
1302 */
1303static int __sctp_setsockopt_connectx(struct sock *sk,
1304 struct sockaddr __user *addrs,
1305 int addrs_size,
1306 sctp_assoc_t *assoc_id)
1307{
1308 int err = 0;
1309 struct sockaddr *kaddrs;
1310
1311 pr_debug("%s: sk:%p addrs:%p addrs_size:%d\n",
1312 __func__, sk, addrs, addrs_size);
1313
1314 if (unlikely(addrs_size <= 0))
1315 return -EINVAL;
1316
1317 /* Check the user passed a healthy pointer. */
1318 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1319 return -EFAULT;
1320
1321 /* Alloc space for the address array in kernel memory. */
1322 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1323 if (unlikely(!kaddrs))
1324 return -ENOMEM;
1325
1326 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1327 err = -EFAULT;
1328 } else {
1329 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1330 }
1331
1332 kfree(kaddrs);
1333
1334 return err;
1335}
1336
1337/*
1338 * This is an older interface. It's kept for backward compatibility
1339 * to the option that doesn't provide association id.
1340 */
1341static int sctp_setsockopt_connectx_old(struct sock *sk,
1342 struct sockaddr __user *addrs,
1343 int addrs_size)
1344{
1345 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1346}
1347
1348/*
1349 * New interface for the API. The since the API is done with a socket
1350 * option, to make it simple we feed back the association id is as a return
1351 * indication to the call. Error is always negative and association id is
1352 * always positive.
1353 */
1354static int sctp_setsockopt_connectx(struct sock *sk,
1355 struct sockaddr __user *addrs,
1356 int addrs_size)
1357{
1358 sctp_assoc_t assoc_id = 0;
1359 int err = 0;
1360
1361 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1362
1363 if (err)
1364 return err;
1365 else
1366 return assoc_id;
1367}
1368
1369/*
1370 * New (hopefully final) interface for the API.
1371 * We use the sctp_getaddrs_old structure so that use-space library
1372 * can avoid any unnecessary allocations. The only different part
1373 * is that we store the actual length of the address buffer into the
1374 * addrs_num structure member. That way we can re-use the existing
1375 * code.
1376 */
1377#ifdef CONFIG_COMPAT
1378struct compat_sctp_getaddrs_old {
1379 sctp_assoc_t assoc_id;
1380 s32 addr_num;
1381 compat_uptr_t addrs; /* struct sockaddr * */
1382};
1383#endif
1384
1385static int sctp_getsockopt_connectx3(struct sock *sk, int len,
1386 char __user *optval,
1387 int __user *optlen)
1388{
1389 struct sctp_getaddrs_old param;
1390 sctp_assoc_t assoc_id = 0;
1391 int err = 0;
1392
1393#ifdef CONFIG_COMPAT
1394 if (is_compat_task()) {
1395 struct compat_sctp_getaddrs_old param32;
1396
1397 if (len < sizeof(param32))
1398 return -EINVAL;
1399 if (copy_from_user(¶m32, optval, sizeof(param32)))
1400 return -EFAULT;
1401
1402 param.assoc_id = param32.assoc_id;
1403 param.addr_num = param32.addr_num;
1404 param.addrs = compat_ptr(param32.addrs);
1405 } else
1406#endif
1407 {
1408 if (len < sizeof(param))
1409 return -EINVAL;
1410 if (copy_from_user(¶m, optval, sizeof(param)))
1411 return -EFAULT;
1412 }
1413
1414 err = __sctp_setsockopt_connectx(sk, (struct sockaddr __user *)
1415 param.addrs, param.addr_num,
1416 &assoc_id);
1417 if (err == 0 || err == -EINPROGRESS) {
1418 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1419 return -EFAULT;
1420 if (put_user(sizeof(assoc_id), optlen))
1421 return -EFAULT;
1422 }
1423
1424 return err;
1425}
1426
1427/* API 3.1.4 close() - UDP Style Syntax
1428 * Applications use close() to perform graceful shutdown (as described in
1429 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1430 * by a UDP-style socket.
1431 *
1432 * The syntax is
1433 *
1434 * ret = close(int sd);
1435 *
1436 * sd - the socket descriptor of the associations to be closed.
1437 *
1438 * To gracefully shutdown a specific association represented by the
1439 * UDP-style socket, an application should use the sendmsg() call,
1440 * passing no user data, but including the appropriate flag in the
1441 * ancillary data (see Section xxxx).
1442 *
1443 * If sd in the close() call is a branched-off socket representing only
1444 * one association, the shutdown is performed on that association only.
1445 *
1446 * 4.1.6 close() - TCP Style Syntax
1447 *
1448 * Applications use close() to gracefully close down an association.
1449 *
1450 * The syntax is:
1451 *
1452 * int close(int sd);
1453 *
1454 * sd - the socket descriptor of the association to be closed.
1455 *
1456 * After an application calls close() on a socket descriptor, no further
1457 * socket operations will succeed on that descriptor.
1458 *
1459 * API 7.1.4 SO_LINGER
1460 *
1461 * An application using the TCP-style socket can use this option to
1462 * perform the SCTP ABORT primitive. The linger option structure is:
1463 *
1464 * struct linger {
1465 * int l_onoff; // option on/off
1466 * int l_linger; // linger time
1467 * };
1468 *
1469 * To enable the option, set l_onoff to 1. If the l_linger value is set
1470 * to 0, calling close() is the same as the ABORT primitive. If the
1471 * value is set to a negative value, the setsockopt() call will return
1472 * an error. If the value is set to a positive value linger_time, the
1473 * close() can be blocked for at most linger_time ms. If the graceful
1474 * shutdown phase does not finish during this period, close() will
1475 * return but the graceful shutdown phase continues in the system.
1476 */
1477static void sctp_close(struct sock *sk, long timeout)
1478{
1479 struct net *net = sock_net(sk);
1480 struct sctp_endpoint *ep;
1481 struct sctp_association *asoc;
1482 struct list_head *pos, *temp;
1483 unsigned int data_was_unread;
1484
1485 pr_debug("%s: sk:%p, timeout:%ld\n", __func__, sk, timeout);
1486
1487 lock_sock(sk);
1488 sk->sk_shutdown = SHUTDOWN_MASK;
1489 sk->sk_state = SCTP_SS_CLOSING;
1490
1491 ep = sctp_sk(sk)->ep;
1492
1493 /* Clean up any skbs sitting on the receive queue. */
1494 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1495 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1496
1497 /* Walk all associations on an endpoint. */
1498 list_for_each_safe(pos, temp, &ep->asocs) {
1499 asoc = list_entry(pos, struct sctp_association, asocs);
1500
1501 if (sctp_style(sk, TCP)) {
1502 /* A closed association can still be in the list if
1503 * it belongs to a TCP-style listening socket that is
1504 * not yet accepted. If so, free it. If not, send an
1505 * ABORT or SHUTDOWN based on the linger options.
1506 */
1507 if (sctp_state(asoc, CLOSED)) {
1508 sctp_unhash_established(asoc);
1509 sctp_association_free(asoc);
1510 continue;
1511 }
1512 }
1513
1514 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1515 !skb_queue_empty(&asoc->ulpq.reasm) ||
1516 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1517 struct sctp_chunk *chunk;
1518
1519 chunk = sctp_make_abort_user(asoc, NULL, 0);
1520 if (chunk)
1521 sctp_primitive_ABORT(net, asoc, chunk);
1522 } else
1523 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1524 }
1525
1526 /* On a TCP-style socket, block for at most linger_time if set. */
1527 if (sctp_style(sk, TCP) && timeout)
1528 sctp_wait_for_close(sk, timeout);
1529
1530 /* This will run the backlog queue. */
1531 release_sock(sk);
1532
1533 /* Supposedly, no process has access to the socket, but
1534 * the net layers still may.
1535 */
1536 local_bh_disable();
1537 bh_lock_sock(sk);
1538
1539 /* Hold the sock, since sk_common_release() will put sock_put()
1540 * and we have just a little more cleanup.
1541 */
1542 sock_hold(sk);
1543 sk_common_release(sk);
1544
1545 bh_unlock_sock(sk);
1546 local_bh_enable();
1547
1548 sock_put(sk);
1549
1550 SCTP_DBG_OBJCNT_DEC(sock);
1551}
1552
1553/* Handle EPIPE error. */
1554static int sctp_error(struct sock *sk, int flags, int err)
1555{
1556 if (err == -EPIPE)
1557 err = sock_error(sk) ? : -EPIPE;
1558 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1559 send_sig(SIGPIPE, current, 0);
1560 return err;
1561}
1562
1563/* API 3.1.3 sendmsg() - UDP Style Syntax
1564 *
1565 * An application uses sendmsg() and recvmsg() calls to transmit data to
1566 * and receive data from its peer.
1567 *
1568 * ssize_t sendmsg(int socket, const struct msghdr *message,
1569 * int flags);
1570 *
1571 * socket - the socket descriptor of the endpoint.
1572 * message - pointer to the msghdr structure which contains a single
1573 * user message and possibly some ancillary data.
1574 *
1575 * See Section 5 for complete description of the data
1576 * structures.
1577 *
1578 * flags - flags sent or received with the user message, see Section
1579 * 5 for complete description of the flags.
1580 *
1581 * Note: This function could use a rewrite especially when explicit
1582 * connect support comes in.
1583 */
1584/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1585
1586static int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1587
1588static int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1589 struct msghdr *msg, size_t msg_len)
1590{
1591 struct net *net = sock_net(sk);
1592 struct sctp_sock *sp;
1593 struct sctp_endpoint *ep;
1594 struct sctp_association *new_asoc = NULL, *asoc = NULL;
1595 struct sctp_transport *transport, *chunk_tp;
1596 struct sctp_chunk *chunk;
1597 union sctp_addr to;
1598 struct sockaddr *msg_name = NULL;
1599 struct sctp_sndrcvinfo default_sinfo;
1600 struct sctp_sndrcvinfo *sinfo;
1601 struct sctp_initmsg *sinit;
1602 sctp_assoc_t associd = 0;
1603 sctp_cmsgs_t cmsgs = { NULL };
1604 int err;
1605 sctp_scope_t scope;
1606 long timeo;
1607 __u16 sinfo_flags = 0;
1608 struct sctp_datamsg *datamsg;
1609 int msg_flags = msg->msg_flags;
1610
1611 err = 0;
1612 sp = sctp_sk(sk);
1613 ep = sp->ep;
1614
1615 pr_debug("%s: sk:%p, msg:%p, msg_len:%zu ep:%p\n", __func__, sk,
1616 msg, msg_len, ep);
1617
1618 /* We cannot send a message over a TCP-style listening socket. */
1619 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1620 err = -EPIPE;
1621 goto out_nounlock;
1622 }
1623
1624 /* Parse out the SCTP CMSGs. */
1625 err = sctp_msghdr_parse(msg, &cmsgs);
1626 if (err) {
1627 pr_debug("%s: msghdr parse err:%x\n", __func__, err);
1628 goto out_nounlock;
1629 }
1630
1631 /* Fetch the destination address for this packet. This
1632 * address only selects the association--it is not necessarily
1633 * the address we will send to.
1634 * For a peeled-off socket, msg_name is ignored.
1635 */
1636 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1637 int msg_namelen = msg->msg_namelen;
1638
1639 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1640 msg_namelen);
1641 if (err)
1642 return err;
1643
1644 if (msg_namelen > sizeof(to))
1645 msg_namelen = sizeof(to);
1646 memcpy(&to, msg->msg_name, msg_namelen);
1647 msg_name = msg->msg_name;
1648 }
1649
1650 sinfo = cmsgs.info;
1651 sinit = cmsgs.init;
1652
1653 /* Did the user specify SNDRCVINFO? */
1654 if (sinfo) {
1655 sinfo_flags = sinfo->sinfo_flags;
1656 associd = sinfo->sinfo_assoc_id;
1657 }
1658
1659 pr_debug("%s: msg_len:%zu, sinfo_flags:0x%x\n", __func__,
1660 msg_len, sinfo_flags);
1661
1662 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1663 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1664 err = -EINVAL;
1665 goto out_nounlock;
1666 }
1667
1668 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1669 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1670 * If SCTP_ABORT is set, the message length could be non zero with
1671 * the msg_iov set to the user abort reason.
1672 */
1673 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1674 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1675 err = -EINVAL;
1676 goto out_nounlock;
1677 }
1678
1679 /* If SCTP_ADDR_OVER is set, there must be an address
1680 * specified in msg_name.
1681 */
1682 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1683 err = -EINVAL;
1684 goto out_nounlock;
1685 }
1686
1687 transport = NULL;
1688
1689 pr_debug("%s: about to look up association\n", __func__);
1690
1691 lock_sock(sk);
1692
1693 /* If a msg_name has been specified, assume this is to be used. */
1694 if (msg_name) {
1695 /* Look for a matching association on the endpoint. */
1696 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1697 if (!asoc) {
1698 /* If we could not find a matching association on the
1699 * endpoint, make sure that it is not a TCP-style
1700 * socket that already has an association or there is
1701 * no peeled-off association on another socket.
1702 */
1703 if ((sctp_style(sk, TCP) &&
1704 sctp_sstate(sk, ESTABLISHED)) ||
1705 sctp_endpoint_is_peeled_off(ep, &to)) {
1706 err = -EADDRNOTAVAIL;
1707 goto out_unlock;
1708 }
1709 }
1710 } else {
1711 asoc = sctp_id2assoc(sk, associd);
1712 if (!asoc) {
1713 err = -EPIPE;
1714 goto out_unlock;
1715 }
1716 }
1717
1718 if (asoc) {
1719 pr_debug("%s: just looked up association:%p\n", __func__, asoc);
1720
1721 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1722 * socket that has an association in CLOSED state. This can
1723 * happen when an accepted socket has an association that is
1724 * already CLOSED.
1725 */
1726 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1727 err = -EPIPE;
1728 goto out_unlock;
1729 }
1730
1731 if (sinfo_flags & SCTP_EOF) {
1732 pr_debug("%s: shutting down association:%p\n",
1733 __func__, asoc);
1734
1735 sctp_primitive_SHUTDOWN(net, asoc, NULL);
1736 err = 0;
1737 goto out_unlock;
1738 }
1739 if (sinfo_flags & SCTP_ABORT) {
1740
1741 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1742 if (!chunk) {
1743 err = -ENOMEM;
1744 goto out_unlock;
1745 }
1746
1747 pr_debug("%s: aborting association:%p\n",
1748 __func__, asoc);
1749
1750 sctp_primitive_ABORT(net, asoc, chunk);
1751 err = 0;
1752 goto out_unlock;
1753 }
1754 }
1755
1756 /* Do we need to create the association? */
1757 if (!asoc) {
1758 pr_debug("%s: there is no association yet\n", __func__);
1759
1760 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1761 err = -EINVAL;
1762 goto out_unlock;
1763 }
1764
1765 /* Check for invalid stream against the stream counts,
1766 * either the default or the user specified stream counts.
1767 */
1768 if (sinfo) {
1769 if (!sinit || !sinit->sinit_num_ostreams) {
1770 /* Check against the defaults. */
1771 if (sinfo->sinfo_stream >=
1772 sp->initmsg.sinit_num_ostreams) {
1773 err = -EINVAL;
1774 goto out_unlock;
1775 }
1776 } else {
1777 /* Check against the requested. */
1778 if (sinfo->sinfo_stream >=
1779 sinit->sinit_num_ostreams) {
1780 err = -EINVAL;
1781 goto out_unlock;
1782 }
1783 }
1784 }
1785
1786 /*
1787 * API 3.1.2 bind() - UDP Style Syntax
1788 * If a bind() or sctp_bindx() is not called prior to a
1789 * sendmsg() call that initiates a new association, the
1790 * system picks an ephemeral port and will choose an address
1791 * set equivalent to binding with a wildcard address.
1792 */
1793 if (!ep->base.bind_addr.port) {
1794 if (sctp_autobind(sk)) {
1795 err = -EAGAIN;
1796 goto out_unlock;
1797 }
1798 } else {
1799 /*
1800 * If an unprivileged user inherits a one-to-many
1801 * style socket with open associations on a privileged
1802 * port, it MAY be permitted to accept new associations,
1803 * but it SHOULD NOT be permitted to open new
1804 * associations.
1805 */
1806 if (ep->base.bind_addr.port < PROT_SOCK &&
1807 !ns_capable(net->user_ns, CAP_NET_BIND_SERVICE)) {
1808 err = -EACCES;
1809 goto out_unlock;
1810 }
1811 }
1812
1813 scope = sctp_scope(&to);
1814 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1815 if (!new_asoc) {
1816 err = -ENOMEM;
1817 goto out_unlock;
1818 }
1819 asoc = new_asoc;
1820 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1821 if (err < 0) {
1822 err = -ENOMEM;
1823 goto out_free;
1824 }
1825
1826 /* If the SCTP_INIT ancillary data is specified, set all
1827 * the association init values accordingly.
1828 */
1829 if (sinit) {
1830 if (sinit->sinit_num_ostreams) {
1831 asoc->c.sinit_num_ostreams =
1832 sinit->sinit_num_ostreams;
1833 }
1834 if (sinit->sinit_max_instreams) {
1835 asoc->c.sinit_max_instreams =
1836 sinit->sinit_max_instreams;
1837 }
1838 if (sinit->sinit_max_attempts) {
1839 asoc->max_init_attempts
1840 = sinit->sinit_max_attempts;
1841 }
1842 if (sinit->sinit_max_init_timeo) {
1843 asoc->max_init_timeo =
1844 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1845 }
1846 }
1847
1848 /* Prime the peer's transport structures. */
1849 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1850 if (!transport) {
1851 err = -ENOMEM;
1852 goto out_free;
1853 }
1854 }
1855
1856 /* ASSERT: we have a valid association at this point. */
1857 pr_debug("%s: we have a valid association\n", __func__);
1858
1859 if (!sinfo) {
1860 /* If the user didn't specify SNDRCVINFO, make up one with
1861 * some defaults.
1862 */
1863 memset(&default_sinfo, 0, sizeof(default_sinfo));
1864 default_sinfo.sinfo_stream = asoc->default_stream;
1865 default_sinfo.sinfo_flags = asoc->default_flags;
1866 default_sinfo.sinfo_ppid = asoc->default_ppid;
1867 default_sinfo.sinfo_context = asoc->default_context;
1868 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1869 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1870 sinfo = &default_sinfo;
1871 }
1872
1873 /* API 7.1.7, the sndbuf size per association bounds the
1874 * maximum size of data that can be sent in a single send call.
1875 */
1876 if (msg_len > sk->sk_sndbuf) {
1877 err = -EMSGSIZE;
1878 goto out_free;
1879 }
1880
1881 if (asoc->pmtu_pending)
1882 sctp_assoc_pending_pmtu(sk, asoc);
1883
1884 /* If fragmentation is disabled and the message length exceeds the
1885 * association fragmentation point, return EMSGSIZE. The I-D
1886 * does not specify what this error is, but this looks like
1887 * a great fit.
1888 */
1889 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1890 err = -EMSGSIZE;
1891 goto out_free;
1892 }
1893
1894 /* Check for invalid stream. */
1895 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1896 err = -EINVAL;
1897 goto out_free;
1898 }
1899
1900 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1901 if (!sctp_wspace(asoc)) {
1902 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1903 if (err)
1904 goto out_free;
1905 }
1906
1907 /* If an address is passed with the sendto/sendmsg call, it is used
1908 * to override the primary destination address in the TCP model, or
1909 * when SCTP_ADDR_OVER flag is set in the UDP model.
1910 */
1911 if ((sctp_style(sk, TCP) && msg_name) ||
1912 (sinfo_flags & SCTP_ADDR_OVER)) {
1913 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1914 if (!chunk_tp) {
1915 err = -EINVAL;
1916 goto out_free;
1917 }
1918 } else
1919 chunk_tp = NULL;
1920
1921 /* Auto-connect, if we aren't connected already. */
1922 if (sctp_state(asoc, CLOSED)) {
1923 err = sctp_primitive_ASSOCIATE(net, asoc, NULL);
1924 if (err < 0)
1925 goto out_free;
1926
1927 pr_debug("%s: we associated primitively\n", __func__);
1928 }
1929
1930 /* Break the message into multiple chunks of maximum size. */
1931 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1932 if (IS_ERR(datamsg)) {
1933 err = PTR_ERR(datamsg);
1934 goto out_free;
1935 }
1936
1937 /* Now send the (possibly) fragmented message. */
1938 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1939 sctp_chunk_hold(chunk);
1940
1941 /* Do accounting for the write space. */
1942 sctp_set_owner_w(chunk);
1943
1944 chunk->transport = chunk_tp;
1945 }
1946
1947 /* Send it to the lower layers. Note: all chunks
1948 * must either fail or succeed. The lower layer
1949 * works that way today. Keep it that way or this
1950 * breaks.
1951 */
1952 err = sctp_primitive_SEND(net, asoc, datamsg);
1953 /* Did the lower layer accept the chunk? */
1954 if (err) {
1955 sctp_datamsg_free(datamsg);
1956 goto out_free;
1957 }
1958
1959 pr_debug("%s: we sent primitively\n", __func__);
1960
1961 sctp_datamsg_put(datamsg);
1962 err = msg_len;
1963
1964 /* If we are already past ASSOCIATE, the lower
1965 * layers are responsible for association cleanup.
1966 */
1967 goto out_unlock;
1968
1969out_free:
1970 if (new_asoc) {
1971 sctp_unhash_established(asoc);
1972 sctp_association_free(asoc);
1973 }
1974out_unlock:
1975 release_sock(sk);
1976
1977out_nounlock:
1978 return sctp_error(sk, msg_flags, err);
1979
1980#if 0
1981do_sock_err:
1982 if (msg_len)
1983 err = msg_len;
1984 else
1985 err = sock_error(sk);
1986 goto out;
1987
1988do_interrupted:
1989 if (msg_len)
1990 err = msg_len;
1991 goto out;
1992#endif /* 0 */
1993}
1994
1995/* This is an extended version of skb_pull() that removes the data from the
1996 * start of a skb even when data is spread across the list of skb's in the
1997 * frag_list. len specifies the total amount of data that needs to be removed.
1998 * when 'len' bytes could be removed from the skb, it returns 0.
1999 * If 'len' exceeds the total skb length, it returns the no. of bytes that
2000 * could not be removed.
2001 */
2002static int sctp_skb_pull(struct sk_buff *skb, int len)
2003{
2004 struct sk_buff *list;
2005 int skb_len = skb_headlen(skb);
2006 int rlen;
2007
2008 if (len <= skb_len) {
2009 __skb_pull(skb, len);
2010 return 0;
2011 }
2012 len -= skb_len;
2013 __skb_pull(skb, skb_len);
2014
2015 skb_walk_frags(skb, list) {
2016 rlen = sctp_skb_pull(list, len);
2017 skb->len -= (len-rlen);
2018 skb->data_len -= (len-rlen);
2019
2020 if (!rlen)
2021 return 0;
2022
2023 len = rlen;
2024 }
2025
2026 return len;
2027}
2028
2029/* API 3.1.3 recvmsg() - UDP Style Syntax
2030 *
2031 * ssize_t recvmsg(int socket, struct msghdr *message,
2032 * int flags);
2033 *
2034 * socket - the socket descriptor of the endpoint.
2035 * message - pointer to the msghdr structure which contains a single
2036 * user message and possibly some ancillary data.
2037 *
2038 * See Section 5 for complete description of the data
2039 * structures.
2040 *
2041 * flags - flags sent or received with the user message, see Section
2042 * 5 for complete description of the flags.
2043 */
2044static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2045
2046static int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2047 struct msghdr *msg, size_t len, int noblock,
2048 int flags, int *addr_len)
2049{
2050 struct sctp_ulpevent *event = NULL;
2051 struct sctp_sock *sp = sctp_sk(sk);
2052 struct sk_buff *skb;
2053 int copied;
2054 int err = 0;
2055 int skb_len;
2056
2057 pr_debug("%s: sk:%p, msghdr:%p, len:%zd, noblock:%d, flags:0x%x, "
2058 "addr_len:%p)\n", __func__, sk, msg, len, noblock, flags,
2059 addr_len);
2060
2061 lock_sock(sk);
2062
2063 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2064 err = -ENOTCONN;
2065 goto out;
2066 }
2067
2068 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2069 if (!skb)
2070 goto out;
2071
2072 /* Get the total length of the skb including any skb's in the
2073 * frag_list.
2074 */
2075 skb_len = skb->len;
2076
2077 copied = skb_len;
2078 if (copied > len)
2079 copied = len;
2080
2081 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2082
2083 event = sctp_skb2event(skb);
2084
2085 if (err)
2086 goto out_free;
2087
2088 sock_recv_ts_and_drops(msg, sk, skb);
2089 if (sctp_ulpevent_is_notification(event)) {
2090 msg->msg_flags |= MSG_NOTIFICATION;
2091 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2092 } else {
2093 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2094 }
2095
2096 /* Check if we allow SCTP_SNDRCVINFO. */
2097 if (sp->subscribe.sctp_data_io_event)
2098 sctp_ulpevent_read_sndrcvinfo(event, msg);
2099#if 0
2100 /* FIXME: we should be calling IP/IPv6 layers. */
2101 if (sk->sk_protinfo.af_inet.cmsg_flags)
2102 ip_cmsg_recv(msg, skb);
2103#endif
2104
2105 err = copied;
2106
2107 /* If skb's length exceeds the user's buffer, update the skb and
2108 * push it back to the receive_queue so that the next call to
2109 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2110 */
2111 if (skb_len > copied) {
2112 msg->msg_flags &= ~MSG_EOR;
2113 if (flags & MSG_PEEK)
2114 goto out_free;
2115 sctp_skb_pull(skb, copied);
2116 skb_queue_head(&sk->sk_receive_queue, skb);
2117
2118 /* When only partial message is copied to the user, increase
2119 * rwnd by that amount. If all the data in the skb is read,
2120 * rwnd is updated when the event is freed.
2121 */
2122 if (!sctp_ulpevent_is_notification(event))
2123 sctp_assoc_rwnd_increase(event->asoc, copied);
2124 goto out;
2125 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2126 (event->msg_flags & MSG_EOR))
2127 msg->msg_flags |= MSG_EOR;
2128 else
2129 msg->msg_flags &= ~MSG_EOR;
2130
2131out_free:
2132 if (flags & MSG_PEEK) {
2133 /* Release the skb reference acquired after peeking the skb in
2134 * sctp_skb_recv_datagram().
2135 */
2136 kfree_skb(skb);
2137 } else {
2138 /* Free the event which includes releasing the reference to
2139 * the owner of the skb, freeing the skb and updating the
2140 * rwnd.
2141 */
2142 sctp_ulpevent_free(event);
2143 }
2144out:
2145 release_sock(sk);
2146 return err;
2147}
2148
2149/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2150 *
2151 * This option is a on/off flag. If enabled no SCTP message
2152 * fragmentation will be performed. Instead if a message being sent
2153 * exceeds the current PMTU size, the message will NOT be sent and
2154 * instead a error will be indicated to the user.
2155 */
2156static int sctp_setsockopt_disable_fragments(struct sock *sk,
2157 char __user *optval,
2158 unsigned int optlen)
2159{
2160 int val;
2161
2162 if (optlen < sizeof(int))
2163 return -EINVAL;
2164
2165 if (get_user(val, (int __user *)optval))
2166 return -EFAULT;
2167
2168 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2169
2170 return 0;
2171}
2172
2173static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2174 unsigned int optlen)
2175{
2176 struct sctp_association *asoc;
2177 struct sctp_ulpevent *event;
2178
2179 if (optlen > sizeof(struct sctp_event_subscribe))
2180 return -EINVAL;
2181 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2182 return -EFAULT;
2183
2184 /*
2185 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2186 * if there is no data to be sent or retransmit, the stack will
2187 * immediately send up this notification.
2188 */
2189 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2190 &sctp_sk(sk)->subscribe)) {
2191 asoc = sctp_id2assoc(sk, 0);
2192
2193 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2194 event = sctp_ulpevent_make_sender_dry_event(asoc,
2195 GFP_ATOMIC);
2196 if (!event)
2197 return -ENOMEM;
2198
2199 sctp_ulpq_tail_event(&asoc->ulpq, event);
2200 }
2201 }
2202
2203 return 0;
2204}
2205
2206/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2207 *
2208 * This socket option is applicable to the UDP-style socket only. When
2209 * set it will cause associations that are idle for more than the
2210 * specified number of seconds to automatically close. An association
2211 * being idle is defined an association that has NOT sent or received
2212 * user data. The special value of '0' indicates that no automatic
2213 * close of any associations should be performed. The option expects an
2214 * integer defining the number of seconds of idle time before an
2215 * association is closed.
2216 */
2217static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2218 unsigned int optlen)
2219{
2220 struct sctp_sock *sp = sctp_sk(sk);
2221 struct net *net = sock_net(sk);
2222
2223 /* Applicable to UDP-style socket only */
2224 if (sctp_style(sk, TCP))
2225 return -EOPNOTSUPP;
2226 if (optlen != sizeof(int))
2227 return -EINVAL;
2228 if (copy_from_user(&sp->autoclose, optval, optlen))
2229 return -EFAULT;
2230
2231 if (sp->autoclose > net->sctp.max_autoclose)
2232 sp->autoclose = net->sctp.max_autoclose;
2233
2234 return 0;
2235}
2236
2237/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2238 *
2239 * Applications can enable or disable heartbeats for any peer address of
2240 * an association, modify an address's heartbeat interval, force a
2241 * heartbeat to be sent immediately, and adjust the address's maximum
2242 * number of retransmissions sent before an address is considered
2243 * unreachable. The following structure is used to access and modify an
2244 * address's parameters:
2245 *
2246 * struct sctp_paddrparams {
2247 * sctp_assoc_t spp_assoc_id;
2248 * struct sockaddr_storage spp_address;
2249 * uint32_t spp_hbinterval;
2250 * uint16_t spp_pathmaxrxt;
2251 * uint32_t spp_pathmtu;
2252 * uint32_t spp_sackdelay;
2253 * uint32_t spp_flags;
2254 * };
2255 *
2256 * spp_assoc_id - (one-to-many style socket) This is filled in the
2257 * application, and identifies the association for
2258 * this query.
2259 * spp_address - This specifies which address is of interest.
2260 * spp_hbinterval - This contains the value of the heartbeat interval,
2261 * in milliseconds. If a value of zero
2262 * is present in this field then no changes are to
2263 * be made to this parameter.
2264 * spp_pathmaxrxt - This contains the maximum number of
2265 * retransmissions before this address shall be
2266 * considered unreachable. If a value of zero
2267 * is present in this field then no changes are to
2268 * be made to this parameter.
2269 * spp_pathmtu - When Path MTU discovery is disabled the value
2270 * specified here will be the "fixed" path mtu.
2271 * Note that if the spp_address field is empty
2272 * then all associations on this address will
2273 * have this fixed path mtu set upon them.
2274 *
2275 * spp_sackdelay - When delayed sack is enabled, this value specifies
2276 * the number of milliseconds that sacks will be delayed
2277 * for. This value will apply to all addresses of an
2278 * association if the spp_address field is empty. Note
2279 * also, that if delayed sack is enabled and this
2280 * value is set to 0, no change is made to the last
2281 * recorded delayed sack timer value.
2282 *
2283 * spp_flags - These flags are used to control various features
2284 * on an association. The flag field may contain
2285 * zero or more of the following options.
2286 *
2287 * SPP_HB_ENABLE - Enable heartbeats on the
2288 * specified address. Note that if the address
2289 * field is empty all addresses for the association
2290 * have heartbeats enabled upon them.
2291 *
2292 * SPP_HB_DISABLE - Disable heartbeats on the
2293 * speicifed address. Note that if the address
2294 * field is empty all addresses for the association
2295 * will have their heartbeats disabled. Note also
2296 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2297 * mutually exclusive, only one of these two should
2298 * be specified. Enabling both fields will have
2299 * undetermined results.
2300 *
2301 * SPP_HB_DEMAND - Request a user initiated heartbeat
2302 * to be made immediately.
2303 *
2304 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2305 * heartbeat delayis to be set to the value of 0
2306 * milliseconds.
2307 *
2308 * SPP_PMTUD_ENABLE - This field will enable PMTU
2309 * discovery upon the specified address. Note that
2310 * if the address feild is empty then all addresses
2311 * on the association are effected.
2312 *
2313 * SPP_PMTUD_DISABLE - This field will disable PMTU
2314 * discovery upon the specified address. Note that
2315 * if the address feild is empty then all addresses
2316 * on the association are effected. Not also that
2317 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2318 * exclusive. Enabling both will have undetermined
2319 * results.
2320 *
2321 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2322 * on delayed sack. The time specified in spp_sackdelay
2323 * is used to specify the sack delay for this address. Note
2324 * that if spp_address is empty then all addresses will
2325 * enable delayed sack and take on the sack delay
2326 * value specified in spp_sackdelay.
2327 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2328 * off delayed sack. If the spp_address field is blank then
2329 * delayed sack is disabled for the entire association. Note
2330 * also that this field is mutually exclusive to
2331 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2332 * results.
2333 */
2334static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2335 struct sctp_transport *trans,
2336 struct sctp_association *asoc,
2337 struct sctp_sock *sp,
2338 int hb_change,
2339 int pmtud_change,
2340 int sackdelay_change)
2341{
2342 int error;
2343
2344 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2345 struct net *net = sock_net(trans->asoc->base.sk);
2346
2347 error = sctp_primitive_REQUESTHEARTBEAT(net, trans->asoc, trans);
2348 if (error)
2349 return error;
2350 }
2351
2352 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2353 * this field is ignored. Note also that a value of zero indicates
2354 * the current setting should be left unchanged.
2355 */
2356 if (params->spp_flags & SPP_HB_ENABLE) {
2357
2358 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2359 * set. This lets us use 0 value when this flag
2360 * is set.
2361 */
2362 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2363 params->spp_hbinterval = 0;
2364
2365 if (params->spp_hbinterval ||
2366 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2367 if (trans) {
2368 trans->hbinterval =
2369 msecs_to_jiffies(params->spp_hbinterval);
2370 } else if (asoc) {
2371 asoc->hbinterval =
2372 msecs_to_jiffies(params->spp_hbinterval);
2373 } else {
2374 sp->hbinterval = params->spp_hbinterval;
2375 }
2376 }
2377 }
2378
2379 if (hb_change) {
2380 if (trans) {
2381 trans->param_flags =
2382 (trans->param_flags & ~SPP_HB) | hb_change;
2383 } else if (asoc) {
2384 asoc->param_flags =
2385 (asoc->param_flags & ~SPP_HB) | hb_change;
2386 } else {
2387 sp->param_flags =
2388 (sp->param_flags & ~SPP_HB) | hb_change;
2389 }
2390 }
2391
2392 /* When Path MTU discovery is disabled the value specified here will
2393 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2394 * include the flag SPP_PMTUD_DISABLE for this field to have any
2395 * effect).
2396 */
2397 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2398 if (trans) {
2399 trans->pathmtu = params->spp_pathmtu;
2400 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2401 } else if (asoc) {
2402 asoc->pathmtu = params->spp_pathmtu;
2403 sctp_frag_point(asoc, params->spp_pathmtu);
2404 } else {
2405 sp->pathmtu = params->spp_pathmtu;
2406 }
2407 }
2408
2409 if (pmtud_change) {
2410 if (trans) {
2411 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2412 (params->spp_flags & SPP_PMTUD_ENABLE);
2413 trans->param_flags =
2414 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2415 if (update) {
2416 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2417 sctp_assoc_sync_pmtu(sctp_opt2sk(sp), asoc);
2418 }
2419 } else if (asoc) {
2420 asoc->param_flags =
2421 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2422 } else {
2423 sp->param_flags =
2424 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2425 }
2426 }
2427
2428 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2429 * value of this field is ignored. Note also that a value of zero
2430 * indicates the current setting should be left unchanged.
2431 */
2432 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2433 if (trans) {
2434 trans->sackdelay =
2435 msecs_to_jiffies(params->spp_sackdelay);
2436 } else if (asoc) {
2437 asoc->sackdelay =
2438 msecs_to_jiffies(params->spp_sackdelay);
2439 } else {
2440 sp->sackdelay = params->spp_sackdelay;
2441 }
2442 }
2443
2444 if (sackdelay_change) {
2445 if (trans) {
2446 trans->param_flags =
2447 (trans->param_flags & ~SPP_SACKDELAY) |
2448 sackdelay_change;
2449 } else if (asoc) {
2450 asoc->param_flags =
2451 (asoc->param_flags & ~SPP_SACKDELAY) |
2452 sackdelay_change;
2453 } else {
2454 sp->param_flags =
2455 (sp->param_flags & ~SPP_SACKDELAY) |
2456 sackdelay_change;
2457 }
2458 }
2459
2460 /* Note that a value of zero indicates the current setting should be
2461 left unchanged.
2462 */
2463 if (params->spp_pathmaxrxt) {
2464 if (trans) {
2465 trans->pathmaxrxt = params->spp_pathmaxrxt;
2466 } else if (asoc) {
2467 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2468 } else {
2469 sp->pathmaxrxt = params->spp_pathmaxrxt;
2470 }
2471 }
2472
2473 return 0;
2474}
2475
2476static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2477 char __user *optval,
2478 unsigned int optlen)
2479{
2480 struct sctp_paddrparams params;
2481 struct sctp_transport *trans = NULL;
2482 struct sctp_association *asoc = NULL;
2483 struct sctp_sock *sp = sctp_sk(sk);
2484 int error;
2485 int hb_change, pmtud_change, sackdelay_change;
2486
2487 if (optlen != sizeof(struct sctp_paddrparams))
2488 return -EINVAL;
2489
2490 if (copy_from_user(¶ms, optval, optlen))
2491 return -EFAULT;
2492
2493 /* Validate flags and value parameters. */
2494 hb_change = params.spp_flags & SPP_HB;
2495 pmtud_change = params.spp_flags & SPP_PMTUD;
2496 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2497
2498 if (hb_change == SPP_HB ||
2499 pmtud_change == SPP_PMTUD ||
2500 sackdelay_change == SPP_SACKDELAY ||
2501 params.spp_sackdelay > 500 ||
2502 (params.spp_pathmtu &&
2503 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2504 return -EINVAL;
2505
2506 /* If an address other than INADDR_ANY is specified, and
2507 * no transport is found, then the request is invalid.
2508 */
2509 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
2510 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2511 params.spp_assoc_id);
2512 if (!trans)
2513 return -EINVAL;
2514 }
2515
2516 /* Get association, if assoc_id != 0 and the socket is a one
2517 * to many style socket, and an association was not found, then
2518 * the id was invalid.
2519 */
2520 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2521 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2522 return -EINVAL;
2523
2524 /* Heartbeat demand can only be sent on a transport or
2525 * association, but not a socket.
2526 */
2527 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2528 return -EINVAL;
2529
2530 /* Process parameters. */
2531 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2532 hb_change, pmtud_change,
2533 sackdelay_change);
2534
2535 if (error)
2536 return error;
2537
2538 /* If changes are for association, also apply parameters to each
2539 * transport.
2540 */
2541 if (!trans && asoc) {
2542 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2543 transports) {
2544 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2545 hb_change, pmtud_change,
2546 sackdelay_change);
2547 }
2548 }
2549
2550 return 0;
2551}
2552
2553static inline __u32 sctp_spp_sackdelay_enable(__u32 param_flags)
2554{
2555 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_ENABLE;
2556}
2557
2558static inline __u32 sctp_spp_sackdelay_disable(__u32 param_flags)
2559{
2560 return (param_flags & ~SPP_SACKDELAY) | SPP_SACKDELAY_DISABLE;
2561}
2562
2563/*
2564 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2565 *
2566 * This option will effect the way delayed acks are performed. This
2567 * option allows you to get or set the delayed ack time, in
2568 * milliseconds. It also allows changing the delayed ack frequency.
2569 * Changing the frequency to 1 disables the delayed sack algorithm. If
2570 * the assoc_id is 0, then this sets or gets the endpoints default
2571 * values. If the assoc_id field is non-zero, then the set or get
2572 * effects the specified association for the one to many model (the
2573 * assoc_id field is ignored by the one to one model). Note that if
2574 * sack_delay or sack_freq are 0 when setting this option, then the
2575 * current values will remain unchanged.
2576 *
2577 * struct sctp_sack_info {
2578 * sctp_assoc_t sack_assoc_id;
2579 * uint32_t sack_delay;
2580 * uint32_t sack_freq;
2581 * };
2582 *
2583 * sack_assoc_id - This parameter, indicates which association the user
2584 * is performing an action upon. Note that if this field's value is
2585 * zero then the endpoints default value is changed (effecting future
2586 * associations only).
2587 *
2588 * sack_delay - This parameter contains the number of milliseconds that
2589 * the user is requesting the delayed ACK timer be set to. Note that
2590 * this value is defined in the standard to be between 200 and 500
2591 * milliseconds.
2592 *
2593 * sack_freq - This parameter contains the number of packets that must
2594 * be received before a sack is sent without waiting for the delay
2595 * timer to expire. The default value for this is 2, setting this
2596 * value to 1 will disable the delayed sack algorithm.
2597 */
2598
2599static int sctp_setsockopt_delayed_ack(struct sock *sk,
2600 char __user *optval, unsigned int optlen)
2601{
2602 struct sctp_sack_info params;
2603 struct sctp_transport *trans = NULL;
2604 struct sctp_association *asoc = NULL;
2605 struct sctp_sock *sp = sctp_sk(sk);
2606
2607 if (optlen == sizeof(struct sctp_sack_info)) {
2608 if (copy_from_user(¶ms, optval, optlen))
2609 return -EFAULT;
2610
2611 if (params.sack_delay == 0 && params.sack_freq == 0)
2612 return 0;
2613 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2614 pr_warn_ratelimited(DEPRECATED
2615 "%s (pid %d) "
2616 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
2617 "Use struct sctp_sack_info instead\n",
2618 current->comm, task_pid_nr(current));
2619 if (copy_from_user(¶ms, optval, optlen))
2620 return -EFAULT;
2621
2622 if (params.sack_delay == 0)
2623 params.sack_freq = 1;
2624 else
2625 params.sack_freq = 0;
2626 } else
2627 return -EINVAL;
2628
2629 /* Validate value parameter. */
2630 if (params.sack_delay > 500)
2631 return -EINVAL;
2632
2633 /* Get association, if sack_assoc_id != 0 and the socket is a one
2634 * to many style socket, and an association was not found, then
2635 * the id was invalid.
2636 */
2637 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2638 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2639 return -EINVAL;
2640
2641 if (params.sack_delay) {
2642 if (asoc) {
2643 asoc->sackdelay =
2644 msecs_to_jiffies(params.sack_delay);
2645 asoc->param_flags =
2646 sctp_spp_sackdelay_enable(asoc->param_flags);
2647 } else {
2648 sp->sackdelay = params.sack_delay;
2649 sp->param_flags =
2650 sctp_spp_sackdelay_enable(sp->param_flags);
2651 }
2652 }
2653
2654 if (params.sack_freq == 1) {
2655 if (asoc) {
2656 asoc->param_flags =
2657 sctp_spp_sackdelay_disable(asoc->param_flags);
2658 } else {
2659 sp->param_flags =
2660 sctp_spp_sackdelay_disable(sp->param_flags);
2661 }
2662 } else if (params.sack_freq > 1) {
2663 if (asoc) {
2664 asoc->sackfreq = params.sack_freq;
2665 asoc->param_flags =
2666 sctp_spp_sackdelay_enable(asoc->param_flags);
2667 } else {
2668 sp->sackfreq = params.sack_freq;
2669 sp->param_flags =
2670 sctp_spp_sackdelay_enable(sp->param_flags);
2671 }
2672 }
2673
2674 /* If change is for association, also apply to each transport. */
2675 if (asoc) {
2676 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2677 transports) {
2678 if (params.sack_delay) {
2679 trans->sackdelay =
2680 msecs_to_jiffies(params.sack_delay);
2681 trans->param_flags =
2682 sctp_spp_sackdelay_enable(trans->param_flags);
2683 }
2684 if (params.sack_freq == 1) {
2685 trans->param_flags =
2686 sctp_spp_sackdelay_disable(trans->param_flags);
2687 } else if (params.sack_freq > 1) {
2688 trans->sackfreq = params.sack_freq;
2689 trans->param_flags =
2690 sctp_spp_sackdelay_enable(trans->param_flags);
2691 }
2692 }
2693 }
2694
2695 return 0;
2696}
2697
2698/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2699 *
2700 * Applications can specify protocol parameters for the default association
2701 * initialization. The option name argument to setsockopt() and getsockopt()
2702 * is SCTP_INITMSG.
2703 *
2704 * Setting initialization parameters is effective only on an unconnected
2705 * socket (for UDP-style sockets only future associations are effected
2706 * by the change). With TCP-style sockets, this option is inherited by
2707 * sockets derived from a listener socket.
2708 */
2709static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2710{
2711 struct sctp_initmsg sinit;
2712 struct sctp_sock *sp = sctp_sk(sk);
2713
2714 if (optlen != sizeof(struct sctp_initmsg))
2715 return -EINVAL;
2716 if (copy_from_user(&sinit, optval, optlen))
2717 return -EFAULT;
2718
2719 if (sinit.sinit_num_ostreams)
2720 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2721 if (sinit.sinit_max_instreams)
2722 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2723 if (sinit.sinit_max_attempts)
2724 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2725 if (sinit.sinit_max_init_timeo)
2726 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2727
2728 return 0;
2729}
2730
2731/*
2732 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2733 *
2734 * Applications that wish to use the sendto() system call may wish to
2735 * specify a default set of parameters that would normally be supplied
2736 * through the inclusion of ancillary data. This socket option allows
2737 * such an application to set the default sctp_sndrcvinfo structure.
2738 * The application that wishes to use this socket option simply passes
2739 * in to this call the sctp_sndrcvinfo structure defined in Section
2740 * 5.2.2) The input parameters accepted by this call include
2741 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2742 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2743 * to this call if the caller is using the UDP model.
2744 */
2745static int sctp_setsockopt_default_send_param(struct sock *sk,
2746 char __user *optval,
2747 unsigned int optlen)
2748{
2749 struct sctp_sndrcvinfo info;
2750 struct sctp_association *asoc;
2751 struct sctp_sock *sp = sctp_sk(sk);
2752
2753 if (optlen != sizeof(struct sctp_sndrcvinfo))
2754 return -EINVAL;
2755 if (copy_from_user(&info, optval, optlen))
2756 return -EFAULT;
2757
2758 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2759 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2760 return -EINVAL;
2761
2762 if (asoc) {
2763 asoc->default_stream = info.sinfo_stream;
2764 asoc->default_flags = info.sinfo_flags;
2765 asoc->default_ppid = info.sinfo_ppid;
2766 asoc->default_context = info.sinfo_context;
2767 asoc->default_timetolive = info.sinfo_timetolive;
2768 } else {
2769 sp->default_stream = info.sinfo_stream;
2770 sp->default_flags = info.sinfo_flags;
2771 sp->default_ppid = info.sinfo_ppid;
2772 sp->default_context = info.sinfo_context;
2773 sp->default_timetolive = info.sinfo_timetolive;
2774 }
2775
2776 return 0;
2777}
2778
2779/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2780 *
2781 * Requests that the local SCTP stack use the enclosed peer address as
2782 * the association primary. The enclosed address must be one of the
2783 * association peer's addresses.
2784 */
2785static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2786 unsigned int optlen)
2787{
2788 struct sctp_prim prim;
2789 struct sctp_transport *trans;
2790
2791 if (optlen != sizeof(struct sctp_prim))
2792 return -EINVAL;
2793
2794 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2795 return -EFAULT;
2796
2797 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2798 if (!trans)
2799 return -EINVAL;
2800
2801 sctp_assoc_set_primary(trans->asoc, trans);
2802
2803 return 0;
2804}
2805
2806/*
2807 * 7.1.5 SCTP_NODELAY
2808 *
2809 * Turn on/off any Nagle-like algorithm. This means that packets are
2810 * generally sent as soon as possible and no unnecessary delays are
2811 * introduced, at the cost of more packets in the network. Expects an
2812 * integer boolean flag.
2813 */
2814static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2815 unsigned int optlen)
2816{
2817 int val;
2818
2819 if (optlen < sizeof(int))
2820 return -EINVAL;
2821 if (get_user(val, (int __user *)optval))
2822 return -EFAULT;
2823
2824 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2825 return 0;
2826}
2827
2828/*
2829 *
2830 * 7.1.1 SCTP_RTOINFO
2831 *
2832 * The protocol parameters used to initialize and bound retransmission
2833 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2834 * and modify these parameters.
2835 * All parameters are time values, in milliseconds. A value of 0, when
2836 * modifying the parameters, indicates that the current value should not
2837 * be changed.
2838 *
2839 */
2840static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2841{
2842 struct sctp_rtoinfo rtoinfo;
2843 struct sctp_association *asoc;
2844 unsigned long rto_min, rto_max;
2845 struct sctp_sock *sp = sctp_sk(sk);
2846
2847 if (optlen != sizeof (struct sctp_rtoinfo))
2848 return -EINVAL;
2849
2850 if (copy_from_user(&rtoinfo, optval, optlen))
2851 return -EFAULT;
2852
2853 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2854
2855 /* Set the values to the specific association */
2856 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2857 return -EINVAL;
2858
2859 rto_max = rtoinfo.srto_max;
2860 rto_min = rtoinfo.srto_min;
2861
2862 if (rto_max)
2863 rto_max = asoc ? msecs_to_jiffies(rto_max) : rto_max;
2864 else
2865 rto_max = asoc ? asoc->rto_max : sp->rtoinfo.srto_max;
2866
2867 if (rto_min)
2868 rto_min = asoc ? msecs_to_jiffies(rto_min) : rto_min;
2869 else
2870 rto_min = asoc ? asoc->rto_min : sp->rtoinfo.srto_min;
2871
2872 if (rto_min > rto_max)
2873 return -EINVAL;
2874
2875 if (asoc) {
2876 if (rtoinfo.srto_initial != 0)
2877 asoc->rto_initial =
2878 msecs_to_jiffies(rtoinfo.srto_initial);
2879 asoc->rto_max = rto_max;
2880 asoc->rto_min = rto_min;
2881 } else {
2882 /* If there is no association or the association-id = 0
2883 * set the values to the endpoint.
2884 */
2885 if (rtoinfo.srto_initial != 0)
2886 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2887 sp->rtoinfo.srto_max = rto_max;
2888 sp->rtoinfo.srto_min = rto_min;
2889 }
2890
2891 return 0;
2892}
2893
2894/*
2895 *
2896 * 7.1.2 SCTP_ASSOCINFO
2897 *
2898 * This option is used to tune the maximum retransmission attempts
2899 * of the association.
2900 * Returns an error if the new association retransmission value is
2901 * greater than the sum of the retransmission value of the peer.
2902 * See [SCTP] for more information.
2903 *
2904 */
2905static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2906{
2907
2908 struct sctp_assocparams assocparams;
2909 struct sctp_association *asoc;
2910
2911 if (optlen != sizeof(struct sctp_assocparams))
2912 return -EINVAL;
2913 if (copy_from_user(&assocparams, optval, optlen))
2914 return -EFAULT;
2915
2916 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2917
2918 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2919 return -EINVAL;
2920
2921 /* Set the values to the specific association */
2922 if (asoc) {
2923 if (assocparams.sasoc_asocmaxrxt != 0) {
2924 __u32 path_sum = 0;
2925 int paths = 0;
2926 struct sctp_transport *peer_addr;
2927
2928 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2929 transports) {
2930 path_sum += peer_addr->pathmaxrxt;
2931 paths++;
2932 }
2933
2934 /* Only validate asocmaxrxt if we have more than
2935 * one path/transport. We do this because path
2936 * retransmissions are only counted when we have more
2937 * then one path.
2938 */
2939 if (paths > 1 &&
2940 assocparams.sasoc_asocmaxrxt > path_sum)
2941 return -EINVAL;
2942
2943 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2944 }
2945
2946 if (assocparams.sasoc_cookie_life != 0)
2947 asoc->cookie_life = ms_to_ktime(assocparams.sasoc_cookie_life);
2948 } else {
2949 /* Set the values to the endpoint */
2950 struct sctp_sock *sp = sctp_sk(sk);
2951
2952 if (assocparams.sasoc_asocmaxrxt != 0)
2953 sp->assocparams.sasoc_asocmaxrxt =
2954 assocparams.sasoc_asocmaxrxt;
2955 if (assocparams.sasoc_cookie_life != 0)
2956 sp->assocparams.sasoc_cookie_life =
2957 assocparams.sasoc_cookie_life;
2958 }
2959 return 0;
2960}
2961
2962/*
2963 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2964 *
2965 * This socket option is a boolean flag which turns on or off mapped V4
2966 * addresses. If this option is turned on and the socket is type
2967 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2968 * If this option is turned off, then no mapping will be done of V4
2969 * addresses and a user will receive both PF_INET6 and PF_INET type
2970 * addresses on the socket.
2971 */
2972static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2973{
2974 int val;
2975 struct sctp_sock *sp = sctp_sk(sk);
2976
2977 if (optlen < sizeof(int))
2978 return -EINVAL;
2979 if (get_user(val, (int __user *)optval))
2980 return -EFAULT;
2981 if (val)
2982 sp->v4mapped = 1;
2983 else
2984 sp->v4mapped = 0;
2985
2986 return 0;
2987}
2988
2989/*
2990 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2991 * This option will get or set the maximum size to put in any outgoing
2992 * SCTP DATA chunk. If a message is larger than this size it will be
2993 * fragmented by SCTP into the specified size. Note that the underlying
2994 * SCTP implementation may fragment into smaller sized chunks when the
2995 * PMTU of the underlying association is smaller than the value set by
2996 * the user. The default value for this option is '0' which indicates
2997 * the user is NOT limiting fragmentation and only the PMTU will effect
2998 * SCTP's choice of DATA chunk size. Note also that values set larger
2999 * than the maximum size of an IP datagram will effectively let SCTP
3000 * control fragmentation (i.e. the same as setting this option to 0).
3001 *
3002 * The following structure is used to access and modify this parameter:
3003 *
3004 * struct sctp_assoc_value {
3005 * sctp_assoc_t assoc_id;
3006 * uint32_t assoc_value;
3007 * };
3008 *
3009 * assoc_id: This parameter is ignored for one-to-one style sockets.
3010 * For one-to-many style sockets this parameter indicates which
3011 * association the user is performing an action upon. Note that if
3012 * this field's value is zero then the endpoints default value is
3013 * changed (effecting future associations only).
3014 * assoc_value: This parameter specifies the maximum size in bytes.
3015 */
3016static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
3017{
3018 struct sctp_assoc_value params;
3019 struct sctp_association *asoc;
3020 struct sctp_sock *sp = sctp_sk(sk);
3021 int val;
3022
3023 if (optlen == sizeof(int)) {
3024 pr_warn_ratelimited(DEPRECATED
3025 "%s (pid %d) "
3026 "Use of int in maxseg socket option.\n"
3027 "Use struct sctp_assoc_value instead\n",
3028 current->comm, task_pid_nr(current));
3029 if (copy_from_user(&val, optval, optlen))
3030 return -EFAULT;
3031 params.assoc_id = 0;
3032 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3033 if (copy_from_user(¶ms, optval, optlen))
3034 return -EFAULT;
3035 val = params.assoc_value;
3036 } else
3037 return -EINVAL;
3038
3039 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
3040 return -EINVAL;
3041
3042 asoc = sctp_id2assoc(sk, params.assoc_id);
3043 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
3044 return -EINVAL;
3045
3046 if (asoc) {
3047 if (val == 0) {
3048 val = asoc->pathmtu;
3049 val -= sp->pf->af->net_header_len;
3050 val -= sizeof(struct sctphdr) +
3051 sizeof(struct sctp_data_chunk);
3052 }
3053 asoc->user_frag = val;
3054 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3055 } else {
3056 sp->user_frag = val;
3057 }
3058
3059 return 0;
3060}
3061
3062
3063/*
3064 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3065 *
3066 * Requests that the peer mark the enclosed address as the association
3067 * primary. The enclosed address must be one of the association's
3068 * locally bound addresses. The following structure is used to make a
3069 * set primary request:
3070 */
3071static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3072 unsigned int optlen)
3073{
3074 struct net *net = sock_net(sk);
3075 struct sctp_sock *sp;
3076 struct sctp_association *asoc = NULL;
3077 struct sctp_setpeerprim prim;
3078 struct sctp_chunk *chunk;
3079 struct sctp_af *af;
3080 int err;
3081
3082 sp = sctp_sk(sk);
3083
3084 if (!net->sctp.addip_enable)
3085 return -EPERM;
3086
3087 if (optlen != sizeof(struct sctp_setpeerprim))
3088 return -EINVAL;
3089
3090 if (copy_from_user(&prim, optval, optlen))
3091 return -EFAULT;
3092
3093 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3094 if (!asoc)
3095 return -EINVAL;
3096
3097 if (!asoc->peer.asconf_capable)
3098 return -EPERM;
3099
3100 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3101 return -EPERM;
3102
3103 if (!sctp_state(asoc, ESTABLISHED))
3104 return -ENOTCONN;
3105
3106 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3107 if (!af)
3108 return -EINVAL;
3109
3110 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3111 return -EADDRNOTAVAIL;
3112
3113 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3114 return -EADDRNOTAVAIL;
3115
3116 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3117 chunk = sctp_make_asconf_set_prim(asoc,
3118 (union sctp_addr *)&prim.sspp_addr);
3119 if (!chunk)
3120 return -ENOMEM;
3121
3122 err = sctp_send_asconf(asoc, chunk);
3123
3124 pr_debug("%s: we set peer primary addr primitively\n", __func__);
3125
3126 return err;
3127}
3128
3129static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3130 unsigned int optlen)
3131{
3132 struct sctp_setadaptation adaptation;
3133
3134 if (optlen != sizeof(struct sctp_setadaptation))
3135 return -EINVAL;
3136 if (copy_from_user(&adaptation, optval, optlen))
3137 return -EFAULT;
3138
3139 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3140
3141 return 0;
3142}
3143
3144/*
3145 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3146 *
3147 * The context field in the sctp_sndrcvinfo structure is normally only
3148 * used when a failed message is retrieved holding the value that was
3149 * sent down on the actual send call. This option allows the setting of
3150 * a default context on an association basis that will be received on
3151 * reading messages from the peer. This is especially helpful in the
3152 * one-2-many model for an application to keep some reference to an
3153 * internal state machine that is processing messages on the
3154 * association. Note that the setting of this value only effects
3155 * received messages from the peer and does not effect the value that is
3156 * saved with outbound messages.
3157 */
3158static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3159 unsigned int optlen)
3160{
3161 struct sctp_assoc_value params;
3162 struct sctp_sock *sp;
3163 struct sctp_association *asoc;
3164
3165 if (optlen != sizeof(struct sctp_assoc_value))
3166 return -EINVAL;
3167 if (copy_from_user(¶ms, optval, optlen))
3168 return -EFAULT;
3169
3170 sp = sctp_sk(sk);
3171
3172 if (params.assoc_id != 0) {
3173 asoc = sctp_id2assoc(sk, params.assoc_id);
3174 if (!asoc)
3175 return -EINVAL;
3176 asoc->default_rcv_context = params.assoc_value;
3177 } else {
3178 sp->default_rcv_context = params.assoc_value;
3179 }
3180
3181 return 0;
3182}
3183
3184/*
3185 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3186 *
3187 * This options will at a minimum specify if the implementation is doing
3188 * fragmented interleave. Fragmented interleave, for a one to many
3189 * socket, is when subsequent calls to receive a message may return
3190 * parts of messages from different associations. Some implementations
3191 * may allow you to turn this value on or off. If so, when turned off,
3192 * no fragment interleave will occur (which will cause a head of line
3193 * blocking amongst multiple associations sharing the same one to many
3194 * socket). When this option is turned on, then each receive call may
3195 * come from a different association (thus the user must receive data
3196 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3197 * association each receive belongs to.
3198 *
3199 * This option takes a boolean value. A non-zero value indicates that
3200 * fragmented interleave is on. A value of zero indicates that
3201 * fragmented interleave is off.
3202 *
3203 * Note that it is important that an implementation that allows this
3204 * option to be turned on, have it off by default. Otherwise an unaware
3205 * application using the one to many model may become confused and act
3206 * incorrectly.
3207 */
3208static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3209 char __user *optval,
3210 unsigned int optlen)
3211{
3212 int val;
3213
3214 if (optlen != sizeof(int))
3215 return -EINVAL;
3216 if (get_user(val, (int __user *)optval))
3217 return -EFAULT;
3218
3219 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3220
3221 return 0;
3222}
3223
3224/*
3225 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3226 * (SCTP_PARTIAL_DELIVERY_POINT)
3227 *
3228 * This option will set or get the SCTP partial delivery point. This
3229 * point is the size of a message where the partial delivery API will be
3230 * invoked to help free up rwnd space for the peer. Setting this to a
3231 * lower value will cause partial deliveries to happen more often. The
3232 * calls argument is an integer that sets or gets the partial delivery
3233 * point. Note also that the call will fail if the user attempts to set
3234 * this value larger than the socket receive buffer size.
3235 *
3236 * Note that any single message having a length smaller than or equal to
3237 * the SCTP partial delivery point will be delivered in one single read
3238 * call as long as the user provided buffer is large enough to hold the
3239 * message.
3240 */
3241static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3242 char __user *optval,
3243 unsigned int optlen)
3244{
3245 u32 val;
3246
3247 if (optlen != sizeof(u32))
3248 return -EINVAL;
3249 if (get_user(val, (int __user *)optval))
3250 return -EFAULT;
3251
3252 /* Note: We double the receive buffer from what the user sets
3253 * it to be, also initial rwnd is based on rcvbuf/2.
3254 */
3255 if (val > (sk->sk_rcvbuf >> 1))
3256 return -EINVAL;
3257
3258 sctp_sk(sk)->pd_point = val;
3259
3260 return 0; /* is this the right error code? */
3261}
3262
3263/*
3264 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3265 *
3266 * This option will allow a user to change the maximum burst of packets
3267 * that can be emitted by this association. Note that the default value
3268 * is 4, and some implementations may restrict this setting so that it
3269 * can only be lowered.
3270 *
3271 * NOTE: This text doesn't seem right. Do this on a socket basis with
3272 * future associations inheriting the socket value.
3273 */
3274static int sctp_setsockopt_maxburst(struct sock *sk,
3275 char __user *optval,
3276 unsigned int optlen)
3277{
3278 struct sctp_assoc_value params;
3279 struct sctp_sock *sp;
3280 struct sctp_association *asoc;
3281 int val;
3282 int assoc_id = 0;
3283
3284 if (optlen == sizeof(int)) {
3285 pr_warn_ratelimited(DEPRECATED
3286 "%s (pid %d) "
3287 "Use of int in max_burst socket option deprecated.\n"
3288 "Use struct sctp_assoc_value instead\n",
3289 current->comm, task_pid_nr(current));
3290 if (copy_from_user(&val, optval, optlen))
3291 return -EFAULT;
3292 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3293 if (copy_from_user(¶ms, optval, optlen))
3294 return -EFAULT;
3295 val = params.assoc_value;
3296 assoc_id = params.assoc_id;
3297 } else
3298 return -EINVAL;
3299
3300 sp = sctp_sk(sk);
3301
3302 if (assoc_id != 0) {
3303 asoc = sctp_id2assoc(sk, assoc_id);
3304 if (!asoc)
3305 return -EINVAL;
3306 asoc->max_burst = val;
3307 } else
3308 sp->max_burst = val;
3309
3310 return 0;
3311}
3312
3313/*
3314 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3315 *
3316 * This set option adds a chunk type that the user is requesting to be
3317 * received only in an authenticated way. Changes to the list of chunks
3318 * will only effect future associations on the socket.
3319 */
3320static int sctp_setsockopt_auth_chunk(struct sock *sk,
3321 char __user *optval,
3322 unsigned int optlen)
3323{
3324 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3325 struct sctp_authchunk val;
3326
3327 if (!ep->auth_enable)
3328 return -EACCES;
3329
3330 if (optlen != sizeof(struct sctp_authchunk))
3331 return -EINVAL;
3332 if (copy_from_user(&val, optval, optlen))
3333 return -EFAULT;
3334
3335 switch (val.sauth_chunk) {
3336 case SCTP_CID_INIT:
3337 case SCTP_CID_INIT_ACK:
3338 case SCTP_CID_SHUTDOWN_COMPLETE:
3339 case SCTP_CID_AUTH:
3340 return -EINVAL;
3341 }
3342
3343 /* add this chunk id to the endpoint */
3344 return sctp_auth_ep_add_chunkid(ep, val.sauth_chunk);
3345}
3346
3347/*
3348 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3349 *
3350 * This option gets or sets the list of HMAC algorithms that the local
3351 * endpoint requires the peer to use.
3352 */
3353static int sctp_setsockopt_hmac_ident(struct sock *sk,
3354 char __user *optval,
3355 unsigned int optlen)
3356{
3357 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3358 struct sctp_hmacalgo *hmacs;
3359 u32 idents;
3360 int err;
3361
3362 if (!ep->auth_enable)
3363 return -EACCES;
3364
3365 if (optlen < sizeof(struct sctp_hmacalgo))
3366 return -EINVAL;
3367
3368 hmacs = memdup_user(optval, optlen);
3369 if (IS_ERR(hmacs))
3370 return PTR_ERR(hmacs);
3371
3372 idents = hmacs->shmac_num_idents;
3373 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3374 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3375 err = -EINVAL;
3376 goto out;
3377 }
3378
3379 err = sctp_auth_ep_set_hmacs(ep, hmacs);
3380out:
3381 kfree(hmacs);
3382 return err;
3383}
3384
3385/*
3386 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3387 *
3388 * This option will set a shared secret key which is used to build an
3389 * association shared key.
3390 */
3391static int sctp_setsockopt_auth_key(struct sock *sk,
3392 char __user *optval,
3393 unsigned int optlen)
3394{
3395 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3396 struct sctp_authkey *authkey;
3397 struct sctp_association *asoc;
3398 int ret;
3399
3400 if (!ep->auth_enable)
3401 return -EACCES;
3402
3403 if (optlen <= sizeof(struct sctp_authkey))
3404 return -EINVAL;
3405
3406 authkey = memdup_user(optval, optlen);
3407 if (IS_ERR(authkey))
3408 return PTR_ERR(authkey);
3409
3410 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3411 ret = -EINVAL;
3412 goto out;
3413 }
3414
3415 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3416 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3417 ret = -EINVAL;
3418 goto out;
3419 }
3420
3421 ret = sctp_auth_set_key(ep, asoc, authkey);
3422out:
3423 kzfree(authkey);
3424 return ret;
3425}
3426
3427/*
3428 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3429 *
3430 * This option will get or set the active shared key to be used to build
3431 * the association shared key.
3432 */
3433static int sctp_setsockopt_active_key(struct sock *sk,
3434 char __user *optval,
3435 unsigned int optlen)
3436{
3437 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3438 struct sctp_authkeyid val;
3439 struct sctp_association *asoc;
3440
3441 if (!ep->auth_enable)
3442 return -EACCES;
3443
3444 if (optlen != sizeof(struct sctp_authkeyid))
3445 return -EINVAL;
3446 if (copy_from_user(&val, optval, optlen))
3447 return -EFAULT;
3448
3449 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3450 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3451 return -EINVAL;
3452
3453 return sctp_auth_set_active_key(ep, asoc, val.scact_keynumber);
3454}
3455
3456/*
3457 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3458 *
3459 * This set option will delete a shared secret key from use.
3460 */
3461static int sctp_setsockopt_del_key(struct sock *sk,
3462 char __user *optval,
3463 unsigned int optlen)
3464{
3465 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
3466 struct sctp_authkeyid val;
3467 struct sctp_association *asoc;
3468
3469 if (!ep->auth_enable)
3470 return -EACCES;
3471
3472 if (optlen != sizeof(struct sctp_authkeyid))
3473 return -EINVAL;
3474 if (copy_from_user(&val, optval, optlen))
3475 return -EFAULT;
3476
3477 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3478 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3479 return -EINVAL;
3480
3481 return sctp_auth_del_key_id(ep, asoc, val.scact_keynumber);
3482
3483}
3484
3485/*
3486 * 8.1.23 SCTP_AUTO_ASCONF
3487 *
3488 * This option will enable or disable the use of the automatic generation of
3489 * ASCONF chunks to add and delete addresses to an existing association. Note
3490 * that this option has two caveats namely: a) it only affects sockets that
3491 * are bound to all addresses available to the SCTP stack, and b) the system
3492 * administrator may have an overriding control that turns the ASCONF feature
3493 * off no matter what setting the socket option may have.
3494 * This option expects an integer boolean flag, where a non-zero value turns on
3495 * the option, and a zero value turns off the option.
3496 * Note. In this implementation, socket operation overrides default parameter
3497 * being set by sysctl as well as FreeBSD implementation
3498 */
3499static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3500 unsigned int optlen)
3501{
3502 int val;
3503 struct sctp_sock *sp = sctp_sk(sk);
3504
3505 if (optlen < sizeof(int))
3506 return -EINVAL;
3507 if (get_user(val, (int __user *)optval))
3508 return -EFAULT;
3509 if (!sctp_is_ep_boundall(sk) && val)
3510 return -EINVAL;
3511 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3512 return 0;
3513
3514 if (val == 0 && sp->do_auto_asconf) {
3515 list_del(&sp->auto_asconf_list);
3516 sp->do_auto_asconf = 0;
3517 } else if (val && !sp->do_auto_asconf) {
3518 list_add_tail(&sp->auto_asconf_list,
3519 &sock_net(sk)->sctp.auto_asconf_splist);
3520 sp->do_auto_asconf = 1;
3521 }
3522 return 0;
3523}
3524
3525
3526/*
3527 * SCTP_PEER_ADDR_THLDS
3528 *
3529 * This option allows us to alter the partially failed threshold for one or all
3530 * transports in an association. See Section 6.1 of:
3531 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
3532 */
3533static int sctp_setsockopt_paddr_thresholds(struct sock *sk,
3534 char __user *optval,
3535 unsigned int optlen)
3536{
3537 struct sctp_paddrthlds val;
3538 struct sctp_transport *trans;
3539 struct sctp_association *asoc;
3540
3541 if (optlen < sizeof(struct sctp_paddrthlds))
3542 return -EINVAL;
3543 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval,
3544 sizeof(struct sctp_paddrthlds)))
3545 return -EFAULT;
3546
3547
3548 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
3549 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
3550 if (!asoc)
3551 return -ENOENT;
3552 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
3553 transports) {
3554 if (val.spt_pathmaxrxt)
3555 trans->pathmaxrxt = val.spt_pathmaxrxt;
3556 trans->pf_retrans = val.spt_pathpfthld;
3557 }
3558
3559 if (val.spt_pathmaxrxt)
3560 asoc->pathmaxrxt = val.spt_pathmaxrxt;
3561 asoc->pf_retrans = val.spt_pathpfthld;
3562 } else {
3563 trans = sctp_addr_id2transport(sk, &val.spt_address,
3564 val.spt_assoc_id);
3565 if (!trans)
3566 return -ENOENT;
3567
3568 if (val.spt_pathmaxrxt)
3569 trans->pathmaxrxt = val.spt_pathmaxrxt;
3570 trans->pf_retrans = val.spt_pathpfthld;
3571 }
3572
3573 return 0;
3574}
3575
3576/* API 6.2 setsockopt(), getsockopt()
3577 *
3578 * Applications use setsockopt() and getsockopt() to set or retrieve
3579 * socket options. Socket options are used to change the default
3580 * behavior of sockets calls. They are described in Section 7.
3581 *
3582 * The syntax is:
3583 *
3584 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3585 * int __user *optlen);
3586 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3587 * int optlen);
3588 *
3589 * sd - the socket descript.
3590 * level - set to IPPROTO_SCTP for all SCTP options.
3591 * optname - the option name.
3592 * optval - the buffer to store the value of the option.
3593 * optlen - the size of the buffer.
3594 */
3595static int sctp_setsockopt(struct sock *sk, int level, int optname,
3596 char __user *optval, unsigned int optlen)
3597{
3598 int retval = 0;
3599
3600 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
3601
3602 /* I can hardly begin to describe how wrong this is. This is
3603 * so broken as to be worse than useless. The API draft
3604 * REALLY is NOT helpful here... I am not convinced that the
3605 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3606 * are at all well-founded.
3607 */
3608 if (level != SOL_SCTP) {
3609 struct sctp_af *af = sctp_sk(sk)->pf->af;
3610 retval = af->setsockopt(sk, level, optname, optval, optlen);
3611 goto out_nounlock;
3612 }
3613
3614 lock_sock(sk);
3615
3616 switch (optname) {
3617 case SCTP_SOCKOPT_BINDX_ADD:
3618 /* 'optlen' is the size of the addresses buffer. */
3619 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3620 optlen, SCTP_BINDX_ADD_ADDR);
3621 break;
3622
3623 case SCTP_SOCKOPT_BINDX_REM:
3624 /* 'optlen' is the size of the addresses buffer. */
3625 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3626 optlen, SCTP_BINDX_REM_ADDR);
3627 break;
3628
3629 case SCTP_SOCKOPT_CONNECTX_OLD:
3630 /* 'optlen' is the size of the addresses buffer. */
3631 retval = sctp_setsockopt_connectx_old(sk,
3632 (struct sockaddr __user *)optval,
3633 optlen);
3634 break;
3635
3636 case SCTP_SOCKOPT_CONNECTX:
3637 /* 'optlen' is the size of the addresses buffer. */
3638 retval = sctp_setsockopt_connectx(sk,
3639 (struct sockaddr __user *)optval,
3640 optlen);
3641 break;
3642
3643 case SCTP_DISABLE_FRAGMENTS:
3644 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3645 break;
3646
3647 case SCTP_EVENTS:
3648 retval = sctp_setsockopt_events(sk, optval, optlen);
3649 break;
3650
3651 case SCTP_AUTOCLOSE:
3652 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3653 break;
3654
3655 case SCTP_PEER_ADDR_PARAMS:
3656 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3657 break;
3658
3659 case SCTP_DELAYED_SACK:
3660 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3661 break;
3662 case SCTP_PARTIAL_DELIVERY_POINT:
3663 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3664 break;
3665
3666 case SCTP_INITMSG:
3667 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3668 break;
3669 case SCTP_DEFAULT_SEND_PARAM:
3670 retval = sctp_setsockopt_default_send_param(sk, optval,
3671 optlen);
3672 break;
3673 case SCTP_PRIMARY_ADDR:
3674 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3675 break;
3676 case SCTP_SET_PEER_PRIMARY_ADDR:
3677 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3678 break;
3679 case SCTP_NODELAY:
3680 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3681 break;
3682 case SCTP_RTOINFO:
3683 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3684 break;
3685 case SCTP_ASSOCINFO:
3686 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3687 break;
3688 case SCTP_I_WANT_MAPPED_V4_ADDR:
3689 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3690 break;
3691 case SCTP_MAXSEG:
3692 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3693 break;
3694 case SCTP_ADAPTATION_LAYER:
3695 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3696 break;
3697 case SCTP_CONTEXT:
3698 retval = sctp_setsockopt_context(sk, optval, optlen);
3699 break;
3700 case SCTP_FRAGMENT_INTERLEAVE:
3701 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3702 break;
3703 case SCTP_MAX_BURST:
3704 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3705 break;
3706 case SCTP_AUTH_CHUNK:
3707 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3708 break;
3709 case SCTP_HMAC_IDENT:
3710 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3711 break;
3712 case SCTP_AUTH_KEY:
3713 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3714 break;
3715 case SCTP_AUTH_ACTIVE_KEY:
3716 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3717 break;
3718 case SCTP_AUTH_DELETE_KEY:
3719 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3720 break;
3721 case SCTP_AUTO_ASCONF:
3722 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3723 break;
3724 case SCTP_PEER_ADDR_THLDS:
3725 retval = sctp_setsockopt_paddr_thresholds(sk, optval, optlen);
3726 break;
3727 default:
3728 retval = -ENOPROTOOPT;
3729 break;
3730 }
3731
3732 release_sock(sk);
3733
3734out_nounlock:
3735 return retval;
3736}
3737
3738/* API 3.1.6 connect() - UDP Style Syntax
3739 *
3740 * An application may use the connect() call in the UDP model to initiate an
3741 * association without sending data.
3742 *
3743 * The syntax is:
3744 *
3745 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3746 *
3747 * sd: the socket descriptor to have a new association added to.
3748 *
3749 * nam: the address structure (either struct sockaddr_in or struct
3750 * sockaddr_in6 defined in RFC2553 [7]).
3751 *
3752 * len: the size of the address.
3753 */
3754static int sctp_connect(struct sock *sk, struct sockaddr *addr,
3755 int addr_len)
3756{
3757 int err = 0;
3758 struct sctp_af *af;
3759
3760 lock_sock(sk);
3761
3762 pr_debug("%s: sk:%p, sockaddr:%p, addr_len:%d\n", __func__, sk,
3763 addr, addr_len);
3764
3765 /* Validate addr_len before calling common connect/connectx routine. */
3766 af = sctp_get_af_specific(addr->sa_family);
3767 if (!af || addr_len < af->sockaddr_len) {
3768 err = -EINVAL;
3769 } else {
3770 /* Pass correct addr len to common routine (so it knows there
3771 * is only one address being passed.
3772 */
3773 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3774 }
3775
3776 release_sock(sk);
3777 return err;
3778}
3779
3780/* FIXME: Write comments. */
3781static int sctp_disconnect(struct sock *sk, int flags)
3782{
3783 return -EOPNOTSUPP; /* STUB */
3784}
3785
3786/* 4.1.4 accept() - TCP Style Syntax
3787 *
3788 * Applications use accept() call to remove an established SCTP
3789 * association from the accept queue of the endpoint. A new socket
3790 * descriptor will be returned from accept() to represent the newly
3791 * formed association.
3792 */
3793static struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3794{
3795 struct sctp_sock *sp;
3796 struct sctp_endpoint *ep;
3797 struct sock *newsk = NULL;
3798 struct sctp_association *asoc;
3799 long timeo;
3800 int error = 0;
3801
3802 lock_sock(sk);
3803
3804 sp = sctp_sk(sk);
3805 ep = sp->ep;
3806
3807 if (!sctp_style(sk, TCP)) {
3808 error = -EOPNOTSUPP;
3809 goto out;
3810 }
3811
3812 if (!sctp_sstate(sk, LISTENING)) {
3813 error = -EINVAL;
3814 goto out;
3815 }
3816
3817 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3818
3819 error = sctp_wait_for_accept(sk, timeo);
3820 if (error)
3821 goto out;
3822
3823 /* We treat the list of associations on the endpoint as the accept
3824 * queue and pick the first association on the list.
3825 */
3826 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3827
3828 newsk = sp->pf->create_accept_sk(sk, asoc);
3829 if (!newsk) {
3830 error = -ENOMEM;
3831 goto out;
3832 }
3833
3834 /* Populate the fields of the newsk from the oldsk and migrate the
3835 * asoc to the newsk.
3836 */
3837 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3838
3839out:
3840 release_sock(sk);
3841 *err = error;
3842 return newsk;
3843}
3844
3845/* The SCTP ioctl handler. */
3846static int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3847{
3848 int rc = -ENOTCONN;
3849
3850 lock_sock(sk);
3851
3852 /*
3853 * SEQPACKET-style sockets in LISTENING state are valid, for
3854 * SCTP, so only discard TCP-style sockets in LISTENING state.
3855 */
3856 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3857 goto out;
3858
3859 switch (cmd) {
3860 case SIOCINQ: {
3861 struct sk_buff *skb;
3862 unsigned int amount = 0;
3863
3864 skb = skb_peek(&sk->sk_receive_queue);
3865 if (skb != NULL) {
3866 /*
3867 * We will only return the amount of this packet since
3868 * that is all that will be read.
3869 */
3870 amount = skb->len;
3871 }
3872 rc = put_user(amount, (int __user *)arg);
3873 break;
3874 }
3875 default:
3876 rc = -ENOIOCTLCMD;
3877 break;
3878 }
3879out:
3880 release_sock(sk);
3881 return rc;
3882}
3883
3884/* This is the function which gets called during socket creation to
3885 * initialized the SCTP-specific portion of the sock.
3886 * The sock structure should already be zero-filled memory.
3887 */
3888static int sctp_init_sock(struct sock *sk)
3889{
3890 struct net *net = sock_net(sk);
3891 struct sctp_sock *sp;
3892
3893 pr_debug("%s: sk:%p\n", __func__, sk);
3894
3895 sp = sctp_sk(sk);
3896
3897 /* Initialize the SCTP per socket area. */
3898 switch (sk->sk_type) {
3899 case SOCK_SEQPACKET:
3900 sp->type = SCTP_SOCKET_UDP;
3901 break;
3902 case SOCK_STREAM:
3903 sp->type = SCTP_SOCKET_TCP;
3904 break;
3905 default:
3906 return -ESOCKTNOSUPPORT;
3907 }
3908
3909 /* Initialize default send parameters. These parameters can be
3910 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3911 */
3912 sp->default_stream = 0;
3913 sp->default_ppid = 0;
3914 sp->default_flags = 0;
3915 sp->default_context = 0;
3916 sp->default_timetolive = 0;
3917
3918 sp->default_rcv_context = 0;
3919 sp->max_burst = net->sctp.max_burst;
3920
3921 sp->sctp_hmac_alg = net->sctp.sctp_hmac_alg;
3922
3923 /* Initialize default setup parameters. These parameters
3924 * can be modified with the SCTP_INITMSG socket option or
3925 * overridden by the SCTP_INIT CMSG.
3926 */
3927 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3928 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3929 sp->initmsg.sinit_max_attempts = net->sctp.max_retrans_init;
3930 sp->initmsg.sinit_max_init_timeo = net->sctp.rto_max;
3931
3932 /* Initialize default RTO related parameters. These parameters can
3933 * be modified for with the SCTP_RTOINFO socket option.
3934 */
3935 sp->rtoinfo.srto_initial = net->sctp.rto_initial;
3936 sp->rtoinfo.srto_max = net->sctp.rto_max;
3937 sp->rtoinfo.srto_min = net->sctp.rto_min;
3938
3939 /* Initialize default association related parameters. These parameters
3940 * can be modified with the SCTP_ASSOCINFO socket option.
3941 */
3942 sp->assocparams.sasoc_asocmaxrxt = net->sctp.max_retrans_association;
3943 sp->assocparams.sasoc_number_peer_destinations = 0;
3944 sp->assocparams.sasoc_peer_rwnd = 0;
3945 sp->assocparams.sasoc_local_rwnd = 0;
3946 sp->assocparams.sasoc_cookie_life = net->sctp.valid_cookie_life;
3947
3948 /* Initialize default event subscriptions. By default, all the
3949 * options are off.
3950 */
3951 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3952
3953 /* Default Peer Address Parameters. These defaults can
3954 * be modified via SCTP_PEER_ADDR_PARAMS
3955 */
3956 sp->hbinterval = net->sctp.hb_interval;
3957 sp->pathmaxrxt = net->sctp.max_retrans_path;
3958 sp->pathmtu = 0; /* allow default discovery */
3959 sp->sackdelay = net->sctp.sack_timeout;
3960 sp->sackfreq = 2;
3961 sp->param_flags = SPP_HB_ENABLE |
3962 SPP_PMTUD_ENABLE |
3963 SPP_SACKDELAY_ENABLE;
3964
3965 /* If enabled no SCTP message fragmentation will be performed.
3966 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3967 */
3968 sp->disable_fragments = 0;
3969
3970 /* Enable Nagle algorithm by default. */
3971 sp->nodelay = 0;
3972
3973 /* Enable by default. */
3974 sp->v4mapped = 1;
3975
3976 /* Auto-close idle associations after the configured
3977 * number of seconds. A value of 0 disables this
3978 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3979 * for UDP-style sockets only.
3980 */
3981 sp->autoclose = 0;
3982
3983 /* User specified fragmentation limit. */
3984 sp->user_frag = 0;
3985
3986 sp->adaptation_ind = 0;
3987
3988 sp->pf = sctp_get_pf_specific(sk->sk_family);
3989
3990 /* Control variables for partial data delivery. */
3991 atomic_set(&sp->pd_mode, 0);
3992 skb_queue_head_init(&sp->pd_lobby);
3993 sp->frag_interleave = 0;
3994
3995 /* Create a per socket endpoint structure. Even if we
3996 * change the data structure relationships, this may still
3997 * be useful for storing pre-connect address information.
3998 */
3999 sp->ep = sctp_endpoint_new(sk, GFP_KERNEL);
4000 if (!sp->ep)
4001 return -ENOMEM;
4002
4003 sp->hmac = NULL;
4004
4005 sk->sk_destruct = sctp_destruct_sock;
4006
4007 SCTP_DBG_OBJCNT_INC(sock);
4008
4009 local_bh_disable();
4010 percpu_counter_inc(&sctp_sockets_allocated);
4011 sock_prot_inuse_add(net, sk->sk_prot, 1);
4012 if (net->sctp.default_auto_asconf) {
4013 list_add_tail(&sp->auto_asconf_list,
4014 &net->sctp.auto_asconf_splist);
4015 sp->do_auto_asconf = 1;
4016 } else
4017 sp->do_auto_asconf = 0;
4018 local_bh_enable();
4019
4020 return 0;
4021}
4022
4023/* Cleanup any SCTP per socket resources. */
4024static void sctp_destroy_sock(struct sock *sk)
4025{
4026 struct sctp_sock *sp;
4027
4028 pr_debug("%s: sk:%p\n", __func__, sk);
4029
4030 /* Release our hold on the endpoint. */
4031 sp = sctp_sk(sk);
4032 /* This could happen during socket init, thus we bail out
4033 * early, since the rest of the below is not setup either.
4034 */
4035 if (sp->ep == NULL)
4036 return;
4037
4038 if (sp->do_auto_asconf) {
4039 sp->do_auto_asconf = 0;
4040 list_del(&sp->auto_asconf_list);
4041 }
4042 sctp_endpoint_free(sp->ep);
4043 local_bh_disable();
4044 percpu_counter_dec(&sctp_sockets_allocated);
4045 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
4046 local_bh_enable();
4047}
4048
4049/* Triggered when there are no references on the socket anymore */
4050static void sctp_destruct_sock(struct sock *sk)
4051{
4052 struct sctp_sock *sp = sctp_sk(sk);
4053
4054 /* Free up the HMAC transform. */
4055 crypto_free_hash(sp->hmac);
4056
4057 inet_sock_destruct(sk);
4058}
4059
4060/* API 4.1.7 shutdown() - TCP Style Syntax
4061 * int shutdown(int socket, int how);
4062 *
4063 * sd - the socket descriptor of the association to be closed.
4064 * how - Specifies the type of shutdown. The values are
4065 * as follows:
4066 * SHUT_RD
4067 * Disables further receive operations. No SCTP
4068 * protocol action is taken.
4069 * SHUT_WR
4070 * Disables further send operations, and initiates
4071 * the SCTP shutdown sequence.
4072 * SHUT_RDWR
4073 * Disables further send and receive operations
4074 * and initiates the SCTP shutdown sequence.
4075 */
4076static void sctp_shutdown(struct sock *sk, int how)
4077{
4078 struct net *net = sock_net(sk);
4079 struct sctp_endpoint *ep;
4080 struct sctp_association *asoc;
4081
4082 if (!sctp_style(sk, TCP))
4083 return;
4084
4085 if (how & SEND_SHUTDOWN) {
4086 ep = sctp_sk(sk)->ep;
4087 if (!list_empty(&ep->asocs)) {
4088 asoc = list_entry(ep->asocs.next,
4089 struct sctp_association, asocs);
4090 sctp_primitive_SHUTDOWN(net, asoc, NULL);
4091 }
4092 }
4093}
4094
4095/* 7.2.1 Association Status (SCTP_STATUS)
4096
4097 * Applications can retrieve current status information about an
4098 * association, including association state, peer receiver window size,
4099 * number of unacked data chunks, and number of data chunks pending
4100 * receipt. This information is read-only.
4101 */
4102static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
4103 char __user *optval,
4104 int __user *optlen)
4105{
4106 struct sctp_status status;
4107 struct sctp_association *asoc = NULL;
4108 struct sctp_transport *transport;
4109 sctp_assoc_t associd;
4110 int retval = 0;
4111
4112 if (len < sizeof(status)) {
4113 retval = -EINVAL;
4114 goto out;
4115 }
4116
4117 len = sizeof(status);
4118 if (copy_from_user(&status, optval, len)) {
4119 retval = -EFAULT;
4120 goto out;
4121 }
4122
4123 associd = status.sstat_assoc_id;
4124 asoc = sctp_id2assoc(sk, associd);
4125 if (!asoc) {
4126 retval = -EINVAL;
4127 goto out;
4128 }
4129
4130 transport = asoc->peer.primary_path;
4131
4132 status.sstat_assoc_id = sctp_assoc2id(asoc);
4133 status.sstat_state = asoc->state;
4134 status.sstat_rwnd = asoc->peer.rwnd;
4135 status.sstat_unackdata = asoc->unack_data;
4136
4137 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4138 status.sstat_instrms = asoc->c.sinit_max_instreams;
4139 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4140 status.sstat_fragmentation_point = asoc->frag_point;
4141 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4142 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4143 transport->af_specific->sockaddr_len);
4144 /* Map ipv4 address into v4-mapped-on-v6 address. */
4145 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4146 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4147 status.sstat_primary.spinfo_state = transport->state;
4148 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4149 status.sstat_primary.spinfo_srtt = transport->srtt;
4150 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4151 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4152
4153 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4154 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4155
4156 if (put_user(len, optlen)) {
4157 retval = -EFAULT;
4158 goto out;
4159 }
4160
4161 pr_debug("%s: len:%d, state:%d, rwnd:%d, assoc_id:%d\n",
4162 __func__, len, status.sstat_state, status.sstat_rwnd,
4163 status.sstat_assoc_id);
4164
4165 if (copy_to_user(optval, &status, len)) {
4166 retval = -EFAULT;
4167 goto out;
4168 }
4169
4170out:
4171 return retval;
4172}
4173
4174
4175/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4176 *
4177 * Applications can retrieve information about a specific peer address
4178 * of an association, including its reachability state, congestion
4179 * window, and retransmission timer values. This information is
4180 * read-only.
4181 */
4182static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4183 char __user *optval,
4184 int __user *optlen)
4185{
4186 struct sctp_paddrinfo pinfo;
4187 struct sctp_transport *transport;
4188 int retval = 0;
4189
4190 if (len < sizeof(pinfo)) {
4191 retval = -EINVAL;
4192 goto out;
4193 }
4194
4195 len = sizeof(pinfo);
4196 if (copy_from_user(&pinfo, optval, len)) {
4197 retval = -EFAULT;
4198 goto out;
4199 }
4200
4201 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4202 pinfo.spinfo_assoc_id);
4203 if (!transport)
4204 return -EINVAL;
4205
4206 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4207 pinfo.spinfo_state = transport->state;
4208 pinfo.spinfo_cwnd = transport->cwnd;
4209 pinfo.spinfo_srtt = transport->srtt;
4210 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4211 pinfo.spinfo_mtu = transport->pathmtu;
4212
4213 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4214 pinfo.spinfo_state = SCTP_ACTIVE;
4215
4216 if (put_user(len, optlen)) {
4217 retval = -EFAULT;
4218 goto out;
4219 }
4220
4221 if (copy_to_user(optval, &pinfo, len)) {
4222 retval = -EFAULT;
4223 goto out;
4224 }
4225
4226out:
4227 return retval;
4228}
4229
4230/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4231 *
4232 * This option is a on/off flag. If enabled no SCTP message
4233 * fragmentation will be performed. Instead if a message being sent
4234 * exceeds the current PMTU size, the message will NOT be sent and
4235 * instead a error will be indicated to the user.
4236 */
4237static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4238 char __user *optval, int __user *optlen)
4239{
4240 int val;
4241
4242 if (len < sizeof(int))
4243 return -EINVAL;
4244
4245 len = sizeof(int);
4246 val = (sctp_sk(sk)->disable_fragments == 1);
4247 if (put_user(len, optlen))
4248 return -EFAULT;
4249 if (copy_to_user(optval, &val, len))
4250 return -EFAULT;
4251 return 0;
4252}
4253
4254/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4255 *
4256 * This socket option is used to specify various notifications and
4257 * ancillary data the user wishes to receive.
4258 */
4259static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4260 int __user *optlen)
4261{
4262 if (len <= 0)
4263 return -EINVAL;
4264 if (len > sizeof(struct sctp_event_subscribe))
4265 len = sizeof(struct sctp_event_subscribe);
4266 if (put_user(len, optlen))
4267 return -EFAULT;
4268 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4269 return -EFAULT;
4270 return 0;
4271}
4272
4273/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4274 *
4275 * This socket option is applicable to the UDP-style socket only. When
4276 * set it will cause associations that are idle for more than the
4277 * specified number of seconds to automatically close. An association
4278 * being idle is defined an association that has NOT sent or received
4279 * user data. The special value of '0' indicates that no automatic
4280 * close of any associations should be performed. The option expects an
4281 * integer defining the number of seconds of idle time before an
4282 * association is closed.
4283 */
4284static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4285{
4286 /* Applicable to UDP-style socket only */
4287 if (sctp_style(sk, TCP))
4288 return -EOPNOTSUPP;
4289 if (len < sizeof(int))
4290 return -EINVAL;
4291 len = sizeof(int);
4292 if (put_user(len, optlen))
4293 return -EFAULT;
4294 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4295 return -EFAULT;
4296 return 0;
4297}
4298
4299/* Helper routine to branch off an association to a new socket. */
4300int sctp_do_peeloff(struct sock *sk, sctp_assoc_t id, struct socket **sockp)
4301{
4302 struct sctp_association *asoc = sctp_id2assoc(sk, id);
4303 struct socket *sock;
4304 struct sctp_af *af;
4305 int err = 0;
4306
4307 if (!asoc)
4308 return -EINVAL;
4309
4310 /* An association cannot be branched off from an already peeled-off
4311 * socket, nor is this supported for tcp style sockets.
4312 */
4313 if (!sctp_style(sk, UDP))
4314 return -EINVAL;
4315
4316 /* Create a new socket. */
4317 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4318 if (err < 0)
4319 return err;
4320
4321 sctp_copy_sock(sock->sk, sk, asoc);
4322
4323 /* Make peeled-off sockets more like 1-1 accepted sockets.
4324 * Set the daddr and initialize id to something more random
4325 */
4326 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4327 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4328
4329 /* Populate the fields of the newsk from the oldsk and migrate the
4330 * asoc to the newsk.
4331 */
4332 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4333
4334 *sockp = sock;
4335
4336 return err;
4337}
4338EXPORT_SYMBOL(sctp_do_peeloff);
4339
4340static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4341{
4342 sctp_peeloff_arg_t peeloff;
4343 struct socket *newsock;
4344 struct file *newfile;
4345 int retval = 0;
4346
4347 if (len < sizeof(sctp_peeloff_arg_t))
4348 return -EINVAL;
4349 len = sizeof(sctp_peeloff_arg_t);
4350 if (copy_from_user(&peeloff, optval, len))
4351 return -EFAULT;
4352
4353 retval = sctp_do_peeloff(sk, peeloff.associd, &newsock);
4354 if (retval < 0)
4355 goto out;
4356
4357 /* Map the socket to an unused fd that can be returned to the user. */
4358 retval = get_unused_fd_flags(0);
4359 if (retval < 0) {
4360 sock_release(newsock);
4361 goto out;
4362 }
4363
4364 newfile = sock_alloc_file(newsock, 0, NULL);
4365 if (unlikely(IS_ERR(newfile))) {
4366 put_unused_fd(retval);
4367 sock_release(newsock);
4368 return PTR_ERR(newfile);
4369 }
4370
4371 pr_debug("%s: sk:%p, newsk:%p, sd:%d\n", __func__, sk, newsock->sk,
4372 retval);
4373
4374 /* Return the fd mapped to the new socket. */
4375 if (put_user(len, optlen)) {
4376 fput(newfile);
4377 put_unused_fd(retval);
4378 return -EFAULT;
4379 }
4380 peeloff.sd = retval;
4381 if (copy_to_user(optval, &peeloff, len)) {
4382 fput(newfile);
4383 put_unused_fd(retval);
4384 return -EFAULT;
4385 }
4386 fd_install(retval, newfile);
4387out:
4388 return retval;
4389}
4390
4391/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4392 *
4393 * Applications can enable or disable heartbeats for any peer address of
4394 * an association, modify an address's heartbeat interval, force a
4395 * heartbeat to be sent immediately, and adjust the address's maximum
4396 * number of retransmissions sent before an address is considered
4397 * unreachable. The following structure is used to access and modify an
4398 * address's parameters:
4399 *
4400 * struct sctp_paddrparams {
4401 * sctp_assoc_t spp_assoc_id;
4402 * struct sockaddr_storage spp_address;
4403 * uint32_t spp_hbinterval;
4404 * uint16_t spp_pathmaxrxt;
4405 * uint32_t spp_pathmtu;
4406 * uint32_t spp_sackdelay;
4407 * uint32_t spp_flags;
4408 * };
4409 *
4410 * spp_assoc_id - (one-to-many style socket) This is filled in the
4411 * application, and identifies the association for
4412 * this query.
4413 * spp_address - This specifies which address is of interest.
4414 * spp_hbinterval - This contains the value of the heartbeat interval,
4415 * in milliseconds. If a value of zero
4416 * is present in this field then no changes are to
4417 * be made to this parameter.
4418 * spp_pathmaxrxt - This contains the maximum number of
4419 * retransmissions before this address shall be
4420 * considered unreachable. If a value of zero
4421 * is present in this field then no changes are to
4422 * be made to this parameter.
4423 * spp_pathmtu - When Path MTU discovery is disabled the value
4424 * specified here will be the "fixed" path mtu.
4425 * Note that if the spp_address field is empty
4426 * then all associations on this address will
4427 * have this fixed path mtu set upon them.
4428 *
4429 * spp_sackdelay - When delayed sack is enabled, this value specifies
4430 * the number of milliseconds that sacks will be delayed
4431 * for. This value will apply to all addresses of an
4432 * association if the spp_address field is empty. Note
4433 * also, that if delayed sack is enabled and this
4434 * value is set to 0, no change is made to the last
4435 * recorded delayed sack timer value.
4436 *
4437 * spp_flags - These flags are used to control various features
4438 * on an association. The flag field may contain
4439 * zero or more of the following options.
4440 *
4441 * SPP_HB_ENABLE - Enable heartbeats on the
4442 * specified address. Note that if the address
4443 * field is empty all addresses for the association
4444 * have heartbeats enabled upon them.
4445 *
4446 * SPP_HB_DISABLE - Disable heartbeats on the
4447 * speicifed address. Note that if the address
4448 * field is empty all addresses for the association
4449 * will have their heartbeats disabled. Note also
4450 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4451 * mutually exclusive, only one of these two should
4452 * be specified. Enabling both fields will have
4453 * undetermined results.
4454 *
4455 * SPP_HB_DEMAND - Request a user initiated heartbeat
4456 * to be made immediately.
4457 *
4458 * SPP_PMTUD_ENABLE - This field will enable PMTU
4459 * discovery upon the specified address. Note that
4460 * if the address feild is empty then all addresses
4461 * on the association are effected.
4462 *
4463 * SPP_PMTUD_DISABLE - This field will disable PMTU
4464 * discovery upon the specified address. Note that
4465 * if the address feild is empty then all addresses
4466 * on the association are effected. Not also that
4467 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4468 * exclusive. Enabling both will have undetermined
4469 * results.
4470 *
4471 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4472 * on delayed sack. The time specified in spp_sackdelay
4473 * is used to specify the sack delay for this address. Note
4474 * that if spp_address is empty then all addresses will
4475 * enable delayed sack and take on the sack delay
4476 * value specified in spp_sackdelay.
4477 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4478 * off delayed sack. If the spp_address field is blank then
4479 * delayed sack is disabled for the entire association. Note
4480 * also that this field is mutually exclusive to
4481 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4482 * results.
4483 */
4484static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4485 char __user *optval, int __user *optlen)
4486{
4487 struct sctp_paddrparams params;
4488 struct sctp_transport *trans = NULL;
4489 struct sctp_association *asoc = NULL;
4490 struct sctp_sock *sp = sctp_sk(sk);
4491
4492 if (len < sizeof(struct sctp_paddrparams))
4493 return -EINVAL;
4494 len = sizeof(struct sctp_paddrparams);
4495 if (copy_from_user(¶ms, optval, len))
4496 return -EFAULT;
4497
4498 /* If an address other than INADDR_ANY is specified, and
4499 * no transport is found, then the request is invalid.
4500 */
4501 if (!sctp_is_any(sk, (union sctp_addr *)¶ms.spp_address)) {
4502 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4503 params.spp_assoc_id);
4504 if (!trans) {
4505 pr_debug("%s: failed no transport\n", __func__);
4506 return -EINVAL;
4507 }
4508 }
4509
4510 /* Get association, if assoc_id != 0 and the socket is a one
4511 * to many style socket, and an association was not found, then
4512 * the id was invalid.
4513 */
4514 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4515 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4516 pr_debug("%s: failed no association\n", __func__);
4517 return -EINVAL;
4518 }
4519
4520 if (trans) {
4521 /* Fetch transport values. */
4522 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4523 params.spp_pathmtu = trans->pathmtu;
4524 params.spp_pathmaxrxt = trans->pathmaxrxt;
4525 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4526
4527 /*draft-11 doesn't say what to return in spp_flags*/
4528 params.spp_flags = trans->param_flags;
4529 } else if (asoc) {
4530 /* Fetch association values. */
4531 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4532 params.spp_pathmtu = asoc->pathmtu;
4533 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4534 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4535
4536 /*draft-11 doesn't say what to return in spp_flags*/
4537 params.spp_flags = asoc->param_flags;
4538 } else {
4539 /* Fetch socket values. */
4540 params.spp_hbinterval = sp->hbinterval;
4541 params.spp_pathmtu = sp->pathmtu;
4542 params.spp_sackdelay = sp->sackdelay;
4543 params.spp_pathmaxrxt = sp->pathmaxrxt;
4544
4545 /*draft-11 doesn't say what to return in spp_flags*/
4546 params.spp_flags = sp->param_flags;
4547 }
4548
4549 if (copy_to_user(optval, ¶ms, len))
4550 return -EFAULT;
4551
4552 if (put_user(len, optlen))
4553 return -EFAULT;
4554
4555 return 0;
4556}
4557
4558/*
4559 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4560 *
4561 * This option will effect the way delayed acks are performed. This
4562 * option allows you to get or set the delayed ack time, in
4563 * milliseconds. It also allows changing the delayed ack frequency.
4564 * Changing the frequency to 1 disables the delayed sack algorithm. If
4565 * the assoc_id is 0, then this sets or gets the endpoints default
4566 * values. If the assoc_id field is non-zero, then the set or get
4567 * effects the specified association for the one to many model (the
4568 * assoc_id field is ignored by the one to one model). Note that if
4569 * sack_delay or sack_freq are 0 when setting this option, then the
4570 * current values will remain unchanged.
4571 *
4572 * struct sctp_sack_info {
4573 * sctp_assoc_t sack_assoc_id;
4574 * uint32_t sack_delay;
4575 * uint32_t sack_freq;
4576 * };
4577 *
4578 * sack_assoc_id - This parameter, indicates which association the user
4579 * is performing an action upon. Note that if this field's value is
4580 * zero then the endpoints default value is changed (effecting future
4581 * associations only).
4582 *
4583 * sack_delay - This parameter contains the number of milliseconds that
4584 * the user is requesting the delayed ACK timer be set to. Note that
4585 * this value is defined in the standard to be between 200 and 500
4586 * milliseconds.
4587 *
4588 * sack_freq - This parameter contains the number of packets that must
4589 * be received before a sack is sent without waiting for the delay
4590 * timer to expire. The default value for this is 2, setting this
4591 * value to 1 will disable the delayed sack algorithm.
4592 */
4593static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4594 char __user *optval,
4595 int __user *optlen)
4596{
4597 struct sctp_sack_info params;
4598 struct sctp_association *asoc = NULL;
4599 struct sctp_sock *sp = sctp_sk(sk);
4600
4601 if (len >= sizeof(struct sctp_sack_info)) {
4602 len = sizeof(struct sctp_sack_info);
4603
4604 if (copy_from_user(¶ms, optval, len))
4605 return -EFAULT;
4606 } else if (len == sizeof(struct sctp_assoc_value)) {
4607 pr_warn_ratelimited(DEPRECATED
4608 "%s (pid %d) "
4609 "Use of struct sctp_assoc_value in delayed_ack socket option.\n"
4610 "Use struct sctp_sack_info instead\n",
4611 current->comm, task_pid_nr(current));
4612 if (copy_from_user(¶ms, optval, len))
4613 return -EFAULT;
4614 } else
4615 return -EINVAL;
4616
4617 /* Get association, if sack_assoc_id != 0 and the socket is a one
4618 * to many style socket, and an association was not found, then
4619 * the id was invalid.
4620 */
4621 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4622 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4623 return -EINVAL;
4624
4625 if (asoc) {
4626 /* Fetch association values. */
4627 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4628 params.sack_delay = jiffies_to_msecs(
4629 asoc->sackdelay);
4630 params.sack_freq = asoc->sackfreq;
4631
4632 } else {
4633 params.sack_delay = 0;
4634 params.sack_freq = 1;
4635 }
4636 } else {
4637 /* Fetch socket values. */
4638 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4639 params.sack_delay = sp->sackdelay;
4640 params.sack_freq = sp->sackfreq;
4641 } else {
4642 params.sack_delay = 0;
4643 params.sack_freq = 1;
4644 }
4645 }
4646
4647 if (copy_to_user(optval, ¶ms, len))
4648 return -EFAULT;
4649
4650 if (put_user(len, optlen))
4651 return -EFAULT;
4652
4653 return 0;
4654}
4655
4656/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4657 *
4658 * Applications can specify protocol parameters for the default association
4659 * initialization. The option name argument to setsockopt() and getsockopt()
4660 * is SCTP_INITMSG.
4661 *
4662 * Setting initialization parameters is effective only on an unconnected
4663 * socket (for UDP-style sockets only future associations are effected
4664 * by the change). With TCP-style sockets, this option is inherited by
4665 * sockets derived from a listener socket.
4666 */
4667static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4668{
4669 if (len < sizeof(struct sctp_initmsg))
4670 return -EINVAL;
4671 len = sizeof(struct sctp_initmsg);
4672 if (put_user(len, optlen))
4673 return -EFAULT;
4674 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4675 return -EFAULT;
4676 return 0;
4677}
4678
4679
4680static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4681 char __user *optval, int __user *optlen)
4682{
4683 struct sctp_association *asoc;
4684 int cnt = 0;
4685 struct sctp_getaddrs getaddrs;
4686 struct sctp_transport *from;
4687 void __user *to;
4688 union sctp_addr temp;
4689 struct sctp_sock *sp = sctp_sk(sk);
4690 int addrlen;
4691 size_t space_left;
4692 int bytes_copied;
4693
4694 if (len < sizeof(struct sctp_getaddrs))
4695 return -EINVAL;
4696
4697 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4698 return -EFAULT;
4699
4700 /* For UDP-style sockets, id specifies the association to query. */
4701 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4702 if (!asoc)
4703 return -EINVAL;
4704
4705 to = optval + offsetof(struct sctp_getaddrs, addrs);
4706 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4707
4708 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4709 transports) {
4710 memcpy(&temp, &from->ipaddr, sizeof(temp));
4711 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4712 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4713 if (space_left < addrlen)
4714 return -ENOMEM;
4715 if (copy_to_user(to, &temp, addrlen))
4716 return -EFAULT;
4717 to += addrlen;
4718 cnt++;
4719 space_left -= addrlen;
4720 }
4721
4722 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4723 return -EFAULT;
4724 bytes_copied = ((char __user *)to) - optval;
4725 if (put_user(bytes_copied, optlen))
4726 return -EFAULT;
4727
4728 return 0;
4729}
4730
4731static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4732 size_t space_left, int *bytes_copied)
4733{
4734 struct sctp_sockaddr_entry *addr;
4735 union sctp_addr temp;
4736 int cnt = 0;
4737 int addrlen;
4738 struct net *net = sock_net(sk);
4739
4740 rcu_read_lock();
4741 list_for_each_entry_rcu(addr, &net->sctp.local_addr_list, list) {
4742 if (!addr->valid)
4743 continue;
4744
4745 if ((PF_INET == sk->sk_family) &&
4746 (AF_INET6 == addr->a.sa.sa_family))
4747 continue;
4748 if ((PF_INET6 == sk->sk_family) &&
4749 inet_v6_ipv6only(sk) &&
4750 (AF_INET == addr->a.sa.sa_family))
4751 continue;
4752 memcpy(&temp, &addr->a, sizeof(temp));
4753 if (!temp.v4.sin_port)
4754 temp.v4.sin_port = htons(port);
4755
4756 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4757 &temp);
4758 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4759 if (space_left < addrlen) {
4760 cnt = -ENOMEM;
4761 break;
4762 }
4763 memcpy(to, &temp, addrlen);
4764
4765 to += addrlen;
4766 cnt++;
4767 space_left -= addrlen;
4768 *bytes_copied += addrlen;
4769 }
4770 rcu_read_unlock();
4771
4772 return cnt;
4773}
4774
4775
4776static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4777 char __user *optval, int __user *optlen)
4778{
4779 struct sctp_bind_addr *bp;
4780 struct sctp_association *asoc;
4781 int cnt = 0;
4782 struct sctp_getaddrs getaddrs;
4783 struct sctp_sockaddr_entry *addr;
4784 void __user *to;
4785 union sctp_addr temp;
4786 struct sctp_sock *sp = sctp_sk(sk);
4787 int addrlen;
4788 int err = 0;
4789 size_t space_left;
4790 int bytes_copied = 0;
4791 void *addrs;
4792 void *buf;
4793
4794 if (len < sizeof(struct sctp_getaddrs))
4795 return -EINVAL;
4796
4797 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4798 return -EFAULT;
4799
4800 /*
4801 * For UDP-style sockets, id specifies the association to query.
4802 * If the id field is set to the value '0' then the locally bound
4803 * addresses are returned without regard to any particular
4804 * association.
4805 */
4806 if (0 == getaddrs.assoc_id) {
4807 bp = &sctp_sk(sk)->ep->base.bind_addr;
4808 } else {
4809 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4810 if (!asoc)
4811 return -EINVAL;
4812 bp = &asoc->base.bind_addr;
4813 }
4814
4815 to = optval + offsetof(struct sctp_getaddrs, addrs);
4816 space_left = len - offsetof(struct sctp_getaddrs, addrs);
4817
4818 addrs = kmalloc(space_left, GFP_KERNEL);
4819 if (!addrs)
4820 return -ENOMEM;
4821
4822 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4823 * addresses from the global local address list.
4824 */
4825 if (sctp_list_single_entry(&bp->address_list)) {
4826 addr = list_entry(bp->address_list.next,
4827 struct sctp_sockaddr_entry, list);
4828 if (sctp_is_any(sk, &addr->a)) {
4829 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4830 space_left, &bytes_copied);
4831 if (cnt < 0) {
4832 err = cnt;
4833 goto out;
4834 }
4835 goto copy_getaddrs;
4836 }
4837 }
4838
4839 buf = addrs;
4840 /* Protection on the bound address list is not needed since
4841 * in the socket option context we hold a socket lock and
4842 * thus the bound address list can't change.
4843 */
4844 list_for_each_entry(addr, &bp->address_list, list) {
4845 memcpy(&temp, &addr->a, sizeof(temp));
4846 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4847 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4848 if (space_left < addrlen) {
4849 err = -ENOMEM; /*fixme: right error?*/
4850 goto out;
4851 }
4852 memcpy(buf, &temp, addrlen);
4853 buf += addrlen;
4854 bytes_copied += addrlen;
4855 cnt++;
4856 space_left -= addrlen;
4857 }
4858
4859copy_getaddrs:
4860 if (copy_to_user(to, addrs, bytes_copied)) {
4861 err = -EFAULT;
4862 goto out;
4863 }
4864 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4865 err = -EFAULT;
4866 goto out;
4867 }
4868 if (put_user(bytes_copied, optlen))
4869 err = -EFAULT;
4870out:
4871 kfree(addrs);
4872 return err;
4873}
4874
4875/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4876 *
4877 * Requests that the local SCTP stack use the enclosed peer address as
4878 * the association primary. The enclosed address must be one of the
4879 * association peer's addresses.
4880 */
4881static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4882 char __user *optval, int __user *optlen)
4883{
4884 struct sctp_prim prim;
4885 struct sctp_association *asoc;
4886 struct sctp_sock *sp = sctp_sk(sk);
4887
4888 if (len < sizeof(struct sctp_prim))
4889 return -EINVAL;
4890
4891 len = sizeof(struct sctp_prim);
4892
4893 if (copy_from_user(&prim, optval, len))
4894 return -EFAULT;
4895
4896 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4897 if (!asoc)
4898 return -EINVAL;
4899
4900 if (!asoc->peer.primary_path)
4901 return -ENOTCONN;
4902
4903 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4904 asoc->peer.primary_path->af_specific->sockaddr_len);
4905
4906 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4907 (union sctp_addr *)&prim.ssp_addr);
4908
4909 if (put_user(len, optlen))
4910 return -EFAULT;
4911 if (copy_to_user(optval, &prim, len))
4912 return -EFAULT;
4913
4914 return 0;
4915}
4916
4917/*
4918 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4919 *
4920 * Requests that the local endpoint set the specified Adaptation Layer
4921 * Indication parameter for all future INIT and INIT-ACK exchanges.
4922 */
4923static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4924 char __user *optval, int __user *optlen)
4925{
4926 struct sctp_setadaptation adaptation;
4927
4928 if (len < sizeof(struct sctp_setadaptation))
4929 return -EINVAL;
4930
4931 len = sizeof(struct sctp_setadaptation);
4932
4933 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4934
4935 if (put_user(len, optlen))
4936 return -EFAULT;
4937 if (copy_to_user(optval, &adaptation, len))
4938 return -EFAULT;
4939
4940 return 0;
4941}
4942
4943/*
4944 *
4945 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4946 *
4947 * Applications that wish to use the sendto() system call may wish to
4948 * specify a default set of parameters that would normally be supplied
4949 * through the inclusion of ancillary data. This socket option allows
4950 * such an application to set the default sctp_sndrcvinfo structure.
4951
4952
4953 * The application that wishes to use this socket option simply passes
4954 * in to this call the sctp_sndrcvinfo structure defined in Section
4955 * 5.2.2) The input parameters accepted by this call include
4956 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4957 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4958 * to this call if the caller is using the UDP model.
4959 *
4960 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4961 */
4962static int sctp_getsockopt_default_send_param(struct sock *sk,
4963 int len, char __user *optval,
4964 int __user *optlen)
4965{
4966 struct sctp_sndrcvinfo info;
4967 struct sctp_association *asoc;
4968 struct sctp_sock *sp = sctp_sk(sk);
4969
4970 if (len < sizeof(struct sctp_sndrcvinfo))
4971 return -EINVAL;
4972
4973 len = sizeof(struct sctp_sndrcvinfo);
4974
4975 if (copy_from_user(&info, optval, len))
4976 return -EFAULT;
4977
4978 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4979 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4980 return -EINVAL;
4981
4982 if (asoc) {
4983 info.sinfo_stream = asoc->default_stream;
4984 info.sinfo_flags = asoc->default_flags;
4985 info.sinfo_ppid = asoc->default_ppid;
4986 info.sinfo_context = asoc->default_context;
4987 info.sinfo_timetolive = asoc->default_timetolive;
4988 } else {
4989 info.sinfo_stream = sp->default_stream;
4990 info.sinfo_flags = sp->default_flags;
4991 info.sinfo_ppid = sp->default_ppid;
4992 info.sinfo_context = sp->default_context;
4993 info.sinfo_timetolive = sp->default_timetolive;
4994 }
4995
4996 if (put_user(len, optlen))
4997 return -EFAULT;
4998 if (copy_to_user(optval, &info, len))
4999 return -EFAULT;
5000
5001 return 0;
5002}
5003
5004/*
5005 *
5006 * 7.1.5 SCTP_NODELAY
5007 *
5008 * Turn on/off any Nagle-like algorithm. This means that packets are
5009 * generally sent as soon as possible and no unnecessary delays are
5010 * introduced, at the cost of more packets in the network. Expects an
5011 * integer boolean flag.
5012 */
5013
5014static int sctp_getsockopt_nodelay(struct sock *sk, int len,
5015 char __user *optval, int __user *optlen)
5016{
5017 int val;
5018
5019 if (len < sizeof(int))
5020 return -EINVAL;
5021
5022 len = sizeof(int);
5023 val = (sctp_sk(sk)->nodelay == 1);
5024 if (put_user(len, optlen))
5025 return -EFAULT;
5026 if (copy_to_user(optval, &val, len))
5027 return -EFAULT;
5028 return 0;
5029}
5030
5031/*
5032 *
5033 * 7.1.1 SCTP_RTOINFO
5034 *
5035 * The protocol parameters used to initialize and bound retransmission
5036 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
5037 * and modify these parameters.
5038 * All parameters are time values, in milliseconds. A value of 0, when
5039 * modifying the parameters, indicates that the current value should not
5040 * be changed.
5041 *
5042 */
5043static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
5044 char __user *optval,
5045 int __user *optlen) {
5046 struct sctp_rtoinfo rtoinfo;
5047 struct sctp_association *asoc;
5048
5049 if (len < sizeof (struct sctp_rtoinfo))
5050 return -EINVAL;
5051
5052 len = sizeof(struct sctp_rtoinfo);
5053
5054 if (copy_from_user(&rtoinfo, optval, len))
5055 return -EFAULT;
5056
5057 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
5058
5059 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
5060 return -EINVAL;
5061
5062 /* Values corresponding to the specific association. */
5063 if (asoc) {
5064 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
5065 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
5066 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
5067 } else {
5068 /* Values corresponding to the endpoint. */
5069 struct sctp_sock *sp = sctp_sk(sk);
5070
5071 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
5072 rtoinfo.srto_max = sp->rtoinfo.srto_max;
5073 rtoinfo.srto_min = sp->rtoinfo.srto_min;
5074 }
5075
5076 if (put_user(len, optlen))
5077 return -EFAULT;
5078
5079 if (copy_to_user(optval, &rtoinfo, len))
5080 return -EFAULT;
5081
5082 return 0;
5083}
5084
5085/*
5086 *
5087 * 7.1.2 SCTP_ASSOCINFO
5088 *
5089 * This option is used to tune the maximum retransmission attempts
5090 * of the association.
5091 * Returns an error if the new association retransmission value is
5092 * greater than the sum of the retransmission value of the peer.
5093 * See [SCTP] for more information.
5094 *
5095 */
5096static int sctp_getsockopt_associnfo(struct sock *sk, int len,
5097 char __user *optval,
5098 int __user *optlen)
5099{
5100
5101 struct sctp_assocparams assocparams;
5102 struct sctp_association *asoc;
5103 struct list_head *pos;
5104 int cnt = 0;
5105
5106 if (len < sizeof (struct sctp_assocparams))
5107 return -EINVAL;
5108
5109 len = sizeof(struct sctp_assocparams);
5110
5111 if (copy_from_user(&assocparams, optval, len))
5112 return -EFAULT;
5113
5114 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
5115
5116 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
5117 return -EINVAL;
5118
5119 /* Values correspoinding to the specific association */
5120 if (asoc) {
5121 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
5122 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
5123 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
5124 assocparams.sasoc_cookie_life = ktime_to_ms(asoc->cookie_life);
5125
5126 list_for_each(pos, &asoc->peer.transport_addr_list) {
5127 cnt++;
5128 }
5129
5130 assocparams.sasoc_number_peer_destinations = cnt;
5131 } else {
5132 /* Values corresponding to the endpoint */
5133 struct sctp_sock *sp = sctp_sk(sk);
5134
5135 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5136 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5137 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5138 assocparams.sasoc_cookie_life =
5139 sp->assocparams.sasoc_cookie_life;
5140 assocparams.sasoc_number_peer_destinations =
5141 sp->assocparams.
5142 sasoc_number_peer_destinations;
5143 }
5144
5145 if (put_user(len, optlen))
5146 return -EFAULT;
5147
5148 if (copy_to_user(optval, &assocparams, len))
5149 return -EFAULT;
5150
5151 return 0;
5152}
5153
5154/*
5155 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5156 *
5157 * This socket option is a boolean flag which turns on or off mapped V4
5158 * addresses. If this option is turned on and the socket is type
5159 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5160 * If this option is turned off, then no mapping will be done of V4
5161 * addresses and a user will receive both PF_INET6 and PF_INET type
5162 * addresses on the socket.
5163 */
5164static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5165 char __user *optval, int __user *optlen)
5166{
5167 int val;
5168 struct sctp_sock *sp = sctp_sk(sk);
5169
5170 if (len < sizeof(int))
5171 return -EINVAL;
5172
5173 len = sizeof(int);
5174 val = sp->v4mapped;
5175 if (put_user(len, optlen))
5176 return -EFAULT;
5177 if (copy_to_user(optval, &val, len))
5178 return -EFAULT;
5179
5180 return 0;
5181}
5182
5183/*
5184 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5185 * (chapter and verse is quoted at sctp_setsockopt_context())
5186 */
5187static int sctp_getsockopt_context(struct sock *sk, int len,
5188 char __user *optval, int __user *optlen)
5189{
5190 struct sctp_assoc_value params;
5191 struct sctp_sock *sp;
5192 struct sctp_association *asoc;
5193
5194 if (len < sizeof(struct sctp_assoc_value))
5195 return -EINVAL;
5196
5197 len = sizeof(struct sctp_assoc_value);
5198
5199 if (copy_from_user(¶ms, optval, len))
5200 return -EFAULT;
5201
5202 sp = sctp_sk(sk);
5203
5204 if (params.assoc_id != 0) {
5205 asoc = sctp_id2assoc(sk, params.assoc_id);
5206 if (!asoc)
5207 return -EINVAL;
5208 params.assoc_value = asoc->default_rcv_context;
5209 } else {
5210 params.assoc_value = sp->default_rcv_context;
5211 }
5212
5213 if (put_user(len, optlen))
5214 return -EFAULT;
5215 if (copy_to_user(optval, ¶ms, len))
5216 return -EFAULT;
5217
5218 return 0;
5219}
5220
5221/*
5222 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5223 * This option will get or set the maximum size to put in any outgoing
5224 * SCTP DATA chunk. If a message is larger than this size it will be
5225 * fragmented by SCTP into the specified size. Note that the underlying
5226 * SCTP implementation may fragment into smaller sized chunks when the
5227 * PMTU of the underlying association is smaller than the value set by
5228 * the user. The default value for this option is '0' which indicates
5229 * the user is NOT limiting fragmentation and only the PMTU will effect
5230 * SCTP's choice of DATA chunk size. Note also that values set larger
5231 * than the maximum size of an IP datagram will effectively let SCTP
5232 * control fragmentation (i.e. the same as setting this option to 0).
5233 *
5234 * The following structure is used to access and modify this parameter:
5235 *
5236 * struct sctp_assoc_value {
5237 * sctp_assoc_t assoc_id;
5238 * uint32_t assoc_value;
5239 * };
5240 *
5241 * assoc_id: This parameter is ignored for one-to-one style sockets.
5242 * For one-to-many style sockets this parameter indicates which
5243 * association the user is performing an action upon. Note that if
5244 * this field's value is zero then the endpoints default value is
5245 * changed (effecting future associations only).
5246 * assoc_value: This parameter specifies the maximum size in bytes.
5247 */
5248static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5249 char __user *optval, int __user *optlen)
5250{
5251 struct sctp_assoc_value params;
5252 struct sctp_association *asoc;
5253
5254 if (len == sizeof(int)) {
5255 pr_warn_ratelimited(DEPRECATED
5256 "%s (pid %d) "
5257 "Use of int in maxseg socket option.\n"
5258 "Use struct sctp_assoc_value instead\n",
5259 current->comm, task_pid_nr(current));
5260 params.assoc_id = 0;
5261 } else if (len >= sizeof(struct sctp_assoc_value)) {
5262 len = sizeof(struct sctp_assoc_value);
5263 if (copy_from_user(¶ms, optval, sizeof(params)))
5264 return -EFAULT;
5265 } else
5266 return -EINVAL;
5267
5268 asoc = sctp_id2assoc(sk, params.assoc_id);
5269 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5270 return -EINVAL;
5271
5272 if (asoc)
5273 params.assoc_value = asoc->frag_point;
5274 else
5275 params.assoc_value = sctp_sk(sk)->user_frag;
5276
5277 if (put_user(len, optlen))
5278 return -EFAULT;
5279 if (len == sizeof(int)) {
5280 if (copy_to_user(optval, ¶ms.assoc_value, len))
5281 return -EFAULT;
5282 } else {
5283 if (copy_to_user(optval, ¶ms, len))
5284 return -EFAULT;
5285 }
5286
5287 return 0;
5288}
5289
5290/*
5291 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5292 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5293 */
5294static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5295 char __user *optval, int __user *optlen)
5296{
5297 int val;
5298
5299 if (len < sizeof(int))
5300 return -EINVAL;
5301
5302 len = sizeof(int);
5303
5304 val = sctp_sk(sk)->frag_interleave;
5305 if (put_user(len, optlen))
5306 return -EFAULT;
5307 if (copy_to_user(optval, &val, len))
5308 return -EFAULT;
5309
5310 return 0;
5311}
5312
5313/*
5314 * 7.1.25. Set or Get the sctp partial delivery point
5315 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5316 */
5317static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5318 char __user *optval,
5319 int __user *optlen)
5320{
5321 u32 val;
5322
5323 if (len < sizeof(u32))
5324 return -EINVAL;
5325
5326 len = sizeof(u32);
5327
5328 val = sctp_sk(sk)->pd_point;
5329 if (put_user(len, optlen))
5330 return -EFAULT;
5331 if (copy_to_user(optval, &val, len))
5332 return -EFAULT;
5333
5334 return 0;
5335}
5336
5337/*
5338 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5339 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5340 */
5341static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5342 char __user *optval,
5343 int __user *optlen)
5344{
5345 struct sctp_assoc_value params;
5346 struct sctp_sock *sp;
5347 struct sctp_association *asoc;
5348
5349 if (len == sizeof(int)) {
5350 pr_warn_ratelimited(DEPRECATED
5351 "%s (pid %d) "
5352 "Use of int in max_burst socket option.\n"
5353 "Use struct sctp_assoc_value instead\n",
5354 current->comm, task_pid_nr(current));
5355 params.assoc_id = 0;
5356 } else if (len >= sizeof(struct sctp_assoc_value)) {
5357 len = sizeof(struct sctp_assoc_value);
5358 if (copy_from_user(¶ms, optval, len))
5359 return -EFAULT;
5360 } else
5361 return -EINVAL;
5362
5363 sp = sctp_sk(sk);
5364
5365 if (params.assoc_id != 0) {
5366 asoc = sctp_id2assoc(sk, params.assoc_id);
5367 if (!asoc)
5368 return -EINVAL;
5369 params.assoc_value = asoc->max_burst;
5370 } else
5371 params.assoc_value = sp->max_burst;
5372
5373 if (len == sizeof(int)) {
5374 if (copy_to_user(optval, ¶ms.assoc_value, len))
5375 return -EFAULT;
5376 } else {
5377 if (copy_to_user(optval, ¶ms, len))
5378 return -EFAULT;
5379 }
5380
5381 return 0;
5382
5383}
5384
5385static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5386 char __user *optval, int __user *optlen)
5387{
5388 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5389 struct sctp_hmacalgo __user *p = (void __user *)optval;
5390 struct sctp_hmac_algo_param *hmacs;
5391 __u16 data_len = 0;
5392 u32 num_idents;
5393
5394 if (!ep->auth_enable)
5395 return -EACCES;
5396
5397 hmacs = ep->auth_hmacs_list;
5398 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5399
5400 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5401 return -EINVAL;
5402
5403 len = sizeof(struct sctp_hmacalgo) + data_len;
5404 num_idents = data_len / sizeof(u16);
5405
5406 if (put_user(len, optlen))
5407 return -EFAULT;
5408 if (put_user(num_idents, &p->shmac_num_idents))
5409 return -EFAULT;
5410 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5411 return -EFAULT;
5412 return 0;
5413}
5414
5415static int sctp_getsockopt_active_key(struct sock *sk, int len,
5416 char __user *optval, int __user *optlen)
5417{
5418 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5419 struct sctp_authkeyid val;
5420 struct sctp_association *asoc;
5421
5422 if (!ep->auth_enable)
5423 return -EACCES;
5424
5425 if (len < sizeof(struct sctp_authkeyid))
5426 return -EINVAL;
5427 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5428 return -EFAULT;
5429
5430 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5431 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5432 return -EINVAL;
5433
5434 if (asoc)
5435 val.scact_keynumber = asoc->active_key_id;
5436 else
5437 val.scact_keynumber = ep->active_key_id;
5438
5439 len = sizeof(struct sctp_authkeyid);
5440 if (put_user(len, optlen))
5441 return -EFAULT;
5442 if (copy_to_user(optval, &val, len))
5443 return -EFAULT;
5444
5445 return 0;
5446}
5447
5448static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5449 char __user *optval, int __user *optlen)
5450{
5451 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5452 struct sctp_authchunks __user *p = (void __user *)optval;
5453 struct sctp_authchunks val;
5454 struct sctp_association *asoc;
5455 struct sctp_chunks_param *ch;
5456 u32 num_chunks = 0;
5457 char __user *to;
5458
5459 if (!ep->auth_enable)
5460 return -EACCES;
5461
5462 if (len < sizeof(struct sctp_authchunks))
5463 return -EINVAL;
5464
5465 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5466 return -EFAULT;
5467
5468 to = p->gauth_chunks;
5469 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5470 if (!asoc)
5471 return -EINVAL;
5472
5473 ch = asoc->peer.peer_chunks;
5474 if (!ch)
5475 goto num;
5476
5477 /* See if the user provided enough room for all the data */
5478 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5479 if (len < num_chunks)
5480 return -EINVAL;
5481
5482 if (copy_to_user(to, ch->chunks, num_chunks))
5483 return -EFAULT;
5484num:
5485 len = sizeof(struct sctp_authchunks) + num_chunks;
5486 if (put_user(len, optlen))
5487 return -EFAULT;
5488 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5489 return -EFAULT;
5490 return 0;
5491}
5492
5493static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5494 char __user *optval, int __user *optlen)
5495{
5496 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5497 struct sctp_authchunks __user *p = (void __user *)optval;
5498 struct sctp_authchunks val;
5499 struct sctp_association *asoc;
5500 struct sctp_chunks_param *ch;
5501 u32 num_chunks = 0;
5502 char __user *to;
5503
5504 if (!ep->auth_enable)
5505 return -EACCES;
5506
5507 if (len < sizeof(struct sctp_authchunks))
5508 return -EINVAL;
5509
5510 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5511 return -EFAULT;
5512
5513 to = p->gauth_chunks;
5514 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5515 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5516 return -EINVAL;
5517
5518 if (asoc)
5519 ch = (struct sctp_chunks_param *)asoc->c.auth_chunks;
5520 else
5521 ch = ep->auth_chunk_list;
5522
5523 if (!ch)
5524 goto num;
5525
5526 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5527 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5528 return -EINVAL;
5529
5530 if (copy_to_user(to, ch->chunks, num_chunks))
5531 return -EFAULT;
5532num:
5533 len = sizeof(struct sctp_authchunks) + num_chunks;
5534 if (put_user(len, optlen))
5535 return -EFAULT;
5536 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5537 return -EFAULT;
5538
5539 return 0;
5540}
5541
5542/*
5543 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5544 * This option gets the current number of associations that are attached
5545 * to a one-to-many style socket. The option value is an uint32_t.
5546 */
5547static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5548 char __user *optval, int __user *optlen)
5549{
5550 struct sctp_sock *sp = sctp_sk(sk);
5551 struct sctp_association *asoc;
5552 u32 val = 0;
5553
5554 if (sctp_style(sk, TCP))
5555 return -EOPNOTSUPP;
5556
5557 if (len < sizeof(u32))
5558 return -EINVAL;
5559
5560 len = sizeof(u32);
5561
5562 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5563 val++;
5564 }
5565
5566 if (put_user(len, optlen))
5567 return -EFAULT;
5568 if (copy_to_user(optval, &val, len))
5569 return -EFAULT;
5570
5571 return 0;
5572}
5573
5574/*
5575 * 8.1.23 SCTP_AUTO_ASCONF
5576 * See the corresponding setsockopt entry as description
5577 */
5578static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5579 char __user *optval, int __user *optlen)
5580{
5581 int val = 0;
5582
5583 if (len < sizeof(int))
5584 return -EINVAL;
5585
5586 len = sizeof(int);
5587 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5588 val = 1;
5589 if (put_user(len, optlen))
5590 return -EFAULT;
5591 if (copy_to_user(optval, &val, len))
5592 return -EFAULT;
5593 return 0;
5594}
5595
5596/*
5597 * 8.2.6. Get the Current Identifiers of Associations
5598 * (SCTP_GET_ASSOC_ID_LIST)
5599 *
5600 * This option gets the current list of SCTP association identifiers of
5601 * the SCTP associations handled by a one-to-many style socket.
5602 */
5603static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5604 char __user *optval, int __user *optlen)
5605{
5606 struct sctp_sock *sp = sctp_sk(sk);
5607 struct sctp_association *asoc;
5608 struct sctp_assoc_ids *ids;
5609 u32 num = 0;
5610
5611 if (sctp_style(sk, TCP))
5612 return -EOPNOTSUPP;
5613
5614 if (len < sizeof(struct sctp_assoc_ids))
5615 return -EINVAL;
5616
5617 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5618 num++;
5619 }
5620
5621 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5622 return -EINVAL;
5623
5624 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5625
5626 ids = kmalloc(len, GFP_KERNEL);
5627 if (unlikely(!ids))
5628 return -ENOMEM;
5629
5630 ids->gaids_number_of_ids = num;
5631 num = 0;
5632 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5633 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5634 }
5635
5636 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5637 kfree(ids);
5638 return -EFAULT;
5639 }
5640
5641 kfree(ids);
5642 return 0;
5643}
5644
5645/*
5646 * SCTP_PEER_ADDR_THLDS
5647 *
5648 * This option allows us to fetch the partially failed threshold for one or all
5649 * transports in an association. See Section 6.1 of:
5650 * http://www.ietf.org/id/draft-nishida-tsvwg-sctp-failover-05.txt
5651 */
5652static int sctp_getsockopt_paddr_thresholds(struct sock *sk,
5653 char __user *optval,
5654 int len,
5655 int __user *optlen)
5656{
5657 struct sctp_paddrthlds val;
5658 struct sctp_transport *trans;
5659 struct sctp_association *asoc;
5660
5661 if (len < sizeof(struct sctp_paddrthlds))
5662 return -EINVAL;
5663 len = sizeof(struct sctp_paddrthlds);
5664 if (copy_from_user(&val, (struct sctp_paddrthlds __user *)optval, len))
5665 return -EFAULT;
5666
5667 if (sctp_is_any(sk, (const union sctp_addr *)&val.spt_address)) {
5668 asoc = sctp_id2assoc(sk, val.spt_assoc_id);
5669 if (!asoc)
5670 return -ENOENT;
5671
5672 val.spt_pathpfthld = asoc->pf_retrans;
5673 val.spt_pathmaxrxt = asoc->pathmaxrxt;
5674 } else {
5675 trans = sctp_addr_id2transport(sk, &val.spt_address,
5676 val.spt_assoc_id);
5677 if (!trans)
5678 return -ENOENT;
5679
5680 val.spt_pathmaxrxt = trans->pathmaxrxt;
5681 val.spt_pathpfthld = trans->pf_retrans;
5682 }
5683
5684 if (put_user(len, optlen) || copy_to_user(optval, &val, len))
5685 return -EFAULT;
5686
5687 return 0;
5688}
5689
5690/*
5691 * SCTP_GET_ASSOC_STATS
5692 *
5693 * This option retrieves local per endpoint statistics. It is modeled
5694 * after OpenSolaris' implementation
5695 */
5696static int sctp_getsockopt_assoc_stats(struct sock *sk, int len,
5697 char __user *optval,
5698 int __user *optlen)
5699{
5700 struct sctp_assoc_stats sas;
5701 struct sctp_association *asoc = NULL;
5702
5703 /* User must provide at least the assoc id */
5704 if (len < sizeof(sctp_assoc_t))
5705 return -EINVAL;
5706
5707 /* Allow the struct to grow and fill in as much as possible */
5708 len = min_t(size_t, len, sizeof(sas));
5709
5710 if (copy_from_user(&sas, optval, len))
5711 return -EFAULT;
5712
5713 asoc = sctp_id2assoc(sk, sas.sas_assoc_id);
5714 if (!asoc)
5715 return -EINVAL;
5716
5717 sas.sas_rtxchunks = asoc->stats.rtxchunks;
5718 sas.sas_gapcnt = asoc->stats.gapcnt;
5719 sas.sas_outofseqtsns = asoc->stats.outofseqtsns;
5720 sas.sas_osacks = asoc->stats.osacks;
5721 sas.sas_isacks = asoc->stats.isacks;
5722 sas.sas_octrlchunks = asoc->stats.octrlchunks;
5723 sas.sas_ictrlchunks = asoc->stats.ictrlchunks;
5724 sas.sas_oodchunks = asoc->stats.oodchunks;
5725 sas.sas_iodchunks = asoc->stats.iodchunks;
5726 sas.sas_ouodchunks = asoc->stats.ouodchunks;
5727 sas.sas_iuodchunks = asoc->stats.iuodchunks;
5728 sas.sas_idupchunks = asoc->stats.idupchunks;
5729 sas.sas_opackets = asoc->stats.opackets;
5730 sas.sas_ipackets = asoc->stats.ipackets;
5731
5732 /* New high max rto observed, will return 0 if not a single
5733 * RTO update took place. obs_rto_ipaddr will be bogus
5734 * in such a case
5735 */
5736 sas.sas_maxrto = asoc->stats.max_obs_rto;
5737 memcpy(&sas.sas_obs_rto_ipaddr, &asoc->stats.obs_rto_ipaddr,
5738 sizeof(struct sockaddr_storage));
5739
5740 /* Mark beginning of a new observation period */
5741 asoc->stats.max_obs_rto = asoc->rto_min;
5742
5743 if (put_user(len, optlen))
5744 return -EFAULT;
5745
5746 pr_debug("%s: len:%d, assoc_id:%d\n", __func__, len, sas.sas_assoc_id);
5747
5748 if (copy_to_user(optval, &sas, len))
5749 return -EFAULT;
5750
5751 return 0;
5752}
5753
5754static int sctp_getsockopt(struct sock *sk, int level, int optname,
5755 char __user *optval, int __user *optlen)
5756{
5757 int retval = 0;
5758 int len;
5759
5760 pr_debug("%s: sk:%p, optname:%d\n", __func__, sk, optname);
5761
5762 /* I can hardly begin to describe how wrong this is. This is
5763 * so broken as to be worse than useless. The API draft
5764 * REALLY is NOT helpful here... I am not convinced that the
5765 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5766 * are at all well-founded.
5767 */
5768 if (level != SOL_SCTP) {
5769 struct sctp_af *af = sctp_sk(sk)->pf->af;
5770
5771 retval = af->getsockopt(sk, level, optname, optval, optlen);
5772 return retval;
5773 }
5774
5775 if (get_user(len, optlen))
5776 return -EFAULT;
5777
5778 lock_sock(sk);
5779
5780 switch (optname) {
5781 case SCTP_STATUS:
5782 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5783 break;
5784 case SCTP_DISABLE_FRAGMENTS:
5785 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5786 optlen);
5787 break;
5788 case SCTP_EVENTS:
5789 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5790 break;
5791 case SCTP_AUTOCLOSE:
5792 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5793 break;
5794 case SCTP_SOCKOPT_PEELOFF:
5795 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5796 break;
5797 case SCTP_PEER_ADDR_PARAMS:
5798 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5799 optlen);
5800 break;
5801 case SCTP_DELAYED_SACK:
5802 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5803 optlen);
5804 break;
5805 case SCTP_INITMSG:
5806 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5807 break;
5808 case SCTP_GET_PEER_ADDRS:
5809 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5810 optlen);
5811 break;
5812 case SCTP_GET_LOCAL_ADDRS:
5813 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5814 optlen);
5815 break;
5816 case SCTP_SOCKOPT_CONNECTX3:
5817 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5818 break;
5819 case SCTP_DEFAULT_SEND_PARAM:
5820 retval = sctp_getsockopt_default_send_param(sk, len,
5821 optval, optlen);
5822 break;
5823 case SCTP_PRIMARY_ADDR:
5824 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5825 break;
5826 case SCTP_NODELAY:
5827 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5828 break;
5829 case SCTP_RTOINFO:
5830 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5831 break;
5832 case SCTP_ASSOCINFO:
5833 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5834 break;
5835 case SCTP_I_WANT_MAPPED_V4_ADDR:
5836 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5837 break;
5838 case SCTP_MAXSEG:
5839 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5840 break;
5841 case SCTP_GET_PEER_ADDR_INFO:
5842 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5843 optlen);
5844 break;
5845 case SCTP_ADAPTATION_LAYER:
5846 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5847 optlen);
5848 break;
5849 case SCTP_CONTEXT:
5850 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5851 break;
5852 case SCTP_FRAGMENT_INTERLEAVE:
5853 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5854 optlen);
5855 break;
5856 case SCTP_PARTIAL_DELIVERY_POINT:
5857 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5858 optlen);
5859 break;
5860 case SCTP_MAX_BURST:
5861 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5862 break;
5863 case SCTP_AUTH_KEY:
5864 case SCTP_AUTH_CHUNK:
5865 case SCTP_AUTH_DELETE_KEY:
5866 retval = -EOPNOTSUPP;
5867 break;
5868 case SCTP_HMAC_IDENT:
5869 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5870 break;
5871 case SCTP_AUTH_ACTIVE_KEY:
5872 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5873 break;
5874 case SCTP_PEER_AUTH_CHUNKS:
5875 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5876 optlen);
5877 break;
5878 case SCTP_LOCAL_AUTH_CHUNKS:
5879 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5880 optlen);
5881 break;
5882 case SCTP_GET_ASSOC_NUMBER:
5883 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5884 break;
5885 case SCTP_GET_ASSOC_ID_LIST:
5886 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5887 break;
5888 case SCTP_AUTO_ASCONF:
5889 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5890 break;
5891 case SCTP_PEER_ADDR_THLDS:
5892 retval = sctp_getsockopt_paddr_thresholds(sk, optval, len, optlen);
5893 break;
5894 case SCTP_GET_ASSOC_STATS:
5895 retval = sctp_getsockopt_assoc_stats(sk, len, optval, optlen);
5896 break;
5897 default:
5898 retval = -ENOPROTOOPT;
5899 break;
5900 }
5901
5902 release_sock(sk);
5903 return retval;
5904}
5905
5906static void sctp_hash(struct sock *sk)
5907{
5908 /* STUB */
5909}
5910
5911static void sctp_unhash(struct sock *sk)
5912{
5913 /* STUB */
5914}
5915
5916/* Check if port is acceptable. Possibly find first available port.
5917 *
5918 * The port hash table (contained in the 'global' SCTP protocol storage
5919 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5920 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5921 * list (the list number is the port number hashed out, so as you
5922 * would expect from a hash function, all the ports in a given list have
5923 * such a number that hashes out to the same list number; you were
5924 * expecting that, right?); so each list has a set of ports, with a
5925 * link to the socket (struct sock) that uses it, the port number and
5926 * a fastreuse flag (FIXME: NPI ipg).
5927 */
5928static struct sctp_bind_bucket *sctp_bucket_create(
5929 struct sctp_bind_hashbucket *head, struct net *, unsigned short snum);
5930
5931static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5932{
5933 struct sctp_bind_hashbucket *head; /* hash list */
5934 struct sctp_bind_bucket *pp;
5935 unsigned short snum;
5936 int ret;
5937
5938 snum = ntohs(addr->v4.sin_port);
5939
5940 pr_debug("%s: begins, snum:%d\n", __func__, snum);
5941
5942 local_bh_disable();
5943
5944 if (snum == 0) {
5945 /* Search for an available port. */
5946 int low, high, remaining, index;
5947 unsigned int rover;
5948
5949 inet_get_local_port_range(sock_net(sk), &low, &high);
5950 remaining = (high - low) + 1;
5951 rover = prandom_u32() % remaining + low;
5952
5953 do {
5954 rover++;
5955 if ((rover < low) || (rover > high))
5956 rover = low;
5957 if (inet_is_reserved_local_port(rover))
5958 continue;
5959 index = sctp_phashfn(sock_net(sk), rover);
5960 head = &sctp_port_hashtable[index];
5961 spin_lock(&head->lock);
5962 sctp_for_each_hentry(pp, &head->chain)
5963 if ((pp->port == rover) &&
5964 net_eq(sock_net(sk), pp->net))
5965 goto next;
5966 break;
5967 next:
5968 spin_unlock(&head->lock);
5969 } while (--remaining > 0);
5970
5971 /* Exhausted local port range during search? */
5972 ret = 1;
5973 if (remaining <= 0)
5974 goto fail;
5975
5976 /* OK, here is the one we will use. HEAD (the port
5977 * hash table list entry) is non-NULL and we hold it's
5978 * mutex.
5979 */
5980 snum = rover;
5981 } else {
5982 /* We are given an specific port number; we verify
5983 * that it is not being used. If it is used, we will
5984 * exahust the search in the hash list corresponding
5985 * to the port number (snum) - we detect that with the
5986 * port iterator, pp being NULL.
5987 */
5988 head = &sctp_port_hashtable[sctp_phashfn(sock_net(sk), snum)];
5989 spin_lock(&head->lock);
5990 sctp_for_each_hentry(pp, &head->chain) {
5991 if ((pp->port == snum) && net_eq(pp->net, sock_net(sk)))
5992 goto pp_found;
5993 }
5994 }
5995 pp = NULL;
5996 goto pp_not_found;
5997pp_found:
5998 if (!hlist_empty(&pp->owner)) {
5999 /* We had a port hash table hit - there is an
6000 * available port (pp != NULL) and it is being
6001 * used by other socket (pp->owner not empty); that other
6002 * socket is going to be sk2.
6003 */
6004 int reuse = sk->sk_reuse;
6005 struct sock *sk2;
6006
6007 pr_debug("%s: found a possible match\n", __func__);
6008
6009 if (pp->fastreuse && sk->sk_reuse &&
6010 sk->sk_state != SCTP_SS_LISTENING)
6011 goto success;
6012
6013 /* Run through the list of sockets bound to the port
6014 * (pp->port) [via the pointers bind_next and
6015 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
6016 * we get the endpoint they describe and run through
6017 * the endpoint's list of IP (v4 or v6) addresses,
6018 * comparing each of the addresses with the address of
6019 * the socket sk. If we find a match, then that means
6020 * that this port/socket (sk) combination are already
6021 * in an endpoint.
6022 */
6023 sk_for_each_bound(sk2, &pp->owner) {
6024 struct sctp_endpoint *ep2;
6025 ep2 = sctp_sk(sk2)->ep;
6026
6027 if (sk == sk2 ||
6028 (reuse && sk2->sk_reuse &&
6029 sk2->sk_state != SCTP_SS_LISTENING))
6030 continue;
6031
6032 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
6033 sctp_sk(sk2), sctp_sk(sk))) {
6034 ret = (long)sk2;
6035 goto fail_unlock;
6036 }
6037 }
6038
6039 pr_debug("%s: found a match\n", __func__);
6040 }
6041pp_not_found:
6042 /* If there was a hash table miss, create a new port. */
6043 ret = 1;
6044 if (!pp && !(pp = sctp_bucket_create(head, sock_net(sk), snum)))
6045 goto fail_unlock;
6046
6047 /* In either case (hit or miss), make sure fastreuse is 1 only
6048 * if sk->sk_reuse is too (that is, if the caller requested
6049 * SO_REUSEADDR on this socket -sk-).
6050 */
6051 if (hlist_empty(&pp->owner)) {
6052 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
6053 pp->fastreuse = 1;
6054 else
6055 pp->fastreuse = 0;
6056 } else if (pp->fastreuse &&
6057 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
6058 pp->fastreuse = 0;
6059
6060 /* We are set, so fill up all the data in the hash table
6061 * entry, tie the socket list information with the rest of the
6062 * sockets FIXME: Blurry, NPI (ipg).
6063 */
6064success:
6065 if (!sctp_sk(sk)->bind_hash) {
6066 inet_sk(sk)->inet_num = snum;
6067 sk_add_bind_node(sk, &pp->owner);
6068 sctp_sk(sk)->bind_hash = pp;
6069 }
6070 ret = 0;
6071
6072fail_unlock:
6073 spin_unlock(&head->lock);
6074
6075fail:
6076 local_bh_enable();
6077 return ret;
6078}
6079
6080/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
6081 * port is requested.
6082 */
6083static int sctp_get_port(struct sock *sk, unsigned short snum)
6084{
6085 union sctp_addr addr;
6086 struct sctp_af *af = sctp_sk(sk)->pf->af;
6087
6088 /* Set up a dummy address struct from the sk. */
6089 af->from_sk(&addr, sk);
6090 addr.v4.sin_port = htons(snum);
6091
6092 /* Note: sk->sk_num gets filled in if ephemeral port request. */
6093 return !!sctp_get_port_local(sk, &addr);
6094}
6095
6096/*
6097 * Move a socket to LISTENING state.
6098 */
6099static int sctp_listen_start(struct sock *sk, int backlog)
6100{
6101 struct sctp_sock *sp = sctp_sk(sk);
6102 struct sctp_endpoint *ep = sp->ep;
6103 struct crypto_hash *tfm = NULL;
6104 char alg[32];
6105
6106 /* Allocate HMAC for generating cookie. */
6107 if (!sp->hmac && sp->sctp_hmac_alg) {
6108 sprintf(alg, "hmac(%s)", sp->sctp_hmac_alg);
6109 tfm = crypto_alloc_hash(alg, 0, CRYPTO_ALG_ASYNC);
6110 if (IS_ERR(tfm)) {
6111 net_info_ratelimited("failed to load transform for %s: %ld\n",
6112 sp->sctp_hmac_alg, PTR_ERR(tfm));
6113 return -ENOSYS;
6114 }
6115 sctp_sk(sk)->hmac = tfm;
6116 }
6117
6118 /*
6119 * If a bind() or sctp_bindx() is not called prior to a listen()
6120 * call that allows new associations to be accepted, the system
6121 * picks an ephemeral port and will choose an address set equivalent
6122 * to binding with a wildcard address.
6123 *
6124 * This is not currently spelled out in the SCTP sockets
6125 * extensions draft, but follows the practice as seen in TCP
6126 * sockets.
6127 *
6128 */
6129 sk->sk_state = SCTP_SS_LISTENING;
6130 if (!ep->base.bind_addr.port) {
6131 if (sctp_autobind(sk))
6132 return -EAGAIN;
6133 } else {
6134 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
6135 sk->sk_state = SCTP_SS_CLOSED;
6136 return -EADDRINUSE;
6137 }
6138 }
6139
6140 sk->sk_max_ack_backlog = backlog;
6141 sctp_hash_endpoint(ep);
6142 return 0;
6143}
6144
6145/*
6146 * 4.1.3 / 5.1.3 listen()
6147 *
6148 * By default, new associations are not accepted for UDP style sockets.
6149 * An application uses listen() to mark a socket as being able to
6150 * accept new associations.
6151 *
6152 * On TCP style sockets, applications use listen() to ready the SCTP
6153 * endpoint for accepting inbound associations.
6154 *
6155 * On both types of endpoints a backlog of '0' disables listening.
6156 *
6157 * Move a socket to LISTENING state.
6158 */
6159int sctp_inet_listen(struct socket *sock, int backlog)
6160{
6161 struct sock *sk = sock->sk;
6162 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
6163 int err = -EINVAL;
6164
6165 if (unlikely(backlog < 0))
6166 return err;
6167
6168 lock_sock(sk);
6169
6170 /* Peeled-off sockets are not allowed to listen(). */
6171 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
6172 goto out;
6173
6174 if (sock->state != SS_UNCONNECTED)
6175 goto out;
6176
6177 /* If backlog is zero, disable listening. */
6178 if (!backlog) {
6179 if (sctp_sstate(sk, CLOSED))
6180 goto out;
6181
6182 err = 0;
6183 sctp_unhash_endpoint(ep);
6184 sk->sk_state = SCTP_SS_CLOSED;
6185 if (sk->sk_reuse)
6186 sctp_sk(sk)->bind_hash->fastreuse = 1;
6187 goto out;
6188 }
6189
6190 /* If we are already listening, just update the backlog */
6191 if (sctp_sstate(sk, LISTENING))
6192 sk->sk_max_ack_backlog = backlog;
6193 else {
6194 err = sctp_listen_start(sk, backlog);
6195 if (err)
6196 goto out;
6197 }
6198
6199 err = 0;
6200out:
6201 release_sock(sk);
6202 return err;
6203}
6204
6205/*
6206 * This function is done by modeling the current datagram_poll() and the
6207 * tcp_poll(). Note that, based on these implementations, we don't
6208 * lock the socket in this function, even though it seems that,
6209 * ideally, locking or some other mechanisms can be used to ensure
6210 * the integrity of the counters (sndbuf and wmem_alloc) used
6211 * in this place. We assume that we don't need locks either until proven
6212 * otherwise.
6213 *
6214 * Another thing to note is that we include the Async I/O support
6215 * here, again, by modeling the current TCP/UDP code. We don't have
6216 * a good way to test with it yet.
6217 */
6218unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
6219{
6220 struct sock *sk = sock->sk;
6221 struct sctp_sock *sp = sctp_sk(sk);
6222 unsigned int mask;
6223
6224 poll_wait(file, sk_sleep(sk), wait);
6225
6226 /* A TCP-style listening socket becomes readable when the accept queue
6227 * is not empty.
6228 */
6229 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
6230 return (!list_empty(&sp->ep->asocs)) ?
6231 (POLLIN | POLLRDNORM) : 0;
6232
6233 mask = 0;
6234
6235 /* Is there any exceptional events? */
6236 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
6237 mask |= POLLERR |
6238 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
6239 if (sk->sk_shutdown & RCV_SHUTDOWN)
6240 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
6241 if (sk->sk_shutdown == SHUTDOWN_MASK)
6242 mask |= POLLHUP;
6243
6244 /* Is it readable? Reconsider this code with TCP-style support. */
6245 if (!skb_queue_empty(&sk->sk_receive_queue))
6246 mask |= POLLIN | POLLRDNORM;
6247
6248 /* The association is either gone or not ready. */
6249 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
6250 return mask;
6251
6252 /* Is it writable? */
6253 if (sctp_writeable(sk)) {
6254 mask |= POLLOUT | POLLWRNORM;
6255 } else {
6256 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
6257 /*
6258 * Since the socket is not locked, the buffer
6259 * might be made available after the writeable check and
6260 * before the bit is set. This could cause a lost I/O
6261 * signal. tcp_poll() has a race breaker for this race
6262 * condition. Based on their implementation, we put
6263 * in the following code to cover it as well.
6264 */
6265 if (sctp_writeable(sk))
6266 mask |= POLLOUT | POLLWRNORM;
6267 }
6268 return mask;
6269}
6270
6271/********************************************************************
6272 * 2nd Level Abstractions
6273 ********************************************************************/
6274
6275static struct sctp_bind_bucket *sctp_bucket_create(
6276 struct sctp_bind_hashbucket *head, struct net *net, unsigned short snum)
6277{
6278 struct sctp_bind_bucket *pp;
6279
6280 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6281 if (pp) {
6282 SCTP_DBG_OBJCNT_INC(bind_bucket);
6283 pp->port = snum;
6284 pp->fastreuse = 0;
6285 INIT_HLIST_HEAD(&pp->owner);
6286 pp->net = net;
6287 hlist_add_head(&pp->node, &head->chain);
6288 }
6289 return pp;
6290}
6291
6292/* Caller must hold hashbucket lock for this tb with local BH disabled */
6293static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6294{
6295 if (pp && hlist_empty(&pp->owner)) {
6296 __hlist_del(&pp->node);
6297 kmem_cache_free(sctp_bucket_cachep, pp);
6298 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6299 }
6300}
6301
6302/* Release this socket's reference to a local port. */
6303static inline void __sctp_put_port(struct sock *sk)
6304{
6305 struct sctp_bind_hashbucket *head =
6306 &sctp_port_hashtable[sctp_phashfn(sock_net(sk),
6307 inet_sk(sk)->inet_num)];
6308 struct sctp_bind_bucket *pp;
6309
6310 spin_lock(&head->lock);
6311 pp = sctp_sk(sk)->bind_hash;
6312 __sk_del_bind_node(sk);
6313 sctp_sk(sk)->bind_hash = NULL;
6314 inet_sk(sk)->inet_num = 0;
6315 sctp_bucket_destroy(pp);
6316 spin_unlock(&head->lock);
6317}
6318
6319void sctp_put_port(struct sock *sk)
6320{
6321 local_bh_disable();
6322 __sctp_put_port(sk);
6323 local_bh_enable();
6324}
6325
6326/*
6327 * The system picks an ephemeral port and choose an address set equivalent
6328 * to binding with a wildcard address.
6329 * One of those addresses will be the primary address for the association.
6330 * This automatically enables the multihoming capability of SCTP.
6331 */
6332static int sctp_autobind(struct sock *sk)
6333{
6334 union sctp_addr autoaddr;
6335 struct sctp_af *af;
6336 __be16 port;
6337
6338 /* Initialize a local sockaddr structure to INADDR_ANY. */
6339 af = sctp_sk(sk)->pf->af;
6340
6341 port = htons(inet_sk(sk)->inet_num);
6342 af->inaddr_any(&autoaddr, port);
6343
6344 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6345}
6346
6347/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6348 *
6349 * From RFC 2292
6350 * 4.2 The cmsghdr Structure *
6351 *
6352 * When ancillary data is sent or received, any number of ancillary data
6353 * objects can be specified by the msg_control and msg_controllen members of
6354 * the msghdr structure, because each object is preceded by
6355 * a cmsghdr structure defining the object's length (the cmsg_len member).
6356 * Historically Berkeley-derived implementations have passed only one object
6357 * at a time, but this API allows multiple objects to be
6358 * passed in a single call to sendmsg() or recvmsg(). The following example
6359 * shows two ancillary data objects in a control buffer.
6360 *
6361 * |<--------------------------- msg_controllen -------------------------->|
6362 * | |
6363 *
6364 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6365 *
6366 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6367 * | | |
6368 *
6369 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6370 *
6371 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6372 * | | | | |
6373 *
6374 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6375 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6376 *
6377 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6378 *
6379 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6380 * ^
6381 * |
6382 *
6383 * msg_control
6384 * points here
6385 */
6386static int sctp_msghdr_parse(const struct msghdr *msg, sctp_cmsgs_t *cmsgs)
6387{
6388 struct cmsghdr *cmsg;
6389 struct msghdr *my_msg = (struct msghdr *)msg;
6390
6391 for (cmsg = CMSG_FIRSTHDR(msg);
6392 cmsg != NULL;
6393 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6394 if (!CMSG_OK(my_msg, cmsg))
6395 return -EINVAL;
6396
6397 /* Should we parse this header or ignore? */
6398 if (cmsg->cmsg_level != IPPROTO_SCTP)
6399 continue;
6400
6401 /* Strictly check lengths following example in SCM code. */
6402 switch (cmsg->cmsg_type) {
6403 case SCTP_INIT:
6404 /* SCTP Socket API Extension
6405 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6406 *
6407 * This cmsghdr structure provides information for
6408 * initializing new SCTP associations with sendmsg().
6409 * The SCTP_INITMSG socket option uses this same data
6410 * structure. This structure is not used for
6411 * recvmsg().
6412 *
6413 * cmsg_level cmsg_type cmsg_data[]
6414 * ------------ ------------ ----------------------
6415 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6416 */
6417 if (cmsg->cmsg_len !=
6418 CMSG_LEN(sizeof(struct sctp_initmsg)))
6419 return -EINVAL;
6420 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6421 break;
6422
6423 case SCTP_SNDRCV:
6424 /* SCTP Socket API Extension
6425 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6426 *
6427 * This cmsghdr structure specifies SCTP options for
6428 * sendmsg() and describes SCTP header information
6429 * about a received message through recvmsg().
6430 *
6431 * cmsg_level cmsg_type cmsg_data[]
6432 * ------------ ------------ ----------------------
6433 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6434 */
6435 if (cmsg->cmsg_len !=
6436 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6437 return -EINVAL;
6438
6439 cmsgs->info =
6440 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6441
6442 /* Minimally, validate the sinfo_flags. */
6443 if (cmsgs->info->sinfo_flags &
6444 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6445 SCTP_ABORT | SCTP_EOF))
6446 return -EINVAL;
6447 break;
6448
6449 default:
6450 return -EINVAL;
6451 }
6452 }
6453 return 0;
6454}
6455
6456/*
6457 * Wait for a packet..
6458 * Note: This function is the same function as in core/datagram.c
6459 * with a few modifications to make lksctp work.
6460 */
6461static int sctp_wait_for_packet(struct sock *sk, int *err, long *timeo_p)
6462{
6463 int error;
6464 DEFINE_WAIT(wait);
6465
6466 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6467
6468 /* Socket errors? */
6469 error = sock_error(sk);
6470 if (error)
6471 goto out;
6472
6473 if (!skb_queue_empty(&sk->sk_receive_queue))
6474 goto ready;
6475
6476 /* Socket shut down? */
6477 if (sk->sk_shutdown & RCV_SHUTDOWN)
6478 goto out;
6479
6480 /* Sequenced packets can come disconnected. If so we report the
6481 * problem.
6482 */
6483 error = -ENOTCONN;
6484
6485 /* Is there a good reason to think that we may receive some data? */
6486 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6487 goto out;
6488
6489 /* Handle signals. */
6490 if (signal_pending(current))
6491 goto interrupted;
6492
6493 /* Let another process have a go. Since we are going to sleep
6494 * anyway. Note: This may cause odd behaviors if the message
6495 * does not fit in the user's buffer, but this seems to be the
6496 * only way to honor MSG_DONTWAIT realistically.
6497 */
6498 release_sock(sk);
6499 *timeo_p = schedule_timeout(*timeo_p);
6500 lock_sock(sk);
6501
6502ready:
6503 finish_wait(sk_sleep(sk), &wait);
6504 return 0;
6505
6506interrupted:
6507 error = sock_intr_errno(*timeo_p);
6508
6509out:
6510 finish_wait(sk_sleep(sk), &wait);
6511 *err = error;
6512 return error;
6513}
6514
6515/* Receive a datagram.
6516 * Note: This is pretty much the same routine as in core/datagram.c
6517 * with a few changes to make lksctp work.
6518 */
6519static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6520 int noblock, int *err)
6521{
6522 int error;
6523 struct sk_buff *skb;
6524 long timeo;
6525
6526 timeo = sock_rcvtimeo(sk, noblock);
6527
6528 pr_debug("%s: timeo:%ld, max:%ld\n", __func__, timeo,
6529 MAX_SCHEDULE_TIMEOUT);
6530
6531 do {
6532 /* Again only user level code calls this function,
6533 * so nothing interrupt level
6534 * will suddenly eat the receive_queue.
6535 *
6536 * Look at current nfs client by the way...
6537 * However, this function was correct in any case. 8)
6538 */
6539 if (flags & MSG_PEEK) {
6540 spin_lock_bh(&sk->sk_receive_queue.lock);
6541 skb = skb_peek(&sk->sk_receive_queue);
6542 if (skb)
6543 atomic_inc(&skb->users);
6544 spin_unlock_bh(&sk->sk_receive_queue.lock);
6545 } else {
6546 skb = skb_dequeue(&sk->sk_receive_queue);
6547 }
6548
6549 if (skb)
6550 return skb;
6551
6552 /* Caller is allowed not to check sk->sk_err before calling. */
6553 error = sock_error(sk);
6554 if (error)
6555 goto no_packet;
6556
6557 if (sk->sk_shutdown & RCV_SHUTDOWN)
6558 break;
6559
6560 /* User doesn't want to wait. */
6561 error = -EAGAIN;
6562 if (!timeo)
6563 goto no_packet;
6564 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6565
6566 return NULL;
6567
6568no_packet:
6569 *err = error;
6570 return NULL;
6571}
6572
6573/* If sndbuf has changed, wake up per association sndbuf waiters. */
6574static void __sctp_write_space(struct sctp_association *asoc)
6575{
6576 struct sock *sk = asoc->base.sk;
6577 struct socket *sock = sk->sk_socket;
6578
6579 if ((sctp_wspace(asoc) > 0) && sock) {
6580 if (waitqueue_active(&asoc->wait))
6581 wake_up_interruptible(&asoc->wait);
6582
6583 if (sctp_writeable(sk)) {
6584 wait_queue_head_t *wq = sk_sleep(sk);
6585
6586 if (wq && waitqueue_active(wq))
6587 wake_up_interruptible(wq);
6588
6589 /* Note that we try to include the Async I/O support
6590 * here by modeling from the current TCP/UDP code.
6591 * We have not tested with it yet.
6592 */
6593 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6594 sock_wake_async(sock,
6595 SOCK_WAKE_SPACE, POLL_OUT);
6596 }
6597 }
6598}
6599
6600static void sctp_wake_up_waiters(struct sock *sk,
6601 struct sctp_association *asoc)
6602{
6603 struct sctp_association *tmp = asoc;
6604
6605 /* We do accounting for the sndbuf space per association,
6606 * so we only need to wake our own association.
6607 */
6608 if (asoc->ep->sndbuf_policy)
6609 return __sctp_write_space(asoc);
6610
6611 /* If association goes down and is just flushing its
6612 * outq, then just normally notify others.
6613 */
6614 if (asoc->base.dead)
6615 return sctp_write_space(sk);
6616
6617 /* Accounting for the sndbuf space is per socket, so we
6618 * need to wake up others, try to be fair and in case of
6619 * other associations, let them have a go first instead
6620 * of just doing a sctp_write_space() call.
6621 *
6622 * Note that we reach sctp_wake_up_waiters() only when
6623 * associations free up queued chunks, thus we are under
6624 * lock and the list of associations on a socket is
6625 * guaranteed not to change.
6626 */
6627 for (tmp = list_next_entry(tmp, asocs); 1;
6628 tmp = list_next_entry(tmp, asocs)) {
6629 /* Manually skip the head element. */
6630 if (&tmp->asocs == &((sctp_sk(sk))->ep->asocs))
6631 continue;
6632 /* Wake up association. */
6633 __sctp_write_space(tmp);
6634 /* We've reached the end. */
6635 if (tmp == asoc)
6636 break;
6637 }
6638}
6639
6640/* Do accounting for the sndbuf space.
6641 * Decrement the used sndbuf space of the corresponding association by the
6642 * data size which was just transmitted(freed).
6643 */
6644static void sctp_wfree(struct sk_buff *skb)
6645{
6646 struct sctp_association *asoc;
6647 struct sctp_chunk *chunk;
6648 struct sock *sk;
6649
6650 /* Get the saved chunk pointer. */
6651 chunk = *((struct sctp_chunk **)(skb->cb));
6652 asoc = chunk->asoc;
6653 sk = asoc->base.sk;
6654 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6655 sizeof(struct sk_buff) +
6656 sizeof(struct sctp_chunk);
6657
6658 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6659
6660 /*
6661 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6662 */
6663 sk->sk_wmem_queued -= skb->truesize;
6664 sk_mem_uncharge(sk, skb->truesize);
6665
6666 sock_wfree(skb);
6667 sctp_wake_up_waiters(sk, asoc);
6668
6669 sctp_association_put(asoc);
6670}
6671
6672/* Do accounting for the receive space on the socket.
6673 * Accounting for the association is done in ulpevent.c
6674 * We set this as a destructor for the cloned data skbs so that
6675 * accounting is done at the correct time.
6676 */
6677void sctp_sock_rfree(struct sk_buff *skb)
6678{
6679 struct sock *sk = skb->sk;
6680 struct sctp_ulpevent *event = sctp_skb2event(skb);
6681
6682 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6683
6684 /*
6685 * Mimic the behavior of sock_rfree
6686 */
6687 sk_mem_uncharge(sk, event->rmem_len);
6688}
6689
6690
6691/* Helper function to wait for space in the sndbuf. */
6692static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6693 size_t msg_len)
6694{
6695 struct sock *sk = asoc->base.sk;
6696 int err = 0;
6697 long current_timeo = *timeo_p;
6698 DEFINE_WAIT(wait);
6699
6700 pr_debug("%s: asoc:%p, timeo:%ld, msg_len:%zu\n", __func__, asoc,
6701 *timeo_p, msg_len);
6702
6703 /* Increment the association's refcnt. */
6704 sctp_association_hold(asoc);
6705
6706 /* Wait on the association specific sndbuf space. */
6707 for (;;) {
6708 prepare_to_wait_exclusive(&asoc->wait, &wait,
6709 TASK_INTERRUPTIBLE);
6710 if (!*timeo_p)
6711 goto do_nonblock;
6712 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6713 asoc->base.dead)
6714 goto do_error;
6715 if (signal_pending(current))
6716 goto do_interrupted;
6717 if (msg_len <= sctp_wspace(asoc))
6718 break;
6719
6720 /* Let another process have a go. Since we are going
6721 * to sleep anyway.
6722 */
6723 release_sock(sk);
6724 current_timeo = schedule_timeout(current_timeo);
6725 BUG_ON(sk != asoc->base.sk);
6726 lock_sock(sk);
6727
6728 *timeo_p = current_timeo;
6729 }
6730
6731out:
6732 finish_wait(&asoc->wait, &wait);
6733
6734 /* Release the association's refcnt. */
6735 sctp_association_put(asoc);
6736
6737 return err;
6738
6739do_error:
6740 err = -EPIPE;
6741 goto out;
6742
6743do_interrupted:
6744 err = sock_intr_errno(*timeo_p);
6745 goto out;
6746
6747do_nonblock:
6748 err = -EAGAIN;
6749 goto out;
6750}
6751
6752void sctp_data_ready(struct sock *sk)
6753{
6754 struct socket_wq *wq;
6755
6756 rcu_read_lock();
6757 wq = rcu_dereference(sk->sk_wq);
6758 if (wq_has_sleeper(wq))
6759 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6760 POLLRDNORM | POLLRDBAND);
6761 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6762 rcu_read_unlock();
6763}
6764
6765/* If socket sndbuf has changed, wake up all per association waiters. */
6766void sctp_write_space(struct sock *sk)
6767{
6768 struct sctp_association *asoc;
6769
6770 /* Wake up the tasks in each wait queue. */
6771 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6772 __sctp_write_space(asoc);
6773 }
6774}
6775
6776/* Is there any sndbuf space available on the socket?
6777 *
6778 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6779 * associations on the same socket. For a UDP-style socket with
6780 * multiple associations, it is possible for it to be "unwriteable"
6781 * prematurely. I assume that this is acceptable because
6782 * a premature "unwriteable" is better than an accidental "writeable" which
6783 * would cause an unwanted block under certain circumstances. For the 1-1
6784 * UDP-style sockets or TCP-style sockets, this code should work.
6785 * - Daisy
6786 */
6787static int sctp_writeable(struct sock *sk)
6788{
6789 int amt = 0;
6790
6791 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6792 if (amt < 0)
6793 amt = 0;
6794 return amt;
6795}
6796
6797/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6798 * returns immediately with EINPROGRESS.
6799 */
6800static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6801{
6802 struct sock *sk = asoc->base.sk;
6803 int err = 0;
6804 long current_timeo = *timeo_p;
6805 DEFINE_WAIT(wait);
6806
6807 pr_debug("%s: asoc:%p, timeo:%ld\n", __func__, asoc, *timeo_p);
6808
6809 /* Increment the association's refcnt. */
6810 sctp_association_hold(asoc);
6811
6812 for (;;) {
6813 prepare_to_wait_exclusive(&asoc->wait, &wait,
6814 TASK_INTERRUPTIBLE);
6815 if (!*timeo_p)
6816 goto do_nonblock;
6817 if (sk->sk_shutdown & RCV_SHUTDOWN)
6818 break;
6819 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6820 asoc->base.dead)
6821 goto do_error;
6822 if (signal_pending(current))
6823 goto do_interrupted;
6824
6825 if (sctp_state(asoc, ESTABLISHED))
6826 break;
6827
6828 /* Let another process have a go. Since we are going
6829 * to sleep anyway.
6830 */
6831 release_sock(sk);
6832 current_timeo = schedule_timeout(current_timeo);
6833 lock_sock(sk);
6834
6835 *timeo_p = current_timeo;
6836 }
6837
6838out:
6839 finish_wait(&asoc->wait, &wait);
6840
6841 /* Release the association's refcnt. */
6842 sctp_association_put(asoc);
6843
6844 return err;
6845
6846do_error:
6847 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6848 err = -ETIMEDOUT;
6849 else
6850 err = -ECONNREFUSED;
6851 goto out;
6852
6853do_interrupted:
6854 err = sock_intr_errno(*timeo_p);
6855 goto out;
6856
6857do_nonblock:
6858 err = -EINPROGRESS;
6859 goto out;
6860}
6861
6862static int sctp_wait_for_accept(struct sock *sk, long timeo)
6863{
6864 struct sctp_endpoint *ep;
6865 int err = 0;
6866 DEFINE_WAIT(wait);
6867
6868 ep = sctp_sk(sk)->ep;
6869
6870
6871 for (;;) {
6872 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6873 TASK_INTERRUPTIBLE);
6874
6875 if (list_empty(&ep->asocs)) {
6876 release_sock(sk);
6877 timeo = schedule_timeout(timeo);
6878 lock_sock(sk);
6879 }
6880
6881 err = -EINVAL;
6882 if (!sctp_sstate(sk, LISTENING))
6883 break;
6884
6885 err = 0;
6886 if (!list_empty(&ep->asocs))
6887 break;
6888
6889 err = sock_intr_errno(timeo);
6890 if (signal_pending(current))
6891 break;
6892
6893 err = -EAGAIN;
6894 if (!timeo)
6895 break;
6896 }
6897
6898 finish_wait(sk_sleep(sk), &wait);
6899
6900 return err;
6901}
6902
6903static void sctp_wait_for_close(struct sock *sk, long timeout)
6904{
6905 DEFINE_WAIT(wait);
6906
6907 do {
6908 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6909 if (list_empty(&sctp_sk(sk)->ep->asocs))
6910 break;
6911 release_sock(sk);
6912 timeout = schedule_timeout(timeout);
6913 lock_sock(sk);
6914 } while (!signal_pending(current) && timeout);
6915
6916 finish_wait(sk_sleep(sk), &wait);
6917}
6918
6919static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6920{
6921 struct sk_buff *frag;
6922
6923 if (!skb->data_len)
6924 goto done;
6925
6926 /* Don't forget the fragments. */
6927 skb_walk_frags(skb, frag)
6928 sctp_skb_set_owner_r_frag(frag, sk);
6929
6930done:
6931 sctp_skb_set_owner_r(skb, sk);
6932}
6933
6934void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6935 struct sctp_association *asoc)
6936{
6937 struct inet_sock *inet = inet_sk(sk);
6938 struct inet_sock *newinet;
6939
6940 newsk->sk_type = sk->sk_type;
6941 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6942 newsk->sk_flags = sk->sk_flags;
6943 newsk->sk_no_check = sk->sk_no_check;
6944 newsk->sk_reuse = sk->sk_reuse;
6945
6946 newsk->sk_shutdown = sk->sk_shutdown;
6947 newsk->sk_destruct = sctp_destruct_sock;
6948 newsk->sk_family = sk->sk_family;
6949 newsk->sk_protocol = IPPROTO_SCTP;
6950 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6951 newsk->sk_sndbuf = sk->sk_sndbuf;
6952 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6953 newsk->sk_lingertime = sk->sk_lingertime;
6954 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6955 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6956
6957 newinet = inet_sk(newsk);
6958
6959 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6960 * getsockname() and getpeername()
6961 */
6962 newinet->inet_sport = inet->inet_sport;
6963 newinet->inet_saddr = inet->inet_saddr;
6964 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6965 newinet->inet_dport = htons(asoc->peer.port);
6966 newinet->pmtudisc = inet->pmtudisc;
6967 newinet->inet_id = asoc->next_tsn ^ jiffies;
6968
6969 newinet->uc_ttl = inet->uc_ttl;
6970 newinet->mc_loop = 1;
6971 newinet->mc_ttl = 1;
6972 newinet->mc_index = 0;
6973 newinet->mc_list = NULL;
6974}
6975
6976/* Populate the fields of the newsk from the oldsk and migrate the assoc
6977 * and its messages to the newsk.
6978 */
6979static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6980 struct sctp_association *assoc,
6981 sctp_socket_type_t type)
6982{
6983 struct sctp_sock *oldsp = sctp_sk(oldsk);
6984 struct sctp_sock *newsp = sctp_sk(newsk);
6985 struct sctp_bind_bucket *pp; /* hash list port iterator */
6986 struct sctp_endpoint *newep = newsp->ep;
6987 struct sk_buff *skb, *tmp;
6988 struct sctp_ulpevent *event;
6989 struct sctp_bind_hashbucket *head;
6990 struct list_head tmplist;
6991
6992 /* Migrate socket buffer sizes and all the socket level options to the
6993 * new socket.
6994 */
6995 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6996 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6997 /* Brute force copy old sctp opt. */
6998 if (oldsp->do_auto_asconf) {
6999 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
7000 inet_sk_copy_descendant(newsk, oldsk);
7001 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
7002 } else
7003 inet_sk_copy_descendant(newsk, oldsk);
7004
7005 /* Restore the ep value that was overwritten with the above structure
7006 * copy.
7007 */
7008 newsp->ep = newep;
7009 newsp->hmac = NULL;
7010
7011 /* Hook this new socket in to the bind_hash list. */
7012 head = &sctp_port_hashtable[sctp_phashfn(sock_net(oldsk),
7013 inet_sk(oldsk)->inet_num)];
7014 local_bh_disable();
7015 spin_lock(&head->lock);
7016 pp = sctp_sk(oldsk)->bind_hash;
7017 sk_add_bind_node(newsk, &pp->owner);
7018 sctp_sk(newsk)->bind_hash = pp;
7019 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
7020 spin_unlock(&head->lock);
7021 local_bh_enable();
7022
7023 /* Copy the bind_addr list from the original endpoint to the new
7024 * endpoint so that we can handle restarts properly
7025 */
7026 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
7027 &oldsp->ep->base.bind_addr, GFP_KERNEL);
7028
7029 /* Move any messages in the old socket's receive queue that are for the
7030 * peeled off association to the new socket's receive queue.
7031 */
7032 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
7033 event = sctp_skb2event(skb);
7034 if (event->asoc == assoc) {
7035 __skb_unlink(skb, &oldsk->sk_receive_queue);
7036 __skb_queue_tail(&newsk->sk_receive_queue, skb);
7037 sctp_skb_set_owner_r_frag(skb, newsk);
7038 }
7039 }
7040
7041 /* Clean up any messages pending delivery due to partial
7042 * delivery. Three cases:
7043 * 1) No partial deliver; no work.
7044 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
7045 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
7046 */
7047 skb_queue_head_init(&newsp->pd_lobby);
7048 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
7049
7050 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
7051 struct sk_buff_head *queue;
7052
7053 /* Decide which queue to move pd_lobby skbs to. */
7054 if (assoc->ulpq.pd_mode) {
7055 queue = &newsp->pd_lobby;
7056 } else
7057 queue = &newsk->sk_receive_queue;
7058
7059 /* Walk through the pd_lobby, looking for skbs that
7060 * need moved to the new socket.
7061 */
7062 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
7063 event = sctp_skb2event(skb);
7064 if (event->asoc == assoc) {
7065 __skb_unlink(skb, &oldsp->pd_lobby);
7066 __skb_queue_tail(queue, skb);
7067 sctp_skb_set_owner_r_frag(skb, newsk);
7068 }
7069 }
7070
7071 /* Clear up any skbs waiting for the partial
7072 * delivery to finish.
7073 */
7074 if (assoc->ulpq.pd_mode)
7075 sctp_clear_pd(oldsk, NULL);
7076
7077 }
7078
7079 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
7080 sctp_skb_set_owner_r_frag(skb, newsk);
7081
7082 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
7083 sctp_skb_set_owner_r_frag(skb, newsk);
7084
7085 /* Set the type of socket to indicate that it is peeled off from the
7086 * original UDP-style socket or created with the accept() call on a
7087 * TCP-style socket..
7088 */
7089 newsp->type = type;
7090
7091 /* Mark the new socket "in-use" by the user so that any packets
7092 * that may arrive on the association after we've moved it are
7093 * queued to the backlog. This prevents a potential race between
7094 * backlog processing on the old socket and new-packet processing
7095 * on the new socket.
7096 *
7097 * The caller has just allocated newsk so we can guarantee that other
7098 * paths won't try to lock it and then oldsk.
7099 */
7100 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
7101 sctp_assoc_migrate(assoc, newsk);
7102
7103 /* If the association on the newsk is already closed before accept()
7104 * is called, set RCV_SHUTDOWN flag.
7105 */
7106 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
7107 newsk->sk_shutdown |= RCV_SHUTDOWN;
7108
7109 newsk->sk_state = SCTP_SS_ESTABLISHED;
7110 release_sock(newsk);
7111}
7112
7113
7114/* This proto struct describes the ULP interface for SCTP. */
7115struct proto sctp_prot = {
7116 .name = "SCTP",
7117 .owner = THIS_MODULE,
7118 .close = sctp_close,
7119 .connect = sctp_connect,
7120 .disconnect = sctp_disconnect,
7121 .accept = sctp_accept,
7122 .ioctl = sctp_ioctl,
7123 .init = sctp_init_sock,
7124 .destroy = sctp_destroy_sock,
7125 .shutdown = sctp_shutdown,
7126 .setsockopt = sctp_setsockopt,
7127 .getsockopt = sctp_getsockopt,
7128 .sendmsg = sctp_sendmsg,
7129 .recvmsg = sctp_recvmsg,
7130 .bind = sctp_bind,
7131 .backlog_rcv = sctp_backlog_rcv,
7132 .hash = sctp_hash,
7133 .unhash = sctp_unhash,
7134 .get_port = sctp_get_port,
7135 .obj_size = sizeof(struct sctp_sock),
7136 .sysctl_mem = sysctl_sctp_mem,
7137 .sysctl_rmem = sysctl_sctp_rmem,
7138 .sysctl_wmem = sysctl_sctp_wmem,
7139 .memory_pressure = &sctp_memory_pressure,
7140 .enter_memory_pressure = sctp_enter_memory_pressure,
7141 .memory_allocated = &sctp_memory_allocated,
7142 .sockets_allocated = &sctp_sockets_allocated,
7143};
7144
7145#if IS_ENABLED(CONFIG_IPV6)
7146
7147struct proto sctpv6_prot = {
7148 .name = "SCTPv6",
7149 .owner = THIS_MODULE,
7150 .close = sctp_close,
7151 .connect = sctp_connect,
7152 .disconnect = sctp_disconnect,
7153 .accept = sctp_accept,
7154 .ioctl = sctp_ioctl,
7155 .init = sctp_init_sock,
7156 .destroy = sctp_destroy_sock,
7157 .shutdown = sctp_shutdown,
7158 .setsockopt = sctp_setsockopt,
7159 .getsockopt = sctp_getsockopt,
7160 .sendmsg = sctp_sendmsg,
7161 .recvmsg = sctp_recvmsg,
7162 .bind = sctp_bind,
7163 .backlog_rcv = sctp_backlog_rcv,
7164 .hash = sctp_hash,
7165 .unhash = sctp_unhash,
7166 .get_port = sctp_get_port,
7167 .obj_size = sizeof(struct sctp6_sock),
7168 .sysctl_mem = sysctl_sctp_mem,
7169 .sysctl_rmem = sysctl_sctp_rmem,
7170 .sysctl_wmem = sysctl_sctp_wmem,
7171 .memory_pressure = &sctp_memory_pressure,
7172 .enter_memory_pressure = sctp_enter_memory_pressure,
7173 .memory_allocated = &sctp_memory_allocated,
7174 .sockets_allocated = &sctp_sockets_allocated,
7175};
7176#endif /* IS_ENABLED(CONFIG_IPV6) */
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
8 *
9 * This file is part of the SCTP kernel implementation
10 *
11 * These functions interface with the sockets layer to implement the
12 * SCTP Extensions for the Sockets API.
13 *
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.
17 *
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.
23 *
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.
29 *
30 * You should have received a copy of the GNU General Public License
31 * along with GNU CC; see the file COPYING. If not, write to
32 * the Free Software Foundation, 59 Temple Place - Suite 330,
33 * Boston, MA 02111-1307, USA.
34 *
35 * Please send any bug reports or fixes you make to the
36 * email address(es):
37 * lksctp developers <lksctp-developers@lists.sourceforge.net>
38 *
39 * Or submit a bug report through the following website:
40 * http://www.sf.net/projects/lksctp
41 *
42 * Written or modified by:
43 * La Monte H.P. Yarroll <piggy@acm.org>
44 * Narasimha Budihal <narsi@refcode.org>
45 * Karl Knutson <karl@athena.chicago.il.us>
46 * Jon Grimm <jgrimm@us.ibm.com>
47 * Xingang Guo <xingang.guo@intel.com>
48 * Daisy Chang <daisyc@us.ibm.com>
49 * Sridhar Samudrala <samudrala@us.ibm.com>
50 * Inaky Perez-Gonzalez <inaky.gonzalez@intel.com>
51 * Ardelle Fan <ardelle.fan@intel.com>
52 * Ryan Layer <rmlayer@us.ibm.com>
53 * Anup Pemmaiah <pemmaiah@cc.usu.edu>
54 * Kevin Gao <kevin.gao@intel.com>
55 *
56 * Any bugs reported given to us we will try to fix... any fixes shared will
57 * be incorporated into the next SCTP release.
58 */
59
60#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
61
62#include <linux/types.h>
63#include <linux/kernel.h>
64#include <linux/wait.h>
65#include <linux/time.h>
66#include <linux/ip.h>
67#include <linux/capability.h>
68#include <linux/fcntl.h>
69#include <linux/poll.h>
70#include <linux/init.h>
71#include <linux/crypto.h>
72#include <linux/slab.h>
73
74#include <net/ip.h>
75#include <net/icmp.h>
76#include <net/route.h>
77#include <net/ipv6.h>
78#include <net/inet_common.h>
79
80#include <linux/socket.h> /* for sa_family_t */
81#include <net/sock.h>
82#include <net/sctp/sctp.h>
83#include <net/sctp/sm.h>
84
85/* WARNING: Please do not remove the SCTP_STATIC attribute to
86 * any of the functions below as they are used to export functions
87 * used by a project regression testsuite.
88 */
89
90/* Forward declarations for internal helper functions. */
91static int sctp_writeable(struct sock *sk);
92static void sctp_wfree(struct sk_buff *skb);
93static int sctp_wait_for_sndbuf(struct sctp_association *, long *timeo_p,
94 size_t msg_len);
95static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p);
96static int sctp_wait_for_connect(struct sctp_association *, long *timeo_p);
97static int sctp_wait_for_accept(struct sock *sk, long timeo);
98static void sctp_wait_for_close(struct sock *sk, long timeo);
99static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
100 union sctp_addr *addr, int len);
101static int sctp_bindx_add(struct sock *, struct sockaddr *, int);
102static int sctp_bindx_rem(struct sock *, struct sockaddr *, int);
103static int sctp_send_asconf_add_ip(struct sock *, struct sockaddr *, int);
104static int sctp_send_asconf_del_ip(struct sock *, struct sockaddr *, int);
105static int sctp_send_asconf(struct sctp_association *asoc,
106 struct sctp_chunk *chunk);
107static int sctp_do_bind(struct sock *, union sctp_addr *, int);
108static int sctp_autobind(struct sock *sk);
109static void sctp_sock_migrate(struct sock *, struct sock *,
110 struct sctp_association *, sctp_socket_type_t);
111static char *sctp_hmac_alg = SCTP_COOKIE_HMAC_ALG;
112
113extern struct kmem_cache *sctp_bucket_cachep;
114extern long sysctl_sctp_mem[3];
115extern int sysctl_sctp_rmem[3];
116extern int sysctl_sctp_wmem[3];
117
118static int sctp_memory_pressure;
119static atomic_long_t sctp_memory_allocated;
120struct percpu_counter sctp_sockets_allocated;
121
122static void sctp_enter_memory_pressure(struct sock *sk)
123{
124 sctp_memory_pressure = 1;
125}
126
127
128/* Get the sndbuf space available at the time on the association. */
129static inline int sctp_wspace(struct sctp_association *asoc)
130{
131 int amt;
132
133 if (asoc->ep->sndbuf_policy)
134 amt = asoc->sndbuf_used;
135 else
136 amt = sk_wmem_alloc_get(asoc->base.sk);
137
138 if (amt >= asoc->base.sk->sk_sndbuf) {
139 if (asoc->base.sk->sk_userlocks & SOCK_SNDBUF_LOCK)
140 amt = 0;
141 else {
142 amt = sk_stream_wspace(asoc->base.sk);
143 if (amt < 0)
144 amt = 0;
145 }
146 } else {
147 amt = asoc->base.sk->sk_sndbuf - amt;
148 }
149 return amt;
150}
151
152/* Increment the used sndbuf space count of the corresponding association by
153 * the size of the outgoing data chunk.
154 * Also, set the skb destructor for sndbuf accounting later.
155 *
156 * Since it is always 1-1 between chunk and skb, and also a new skb is always
157 * allocated for chunk bundling in sctp_packet_transmit(), we can use the
158 * destructor in the data chunk skb for the purpose of the sndbuf space
159 * tracking.
160 */
161static inline void sctp_set_owner_w(struct sctp_chunk *chunk)
162{
163 struct sctp_association *asoc = chunk->asoc;
164 struct sock *sk = asoc->base.sk;
165
166 /* The sndbuf space is tracked per association. */
167 sctp_association_hold(asoc);
168
169 skb_set_owner_w(chunk->skb, sk);
170
171 chunk->skb->destructor = sctp_wfree;
172 /* Save the chunk pointer in skb for sctp_wfree to use later. */
173 *((struct sctp_chunk **)(chunk->skb->cb)) = chunk;
174
175 asoc->sndbuf_used += SCTP_DATA_SNDSIZE(chunk) +
176 sizeof(struct sk_buff) +
177 sizeof(struct sctp_chunk);
178
179 atomic_add(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
180 sk->sk_wmem_queued += chunk->skb->truesize;
181 sk_mem_charge(sk, chunk->skb->truesize);
182}
183
184/* Verify that this is a valid address. */
185static inline int sctp_verify_addr(struct sock *sk, union sctp_addr *addr,
186 int len)
187{
188 struct sctp_af *af;
189
190 /* Verify basic sockaddr. */
191 af = sctp_sockaddr_af(sctp_sk(sk), addr, len);
192 if (!af)
193 return -EINVAL;
194
195 /* Is this a valid SCTP address? */
196 if (!af->addr_valid(addr, sctp_sk(sk), NULL))
197 return -EINVAL;
198
199 if (!sctp_sk(sk)->pf->send_verify(sctp_sk(sk), (addr)))
200 return -EINVAL;
201
202 return 0;
203}
204
205/* Look up the association by its id. If this is not a UDP-style
206 * socket, the ID field is always ignored.
207 */
208struct sctp_association *sctp_id2assoc(struct sock *sk, sctp_assoc_t id)
209{
210 struct sctp_association *asoc = NULL;
211
212 /* If this is not a UDP-style socket, assoc id should be ignored. */
213 if (!sctp_style(sk, UDP)) {
214 /* Return NULL if the socket state is not ESTABLISHED. It
215 * could be a TCP-style listening socket or a socket which
216 * hasn't yet called connect() to establish an association.
217 */
218 if (!sctp_sstate(sk, ESTABLISHED))
219 return NULL;
220
221 /* Get the first and the only association from the list. */
222 if (!list_empty(&sctp_sk(sk)->ep->asocs))
223 asoc = list_entry(sctp_sk(sk)->ep->asocs.next,
224 struct sctp_association, asocs);
225 return asoc;
226 }
227
228 /* Otherwise this is a UDP-style socket. */
229 if (!id || (id == (sctp_assoc_t)-1))
230 return NULL;
231
232 spin_lock_bh(&sctp_assocs_id_lock);
233 asoc = (struct sctp_association *)idr_find(&sctp_assocs_id, (int)id);
234 spin_unlock_bh(&sctp_assocs_id_lock);
235
236 if (!asoc || (asoc->base.sk != sk) || asoc->base.dead)
237 return NULL;
238
239 return asoc;
240}
241
242/* Look up the transport from an address and an assoc id. If both address and
243 * id are specified, the associations matching the address and the id should be
244 * the same.
245 */
246static struct sctp_transport *sctp_addr_id2transport(struct sock *sk,
247 struct sockaddr_storage *addr,
248 sctp_assoc_t id)
249{
250 struct sctp_association *addr_asoc = NULL, *id_asoc = NULL;
251 struct sctp_transport *transport;
252 union sctp_addr *laddr = (union sctp_addr *)addr;
253
254 addr_asoc = sctp_endpoint_lookup_assoc(sctp_sk(sk)->ep,
255 laddr,
256 &transport);
257
258 if (!addr_asoc)
259 return NULL;
260
261 id_asoc = sctp_id2assoc(sk, id);
262 if (id_asoc && (id_asoc != addr_asoc))
263 return NULL;
264
265 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
266 (union sctp_addr *)addr);
267
268 return transport;
269}
270
271/* API 3.1.2 bind() - UDP Style Syntax
272 * The syntax of bind() is,
273 *
274 * ret = bind(int sd, struct sockaddr *addr, int addrlen);
275 *
276 * sd - the socket descriptor returned by socket().
277 * addr - the address structure (struct sockaddr_in or struct
278 * sockaddr_in6 [RFC 2553]),
279 * addr_len - the size of the address structure.
280 */
281SCTP_STATIC int sctp_bind(struct sock *sk, struct sockaddr *addr, int addr_len)
282{
283 int retval = 0;
284
285 sctp_lock_sock(sk);
286
287 SCTP_DEBUG_PRINTK("sctp_bind(sk: %p, addr: %p, addr_len: %d)\n",
288 sk, addr, addr_len);
289
290 /* Disallow binding twice. */
291 if (!sctp_sk(sk)->ep->base.bind_addr.port)
292 retval = sctp_do_bind(sk, (union sctp_addr *)addr,
293 addr_len);
294 else
295 retval = -EINVAL;
296
297 sctp_release_sock(sk);
298
299 return retval;
300}
301
302static long sctp_get_port_local(struct sock *, union sctp_addr *);
303
304/* Verify this is a valid sockaddr. */
305static struct sctp_af *sctp_sockaddr_af(struct sctp_sock *opt,
306 union sctp_addr *addr, int len)
307{
308 struct sctp_af *af;
309
310 /* Check minimum size. */
311 if (len < sizeof (struct sockaddr))
312 return NULL;
313
314 /* V4 mapped address are really of AF_INET family */
315 if (addr->sa.sa_family == AF_INET6 &&
316 ipv6_addr_v4mapped(&addr->v6.sin6_addr)) {
317 if (!opt->pf->af_supported(AF_INET, opt))
318 return NULL;
319 } else {
320 /* Does this PF support this AF? */
321 if (!opt->pf->af_supported(addr->sa.sa_family, opt))
322 return NULL;
323 }
324
325 /* If we get this far, af is valid. */
326 af = sctp_get_af_specific(addr->sa.sa_family);
327
328 if (len < af->sockaddr_len)
329 return NULL;
330
331 return af;
332}
333
334/* Bind a local address either to an endpoint or to an association. */
335SCTP_STATIC int sctp_do_bind(struct sock *sk, union sctp_addr *addr, int len)
336{
337 struct sctp_sock *sp = sctp_sk(sk);
338 struct sctp_endpoint *ep = sp->ep;
339 struct sctp_bind_addr *bp = &ep->base.bind_addr;
340 struct sctp_af *af;
341 unsigned short snum;
342 int ret = 0;
343
344 /* Common sockaddr verification. */
345 af = sctp_sockaddr_af(sp, addr, len);
346 if (!af) {
347 SCTP_DEBUG_PRINTK("sctp_do_bind(sk: %p, newaddr: %p, len: %d) EINVAL\n",
348 sk, addr, len);
349 return -EINVAL;
350 }
351
352 snum = ntohs(addr->v4.sin_port);
353
354 SCTP_DEBUG_PRINTK_IPADDR("sctp_do_bind(sk: %p, new addr: ",
355 ", port: %d, new port: %d, len: %d)\n",
356 sk,
357 addr,
358 bp->port, snum,
359 len);
360
361 /* PF specific bind() address verification. */
362 if (!sp->pf->bind_verify(sp, addr))
363 return -EADDRNOTAVAIL;
364
365 /* We must either be unbound, or bind to the same port.
366 * It's OK to allow 0 ports if we are already bound.
367 * We'll just inhert an already bound port in this case
368 */
369 if (bp->port) {
370 if (!snum)
371 snum = bp->port;
372 else if (snum != bp->port) {
373 SCTP_DEBUG_PRINTK("sctp_do_bind:"
374 " New port %d does not match existing port "
375 "%d.\n", snum, bp->port);
376 return -EINVAL;
377 }
378 }
379
380 if (snum && snum < PROT_SOCK && !capable(CAP_NET_BIND_SERVICE))
381 return -EACCES;
382
383 /* See if the address matches any of the addresses we may have
384 * already bound before checking against other endpoints.
385 */
386 if (sctp_bind_addr_match(bp, addr, sp))
387 return -EINVAL;
388
389 /* Make sure we are allowed to bind here.
390 * The function sctp_get_port_local() does duplicate address
391 * detection.
392 */
393 addr->v4.sin_port = htons(snum);
394 if ((ret = sctp_get_port_local(sk, addr))) {
395 return -EADDRINUSE;
396 }
397
398 /* Refresh ephemeral port. */
399 if (!bp->port)
400 bp->port = inet_sk(sk)->inet_num;
401
402 /* Add the address to the bind address list.
403 * Use GFP_ATOMIC since BHs will be disabled.
404 */
405 ret = sctp_add_bind_addr(bp, addr, SCTP_ADDR_SRC, GFP_ATOMIC);
406
407 /* Copy back into socket for getsockname() use. */
408 if (!ret) {
409 inet_sk(sk)->inet_sport = htons(inet_sk(sk)->inet_num);
410 af->to_sk_saddr(addr, sk);
411 }
412
413 return ret;
414}
415
416 /* ADDIP Section 4.1.1 Congestion Control of ASCONF Chunks
417 *
418 * R1) One and only one ASCONF Chunk MAY be in transit and unacknowledged
419 * at any one time. If a sender, after sending an ASCONF chunk, decides
420 * it needs to transfer another ASCONF Chunk, it MUST wait until the
421 * ASCONF-ACK Chunk returns from the previous ASCONF Chunk before sending a
422 * subsequent ASCONF. Note this restriction binds each side, so at any
423 * time two ASCONF may be in-transit on any given association (one sent
424 * from each endpoint).
425 */
426static int sctp_send_asconf(struct sctp_association *asoc,
427 struct sctp_chunk *chunk)
428{
429 int retval = 0;
430
431 /* If there is an outstanding ASCONF chunk, queue it for later
432 * transmission.
433 */
434 if (asoc->addip_last_asconf) {
435 list_add_tail(&chunk->list, &asoc->addip_chunk_list);
436 goto out;
437 }
438
439 /* Hold the chunk until an ASCONF_ACK is received. */
440 sctp_chunk_hold(chunk);
441 retval = sctp_primitive_ASCONF(asoc, chunk);
442 if (retval)
443 sctp_chunk_free(chunk);
444 else
445 asoc->addip_last_asconf = chunk;
446
447out:
448 return retval;
449}
450
451/* Add a list of addresses as bind addresses to local endpoint or
452 * association.
453 *
454 * Basically run through each address specified in the addrs/addrcnt
455 * array/length pair, determine if it is IPv6 or IPv4 and call
456 * sctp_do_bind() on it.
457 *
458 * If any of them fails, then the operation will be reversed and the
459 * ones that were added will be removed.
460 *
461 * Only sctp_setsockopt_bindx() is supposed to call this function.
462 */
463static int sctp_bindx_add(struct sock *sk, struct sockaddr *addrs, int addrcnt)
464{
465 int cnt;
466 int retval = 0;
467 void *addr_buf;
468 struct sockaddr *sa_addr;
469 struct sctp_af *af;
470
471 SCTP_DEBUG_PRINTK("sctp_bindx_add (sk: %p, addrs: %p, addrcnt: %d)\n",
472 sk, addrs, addrcnt);
473
474 addr_buf = addrs;
475 for (cnt = 0; cnt < addrcnt; cnt++) {
476 /* The list may contain either IPv4 or IPv6 address;
477 * determine the address length for walking thru the list.
478 */
479 sa_addr = addr_buf;
480 af = sctp_get_af_specific(sa_addr->sa_family);
481 if (!af) {
482 retval = -EINVAL;
483 goto err_bindx_add;
484 }
485
486 retval = sctp_do_bind(sk, (union sctp_addr *)sa_addr,
487 af->sockaddr_len);
488
489 addr_buf += af->sockaddr_len;
490
491err_bindx_add:
492 if (retval < 0) {
493 /* Failed. Cleanup the ones that have been added */
494 if (cnt > 0)
495 sctp_bindx_rem(sk, addrs, cnt);
496 return retval;
497 }
498 }
499
500 return retval;
501}
502
503/* Send an ASCONF chunk with Add IP address parameters to all the peers of the
504 * associations that are part of the endpoint indicating that a list of local
505 * addresses are added to the endpoint.
506 *
507 * If any of the addresses is already in the bind address list of the
508 * association, we do not send the chunk for that association. But it will not
509 * affect other associations.
510 *
511 * Only sctp_setsockopt_bindx() is supposed to call this function.
512 */
513static int sctp_send_asconf_add_ip(struct sock *sk,
514 struct sockaddr *addrs,
515 int addrcnt)
516{
517 struct sctp_sock *sp;
518 struct sctp_endpoint *ep;
519 struct sctp_association *asoc;
520 struct sctp_bind_addr *bp;
521 struct sctp_chunk *chunk;
522 struct sctp_sockaddr_entry *laddr;
523 union sctp_addr *addr;
524 union sctp_addr saveaddr;
525 void *addr_buf;
526 struct sctp_af *af;
527 struct list_head *p;
528 int i;
529 int retval = 0;
530
531 if (!sctp_addip_enable)
532 return retval;
533
534 sp = sctp_sk(sk);
535 ep = sp->ep;
536
537 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
538 __func__, sk, addrs, addrcnt);
539
540 list_for_each_entry(asoc, &ep->asocs, asocs) {
541
542 if (!asoc->peer.asconf_capable)
543 continue;
544
545 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_ADD_IP)
546 continue;
547
548 if (!sctp_state(asoc, ESTABLISHED))
549 continue;
550
551 /* Check if any address in the packed array of addresses is
552 * in the bind address list of the association. If so,
553 * do not send the asconf chunk to its peer, but continue with
554 * other associations.
555 */
556 addr_buf = addrs;
557 for (i = 0; i < addrcnt; i++) {
558 addr = addr_buf;
559 af = sctp_get_af_specific(addr->v4.sin_family);
560 if (!af) {
561 retval = -EINVAL;
562 goto out;
563 }
564
565 if (sctp_assoc_lookup_laddr(asoc, addr))
566 break;
567
568 addr_buf += af->sockaddr_len;
569 }
570 if (i < addrcnt)
571 continue;
572
573 /* Use the first valid address in bind addr list of
574 * association as Address Parameter of ASCONF CHUNK.
575 */
576 bp = &asoc->base.bind_addr;
577 p = bp->address_list.next;
578 laddr = list_entry(p, struct sctp_sockaddr_entry, list);
579 chunk = sctp_make_asconf_update_ip(asoc, &laddr->a, addrs,
580 addrcnt, SCTP_PARAM_ADD_IP);
581 if (!chunk) {
582 retval = -ENOMEM;
583 goto out;
584 }
585
586 /* Add the new addresses to the bind address list with
587 * use_as_src set to 0.
588 */
589 addr_buf = addrs;
590 for (i = 0; i < addrcnt; i++) {
591 addr = addr_buf;
592 af = sctp_get_af_specific(addr->v4.sin_family);
593 memcpy(&saveaddr, addr, af->sockaddr_len);
594 retval = sctp_add_bind_addr(bp, &saveaddr,
595 SCTP_ADDR_NEW, GFP_ATOMIC);
596 addr_buf += af->sockaddr_len;
597 }
598 if (asoc->src_out_of_asoc_ok) {
599 struct sctp_transport *trans;
600
601 list_for_each_entry(trans,
602 &asoc->peer.transport_addr_list, transports) {
603 /* Clear the source and route cache */
604 dst_release(trans->dst);
605 trans->cwnd = min(4*asoc->pathmtu, max_t(__u32,
606 2*asoc->pathmtu, 4380));
607 trans->ssthresh = asoc->peer.i.a_rwnd;
608 trans->rto = asoc->rto_initial;
609 trans->rtt = trans->srtt = trans->rttvar = 0;
610 sctp_transport_route(trans, NULL,
611 sctp_sk(asoc->base.sk));
612 }
613 }
614 retval = sctp_send_asconf(asoc, chunk);
615 }
616
617out:
618 return retval;
619}
620
621/* Remove a list of addresses from bind addresses list. Do not remove the
622 * last address.
623 *
624 * Basically run through each address specified in the addrs/addrcnt
625 * array/length pair, determine if it is IPv6 or IPv4 and call
626 * sctp_del_bind() on it.
627 *
628 * If any of them fails, then the operation will be reversed and the
629 * ones that were removed will be added back.
630 *
631 * At least one address has to be left; if only one address is
632 * available, the operation will return -EBUSY.
633 *
634 * Only sctp_setsockopt_bindx() is supposed to call this function.
635 */
636static int sctp_bindx_rem(struct sock *sk, struct sockaddr *addrs, int addrcnt)
637{
638 struct sctp_sock *sp = sctp_sk(sk);
639 struct sctp_endpoint *ep = sp->ep;
640 int cnt;
641 struct sctp_bind_addr *bp = &ep->base.bind_addr;
642 int retval = 0;
643 void *addr_buf;
644 union sctp_addr *sa_addr;
645 struct sctp_af *af;
646
647 SCTP_DEBUG_PRINTK("sctp_bindx_rem (sk: %p, addrs: %p, addrcnt: %d)\n",
648 sk, addrs, addrcnt);
649
650 addr_buf = addrs;
651 for (cnt = 0; cnt < addrcnt; cnt++) {
652 /* If the bind address list is empty or if there is only one
653 * bind address, there is nothing more to be removed (we need
654 * at least one address here).
655 */
656 if (list_empty(&bp->address_list) ||
657 (sctp_list_single_entry(&bp->address_list))) {
658 retval = -EBUSY;
659 goto err_bindx_rem;
660 }
661
662 sa_addr = addr_buf;
663 af = sctp_get_af_specific(sa_addr->sa.sa_family);
664 if (!af) {
665 retval = -EINVAL;
666 goto err_bindx_rem;
667 }
668
669 if (!af->addr_valid(sa_addr, sp, NULL)) {
670 retval = -EADDRNOTAVAIL;
671 goto err_bindx_rem;
672 }
673
674 if (sa_addr->v4.sin_port &&
675 sa_addr->v4.sin_port != htons(bp->port)) {
676 retval = -EINVAL;
677 goto err_bindx_rem;
678 }
679
680 if (!sa_addr->v4.sin_port)
681 sa_addr->v4.sin_port = htons(bp->port);
682
683 /* FIXME - There is probably a need to check if sk->sk_saddr and
684 * sk->sk_rcv_addr are currently set to one of the addresses to
685 * be removed. This is something which needs to be looked into
686 * when we are fixing the outstanding issues with multi-homing
687 * socket routing and failover schemes. Refer to comments in
688 * sctp_do_bind(). -daisy
689 */
690 retval = sctp_del_bind_addr(bp, sa_addr);
691
692 addr_buf += af->sockaddr_len;
693err_bindx_rem:
694 if (retval < 0) {
695 /* Failed. Add the ones that has been removed back */
696 if (cnt > 0)
697 sctp_bindx_add(sk, addrs, cnt);
698 return retval;
699 }
700 }
701
702 return retval;
703}
704
705/* Send an ASCONF chunk with Delete IP address parameters to all the peers of
706 * the associations that are part of the endpoint indicating that a list of
707 * local addresses are removed from the endpoint.
708 *
709 * If any of the addresses is already in the bind address list of the
710 * association, we do not send the chunk for that association. But it will not
711 * affect other associations.
712 *
713 * Only sctp_setsockopt_bindx() is supposed to call this function.
714 */
715static int sctp_send_asconf_del_ip(struct sock *sk,
716 struct sockaddr *addrs,
717 int addrcnt)
718{
719 struct sctp_sock *sp;
720 struct sctp_endpoint *ep;
721 struct sctp_association *asoc;
722 struct sctp_transport *transport;
723 struct sctp_bind_addr *bp;
724 struct sctp_chunk *chunk;
725 union sctp_addr *laddr;
726 void *addr_buf;
727 struct sctp_af *af;
728 struct sctp_sockaddr_entry *saddr;
729 int i;
730 int retval = 0;
731 int stored = 0;
732
733 chunk = NULL;
734 if (!sctp_addip_enable)
735 return retval;
736
737 sp = sctp_sk(sk);
738 ep = sp->ep;
739
740 SCTP_DEBUG_PRINTK("%s: (sk: %p, addrs: %p, addrcnt: %d)\n",
741 __func__, sk, addrs, addrcnt);
742
743 list_for_each_entry(asoc, &ep->asocs, asocs) {
744
745 if (!asoc->peer.asconf_capable)
746 continue;
747
748 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_DEL_IP)
749 continue;
750
751 if (!sctp_state(asoc, ESTABLISHED))
752 continue;
753
754 /* Check if any address in the packed array of addresses is
755 * not present in the bind address list of the association.
756 * If so, do not send the asconf chunk to its peer, but
757 * continue with other associations.
758 */
759 addr_buf = addrs;
760 for (i = 0; i < addrcnt; i++) {
761 laddr = addr_buf;
762 af = sctp_get_af_specific(laddr->v4.sin_family);
763 if (!af) {
764 retval = -EINVAL;
765 goto out;
766 }
767
768 if (!sctp_assoc_lookup_laddr(asoc, laddr))
769 break;
770
771 addr_buf += af->sockaddr_len;
772 }
773 if (i < addrcnt)
774 continue;
775
776 /* Find one address in the association's bind address list
777 * that is not in the packed array of addresses. This is to
778 * make sure that we do not delete all the addresses in the
779 * association.
780 */
781 bp = &asoc->base.bind_addr;
782 laddr = sctp_find_unmatch_addr(bp, (union sctp_addr *)addrs,
783 addrcnt, sp);
784 if ((laddr == NULL) && (addrcnt == 1)) {
785 if (asoc->asconf_addr_del_pending)
786 continue;
787 asoc->asconf_addr_del_pending =
788 kzalloc(sizeof(union sctp_addr), GFP_ATOMIC);
789 if (asoc->asconf_addr_del_pending == NULL) {
790 retval = -ENOMEM;
791 goto out;
792 }
793 asoc->asconf_addr_del_pending->sa.sa_family =
794 addrs->sa_family;
795 asoc->asconf_addr_del_pending->v4.sin_port =
796 htons(bp->port);
797 if (addrs->sa_family == AF_INET) {
798 struct sockaddr_in *sin;
799
800 sin = (struct sockaddr_in *)addrs;
801 asoc->asconf_addr_del_pending->v4.sin_addr.s_addr = sin->sin_addr.s_addr;
802 } else if (addrs->sa_family == AF_INET6) {
803 struct sockaddr_in6 *sin6;
804
805 sin6 = (struct sockaddr_in6 *)addrs;
806 ipv6_addr_copy(&asoc->asconf_addr_del_pending->v6.sin6_addr, &sin6->sin6_addr);
807 }
808 SCTP_DEBUG_PRINTK_IPADDR("send_asconf_del_ip: keep the last address asoc: %p ",
809 " at %p\n", asoc, asoc->asconf_addr_del_pending,
810 asoc->asconf_addr_del_pending);
811 asoc->src_out_of_asoc_ok = 1;
812 stored = 1;
813 goto skip_mkasconf;
814 }
815
816 /* We do not need RCU protection throughout this loop
817 * because this is done under a socket lock from the
818 * setsockopt call.
819 */
820 chunk = sctp_make_asconf_update_ip(asoc, laddr, addrs, addrcnt,
821 SCTP_PARAM_DEL_IP);
822 if (!chunk) {
823 retval = -ENOMEM;
824 goto out;
825 }
826
827skip_mkasconf:
828 /* Reset use_as_src flag for the addresses in the bind address
829 * list that are to be deleted.
830 */
831 addr_buf = addrs;
832 for (i = 0; i < addrcnt; i++) {
833 laddr = addr_buf;
834 af = sctp_get_af_specific(laddr->v4.sin_family);
835 list_for_each_entry(saddr, &bp->address_list, list) {
836 if (sctp_cmp_addr_exact(&saddr->a, laddr))
837 saddr->state = SCTP_ADDR_DEL;
838 }
839 addr_buf += af->sockaddr_len;
840 }
841
842 /* Update the route and saddr entries for all the transports
843 * as some of the addresses in the bind address list are
844 * about to be deleted and cannot be used as source addresses.
845 */
846 list_for_each_entry(transport, &asoc->peer.transport_addr_list,
847 transports) {
848 dst_release(transport->dst);
849 sctp_transport_route(transport, NULL,
850 sctp_sk(asoc->base.sk));
851 }
852
853 if (stored)
854 /* We don't need to transmit ASCONF */
855 continue;
856 retval = sctp_send_asconf(asoc, chunk);
857 }
858out:
859 return retval;
860}
861
862/* set addr events to assocs in the endpoint. ep and addr_wq must be locked */
863int sctp_asconf_mgmt(struct sctp_sock *sp, struct sctp_sockaddr_entry *addrw)
864{
865 struct sock *sk = sctp_opt2sk(sp);
866 union sctp_addr *addr;
867 struct sctp_af *af;
868
869 /* It is safe to write port space in caller. */
870 addr = &addrw->a;
871 addr->v4.sin_port = htons(sp->ep->base.bind_addr.port);
872 af = sctp_get_af_specific(addr->sa.sa_family);
873 if (!af)
874 return -EINVAL;
875 if (sctp_verify_addr(sk, addr, af->sockaddr_len))
876 return -EINVAL;
877
878 if (addrw->state == SCTP_ADDR_NEW)
879 return sctp_send_asconf_add_ip(sk, (struct sockaddr *)addr, 1);
880 else
881 return sctp_send_asconf_del_ip(sk, (struct sockaddr *)addr, 1);
882}
883
884/* Helper for tunneling sctp_bindx() requests through sctp_setsockopt()
885 *
886 * API 8.1
887 * int sctp_bindx(int sd, struct sockaddr *addrs, int addrcnt,
888 * int flags);
889 *
890 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
891 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
892 * or IPv6 addresses.
893 *
894 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
895 * Section 3.1.2 for this usage.
896 *
897 * addrs is a pointer to an array of one or more socket addresses. Each
898 * address is contained in its appropriate structure (i.e. struct
899 * sockaddr_in or struct sockaddr_in6) the family of the address type
900 * must be used to distinguish the address length (note that this
901 * representation is termed a "packed array" of addresses). The caller
902 * specifies the number of addresses in the array with addrcnt.
903 *
904 * On success, sctp_bindx() returns 0. On failure, sctp_bindx() returns
905 * -1, and sets errno to the appropriate error code.
906 *
907 * For SCTP, the port given in each socket address must be the same, or
908 * sctp_bindx() will fail, setting errno to EINVAL.
909 *
910 * The flags parameter is formed from the bitwise OR of zero or more of
911 * the following currently defined flags:
912 *
913 * SCTP_BINDX_ADD_ADDR
914 *
915 * SCTP_BINDX_REM_ADDR
916 *
917 * SCTP_BINDX_ADD_ADDR directs SCTP to add the given addresses to the
918 * association, and SCTP_BINDX_REM_ADDR directs SCTP to remove the given
919 * addresses from the association. The two flags are mutually exclusive;
920 * if both are given, sctp_bindx() will fail with EINVAL. A caller may
921 * not remove all addresses from an association; sctp_bindx() will
922 * reject such an attempt with EINVAL.
923 *
924 * An application can use sctp_bindx(SCTP_BINDX_ADD_ADDR) to associate
925 * additional addresses with an endpoint after calling bind(). Or use
926 * sctp_bindx(SCTP_BINDX_REM_ADDR) to remove some addresses a listening
927 * socket is associated with so that no new association accepted will be
928 * associated with those addresses. If the endpoint supports dynamic
929 * address a SCTP_BINDX_REM_ADDR or SCTP_BINDX_ADD_ADDR may cause a
930 * endpoint to send the appropriate message to the peer to change the
931 * peers address lists.
932 *
933 * Adding and removing addresses from a connected association is
934 * optional functionality. Implementations that do not support this
935 * functionality should return EOPNOTSUPP.
936 *
937 * Basically do nothing but copying the addresses from user to kernel
938 * land and invoking either sctp_bindx_add() or sctp_bindx_rem() on the sk.
939 * This is used for tunneling the sctp_bindx() request through sctp_setsockopt()
940 * from userspace.
941 *
942 * We don't use copy_from_user() for optimization: we first do the
943 * sanity checks (buffer size -fast- and access check-healthy
944 * pointer); if all of those succeed, then we can alloc the memory
945 * (expensive operation) needed to copy the data to kernel. Then we do
946 * the copying without checking the user space area
947 * (__copy_from_user()).
948 *
949 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
950 * it.
951 *
952 * sk The sk of the socket
953 * addrs The pointer to the addresses in user land
954 * addrssize Size of the addrs buffer
955 * op Operation to perform (add or remove, see the flags of
956 * sctp_bindx)
957 *
958 * Returns 0 if ok, <0 errno code on error.
959 */
960SCTP_STATIC int sctp_setsockopt_bindx(struct sock* sk,
961 struct sockaddr __user *addrs,
962 int addrs_size, int op)
963{
964 struct sockaddr *kaddrs;
965 int err;
966 int addrcnt = 0;
967 int walk_size = 0;
968 struct sockaddr *sa_addr;
969 void *addr_buf;
970 struct sctp_af *af;
971
972 SCTP_DEBUG_PRINTK("sctp_setsocktopt_bindx: sk %p addrs %p"
973 " addrs_size %d opt %d\n", sk, addrs, addrs_size, op);
974
975 if (unlikely(addrs_size <= 0))
976 return -EINVAL;
977
978 /* Check the user passed a healthy pointer. */
979 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
980 return -EFAULT;
981
982 /* Alloc space for the address array in kernel memory. */
983 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
984 if (unlikely(!kaddrs))
985 return -ENOMEM;
986
987 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
988 kfree(kaddrs);
989 return -EFAULT;
990 }
991
992 /* Walk through the addrs buffer and count the number of addresses. */
993 addr_buf = kaddrs;
994 while (walk_size < addrs_size) {
995 if (walk_size + sizeof(sa_family_t) > addrs_size) {
996 kfree(kaddrs);
997 return -EINVAL;
998 }
999
1000 sa_addr = addr_buf;
1001 af = sctp_get_af_specific(sa_addr->sa_family);
1002
1003 /* If the address family is not supported or if this address
1004 * causes the address buffer to overflow return EINVAL.
1005 */
1006 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1007 kfree(kaddrs);
1008 return -EINVAL;
1009 }
1010 addrcnt++;
1011 addr_buf += af->sockaddr_len;
1012 walk_size += af->sockaddr_len;
1013 }
1014
1015 /* Do the work. */
1016 switch (op) {
1017 case SCTP_BINDX_ADD_ADDR:
1018 err = sctp_bindx_add(sk, kaddrs, addrcnt);
1019 if (err)
1020 goto out;
1021 err = sctp_send_asconf_add_ip(sk, kaddrs, addrcnt);
1022 break;
1023
1024 case SCTP_BINDX_REM_ADDR:
1025 err = sctp_bindx_rem(sk, kaddrs, addrcnt);
1026 if (err)
1027 goto out;
1028 err = sctp_send_asconf_del_ip(sk, kaddrs, addrcnt);
1029 break;
1030
1031 default:
1032 err = -EINVAL;
1033 break;
1034 }
1035
1036out:
1037 kfree(kaddrs);
1038
1039 return err;
1040}
1041
1042/* __sctp_connect(struct sock* sk, struct sockaddr *kaddrs, int addrs_size)
1043 *
1044 * Common routine for handling connect() and sctp_connectx().
1045 * Connect will come in with just a single address.
1046 */
1047static int __sctp_connect(struct sock* sk,
1048 struct sockaddr *kaddrs,
1049 int addrs_size,
1050 sctp_assoc_t *assoc_id)
1051{
1052 struct sctp_sock *sp;
1053 struct sctp_endpoint *ep;
1054 struct sctp_association *asoc = NULL;
1055 struct sctp_association *asoc2;
1056 struct sctp_transport *transport;
1057 union sctp_addr to;
1058 struct sctp_af *af;
1059 sctp_scope_t scope;
1060 long timeo;
1061 int err = 0;
1062 int addrcnt = 0;
1063 int walk_size = 0;
1064 union sctp_addr *sa_addr = NULL;
1065 void *addr_buf;
1066 unsigned short port;
1067 unsigned int f_flags = 0;
1068
1069 sp = sctp_sk(sk);
1070 ep = sp->ep;
1071
1072 /* connect() cannot be done on a socket that is already in ESTABLISHED
1073 * state - UDP-style peeled off socket or a TCP-style socket that
1074 * is already connected.
1075 * It cannot be done even on a TCP-style listening socket.
1076 */
1077 if (sctp_sstate(sk, ESTABLISHED) ||
1078 (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))) {
1079 err = -EISCONN;
1080 goto out_free;
1081 }
1082
1083 /* Walk through the addrs buffer and count the number of addresses. */
1084 addr_buf = kaddrs;
1085 while (walk_size < addrs_size) {
1086 if (walk_size + sizeof(sa_family_t) > addrs_size) {
1087 err = -EINVAL;
1088 goto out_free;
1089 }
1090
1091 sa_addr = addr_buf;
1092 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1093
1094 /* If the address family is not supported or if this address
1095 * causes the address buffer to overflow return EINVAL.
1096 */
1097 if (!af || (walk_size + af->sockaddr_len) > addrs_size) {
1098 err = -EINVAL;
1099 goto out_free;
1100 }
1101
1102 port = ntohs(sa_addr->v4.sin_port);
1103
1104 /* Save current address so we can work with it */
1105 memcpy(&to, sa_addr, af->sockaddr_len);
1106
1107 err = sctp_verify_addr(sk, &to, af->sockaddr_len);
1108 if (err)
1109 goto out_free;
1110
1111 /* Make sure the destination port is correctly set
1112 * in all addresses.
1113 */
1114 if (asoc && asoc->peer.port && asoc->peer.port != port)
1115 goto out_free;
1116
1117
1118 /* Check if there already is a matching association on the
1119 * endpoint (other than the one created here).
1120 */
1121 asoc2 = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1122 if (asoc2 && asoc2 != asoc) {
1123 if (asoc2->state >= SCTP_STATE_ESTABLISHED)
1124 err = -EISCONN;
1125 else
1126 err = -EALREADY;
1127 goto out_free;
1128 }
1129
1130 /* If we could not find a matching association on the endpoint,
1131 * make sure that there is no peeled-off association matching
1132 * the peer address even on another socket.
1133 */
1134 if (sctp_endpoint_is_peeled_off(ep, &to)) {
1135 err = -EADDRNOTAVAIL;
1136 goto out_free;
1137 }
1138
1139 if (!asoc) {
1140 /* If a bind() or sctp_bindx() is not called prior to
1141 * an sctp_connectx() call, the system picks an
1142 * ephemeral port and will choose an address set
1143 * equivalent to binding with a wildcard address.
1144 */
1145 if (!ep->base.bind_addr.port) {
1146 if (sctp_autobind(sk)) {
1147 err = -EAGAIN;
1148 goto out_free;
1149 }
1150 } else {
1151 /*
1152 * If an unprivileged user inherits a 1-many
1153 * style socket with open associations on a
1154 * privileged port, it MAY be permitted to
1155 * accept new associations, but it SHOULD NOT
1156 * be permitted to open new associations.
1157 */
1158 if (ep->base.bind_addr.port < PROT_SOCK &&
1159 !capable(CAP_NET_BIND_SERVICE)) {
1160 err = -EACCES;
1161 goto out_free;
1162 }
1163 }
1164
1165 scope = sctp_scope(&to);
1166 asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1167 if (!asoc) {
1168 err = -ENOMEM;
1169 goto out_free;
1170 }
1171
1172 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope,
1173 GFP_KERNEL);
1174 if (err < 0) {
1175 goto out_free;
1176 }
1177
1178 }
1179
1180 /* Prime the peer's transport structures. */
1181 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL,
1182 SCTP_UNKNOWN);
1183 if (!transport) {
1184 err = -ENOMEM;
1185 goto out_free;
1186 }
1187
1188 addrcnt++;
1189 addr_buf += af->sockaddr_len;
1190 walk_size += af->sockaddr_len;
1191 }
1192
1193 /* In case the user of sctp_connectx() wants an association
1194 * id back, assign one now.
1195 */
1196 if (assoc_id) {
1197 err = sctp_assoc_set_id(asoc, GFP_KERNEL);
1198 if (err < 0)
1199 goto out_free;
1200 }
1201
1202 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1203 if (err < 0) {
1204 goto out_free;
1205 }
1206
1207 /* Initialize sk's dport and daddr for getpeername() */
1208 inet_sk(sk)->inet_dport = htons(asoc->peer.port);
1209 af = sctp_get_af_specific(sa_addr->sa.sa_family);
1210 af->to_sk_daddr(sa_addr, sk);
1211 sk->sk_err = 0;
1212
1213 /* in-kernel sockets don't generally have a file allocated to them
1214 * if all they do is call sock_create_kern().
1215 */
1216 if (sk->sk_socket->file)
1217 f_flags = sk->sk_socket->file->f_flags;
1218
1219 timeo = sock_sndtimeo(sk, f_flags & O_NONBLOCK);
1220
1221 err = sctp_wait_for_connect(asoc, &timeo);
1222 if ((err == 0 || err == -EINPROGRESS) && assoc_id)
1223 *assoc_id = asoc->assoc_id;
1224
1225 /* Don't free association on exit. */
1226 asoc = NULL;
1227
1228out_free:
1229
1230 SCTP_DEBUG_PRINTK("About to exit __sctp_connect() free asoc: %p"
1231 " kaddrs: %p err: %d\n",
1232 asoc, kaddrs, err);
1233 if (asoc)
1234 sctp_association_free(asoc);
1235 return err;
1236}
1237
1238/* Helper for tunneling sctp_connectx() requests through sctp_setsockopt()
1239 *
1240 * API 8.9
1241 * int sctp_connectx(int sd, struct sockaddr *addrs, int addrcnt,
1242 * sctp_assoc_t *asoc);
1243 *
1244 * If sd is an IPv4 socket, the addresses passed must be IPv4 addresses.
1245 * If the sd is an IPv6 socket, the addresses passed can either be IPv4
1246 * or IPv6 addresses.
1247 *
1248 * A single address may be specified as INADDR_ANY or IN6ADDR_ANY, see
1249 * Section 3.1.2 for this usage.
1250 *
1251 * addrs is a pointer to an array of one or more socket addresses. Each
1252 * address is contained in its appropriate structure (i.e. struct
1253 * sockaddr_in or struct sockaddr_in6) the family of the address type
1254 * must be used to distengish the address length (note that this
1255 * representation is termed a "packed array" of addresses). The caller
1256 * specifies the number of addresses in the array with addrcnt.
1257 *
1258 * On success, sctp_connectx() returns 0. It also sets the assoc_id to
1259 * the association id of the new association. On failure, sctp_connectx()
1260 * returns -1, and sets errno to the appropriate error code. The assoc_id
1261 * is not touched by the kernel.
1262 *
1263 * For SCTP, the port given in each socket address must be the same, or
1264 * sctp_connectx() will fail, setting errno to EINVAL.
1265 *
1266 * An application can use sctp_connectx to initiate an association with
1267 * an endpoint that is multi-homed. Much like sctp_bindx() this call
1268 * allows a caller to specify multiple addresses at which a peer can be
1269 * reached. The way the SCTP stack uses the list of addresses to set up
1270 * the association is implementation dependent. This function only
1271 * specifies that the stack will try to make use of all the addresses in
1272 * the list when needed.
1273 *
1274 * Note that the list of addresses passed in is only used for setting up
1275 * the association. It does not necessarily equal the set of addresses
1276 * the peer uses for the resulting association. If the caller wants to
1277 * find out the set of peer addresses, it must use sctp_getpaddrs() to
1278 * retrieve them after the association has been set up.
1279 *
1280 * Basically do nothing but copying the addresses from user to kernel
1281 * land and invoking either sctp_connectx(). This is used for tunneling
1282 * the sctp_connectx() request through sctp_setsockopt() from userspace.
1283 *
1284 * We don't use copy_from_user() for optimization: we first do the
1285 * sanity checks (buffer size -fast- and access check-healthy
1286 * pointer); if all of those succeed, then we can alloc the memory
1287 * (expensive operation) needed to copy the data to kernel. Then we do
1288 * the copying without checking the user space area
1289 * (__copy_from_user()).
1290 *
1291 * On exit there is no need to do sockfd_put(), sys_setsockopt() does
1292 * it.
1293 *
1294 * sk The sk of the socket
1295 * addrs The pointer to the addresses in user land
1296 * addrssize Size of the addrs buffer
1297 *
1298 * Returns >=0 if ok, <0 errno code on error.
1299 */
1300SCTP_STATIC int __sctp_setsockopt_connectx(struct sock* sk,
1301 struct sockaddr __user *addrs,
1302 int addrs_size,
1303 sctp_assoc_t *assoc_id)
1304{
1305 int err = 0;
1306 struct sockaddr *kaddrs;
1307
1308 SCTP_DEBUG_PRINTK("%s - sk %p addrs %p addrs_size %d\n",
1309 __func__, sk, addrs, addrs_size);
1310
1311 if (unlikely(addrs_size <= 0))
1312 return -EINVAL;
1313
1314 /* Check the user passed a healthy pointer. */
1315 if (unlikely(!access_ok(VERIFY_READ, addrs, addrs_size)))
1316 return -EFAULT;
1317
1318 /* Alloc space for the address array in kernel memory. */
1319 kaddrs = kmalloc(addrs_size, GFP_KERNEL);
1320 if (unlikely(!kaddrs))
1321 return -ENOMEM;
1322
1323 if (__copy_from_user(kaddrs, addrs, addrs_size)) {
1324 err = -EFAULT;
1325 } else {
1326 err = __sctp_connect(sk, kaddrs, addrs_size, assoc_id);
1327 }
1328
1329 kfree(kaddrs);
1330
1331 return err;
1332}
1333
1334/*
1335 * This is an older interface. It's kept for backward compatibility
1336 * to the option that doesn't provide association id.
1337 */
1338SCTP_STATIC int sctp_setsockopt_connectx_old(struct sock* sk,
1339 struct sockaddr __user *addrs,
1340 int addrs_size)
1341{
1342 return __sctp_setsockopt_connectx(sk, addrs, addrs_size, NULL);
1343}
1344
1345/*
1346 * New interface for the API. The since the API is done with a socket
1347 * option, to make it simple we feed back the association id is as a return
1348 * indication to the call. Error is always negative and association id is
1349 * always positive.
1350 */
1351SCTP_STATIC int sctp_setsockopt_connectx(struct sock* sk,
1352 struct sockaddr __user *addrs,
1353 int addrs_size)
1354{
1355 sctp_assoc_t assoc_id = 0;
1356 int err = 0;
1357
1358 err = __sctp_setsockopt_connectx(sk, addrs, addrs_size, &assoc_id);
1359
1360 if (err)
1361 return err;
1362 else
1363 return assoc_id;
1364}
1365
1366/*
1367 * New (hopefully final) interface for the API.
1368 * We use the sctp_getaddrs_old structure so that use-space library
1369 * can avoid any unnecessary allocations. The only defferent part
1370 * is that we store the actual length of the address buffer into the
1371 * addrs_num structure member. That way we can re-use the existing
1372 * code.
1373 */
1374SCTP_STATIC int sctp_getsockopt_connectx3(struct sock* sk, int len,
1375 char __user *optval,
1376 int __user *optlen)
1377{
1378 struct sctp_getaddrs_old param;
1379 sctp_assoc_t assoc_id = 0;
1380 int err = 0;
1381
1382 if (len < sizeof(param))
1383 return -EINVAL;
1384
1385 if (copy_from_user(¶m, optval, sizeof(param)))
1386 return -EFAULT;
1387
1388 err = __sctp_setsockopt_connectx(sk,
1389 (struct sockaddr __user *)param.addrs,
1390 param.addr_num, &assoc_id);
1391
1392 if (err == 0 || err == -EINPROGRESS) {
1393 if (copy_to_user(optval, &assoc_id, sizeof(assoc_id)))
1394 return -EFAULT;
1395 if (put_user(sizeof(assoc_id), optlen))
1396 return -EFAULT;
1397 }
1398
1399 return err;
1400}
1401
1402/* API 3.1.4 close() - UDP Style Syntax
1403 * Applications use close() to perform graceful shutdown (as described in
1404 * Section 10.1 of [SCTP]) on ALL the associations currently represented
1405 * by a UDP-style socket.
1406 *
1407 * The syntax is
1408 *
1409 * ret = close(int sd);
1410 *
1411 * sd - the socket descriptor of the associations to be closed.
1412 *
1413 * To gracefully shutdown a specific association represented by the
1414 * UDP-style socket, an application should use the sendmsg() call,
1415 * passing no user data, but including the appropriate flag in the
1416 * ancillary data (see Section xxxx).
1417 *
1418 * If sd in the close() call is a branched-off socket representing only
1419 * one association, the shutdown is performed on that association only.
1420 *
1421 * 4.1.6 close() - TCP Style Syntax
1422 *
1423 * Applications use close() to gracefully close down an association.
1424 *
1425 * The syntax is:
1426 *
1427 * int close(int sd);
1428 *
1429 * sd - the socket descriptor of the association to be closed.
1430 *
1431 * After an application calls close() on a socket descriptor, no further
1432 * socket operations will succeed on that descriptor.
1433 *
1434 * API 7.1.4 SO_LINGER
1435 *
1436 * An application using the TCP-style socket can use this option to
1437 * perform the SCTP ABORT primitive. The linger option structure is:
1438 *
1439 * struct linger {
1440 * int l_onoff; // option on/off
1441 * int l_linger; // linger time
1442 * };
1443 *
1444 * To enable the option, set l_onoff to 1. If the l_linger value is set
1445 * to 0, calling close() is the same as the ABORT primitive. If the
1446 * value is set to a negative value, the setsockopt() call will return
1447 * an error. If the value is set to a positive value linger_time, the
1448 * close() can be blocked for at most linger_time ms. If the graceful
1449 * shutdown phase does not finish during this period, close() will
1450 * return but the graceful shutdown phase continues in the system.
1451 */
1452SCTP_STATIC void sctp_close(struct sock *sk, long timeout)
1453{
1454 struct sctp_endpoint *ep;
1455 struct sctp_association *asoc;
1456 struct list_head *pos, *temp;
1457 unsigned int data_was_unread;
1458
1459 SCTP_DEBUG_PRINTK("sctp_close(sk: 0x%p, timeout:%ld)\n", sk, timeout);
1460
1461 sctp_lock_sock(sk);
1462 sk->sk_shutdown = SHUTDOWN_MASK;
1463 sk->sk_state = SCTP_SS_CLOSING;
1464
1465 ep = sctp_sk(sk)->ep;
1466
1467 /* Clean up any skbs sitting on the receive queue. */
1468 data_was_unread = sctp_queue_purge_ulpevents(&sk->sk_receive_queue);
1469 data_was_unread += sctp_queue_purge_ulpevents(&sctp_sk(sk)->pd_lobby);
1470
1471 /* Walk all associations on an endpoint. */
1472 list_for_each_safe(pos, temp, &ep->asocs) {
1473 asoc = list_entry(pos, struct sctp_association, asocs);
1474
1475 if (sctp_style(sk, TCP)) {
1476 /* A closed association can still be in the list if
1477 * it belongs to a TCP-style listening socket that is
1478 * not yet accepted. If so, free it. If not, send an
1479 * ABORT or SHUTDOWN based on the linger options.
1480 */
1481 if (sctp_state(asoc, CLOSED)) {
1482 sctp_unhash_established(asoc);
1483 sctp_association_free(asoc);
1484 continue;
1485 }
1486 }
1487
1488 if (data_was_unread || !skb_queue_empty(&asoc->ulpq.lobby) ||
1489 !skb_queue_empty(&asoc->ulpq.reasm) ||
1490 (sock_flag(sk, SOCK_LINGER) && !sk->sk_lingertime)) {
1491 struct sctp_chunk *chunk;
1492
1493 chunk = sctp_make_abort_user(asoc, NULL, 0);
1494 if (chunk)
1495 sctp_primitive_ABORT(asoc, chunk);
1496 } else
1497 sctp_primitive_SHUTDOWN(asoc, NULL);
1498 }
1499
1500 /* On a TCP-style socket, block for at most linger_time if set. */
1501 if (sctp_style(sk, TCP) && timeout)
1502 sctp_wait_for_close(sk, timeout);
1503
1504 /* This will run the backlog queue. */
1505 sctp_release_sock(sk);
1506
1507 /* Supposedly, no process has access to the socket, but
1508 * the net layers still may.
1509 */
1510 sctp_local_bh_disable();
1511 sctp_bh_lock_sock(sk);
1512
1513 /* Hold the sock, since sk_common_release() will put sock_put()
1514 * and we have just a little more cleanup.
1515 */
1516 sock_hold(sk);
1517 sk_common_release(sk);
1518
1519 sctp_bh_unlock_sock(sk);
1520 sctp_local_bh_enable();
1521
1522 sock_put(sk);
1523
1524 SCTP_DBG_OBJCNT_DEC(sock);
1525}
1526
1527/* Handle EPIPE error. */
1528static int sctp_error(struct sock *sk, int flags, int err)
1529{
1530 if (err == -EPIPE)
1531 err = sock_error(sk) ? : -EPIPE;
1532 if (err == -EPIPE && !(flags & MSG_NOSIGNAL))
1533 send_sig(SIGPIPE, current, 0);
1534 return err;
1535}
1536
1537/* API 3.1.3 sendmsg() - UDP Style Syntax
1538 *
1539 * An application uses sendmsg() and recvmsg() calls to transmit data to
1540 * and receive data from its peer.
1541 *
1542 * ssize_t sendmsg(int socket, const struct msghdr *message,
1543 * int flags);
1544 *
1545 * socket - the socket descriptor of the endpoint.
1546 * message - pointer to the msghdr structure which contains a single
1547 * user message and possibly some ancillary data.
1548 *
1549 * See Section 5 for complete description of the data
1550 * structures.
1551 *
1552 * flags - flags sent or received with the user message, see Section
1553 * 5 for complete description of the flags.
1554 *
1555 * Note: This function could use a rewrite especially when explicit
1556 * connect support comes in.
1557 */
1558/* BUG: We do not implement the equivalent of sk_stream_wait_memory(). */
1559
1560SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *, sctp_cmsgs_t *);
1561
1562SCTP_STATIC int sctp_sendmsg(struct kiocb *iocb, struct sock *sk,
1563 struct msghdr *msg, size_t msg_len)
1564{
1565 struct sctp_sock *sp;
1566 struct sctp_endpoint *ep;
1567 struct sctp_association *new_asoc=NULL, *asoc=NULL;
1568 struct sctp_transport *transport, *chunk_tp;
1569 struct sctp_chunk *chunk;
1570 union sctp_addr to;
1571 struct sockaddr *msg_name = NULL;
1572 struct sctp_sndrcvinfo default_sinfo;
1573 struct sctp_sndrcvinfo *sinfo;
1574 struct sctp_initmsg *sinit;
1575 sctp_assoc_t associd = 0;
1576 sctp_cmsgs_t cmsgs = { NULL };
1577 int err;
1578 sctp_scope_t scope;
1579 long timeo;
1580 __u16 sinfo_flags = 0;
1581 struct sctp_datamsg *datamsg;
1582 int msg_flags = msg->msg_flags;
1583
1584 SCTP_DEBUG_PRINTK("sctp_sendmsg(sk: %p, msg: %p, msg_len: %zu)\n",
1585 sk, msg, msg_len);
1586
1587 err = 0;
1588 sp = sctp_sk(sk);
1589 ep = sp->ep;
1590
1591 SCTP_DEBUG_PRINTK("Using endpoint: %p.\n", ep);
1592
1593 /* We cannot send a message over a TCP-style listening socket. */
1594 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING)) {
1595 err = -EPIPE;
1596 goto out_nounlock;
1597 }
1598
1599 /* Parse out the SCTP CMSGs. */
1600 err = sctp_msghdr_parse(msg, &cmsgs);
1601
1602 if (err) {
1603 SCTP_DEBUG_PRINTK("msghdr parse err = %x\n", err);
1604 goto out_nounlock;
1605 }
1606
1607 /* Fetch the destination address for this packet. This
1608 * address only selects the association--it is not necessarily
1609 * the address we will send to.
1610 * For a peeled-off socket, msg_name is ignored.
1611 */
1612 if (!sctp_style(sk, UDP_HIGH_BANDWIDTH) && msg->msg_name) {
1613 int msg_namelen = msg->msg_namelen;
1614
1615 err = sctp_verify_addr(sk, (union sctp_addr *)msg->msg_name,
1616 msg_namelen);
1617 if (err)
1618 return err;
1619
1620 if (msg_namelen > sizeof(to))
1621 msg_namelen = sizeof(to);
1622 memcpy(&to, msg->msg_name, msg_namelen);
1623 msg_name = msg->msg_name;
1624 }
1625
1626 sinfo = cmsgs.info;
1627 sinit = cmsgs.init;
1628
1629 /* Did the user specify SNDRCVINFO? */
1630 if (sinfo) {
1631 sinfo_flags = sinfo->sinfo_flags;
1632 associd = sinfo->sinfo_assoc_id;
1633 }
1634
1635 SCTP_DEBUG_PRINTK("msg_len: %zu, sinfo_flags: 0x%x\n",
1636 msg_len, sinfo_flags);
1637
1638 /* SCTP_EOF or SCTP_ABORT cannot be set on a TCP-style socket. */
1639 if (sctp_style(sk, TCP) && (sinfo_flags & (SCTP_EOF | SCTP_ABORT))) {
1640 err = -EINVAL;
1641 goto out_nounlock;
1642 }
1643
1644 /* If SCTP_EOF is set, no data can be sent. Disallow sending zero
1645 * length messages when SCTP_EOF|SCTP_ABORT is not set.
1646 * If SCTP_ABORT is set, the message length could be non zero with
1647 * the msg_iov set to the user abort reason.
1648 */
1649 if (((sinfo_flags & SCTP_EOF) && (msg_len > 0)) ||
1650 (!(sinfo_flags & (SCTP_EOF|SCTP_ABORT)) && (msg_len == 0))) {
1651 err = -EINVAL;
1652 goto out_nounlock;
1653 }
1654
1655 /* If SCTP_ADDR_OVER is set, there must be an address
1656 * specified in msg_name.
1657 */
1658 if ((sinfo_flags & SCTP_ADDR_OVER) && (!msg->msg_name)) {
1659 err = -EINVAL;
1660 goto out_nounlock;
1661 }
1662
1663 transport = NULL;
1664
1665 SCTP_DEBUG_PRINTK("About to look up association.\n");
1666
1667 sctp_lock_sock(sk);
1668
1669 /* If a msg_name has been specified, assume this is to be used. */
1670 if (msg_name) {
1671 /* Look for a matching association on the endpoint. */
1672 asoc = sctp_endpoint_lookup_assoc(ep, &to, &transport);
1673 if (!asoc) {
1674 /* If we could not find a matching association on the
1675 * endpoint, make sure that it is not a TCP-style
1676 * socket that already has an association or there is
1677 * no peeled-off association on another socket.
1678 */
1679 if ((sctp_style(sk, TCP) &&
1680 sctp_sstate(sk, ESTABLISHED)) ||
1681 sctp_endpoint_is_peeled_off(ep, &to)) {
1682 err = -EADDRNOTAVAIL;
1683 goto out_unlock;
1684 }
1685 }
1686 } else {
1687 asoc = sctp_id2assoc(sk, associd);
1688 if (!asoc) {
1689 err = -EPIPE;
1690 goto out_unlock;
1691 }
1692 }
1693
1694 if (asoc) {
1695 SCTP_DEBUG_PRINTK("Just looked up association: %p.\n", asoc);
1696
1697 /* We cannot send a message on a TCP-style SCTP_SS_ESTABLISHED
1698 * socket that has an association in CLOSED state. This can
1699 * happen when an accepted socket has an association that is
1700 * already CLOSED.
1701 */
1702 if (sctp_state(asoc, CLOSED) && sctp_style(sk, TCP)) {
1703 err = -EPIPE;
1704 goto out_unlock;
1705 }
1706
1707 if (sinfo_flags & SCTP_EOF) {
1708 SCTP_DEBUG_PRINTK("Shutting down association: %p\n",
1709 asoc);
1710 sctp_primitive_SHUTDOWN(asoc, NULL);
1711 err = 0;
1712 goto out_unlock;
1713 }
1714 if (sinfo_flags & SCTP_ABORT) {
1715
1716 chunk = sctp_make_abort_user(asoc, msg, msg_len);
1717 if (!chunk) {
1718 err = -ENOMEM;
1719 goto out_unlock;
1720 }
1721
1722 SCTP_DEBUG_PRINTK("Aborting association: %p\n", asoc);
1723 sctp_primitive_ABORT(asoc, chunk);
1724 err = 0;
1725 goto out_unlock;
1726 }
1727 }
1728
1729 /* Do we need to create the association? */
1730 if (!asoc) {
1731 SCTP_DEBUG_PRINTK("There is no association yet.\n");
1732
1733 if (sinfo_flags & (SCTP_EOF | SCTP_ABORT)) {
1734 err = -EINVAL;
1735 goto out_unlock;
1736 }
1737
1738 /* Check for invalid stream against the stream counts,
1739 * either the default or the user specified stream counts.
1740 */
1741 if (sinfo) {
1742 if (!sinit || (sinit && !sinit->sinit_num_ostreams)) {
1743 /* Check against the defaults. */
1744 if (sinfo->sinfo_stream >=
1745 sp->initmsg.sinit_num_ostreams) {
1746 err = -EINVAL;
1747 goto out_unlock;
1748 }
1749 } else {
1750 /* Check against the requested. */
1751 if (sinfo->sinfo_stream >=
1752 sinit->sinit_num_ostreams) {
1753 err = -EINVAL;
1754 goto out_unlock;
1755 }
1756 }
1757 }
1758
1759 /*
1760 * API 3.1.2 bind() - UDP Style Syntax
1761 * If a bind() or sctp_bindx() is not called prior to a
1762 * sendmsg() call that initiates a new association, the
1763 * system picks an ephemeral port and will choose an address
1764 * set equivalent to binding with a wildcard address.
1765 */
1766 if (!ep->base.bind_addr.port) {
1767 if (sctp_autobind(sk)) {
1768 err = -EAGAIN;
1769 goto out_unlock;
1770 }
1771 } else {
1772 /*
1773 * If an unprivileged user inherits a one-to-many
1774 * style socket with open associations on a privileged
1775 * port, it MAY be permitted to accept new associations,
1776 * but it SHOULD NOT be permitted to open new
1777 * associations.
1778 */
1779 if (ep->base.bind_addr.port < PROT_SOCK &&
1780 !capable(CAP_NET_BIND_SERVICE)) {
1781 err = -EACCES;
1782 goto out_unlock;
1783 }
1784 }
1785
1786 scope = sctp_scope(&to);
1787 new_asoc = sctp_association_new(ep, sk, scope, GFP_KERNEL);
1788 if (!new_asoc) {
1789 err = -ENOMEM;
1790 goto out_unlock;
1791 }
1792 asoc = new_asoc;
1793 err = sctp_assoc_set_bind_addr_from_ep(asoc, scope, GFP_KERNEL);
1794 if (err < 0) {
1795 err = -ENOMEM;
1796 goto out_free;
1797 }
1798
1799 /* If the SCTP_INIT ancillary data is specified, set all
1800 * the association init values accordingly.
1801 */
1802 if (sinit) {
1803 if (sinit->sinit_num_ostreams) {
1804 asoc->c.sinit_num_ostreams =
1805 sinit->sinit_num_ostreams;
1806 }
1807 if (sinit->sinit_max_instreams) {
1808 asoc->c.sinit_max_instreams =
1809 sinit->sinit_max_instreams;
1810 }
1811 if (sinit->sinit_max_attempts) {
1812 asoc->max_init_attempts
1813 = sinit->sinit_max_attempts;
1814 }
1815 if (sinit->sinit_max_init_timeo) {
1816 asoc->max_init_timeo =
1817 msecs_to_jiffies(sinit->sinit_max_init_timeo);
1818 }
1819 }
1820
1821 /* Prime the peer's transport structures. */
1822 transport = sctp_assoc_add_peer(asoc, &to, GFP_KERNEL, SCTP_UNKNOWN);
1823 if (!transport) {
1824 err = -ENOMEM;
1825 goto out_free;
1826 }
1827 }
1828
1829 /* ASSERT: we have a valid association at this point. */
1830 SCTP_DEBUG_PRINTK("We have a valid association.\n");
1831
1832 if (!sinfo) {
1833 /* If the user didn't specify SNDRCVINFO, make up one with
1834 * some defaults.
1835 */
1836 memset(&default_sinfo, 0, sizeof(default_sinfo));
1837 default_sinfo.sinfo_stream = asoc->default_stream;
1838 default_sinfo.sinfo_flags = asoc->default_flags;
1839 default_sinfo.sinfo_ppid = asoc->default_ppid;
1840 default_sinfo.sinfo_context = asoc->default_context;
1841 default_sinfo.sinfo_timetolive = asoc->default_timetolive;
1842 default_sinfo.sinfo_assoc_id = sctp_assoc2id(asoc);
1843 sinfo = &default_sinfo;
1844 }
1845
1846 /* API 7.1.7, the sndbuf size per association bounds the
1847 * maximum size of data that can be sent in a single send call.
1848 */
1849 if (msg_len > sk->sk_sndbuf) {
1850 err = -EMSGSIZE;
1851 goto out_free;
1852 }
1853
1854 if (asoc->pmtu_pending)
1855 sctp_assoc_pending_pmtu(asoc);
1856
1857 /* If fragmentation is disabled and the message length exceeds the
1858 * association fragmentation point, return EMSGSIZE. The I-D
1859 * does not specify what this error is, but this looks like
1860 * a great fit.
1861 */
1862 if (sctp_sk(sk)->disable_fragments && (msg_len > asoc->frag_point)) {
1863 err = -EMSGSIZE;
1864 goto out_free;
1865 }
1866
1867 /* Check for invalid stream. */
1868 if (sinfo->sinfo_stream >= asoc->c.sinit_num_ostreams) {
1869 err = -EINVAL;
1870 goto out_free;
1871 }
1872
1873 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1874 if (!sctp_wspace(asoc)) {
1875 err = sctp_wait_for_sndbuf(asoc, &timeo, msg_len);
1876 if (err)
1877 goto out_free;
1878 }
1879
1880 /* If an address is passed with the sendto/sendmsg call, it is used
1881 * to override the primary destination address in the TCP model, or
1882 * when SCTP_ADDR_OVER flag is set in the UDP model.
1883 */
1884 if ((sctp_style(sk, TCP) && msg_name) ||
1885 (sinfo_flags & SCTP_ADDR_OVER)) {
1886 chunk_tp = sctp_assoc_lookup_paddr(asoc, &to);
1887 if (!chunk_tp) {
1888 err = -EINVAL;
1889 goto out_free;
1890 }
1891 } else
1892 chunk_tp = NULL;
1893
1894 /* Auto-connect, if we aren't connected already. */
1895 if (sctp_state(asoc, CLOSED)) {
1896 err = sctp_primitive_ASSOCIATE(asoc, NULL);
1897 if (err < 0)
1898 goto out_free;
1899 SCTP_DEBUG_PRINTK("We associated primitively.\n");
1900 }
1901
1902 /* Break the message into multiple chunks of maximum size. */
1903 datamsg = sctp_datamsg_from_user(asoc, sinfo, msg, msg_len);
1904 if (!datamsg) {
1905 err = -ENOMEM;
1906 goto out_free;
1907 }
1908
1909 /* Now send the (possibly) fragmented message. */
1910 list_for_each_entry(chunk, &datamsg->chunks, frag_list) {
1911 sctp_chunk_hold(chunk);
1912
1913 /* Do accounting for the write space. */
1914 sctp_set_owner_w(chunk);
1915
1916 chunk->transport = chunk_tp;
1917 }
1918
1919 /* Send it to the lower layers. Note: all chunks
1920 * must either fail or succeed. The lower layer
1921 * works that way today. Keep it that way or this
1922 * breaks.
1923 */
1924 err = sctp_primitive_SEND(asoc, datamsg);
1925 /* Did the lower layer accept the chunk? */
1926 if (err)
1927 sctp_datamsg_free(datamsg);
1928 else
1929 sctp_datamsg_put(datamsg);
1930
1931 SCTP_DEBUG_PRINTK("We sent primitively.\n");
1932
1933 if (err)
1934 goto out_free;
1935 else
1936 err = msg_len;
1937
1938 /* If we are already past ASSOCIATE, the lower
1939 * layers are responsible for association cleanup.
1940 */
1941 goto out_unlock;
1942
1943out_free:
1944 if (new_asoc)
1945 sctp_association_free(asoc);
1946out_unlock:
1947 sctp_release_sock(sk);
1948
1949out_nounlock:
1950 return sctp_error(sk, msg_flags, err);
1951
1952#if 0
1953do_sock_err:
1954 if (msg_len)
1955 err = msg_len;
1956 else
1957 err = sock_error(sk);
1958 goto out;
1959
1960do_interrupted:
1961 if (msg_len)
1962 err = msg_len;
1963 goto out;
1964#endif /* 0 */
1965}
1966
1967/* This is an extended version of skb_pull() that removes the data from the
1968 * start of a skb even when data is spread across the list of skb's in the
1969 * frag_list. len specifies the total amount of data that needs to be removed.
1970 * when 'len' bytes could be removed from the skb, it returns 0.
1971 * If 'len' exceeds the total skb length, it returns the no. of bytes that
1972 * could not be removed.
1973 */
1974static int sctp_skb_pull(struct sk_buff *skb, int len)
1975{
1976 struct sk_buff *list;
1977 int skb_len = skb_headlen(skb);
1978 int rlen;
1979
1980 if (len <= skb_len) {
1981 __skb_pull(skb, len);
1982 return 0;
1983 }
1984 len -= skb_len;
1985 __skb_pull(skb, skb_len);
1986
1987 skb_walk_frags(skb, list) {
1988 rlen = sctp_skb_pull(list, len);
1989 skb->len -= (len-rlen);
1990 skb->data_len -= (len-rlen);
1991
1992 if (!rlen)
1993 return 0;
1994
1995 len = rlen;
1996 }
1997
1998 return len;
1999}
2000
2001/* API 3.1.3 recvmsg() - UDP Style Syntax
2002 *
2003 * ssize_t recvmsg(int socket, struct msghdr *message,
2004 * int flags);
2005 *
2006 * socket - the socket descriptor of the endpoint.
2007 * message - pointer to the msghdr structure which contains a single
2008 * user message and possibly some ancillary data.
2009 *
2010 * See Section 5 for complete description of the data
2011 * structures.
2012 *
2013 * flags - flags sent or received with the user message, see Section
2014 * 5 for complete description of the flags.
2015 */
2016static struct sk_buff *sctp_skb_recv_datagram(struct sock *, int, int, int *);
2017
2018SCTP_STATIC int sctp_recvmsg(struct kiocb *iocb, struct sock *sk,
2019 struct msghdr *msg, size_t len, int noblock,
2020 int flags, int *addr_len)
2021{
2022 struct sctp_ulpevent *event = NULL;
2023 struct sctp_sock *sp = sctp_sk(sk);
2024 struct sk_buff *skb;
2025 int copied;
2026 int err = 0;
2027 int skb_len;
2028
2029 SCTP_DEBUG_PRINTK("sctp_recvmsg(%s: %p, %s: %p, %s: %zd, %s: %d, %s: "
2030 "0x%x, %s: %p)\n", "sk", sk, "msghdr", msg,
2031 "len", len, "knoblauch", noblock,
2032 "flags", flags, "addr_len", addr_len);
2033
2034 sctp_lock_sock(sk);
2035
2036 if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED)) {
2037 err = -ENOTCONN;
2038 goto out;
2039 }
2040
2041 skb = sctp_skb_recv_datagram(sk, flags, noblock, &err);
2042 if (!skb)
2043 goto out;
2044
2045 /* Get the total length of the skb including any skb's in the
2046 * frag_list.
2047 */
2048 skb_len = skb->len;
2049
2050 copied = skb_len;
2051 if (copied > len)
2052 copied = len;
2053
2054 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2055
2056 event = sctp_skb2event(skb);
2057
2058 if (err)
2059 goto out_free;
2060
2061 sock_recv_ts_and_drops(msg, sk, skb);
2062 if (sctp_ulpevent_is_notification(event)) {
2063 msg->msg_flags |= MSG_NOTIFICATION;
2064 sp->pf->event_msgname(event, msg->msg_name, addr_len);
2065 } else {
2066 sp->pf->skb_msgname(skb, msg->msg_name, addr_len);
2067 }
2068
2069 /* Check if we allow SCTP_SNDRCVINFO. */
2070 if (sp->subscribe.sctp_data_io_event)
2071 sctp_ulpevent_read_sndrcvinfo(event, msg);
2072#if 0
2073 /* FIXME: we should be calling IP/IPv6 layers. */
2074 if (sk->sk_protinfo.af_inet.cmsg_flags)
2075 ip_cmsg_recv(msg, skb);
2076#endif
2077
2078 err = copied;
2079
2080 /* If skb's length exceeds the user's buffer, update the skb and
2081 * push it back to the receive_queue so that the next call to
2082 * recvmsg() will return the remaining data. Don't set MSG_EOR.
2083 */
2084 if (skb_len > copied) {
2085 msg->msg_flags &= ~MSG_EOR;
2086 if (flags & MSG_PEEK)
2087 goto out_free;
2088 sctp_skb_pull(skb, copied);
2089 skb_queue_head(&sk->sk_receive_queue, skb);
2090
2091 /* When only partial message is copied to the user, increase
2092 * rwnd by that amount. If all the data in the skb is read,
2093 * rwnd is updated when the event is freed.
2094 */
2095 if (!sctp_ulpevent_is_notification(event))
2096 sctp_assoc_rwnd_increase(event->asoc, copied);
2097 goto out;
2098 } else if ((event->msg_flags & MSG_NOTIFICATION) ||
2099 (event->msg_flags & MSG_EOR))
2100 msg->msg_flags |= MSG_EOR;
2101 else
2102 msg->msg_flags &= ~MSG_EOR;
2103
2104out_free:
2105 if (flags & MSG_PEEK) {
2106 /* Release the skb reference acquired after peeking the skb in
2107 * sctp_skb_recv_datagram().
2108 */
2109 kfree_skb(skb);
2110 } else {
2111 /* Free the event which includes releasing the reference to
2112 * the owner of the skb, freeing the skb and updating the
2113 * rwnd.
2114 */
2115 sctp_ulpevent_free(event);
2116 }
2117out:
2118 sctp_release_sock(sk);
2119 return err;
2120}
2121
2122/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
2123 *
2124 * This option is a on/off flag. If enabled no SCTP message
2125 * fragmentation will be performed. Instead if a message being sent
2126 * exceeds the current PMTU size, the message will NOT be sent and
2127 * instead a error will be indicated to the user.
2128 */
2129static int sctp_setsockopt_disable_fragments(struct sock *sk,
2130 char __user *optval,
2131 unsigned int optlen)
2132{
2133 int val;
2134
2135 if (optlen < sizeof(int))
2136 return -EINVAL;
2137
2138 if (get_user(val, (int __user *)optval))
2139 return -EFAULT;
2140
2141 sctp_sk(sk)->disable_fragments = (val == 0) ? 0 : 1;
2142
2143 return 0;
2144}
2145
2146static int sctp_setsockopt_events(struct sock *sk, char __user *optval,
2147 unsigned int optlen)
2148{
2149 struct sctp_association *asoc;
2150 struct sctp_ulpevent *event;
2151
2152 if (optlen > sizeof(struct sctp_event_subscribe))
2153 return -EINVAL;
2154 if (copy_from_user(&sctp_sk(sk)->subscribe, optval, optlen))
2155 return -EFAULT;
2156
2157 /*
2158 * At the time when a user app subscribes to SCTP_SENDER_DRY_EVENT,
2159 * if there is no data to be sent or retransmit, the stack will
2160 * immediately send up this notification.
2161 */
2162 if (sctp_ulpevent_type_enabled(SCTP_SENDER_DRY_EVENT,
2163 &sctp_sk(sk)->subscribe)) {
2164 asoc = sctp_id2assoc(sk, 0);
2165
2166 if (asoc && sctp_outq_is_empty(&asoc->outqueue)) {
2167 event = sctp_ulpevent_make_sender_dry_event(asoc,
2168 GFP_ATOMIC);
2169 if (!event)
2170 return -ENOMEM;
2171
2172 sctp_ulpq_tail_event(&asoc->ulpq, event);
2173 }
2174 }
2175
2176 return 0;
2177}
2178
2179/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
2180 *
2181 * This socket option is applicable to the UDP-style socket only. When
2182 * set it will cause associations that are idle for more than the
2183 * specified number of seconds to automatically close. An association
2184 * being idle is defined an association that has NOT sent or received
2185 * user data. The special value of '0' indicates that no automatic
2186 * close of any associations should be performed. The option expects an
2187 * integer defining the number of seconds of idle time before an
2188 * association is closed.
2189 */
2190static int sctp_setsockopt_autoclose(struct sock *sk, char __user *optval,
2191 unsigned int optlen)
2192{
2193 struct sctp_sock *sp = sctp_sk(sk);
2194
2195 /* Applicable to UDP-style socket only */
2196 if (sctp_style(sk, TCP))
2197 return -EOPNOTSUPP;
2198 if (optlen != sizeof(int))
2199 return -EINVAL;
2200 if (copy_from_user(&sp->autoclose, optval, optlen))
2201 return -EFAULT;
2202 /* make sure it won't exceed MAX_SCHEDULE_TIMEOUT */
2203 sp->autoclose = min_t(long, sp->autoclose, MAX_SCHEDULE_TIMEOUT / HZ);
2204
2205 return 0;
2206}
2207
2208/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
2209 *
2210 * Applications can enable or disable heartbeats for any peer address of
2211 * an association, modify an address's heartbeat interval, force a
2212 * heartbeat to be sent immediately, and adjust the address's maximum
2213 * number of retransmissions sent before an address is considered
2214 * unreachable. The following structure is used to access and modify an
2215 * address's parameters:
2216 *
2217 * struct sctp_paddrparams {
2218 * sctp_assoc_t spp_assoc_id;
2219 * struct sockaddr_storage spp_address;
2220 * uint32_t spp_hbinterval;
2221 * uint16_t spp_pathmaxrxt;
2222 * uint32_t spp_pathmtu;
2223 * uint32_t spp_sackdelay;
2224 * uint32_t spp_flags;
2225 * };
2226 *
2227 * spp_assoc_id - (one-to-many style socket) This is filled in the
2228 * application, and identifies the association for
2229 * this query.
2230 * spp_address - This specifies which address is of interest.
2231 * spp_hbinterval - This contains the value of the heartbeat interval,
2232 * in milliseconds. If a value of zero
2233 * is present in this field then no changes are to
2234 * be made to this parameter.
2235 * spp_pathmaxrxt - This contains the maximum number of
2236 * retransmissions before this address shall be
2237 * considered unreachable. If a value of zero
2238 * is present in this field then no changes are to
2239 * be made to this parameter.
2240 * spp_pathmtu - When Path MTU discovery is disabled the value
2241 * specified here will be the "fixed" path mtu.
2242 * Note that if the spp_address field is empty
2243 * then all associations on this address will
2244 * have this fixed path mtu set upon them.
2245 *
2246 * spp_sackdelay - When delayed sack is enabled, this value specifies
2247 * the number of milliseconds that sacks will be delayed
2248 * for. This value will apply to all addresses of an
2249 * association if the spp_address field is empty. Note
2250 * also, that if delayed sack is enabled and this
2251 * value is set to 0, no change is made to the last
2252 * recorded delayed sack timer value.
2253 *
2254 * spp_flags - These flags are used to control various features
2255 * on an association. The flag field may contain
2256 * zero or more of the following options.
2257 *
2258 * SPP_HB_ENABLE - Enable heartbeats on the
2259 * specified address. Note that if the address
2260 * field is empty all addresses for the association
2261 * have heartbeats enabled upon them.
2262 *
2263 * SPP_HB_DISABLE - Disable heartbeats on the
2264 * speicifed address. Note that if the address
2265 * field is empty all addresses for the association
2266 * will have their heartbeats disabled. Note also
2267 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
2268 * mutually exclusive, only one of these two should
2269 * be specified. Enabling both fields will have
2270 * undetermined results.
2271 *
2272 * SPP_HB_DEMAND - Request a user initiated heartbeat
2273 * to be made immediately.
2274 *
2275 * SPP_HB_TIME_IS_ZERO - Specify's that the time for
2276 * heartbeat delayis to be set to the value of 0
2277 * milliseconds.
2278 *
2279 * SPP_PMTUD_ENABLE - This field will enable PMTU
2280 * discovery upon the specified address. Note that
2281 * if the address feild is empty then all addresses
2282 * on the association are effected.
2283 *
2284 * SPP_PMTUD_DISABLE - This field will disable PMTU
2285 * discovery upon the specified address. Note that
2286 * if the address feild is empty then all addresses
2287 * on the association are effected. Not also that
2288 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
2289 * exclusive. Enabling both will have undetermined
2290 * results.
2291 *
2292 * SPP_SACKDELAY_ENABLE - Setting this flag turns
2293 * on delayed sack. The time specified in spp_sackdelay
2294 * is used to specify the sack delay for this address. Note
2295 * that if spp_address is empty then all addresses will
2296 * enable delayed sack and take on the sack delay
2297 * value specified in spp_sackdelay.
2298 * SPP_SACKDELAY_DISABLE - Setting this flag turns
2299 * off delayed sack. If the spp_address field is blank then
2300 * delayed sack is disabled for the entire association. Note
2301 * also that this field is mutually exclusive to
2302 * SPP_SACKDELAY_ENABLE, setting both will have undefined
2303 * results.
2304 */
2305static int sctp_apply_peer_addr_params(struct sctp_paddrparams *params,
2306 struct sctp_transport *trans,
2307 struct sctp_association *asoc,
2308 struct sctp_sock *sp,
2309 int hb_change,
2310 int pmtud_change,
2311 int sackdelay_change)
2312{
2313 int error;
2314
2315 if (params->spp_flags & SPP_HB_DEMAND && trans) {
2316 error = sctp_primitive_REQUESTHEARTBEAT (trans->asoc, trans);
2317 if (error)
2318 return error;
2319 }
2320
2321 /* Note that unless the spp_flag is set to SPP_HB_ENABLE the value of
2322 * this field is ignored. Note also that a value of zero indicates
2323 * the current setting should be left unchanged.
2324 */
2325 if (params->spp_flags & SPP_HB_ENABLE) {
2326
2327 /* Re-zero the interval if the SPP_HB_TIME_IS_ZERO is
2328 * set. This lets us use 0 value when this flag
2329 * is set.
2330 */
2331 if (params->spp_flags & SPP_HB_TIME_IS_ZERO)
2332 params->spp_hbinterval = 0;
2333
2334 if (params->spp_hbinterval ||
2335 (params->spp_flags & SPP_HB_TIME_IS_ZERO)) {
2336 if (trans) {
2337 trans->hbinterval =
2338 msecs_to_jiffies(params->spp_hbinterval);
2339 } else if (asoc) {
2340 asoc->hbinterval =
2341 msecs_to_jiffies(params->spp_hbinterval);
2342 } else {
2343 sp->hbinterval = params->spp_hbinterval;
2344 }
2345 }
2346 }
2347
2348 if (hb_change) {
2349 if (trans) {
2350 trans->param_flags =
2351 (trans->param_flags & ~SPP_HB) | hb_change;
2352 } else if (asoc) {
2353 asoc->param_flags =
2354 (asoc->param_flags & ~SPP_HB) | hb_change;
2355 } else {
2356 sp->param_flags =
2357 (sp->param_flags & ~SPP_HB) | hb_change;
2358 }
2359 }
2360
2361 /* When Path MTU discovery is disabled the value specified here will
2362 * be the "fixed" path mtu (i.e. the value of the spp_flags field must
2363 * include the flag SPP_PMTUD_DISABLE for this field to have any
2364 * effect).
2365 */
2366 if ((params->spp_flags & SPP_PMTUD_DISABLE) && params->spp_pathmtu) {
2367 if (trans) {
2368 trans->pathmtu = params->spp_pathmtu;
2369 sctp_assoc_sync_pmtu(asoc);
2370 } else if (asoc) {
2371 asoc->pathmtu = params->spp_pathmtu;
2372 sctp_frag_point(asoc, params->spp_pathmtu);
2373 } else {
2374 sp->pathmtu = params->spp_pathmtu;
2375 }
2376 }
2377
2378 if (pmtud_change) {
2379 if (trans) {
2380 int update = (trans->param_flags & SPP_PMTUD_DISABLE) &&
2381 (params->spp_flags & SPP_PMTUD_ENABLE);
2382 trans->param_flags =
2383 (trans->param_flags & ~SPP_PMTUD) | pmtud_change;
2384 if (update) {
2385 sctp_transport_pmtu(trans, sctp_opt2sk(sp));
2386 sctp_assoc_sync_pmtu(asoc);
2387 }
2388 } else if (asoc) {
2389 asoc->param_flags =
2390 (asoc->param_flags & ~SPP_PMTUD) | pmtud_change;
2391 } else {
2392 sp->param_flags =
2393 (sp->param_flags & ~SPP_PMTUD) | pmtud_change;
2394 }
2395 }
2396
2397 /* Note that unless the spp_flag is set to SPP_SACKDELAY_ENABLE the
2398 * value of this field is ignored. Note also that a value of zero
2399 * indicates the current setting should be left unchanged.
2400 */
2401 if ((params->spp_flags & SPP_SACKDELAY_ENABLE) && params->spp_sackdelay) {
2402 if (trans) {
2403 trans->sackdelay =
2404 msecs_to_jiffies(params->spp_sackdelay);
2405 } else if (asoc) {
2406 asoc->sackdelay =
2407 msecs_to_jiffies(params->spp_sackdelay);
2408 } else {
2409 sp->sackdelay = params->spp_sackdelay;
2410 }
2411 }
2412
2413 if (sackdelay_change) {
2414 if (trans) {
2415 trans->param_flags =
2416 (trans->param_flags & ~SPP_SACKDELAY) |
2417 sackdelay_change;
2418 } else if (asoc) {
2419 asoc->param_flags =
2420 (asoc->param_flags & ~SPP_SACKDELAY) |
2421 sackdelay_change;
2422 } else {
2423 sp->param_flags =
2424 (sp->param_flags & ~SPP_SACKDELAY) |
2425 sackdelay_change;
2426 }
2427 }
2428
2429 /* Note that a value of zero indicates the current setting should be
2430 left unchanged.
2431 */
2432 if (params->spp_pathmaxrxt) {
2433 if (trans) {
2434 trans->pathmaxrxt = params->spp_pathmaxrxt;
2435 } else if (asoc) {
2436 asoc->pathmaxrxt = params->spp_pathmaxrxt;
2437 } else {
2438 sp->pathmaxrxt = params->spp_pathmaxrxt;
2439 }
2440 }
2441
2442 return 0;
2443}
2444
2445static int sctp_setsockopt_peer_addr_params(struct sock *sk,
2446 char __user *optval,
2447 unsigned int optlen)
2448{
2449 struct sctp_paddrparams params;
2450 struct sctp_transport *trans = NULL;
2451 struct sctp_association *asoc = NULL;
2452 struct sctp_sock *sp = sctp_sk(sk);
2453 int error;
2454 int hb_change, pmtud_change, sackdelay_change;
2455
2456 if (optlen != sizeof(struct sctp_paddrparams))
2457 return - EINVAL;
2458
2459 if (copy_from_user(¶ms, optval, optlen))
2460 return -EFAULT;
2461
2462 /* Validate flags and value parameters. */
2463 hb_change = params.spp_flags & SPP_HB;
2464 pmtud_change = params.spp_flags & SPP_PMTUD;
2465 sackdelay_change = params.spp_flags & SPP_SACKDELAY;
2466
2467 if (hb_change == SPP_HB ||
2468 pmtud_change == SPP_PMTUD ||
2469 sackdelay_change == SPP_SACKDELAY ||
2470 params.spp_sackdelay > 500 ||
2471 (params.spp_pathmtu &&
2472 params.spp_pathmtu < SCTP_DEFAULT_MINSEGMENT))
2473 return -EINVAL;
2474
2475 /* If an address other than INADDR_ANY is specified, and
2476 * no transport is found, then the request is invalid.
2477 */
2478 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
2479 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
2480 params.spp_assoc_id);
2481 if (!trans)
2482 return -EINVAL;
2483 }
2484
2485 /* Get association, if assoc_id != 0 and the socket is a one
2486 * to many style socket, and an association was not found, then
2487 * the id was invalid.
2488 */
2489 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
2490 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP))
2491 return -EINVAL;
2492
2493 /* Heartbeat demand can only be sent on a transport or
2494 * association, but not a socket.
2495 */
2496 if (params.spp_flags & SPP_HB_DEMAND && !trans && !asoc)
2497 return -EINVAL;
2498
2499 /* Process parameters. */
2500 error = sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2501 hb_change, pmtud_change,
2502 sackdelay_change);
2503
2504 if (error)
2505 return error;
2506
2507 /* If changes are for association, also apply parameters to each
2508 * transport.
2509 */
2510 if (!trans && asoc) {
2511 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2512 transports) {
2513 sctp_apply_peer_addr_params(¶ms, trans, asoc, sp,
2514 hb_change, pmtud_change,
2515 sackdelay_change);
2516 }
2517 }
2518
2519 return 0;
2520}
2521
2522/*
2523 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
2524 *
2525 * This option will effect the way delayed acks are performed. This
2526 * option allows you to get or set the delayed ack time, in
2527 * milliseconds. It also allows changing the delayed ack frequency.
2528 * Changing the frequency to 1 disables the delayed sack algorithm. If
2529 * the assoc_id is 0, then this sets or gets the endpoints default
2530 * values. If the assoc_id field is non-zero, then the set or get
2531 * effects the specified association for the one to many model (the
2532 * assoc_id field is ignored by the one to one model). Note that if
2533 * sack_delay or sack_freq are 0 when setting this option, then the
2534 * current values will remain unchanged.
2535 *
2536 * struct sctp_sack_info {
2537 * sctp_assoc_t sack_assoc_id;
2538 * uint32_t sack_delay;
2539 * uint32_t sack_freq;
2540 * };
2541 *
2542 * sack_assoc_id - This parameter, indicates which association the user
2543 * is performing an action upon. Note that if this field's value is
2544 * zero then the endpoints default value is changed (effecting future
2545 * associations only).
2546 *
2547 * sack_delay - This parameter contains the number of milliseconds that
2548 * the user is requesting the delayed ACK timer be set to. Note that
2549 * this value is defined in the standard to be between 200 and 500
2550 * milliseconds.
2551 *
2552 * sack_freq - This parameter contains the number of packets that must
2553 * be received before a sack is sent without waiting for the delay
2554 * timer to expire. The default value for this is 2, setting this
2555 * value to 1 will disable the delayed sack algorithm.
2556 */
2557
2558static int sctp_setsockopt_delayed_ack(struct sock *sk,
2559 char __user *optval, unsigned int optlen)
2560{
2561 struct sctp_sack_info params;
2562 struct sctp_transport *trans = NULL;
2563 struct sctp_association *asoc = NULL;
2564 struct sctp_sock *sp = sctp_sk(sk);
2565
2566 if (optlen == sizeof(struct sctp_sack_info)) {
2567 if (copy_from_user(¶ms, optval, optlen))
2568 return -EFAULT;
2569
2570 if (params.sack_delay == 0 && params.sack_freq == 0)
2571 return 0;
2572 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2573 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
2574 pr_warn("Use struct sctp_sack_info instead\n");
2575 if (copy_from_user(¶ms, optval, optlen))
2576 return -EFAULT;
2577
2578 if (params.sack_delay == 0)
2579 params.sack_freq = 1;
2580 else
2581 params.sack_freq = 0;
2582 } else
2583 return - EINVAL;
2584
2585 /* Validate value parameter. */
2586 if (params.sack_delay > 500)
2587 return -EINVAL;
2588
2589 /* Get association, if sack_assoc_id != 0 and the socket is a one
2590 * to many style socket, and an association was not found, then
2591 * the id was invalid.
2592 */
2593 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
2594 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
2595 return -EINVAL;
2596
2597 if (params.sack_delay) {
2598 if (asoc) {
2599 asoc->sackdelay =
2600 msecs_to_jiffies(params.sack_delay);
2601 asoc->param_flags =
2602 (asoc->param_flags & ~SPP_SACKDELAY) |
2603 SPP_SACKDELAY_ENABLE;
2604 } else {
2605 sp->sackdelay = params.sack_delay;
2606 sp->param_flags =
2607 (sp->param_flags & ~SPP_SACKDELAY) |
2608 SPP_SACKDELAY_ENABLE;
2609 }
2610 }
2611
2612 if (params.sack_freq == 1) {
2613 if (asoc) {
2614 asoc->param_flags =
2615 (asoc->param_flags & ~SPP_SACKDELAY) |
2616 SPP_SACKDELAY_DISABLE;
2617 } else {
2618 sp->param_flags =
2619 (sp->param_flags & ~SPP_SACKDELAY) |
2620 SPP_SACKDELAY_DISABLE;
2621 }
2622 } else if (params.sack_freq > 1) {
2623 if (asoc) {
2624 asoc->sackfreq = params.sack_freq;
2625 asoc->param_flags =
2626 (asoc->param_flags & ~SPP_SACKDELAY) |
2627 SPP_SACKDELAY_ENABLE;
2628 } else {
2629 sp->sackfreq = params.sack_freq;
2630 sp->param_flags =
2631 (sp->param_flags & ~SPP_SACKDELAY) |
2632 SPP_SACKDELAY_ENABLE;
2633 }
2634 }
2635
2636 /* If change is for association, also apply to each transport. */
2637 if (asoc) {
2638 list_for_each_entry(trans, &asoc->peer.transport_addr_list,
2639 transports) {
2640 if (params.sack_delay) {
2641 trans->sackdelay =
2642 msecs_to_jiffies(params.sack_delay);
2643 trans->param_flags =
2644 (trans->param_flags & ~SPP_SACKDELAY) |
2645 SPP_SACKDELAY_ENABLE;
2646 }
2647 if (params.sack_freq == 1) {
2648 trans->param_flags =
2649 (trans->param_flags & ~SPP_SACKDELAY) |
2650 SPP_SACKDELAY_DISABLE;
2651 } else if (params.sack_freq > 1) {
2652 trans->sackfreq = params.sack_freq;
2653 trans->param_flags =
2654 (trans->param_flags & ~SPP_SACKDELAY) |
2655 SPP_SACKDELAY_ENABLE;
2656 }
2657 }
2658 }
2659
2660 return 0;
2661}
2662
2663/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
2664 *
2665 * Applications can specify protocol parameters for the default association
2666 * initialization. The option name argument to setsockopt() and getsockopt()
2667 * is SCTP_INITMSG.
2668 *
2669 * Setting initialization parameters is effective only on an unconnected
2670 * socket (for UDP-style sockets only future associations are effected
2671 * by the change). With TCP-style sockets, this option is inherited by
2672 * sockets derived from a listener socket.
2673 */
2674static int sctp_setsockopt_initmsg(struct sock *sk, char __user *optval, unsigned int optlen)
2675{
2676 struct sctp_initmsg sinit;
2677 struct sctp_sock *sp = sctp_sk(sk);
2678
2679 if (optlen != sizeof(struct sctp_initmsg))
2680 return -EINVAL;
2681 if (copy_from_user(&sinit, optval, optlen))
2682 return -EFAULT;
2683
2684 if (sinit.sinit_num_ostreams)
2685 sp->initmsg.sinit_num_ostreams = sinit.sinit_num_ostreams;
2686 if (sinit.sinit_max_instreams)
2687 sp->initmsg.sinit_max_instreams = sinit.sinit_max_instreams;
2688 if (sinit.sinit_max_attempts)
2689 sp->initmsg.sinit_max_attempts = sinit.sinit_max_attempts;
2690 if (sinit.sinit_max_init_timeo)
2691 sp->initmsg.sinit_max_init_timeo = sinit.sinit_max_init_timeo;
2692
2693 return 0;
2694}
2695
2696/*
2697 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
2698 *
2699 * Applications that wish to use the sendto() system call may wish to
2700 * specify a default set of parameters that would normally be supplied
2701 * through the inclusion of ancillary data. This socket option allows
2702 * such an application to set the default sctp_sndrcvinfo structure.
2703 * The application that wishes to use this socket option simply passes
2704 * in to this call the sctp_sndrcvinfo structure defined in Section
2705 * 5.2.2) The input parameters accepted by this call include
2706 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
2707 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
2708 * to this call if the caller is using the UDP model.
2709 */
2710static int sctp_setsockopt_default_send_param(struct sock *sk,
2711 char __user *optval,
2712 unsigned int optlen)
2713{
2714 struct sctp_sndrcvinfo info;
2715 struct sctp_association *asoc;
2716 struct sctp_sock *sp = sctp_sk(sk);
2717
2718 if (optlen != sizeof(struct sctp_sndrcvinfo))
2719 return -EINVAL;
2720 if (copy_from_user(&info, optval, optlen))
2721 return -EFAULT;
2722
2723 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
2724 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
2725 return -EINVAL;
2726
2727 if (asoc) {
2728 asoc->default_stream = info.sinfo_stream;
2729 asoc->default_flags = info.sinfo_flags;
2730 asoc->default_ppid = info.sinfo_ppid;
2731 asoc->default_context = info.sinfo_context;
2732 asoc->default_timetolive = info.sinfo_timetolive;
2733 } else {
2734 sp->default_stream = info.sinfo_stream;
2735 sp->default_flags = info.sinfo_flags;
2736 sp->default_ppid = info.sinfo_ppid;
2737 sp->default_context = info.sinfo_context;
2738 sp->default_timetolive = info.sinfo_timetolive;
2739 }
2740
2741 return 0;
2742}
2743
2744/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
2745 *
2746 * Requests that the local SCTP stack use the enclosed peer address as
2747 * the association primary. The enclosed address must be one of the
2748 * association peer's addresses.
2749 */
2750static int sctp_setsockopt_primary_addr(struct sock *sk, char __user *optval,
2751 unsigned int optlen)
2752{
2753 struct sctp_prim prim;
2754 struct sctp_transport *trans;
2755
2756 if (optlen != sizeof(struct sctp_prim))
2757 return -EINVAL;
2758
2759 if (copy_from_user(&prim, optval, sizeof(struct sctp_prim)))
2760 return -EFAULT;
2761
2762 trans = sctp_addr_id2transport(sk, &prim.ssp_addr, prim.ssp_assoc_id);
2763 if (!trans)
2764 return -EINVAL;
2765
2766 sctp_assoc_set_primary(trans->asoc, trans);
2767
2768 return 0;
2769}
2770
2771/*
2772 * 7.1.5 SCTP_NODELAY
2773 *
2774 * Turn on/off any Nagle-like algorithm. This means that packets are
2775 * generally sent as soon as possible and no unnecessary delays are
2776 * introduced, at the cost of more packets in the network. Expects an
2777 * integer boolean flag.
2778 */
2779static int sctp_setsockopt_nodelay(struct sock *sk, char __user *optval,
2780 unsigned int optlen)
2781{
2782 int val;
2783
2784 if (optlen < sizeof(int))
2785 return -EINVAL;
2786 if (get_user(val, (int __user *)optval))
2787 return -EFAULT;
2788
2789 sctp_sk(sk)->nodelay = (val == 0) ? 0 : 1;
2790 return 0;
2791}
2792
2793/*
2794 *
2795 * 7.1.1 SCTP_RTOINFO
2796 *
2797 * The protocol parameters used to initialize and bound retransmission
2798 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
2799 * and modify these parameters.
2800 * All parameters are time values, in milliseconds. A value of 0, when
2801 * modifying the parameters, indicates that the current value should not
2802 * be changed.
2803 *
2804 */
2805static int sctp_setsockopt_rtoinfo(struct sock *sk, char __user *optval, unsigned int optlen)
2806{
2807 struct sctp_rtoinfo rtoinfo;
2808 struct sctp_association *asoc;
2809
2810 if (optlen != sizeof (struct sctp_rtoinfo))
2811 return -EINVAL;
2812
2813 if (copy_from_user(&rtoinfo, optval, optlen))
2814 return -EFAULT;
2815
2816 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
2817
2818 /* Set the values to the specific association */
2819 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
2820 return -EINVAL;
2821
2822 if (asoc) {
2823 if (rtoinfo.srto_initial != 0)
2824 asoc->rto_initial =
2825 msecs_to_jiffies(rtoinfo.srto_initial);
2826 if (rtoinfo.srto_max != 0)
2827 asoc->rto_max = msecs_to_jiffies(rtoinfo.srto_max);
2828 if (rtoinfo.srto_min != 0)
2829 asoc->rto_min = msecs_to_jiffies(rtoinfo.srto_min);
2830 } else {
2831 /* If there is no association or the association-id = 0
2832 * set the values to the endpoint.
2833 */
2834 struct sctp_sock *sp = sctp_sk(sk);
2835
2836 if (rtoinfo.srto_initial != 0)
2837 sp->rtoinfo.srto_initial = rtoinfo.srto_initial;
2838 if (rtoinfo.srto_max != 0)
2839 sp->rtoinfo.srto_max = rtoinfo.srto_max;
2840 if (rtoinfo.srto_min != 0)
2841 sp->rtoinfo.srto_min = rtoinfo.srto_min;
2842 }
2843
2844 return 0;
2845}
2846
2847/*
2848 *
2849 * 7.1.2 SCTP_ASSOCINFO
2850 *
2851 * This option is used to tune the maximum retransmission attempts
2852 * of the association.
2853 * Returns an error if the new association retransmission value is
2854 * greater than the sum of the retransmission value of the peer.
2855 * See [SCTP] for more information.
2856 *
2857 */
2858static int sctp_setsockopt_associnfo(struct sock *sk, char __user *optval, unsigned int optlen)
2859{
2860
2861 struct sctp_assocparams assocparams;
2862 struct sctp_association *asoc;
2863
2864 if (optlen != sizeof(struct sctp_assocparams))
2865 return -EINVAL;
2866 if (copy_from_user(&assocparams, optval, optlen))
2867 return -EFAULT;
2868
2869 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
2870
2871 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
2872 return -EINVAL;
2873
2874 /* Set the values to the specific association */
2875 if (asoc) {
2876 if (assocparams.sasoc_asocmaxrxt != 0) {
2877 __u32 path_sum = 0;
2878 int paths = 0;
2879 struct sctp_transport *peer_addr;
2880
2881 list_for_each_entry(peer_addr, &asoc->peer.transport_addr_list,
2882 transports) {
2883 path_sum += peer_addr->pathmaxrxt;
2884 paths++;
2885 }
2886
2887 /* Only validate asocmaxrxt if we have more than
2888 * one path/transport. We do this because path
2889 * retransmissions are only counted when we have more
2890 * then one path.
2891 */
2892 if (paths > 1 &&
2893 assocparams.sasoc_asocmaxrxt > path_sum)
2894 return -EINVAL;
2895
2896 asoc->max_retrans = assocparams.sasoc_asocmaxrxt;
2897 }
2898
2899 if (assocparams.sasoc_cookie_life != 0) {
2900 asoc->cookie_life.tv_sec =
2901 assocparams.sasoc_cookie_life / 1000;
2902 asoc->cookie_life.tv_usec =
2903 (assocparams.sasoc_cookie_life % 1000)
2904 * 1000;
2905 }
2906 } else {
2907 /* Set the values to the endpoint */
2908 struct sctp_sock *sp = sctp_sk(sk);
2909
2910 if (assocparams.sasoc_asocmaxrxt != 0)
2911 sp->assocparams.sasoc_asocmaxrxt =
2912 assocparams.sasoc_asocmaxrxt;
2913 if (assocparams.sasoc_cookie_life != 0)
2914 sp->assocparams.sasoc_cookie_life =
2915 assocparams.sasoc_cookie_life;
2916 }
2917 return 0;
2918}
2919
2920/*
2921 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
2922 *
2923 * This socket option is a boolean flag which turns on or off mapped V4
2924 * addresses. If this option is turned on and the socket is type
2925 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
2926 * If this option is turned off, then no mapping will be done of V4
2927 * addresses and a user will receive both PF_INET6 and PF_INET type
2928 * addresses on the socket.
2929 */
2930static int sctp_setsockopt_mappedv4(struct sock *sk, char __user *optval, unsigned int optlen)
2931{
2932 int val;
2933 struct sctp_sock *sp = sctp_sk(sk);
2934
2935 if (optlen < sizeof(int))
2936 return -EINVAL;
2937 if (get_user(val, (int __user *)optval))
2938 return -EFAULT;
2939 if (val)
2940 sp->v4mapped = 1;
2941 else
2942 sp->v4mapped = 0;
2943
2944 return 0;
2945}
2946
2947/*
2948 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
2949 * This option will get or set the maximum size to put in any outgoing
2950 * SCTP DATA chunk. If a message is larger than this size it will be
2951 * fragmented by SCTP into the specified size. Note that the underlying
2952 * SCTP implementation may fragment into smaller sized chunks when the
2953 * PMTU of the underlying association is smaller than the value set by
2954 * the user. The default value for this option is '0' which indicates
2955 * the user is NOT limiting fragmentation and only the PMTU will effect
2956 * SCTP's choice of DATA chunk size. Note also that values set larger
2957 * than the maximum size of an IP datagram will effectively let SCTP
2958 * control fragmentation (i.e. the same as setting this option to 0).
2959 *
2960 * The following structure is used to access and modify this parameter:
2961 *
2962 * struct sctp_assoc_value {
2963 * sctp_assoc_t assoc_id;
2964 * uint32_t assoc_value;
2965 * };
2966 *
2967 * assoc_id: This parameter is ignored for one-to-one style sockets.
2968 * For one-to-many style sockets this parameter indicates which
2969 * association the user is performing an action upon. Note that if
2970 * this field's value is zero then the endpoints default value is
2971 * changed (effecting future associations only).
2972 * assoc_value: This parameter specifies the maximum size in bytes.
2973 */
2974static int sctp_setsockopt_maxseg(struct sock *sk, char __user *optval, unsigned int optlen)
2975{
2976 struct sctp_assoc_value params;
2977 struct sctp_association *asoc;
2978 struct sctp_sock *sp = sctp_sk(sk);
2979 int val;
2980
2981 if (optlen == sizeof(int)) {
2982 pr_warn("Use of int in maxseg socket option deprecated\n");
2983 pr_warn("Use struct sctp_assoc_value instead\n");
2984 if (copy_from_user(&val, optval, optlen))
2985 return -EFAULT;
2986 params.assoc_id = 0;
2987 } else if (optlen == sizeof(struct sctp_assoc_value)) {
2988 if (copy_from_user(¶ms, optval, optlen))
2989 return -EFAULT;
2990 val = params.assoc_value;
2991 } else
2992 return -EINVAL;
2993
2994 if ((val != 0) && ((val < 8) || (val > SCTP_MAX_CHUNK_LEN)))
2995 return -EINVAL;
2996
2997 asoc = sctp_id2assoc(sk, params.assoc_id);
2998 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
2999 return -EINVAL;
3000
3001 if (asoc) {
3002 if (val == 0) {
3003 val = asoc->pathmtu;
3004 val -= sp->pf->af->net_header_len;
3005 val -= sizeof(struct sctphdr) +
3006 sizeof(struct sctp_data_chunk);
3007 }
3008 asoc->user_frag = val;
3009 asoc->frag_point = sctp_frag_point(asoc, asoc->pathmtu);
3010 } else {
3011 sp->user_frag = val;
3012 }
3013
3014 return 0;
3015}
3016
3017
3018/*
3019 * 7.1.9 Set Peer Primary Address (SCTP_SET_PEER_PRIMARY_ADDR)
3020 *
3021 * Requests that the peer mark the enclosed address as the association
3022 * primary. The enclosed address must be one of the association's
3023 * locally bound addresses. The following structure is used to make a
3024 * set primary request:
3025 */
3026static int sctp_setsockopt_peer_primary_addr(struct sock *sk, char __user *optval,
3027 unsigned int optlen)
3028{
3029 struct sctp_sock *sp;
3030 struct sctp_association *asoc = NULL;
3031 struct sctp_setpeerprim prim;
3032 struct sctp_chunk *chunk;
3033 struct sctp_af *af;
3034 int err;
3035
3036 sp = sctp_sk(sk);
3037
3038 if (!sctp_addip_enable)
3039 return -EPERM;
3040
3041 if (optlen != sizeof(struct sctp_setpeerprim))
3042 return -EINVAL;
3043
3044 if (copy_from_user(&prim, optval, optlen))
3045 return -EFAULT;
3046
3047 asoc = sctp_id2assoc(sk, prim.sspp_assoc_id);
3048 if (!asoc)
3049 return -EINVAL;
3050
3051 if (!asoc->peer.asconf_capable)
3052 return -EPERM;
3053
3054 if (asoc->peer.addip_disabled_mask & SCTP_PARAM_SET_PRIMARY)
3055 return -EPERM;
3056
3057 if (!sctp_state(asoc, ESTABLISHED))
3058 return -ENOTCONN;
3059
3060 af = sctp_get_af_specific(prim.sspp_addr.ss_family);
3061 if (!af)
3062 return -EINVAL;
3063
3064 if (!af->addr_valid((union sctp_addr *)&prim.sspp_addr, sp, NULL))
3065 return -EADDRNOTAVAIL;
3066
3067 if (!sctp_assoc_lookup_laddr(asoc, (union sctp_addr *)&prim.sspp_addr))
3068 return -EADDRNOTAVAIL;
3069
3070 /* Create an ASCONF chunk with SET_PRIMARY parameter */
3071 chunk = sctp_make_asconf_set_prim(asoc,
3072 (union sctp_addr *)&prim.sspp_addr);
3073 if (!chunk)
3074 return -ENOMEM;
3075
3076 err = sctp_send_asconf(asoc, chunk);
3077
3078 SCTP_DEBUG_PRINTK("We set peer primary addr primitively.\n");
3079
3080 return err;
3081}
3082
3083static int sctp_setsockopt_adaptation_layer(struct sock *sk, char __user *optval,
3084 unsigned int optlen)
3085{
3086 struct sctp_setadaptation adaptation;
3087
3088 if (optlen != sizeof(struct sctp_setadaptation))
3089 return -EINVAL;
3090 if (copy_from_user(&adaptation, optval, optlen))
3091 return -EFAULT;
3092
3093 sctp_sk(sk)->adaptation_ind = adaptation.ssb_adaptation_ind;
3094
3095 return 0;
3096}
3097
3098/*
3099 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
3100 *
3101 * The context field in the sctp_sndrcvinfo structure is normally only
3102 * used when a failed message is retrieved holding the value that was
3103 * sent down on the actual send call. This option allows the setting of
3104 * a default context on an association basis that will be received on
3105 * reading messages from the peer. This is especially helpful in the
3106 * one-2-many model for an application to keep some reference to an
3107 * internal state machine that is processing messages on the
3108 * association. Note that the setting of this value only effects
3109 * received messages from the peer and does not effect the value that is
3110 * saved with outbound messages.
3111 */
3112static int sctp_setsockopt_context(struct sock *sk, char __user *optval,
3113 unsigned int optlen)
3114{
3115 struct sctp_assoc_value params;
3116 struct sctp_sock *sp;
3117 struct sctp_association *asoc;
3118
3119 if (optlen != sizeof(struct sctp_assoc_value))
3120 return -EINVAL;
3121 if (copy_from_user(¶ms, optval, optlen))
3122 return -EFAULT;
3123
3124 sp = sctp_sk(sk);
3125
3126 if (params.assoc_id != 0) {
3127 asoc = sctp_id2assoc(sk, params.assoc_id);
3128 if (!asoc)
3129 return -EINVAL;
3130 asoc->default_rcv_context = params.assoc_value;
3131 } else {
3132 sp->default_rcv_context = params.assoc_value;
3133 }
3134
3135 return 0;
3136}
3137
3138/*
3139 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
3140 *
3141 * This options will at a minimum specify if the implementation is doing
3142 * fragmented interleave. Fragmented interleave, for a one to many
3143 * socket, is when subsequent calls to receive a message may return
3144 * parts of messages from different associations. Some implementations
3145 * may allow you to turn this value on or off. If so, when turned off,
3146 * no fragment interleave will occur (which will cause a head of line
3147 * blocking amongst multiple associations sharing the same one to many
3148 * socket). When this option is turned on, then each receive call may
3149 * come from a different association (thus the user must receive data
3150 * with the extended calls (e.g. sctp_recvmsg) to keep track of which
3151 * association each receive belongs to.
3152 *
3153 * This option takes a boolean value. A non-zero value indicates that
3154 * fragmented interleave is on. A value of zero indicates that
3155 * fragmented interleave is off.
3156 *
3157 * Note that it is important that an implementation that allows this
3158 * option to be turned on, have it off by default. Otherwise an unaware
3159 * application using the one to many model may become confused and act
3160 * incorrectly.
3161 */
3162static int sctp_setsockopt_fragment_interleave(struct sock *sk,
3163 char __user *optval,
3164 unsigned int optlen)
3165{
3166 int val;
3167
3168 if (optlen != sizeof(int))
3169 return -EINVAL;
3170 if (get_user(val, (int __user *)optval))
3171 return -EFAULT;
3172
3173 sctp_sk(sk)->frag_interleave = (val == 0) ? 0 : 1;
3174
3175 return 0;
3176}
3177
3178/*
3179 * 8.1.21. Set or Get the SCTP Partial Delivery Point
3180 * (SCTP_PARTIAL_DELIVERY_POINT)
3181 *
3182 * This option will set or get the SCTP partial delivery point. This
3183 * point is the size of a message where the partial delivery API will be
3184 * invoked to help free up rwnd space for the peer. Setting this to a
3185 * lower value will cause partial deliveries to happen more often. The
3186 * calls argument is an integer that sets or gets the partial delivery
3187 * point. Note also that the call will fail if the user attempts to set
3188 * this value larger than the socket receive buffer size.
3189 *
3190 * Note that any single message having a length smaller than or equal to
3191 * the SCTP partial delivery point will be delivered in one single read
3192 * call as long as the user provided buffer is large enough to hold the
3193 * message.
3194 */
3195static int sctp_setsockopt_partial_delivery_point(struct sock *sk,
3196 char __user *optval,
3197 unsigned int optlen)
3198{
3199 u32 val;
3200
3201 if (optlen != sizeof(u32))
3202 return -EINVAL;
3203 if (get_user(val, (int __user *)optval))
3204 return -EFAULT;
3205
3206 /* Note: We double the receive buffer from what the user sets
3207 * it to be, also initial rwnd is based on rcvbuf/2.
3208 */
3209 if (val > (sk->sk_rcvbuf >> 1))
3210 return -EINVAL;
3211
3212 sctp_sk(sk)->pd_point = val;
3213
3214 return 0; /* is this the right error code? */
3215}
3216
3217/*
3218 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
3219 *
3220 * This option will allow a user to change the maximum burst of packets
3221 * that can be emitted by this association. Note that the default value
3222 * is 4, and some implementations may restrict this setting so that it
3223 * can only be lowered.
3224 *
3225 * NOTE: This text doesn't seem right. Do this on a socket basis with
3226 * future associations inheriting the socket value.
3227 */
3228static int sctp_setsockopt_maxburst(struct sock *sk,
3229 char __user *optval,
3230 unsigned int optlen)
3231{
3232 struct sctp_assoc_value params;
3233 struct sctp_sock *sp;
3234 struct sctp_association *asoc;
3235 int val;
3236 int assoc_id = 0;
3237
3238 if (optlen == sizeof(int)) {
3239 pr_warn("Use of int in max_burst socket option deprecated\n");
3240 pr_warn("Use struct sctp_assoc_value instead\n");
3241 if (copy_from_user(&val, optval, optlen))
3242 return -EFAULT;
3243 } else if (optlen == sizeof(struct sctp_assoc_value)) {
3244 if (copy_from_user(¶ms, optval, optlen))
3245 return -EFAULT;
3246 val = params.assoc_value;
3247 assoc_id = params.assoc_id;
3248 } else
3249 return -EINVAL;
3250
3251 sp = sctp_sk(sk);
3252
3253 if (assoc_id != 0) {
3254 asoc = sctp_id2assoc(sk, assoc_id);
3255 if (!asoc)
3256 return -EINVAL;
3257 asoc->max_burst = val;
3258 } else
3259 sp->max_burst = val;
3260
3261 return 0;
3262}
3263
3264/*
3265 * 7.1.18. Add a chunk that must be authenticated (SCTP_AUTH_CHUNK)
3266 *
3267 * This set option adds a chunk type that the user is requesting to be
3268 * received only in an authenticated way. Changes to the list of chunks
3269 * will only effect future associations on the socket.
3270 */
3271static int sctp_setsockopt_auth_chunk(struct sock *sk,
3272 char __user *optval,
3273 unsigned int optlen)
3274{
3275 struct sctp_authchunk val;
3276
3277 if (!sctp_auth_enable)
3278 return -EACCES;
3279
3280 if (optlen != sizeof(struct sctp_authchunk))
3281 return -EINVAL;
3282 if (copy_from_user(&val, optval, optlen))
3283 return -EFAULT;
3284
3285 switch (val.sauth_chunk) {
3286 case SCTP_CID_INIT:
3287 case SCTP_CID_INIT_ACK:
3288 case SCTP_CID_SHUTDOWN_COMPLETE:
3289 case SCTP_CID_AUTH:
3290 return -EINVAL;
3291 }
3292
3293 /* add this chunk id to the endpoint */
3294 return sctp_auth_ep_add_chunkid(sctp_sk(sk)->ep, val.sauth_chunk);
3295}
3296
3297/*
3298 * 7.1.19. Get or set the list of supported HMAC Identifiers (SCTP_HMAC_IDENT)
3299 *
3300 * This option gets or sets the list of HMAC algorithms that the local
3301 * endpoint requires the peer to use.
3302 */
3303static int sctp_setsockopt_hmac_ident(struct sock *sk,
3304 char __user *optval,
3305 unsigned int optlen)
3306{
3307 struct sctp_hmacalgo *hmacs;
3308 u32 idents;
3309 int err;
3310
3311 if (!sctp_auth_enable)
3312 return -EACCES;
3313
3314 if (optlen < sizeof(struct sctp_hmacalgo))
3315 return -EINVAL;
3316
3317 hmacs= memdup_user(optval, optlen);
3318 if (IS_ERR(hmacs))
3319 return PTR_ERR(hmacs);
3320
3321 idents = hmacs->shmac_num_idents;
3322 if (idents == 0 || idents > SCTP_AUTH_NUM_HMACS ||
3323 (idents * sizeof(u16)) > (optlen - sizeof(struct sctp_hmacalgo))) {
3324 err = -EINVAL;
3325 goto out;
3326 }
3327
3328 err = sctp_auth_ep_set_hmacs(sctp_sk(sk)->ep, hmacs);
3329out:
3330 kfree(hmacs);
3331 return err;
3332}
3333
3334/*
3335 * 7.1.20. Set a shared key (SCTP_AUTH_KEY)
3336 *
3337 * This option will set a shared secret key which is used to build an
3338 * association shared key.
3339 */
3340static int sctp_setsockopt_auth_key(struct sock *sk,
3341 char __user *optval,
3342 unsigned int optlen)
3343{
3344 struct sctp_authkey *authkey;
3345 struct sctp_association *asoc;
3346 int ret;
3347
3348 if (!sctp_auth_enable)
3349 return -EACCES;
3350
3351 if (optlen <= sizeof(struct sctp_authkey))
3352 return -EINVAL;
3353
3354 authkey= memdup_user(optval, optlen);
3355 if (IS_ERR(authkey))
3356 return PTR_ERR(authkey);
3357
3358 if (authkey->sca_keylength > optlen - sizeof(struct sctp_authkey)) {
3359 ret = -EINVAL;
3360 goto out;
3361 }
3362
3363 asoc = sctp_id2assoc(sk, authkey->sca_assoc_id);
3364 if (!asoc && authkey->sca_assoc_id && sctp_style(sk, UDP)) {
3365 ret = -EINVAL;
3366 goto out;
3367 }
3368
3369 ret = sctp_auth_set_key(sctp_sk(sk)->ep, asoc, authkey);
3370out:
3371 kfree(authkey);
3372 return ret;
3373}
3374
3375/*
3376 * 7.1.21. Get or set the active shared key (SCTP_AUTH_ACTIVE_KEY)
3377 *
3378 * This option will get or set the active shared key to be used to build
3379 * the association shared key.
3380 */
3381static int sctp_setsockopt_active_key(struct sock *sk,
3382 char __user *optval,
3383 unsigned int optlen)
3384{
3385 struct sctp_authkeyid val;
3386 struct sctp_association *asoc;
3387
3388 if (!sctp_auth_enable)
3389 return -EACCES;
3390
3391 if (optlen != sizeof(struct sctp_authkeyid))
3392 return -EINVAL;
3393 if (copy_from_user(&val, optval, optlen))
3394 return -EFAULT;
3395
3396 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3397 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3398 return -EINVAL;
3399
3400 return sctp_auth_set_active_key(sctp_sk(sk)->ep, asoc,
3401 val.scact_keynumber);
3402}
3403
3404/*
3405 * 7.1.22. Delete a shared key (SCTP_AUTH_DELETE_KEY)
3406 *
3407 * This set option will delete a shared secret key from use.
3408 */
3409static int sctp_setsockopt_del_key(struct sock *sk,
3410 char __user *optval,
3411 unsigned int optlen)
3412{
3413 struct sctp_authkeyid val;
3414 struct sctp_association *asoc;
3415
3416 if (!sctp_auth_enable)
3417 return -EACCES;
3418
3419 if (optlen != sizeof(struct sctp_authkeyid))
3420 return -EINVAL;
3421 if (copy_from_user(&val, optval, optlen))
3422 return -EFAULT;
3423
3424 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
3425 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
3426 return -EINVAL;
3427
3428 return sctp_auth_del_key_id(sctp_sk(sk)->ep, asoc,
3429 val.scact_keynumber);
3430
3431}
3432
3433/*
3434 * 8.1.23 SCTP_AUTO_ASCONF
3435 *
3436 * This option will enable or disable the use of the automatic generation of
3437 * ASCONF chunks to add and delete addresses to an existing association. Note
3438 * that this option has two caveats namely: a) it only affects sockets that
3439 * are bound to all addresses available to the SCTP stack, and b) the system
3440 * administrator may have an overriding control that turns the ASCONF feature
3441 * off no matter what setting the socket option may have.
3442 * This option expects an integer boolean flag, where a non-zero value turns on
3443 * the option, and a zero value turns off the option.
3444 * Note. In this implementation, socket operation overrides default parameter
3445 * being set by sysctl as well as FreeBSD implementation
3446 */
3447static int sctp_setsockopt_auto_asconf(struct sock *sk, char __user *optval,
3448 unsigned int optlen)
3449{
3450 int val;
3451 struct sctp_sock *sp = sctp_sk(sk);
3452
3453 if (optlen < sizeof(int))
3454 return -EINVAL;
3455 if (get_user(val, (int __user *)optval))
3456 return -EFAULT;
3457 if (!sctp_is_ep_boundall(sk) && val)
3458 return -EINVAL;
3459 if ((val && sp->do_auto_asconf) || (!val && !sp->do_auto_asconf))
3460 return 0;
3461
3462 if (val == 0 && sp->do_auto_asconf) {
3463 list_del(&sp->auto_asconf_list);
3464 sp->do_auto_asconf = 0;
3465 } else if (val && !sp->do_auto_asconf) {
3466 list_add_tail(&sp->auto_asconf_list,
3467 &sctp_auto_asconf_splist);
3468 sp->do_auto_asconf = 1;
3469 }
3470 return 0;
3471}
3472
3473
3474/* API 6.2 setsockopt(), getsockopt()
3475 *
3476 * Applications use setsockopt() and getsockopt() to set or retrieve
3477 * socket options. Socket options are used to change the default
3478 * behavior of sockets calls. They are described in Section 7.
3479 *
3480 * The syntax is:
3481 *
3482 * ret = getsockopt(int sd, int level, int optname, void __user *optval,
3483 * int __user *optlen);
3484 * ret = setsockopt(int sd, int level, int optname, const void __user *optval,
3485 * int optlen);
3486 *
3487 * sd - the socket descript.
3488 * level - set to IPPROTO_SCTP for all SCTP options.
3489 * optname - the option name.
3490 * optval - the buffer to store the value of the option.
3491 * optlen - the size of the buffer.
3492 */
3493SCTP_STATIC int sctp_setsockopt(struct sock *sk, int level, int optname,
3494 char __user *optval, unsigned int optlen)
3495{
3496 int retval = 0;
3497
3498 SCTP_DEBUG_PRINTK("sctp_setsockopt(sk: %p... optname: %d)\n",
3499 sk, optname);
3500
3501 /* I can hardly begin to describe how wrong this is. This is
3502 * so broken as to be worse than useless. The API draft
3503 * REALLY is NOT helpful here... I am not convinced that the
3504 * semantics of setsockopt() with a level OTHER THAN SOL_SCTP
3505 * are at all well-founded.
3506 */
3507 if (level != SOL_SCTP) {
3508 struct sctp_af *af = sctp_sk(sk)->pf->af;
3509 retval = af->setsockopt(sk, level, optname, optval, optlen);
3510 goto out_nounlock;
3511 }
3512
3513 sctp_lock_sock(sk);
3514
3515 switch (optname) {
3516 case SCTP_SOCKOPT_BINDX_ADD:
3517 /* 'optlen' is the size of the addresses buffer. */
3518 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3519 optlen, SCTP_BINDX_ADD_ADDR);
3520 break;
3521
3522 case SCTP_SOCKOPT_BINDX_REM:
3523 /* 'optlen' is the size of the addresses buffer. */
3524 retval = sctp_setsockopt_bindx(sk, (struct sockaddr __user *)optval,
3525 optlen, SCTP_BINDX_REM_ADDR);
3526 break;
3527
3528 case SCTP_SOCKOPT_CONNECTX_OLD:
3529 /* 'optlen' is the size of the addresses buffer. */
3530 retval = sctp_setsockopt_connectx_old(sk,
3531 (struct sockaddr __user *)optval,
3532 optlen);
3533 break;
3534
3535 case SCTP_SOCKOPT_CONNECTX:
3536 /* 'optlen' is the size of the addresses buffer. */
3537 retval = sctp_setsockopt_connectx(sk,
3538 (struct sockaddr __user *)optval,
3539 optlen);
3540 break;
3541
3542 case SCTP_DISABLE_FRAGMENTS:
3543 retval = sctp_setsockopt_disable_fragments(sk, optval, optlen);
3544 break;
3545
3546 case SCTP_EVENTS:
3547 retval = sctp_setsockopt_events(sk, optval, optlen);
3548 break;
3549
3550 case SCTP_AUTOCLOSE:
3551 retval = sctp_setsockopt_autoclose(sk, optval, optlen);
3552 break;
3553
3554 case SCTP_PEER_ADDR_PARAMS:
3555 retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
3556 break;
3557
3558 case SCTP_DELAYED_SACK:
3559 retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
3560 break;
3561 case SCTP_PARTIAL_DELIVERY_POINT:
3562 retval = sctp_setsockopt_partial_delivery_point(sk, optval, optlen);
3563 break;
3564
3565 case SCTP_INITMSG:
3566 retval = sctp_setsockopt_initmsg(sk, optval, optlen);
3567 break;
3568 case SCTP_DEFAULT_SEND_PARAM:
3569 retval = sctp_setsockopt_default_send_param(sk, optval,
3570 optlen);
3571 break;
3572 case SCTP_PRIMARY_ADDR:
3573 retval = sctp_setsockopt_primary_addr(sk, optval, optlen);
3574 break;
3575 case SCTP_SET_PEER_PRIMARY_ADDR:
3576 retval = sctp_setsockopt_peer_primary_addr(sk, optval, optlen);
3577 break;
3578 case SCTP_NODELAY:
3579 retval = sctp_setsockopt_nodelay(sk, optval, optlen);
3580 break;
3581 case SCTP_RTOINFO:
3582 retval = sctp_setsockopt_rtoinfo(sk, optval, optlen);
3583 break;
3584 case SCTP_ASSOCINFO:
3585 retval = sctp_setsockopt_associnfo(sk, optval, optlen);
3586 break;
3587 case SCTP_I_WANT_MAPPED_V4_ADDR:
3588 retval = sctp_setsockopt_mappedv4(sk, optval, optlen);
3589 break;
3590 case SCTP_MAXSEG:
3591 retval = sctp_setsockopt_maxseg(sk, optval, optlen);
3592 break;
3593 case SCTP_ADAPTATION_LAYER:
3594 retval = sctp_setsockopt_adaptation_layer(sk, optval, optlen);
3595 break;
3596 case SCTP_CONTEXT:
3597 retval = sctp_setsockopt_context(sk, optval, optlen);
3598 break;
3599 case SCTP_FRAGMENT_INTERLEAVE:
3600 retval = sctp_setsockopt_fragment_interleave(sk, optval, optlen);
3601 break;
3602 case SCTP_MAX_BURST:
3603 retval = sctp_setsockopt_maxburst(sk, optval, optlen);
3604 break;
3605 case SCTP_AUTH_CHUNK:
3606 retval = sctp_setsockopt_auth_chunk(sk, optval, optlen);
3607 break;
3608 case SCTP_HMAC_IDENT:
3609 retval = sctp_setsockopt_hmac_ident(sk, optval, optlen);
3610 break;
3611 case SCTP_AUTH_KEY:
3612 retval = sctp_setsockopt_auth_key(sk, optval, optlen);
3613 break;
3614 case SCTP_AUTH_ACTIVE_KEY:
3615 retval = sctp_setsockopt_active_key(sk, optval, optlen);
3616 break;
3617 case SCTP_AUTH_DELETE_KEY:
3618 retval = sctp_setsockopt_del_key(sk, optval, optlen);
3619 break;
3620 case SCTP_AUTO_ASCONF:
3621 retval = sctp_setsockopt_auto_asconf(sk, optval, optlen);
3622 break;
3623 default:
3624 retval = -ENOPROTOOPT;
3625 break;
3626 }
3627
3628 sctp_release_sock(sk);
3629
3630out_nounlock:
3631 return retval;
3632}
3633
3634/* API 3.1.6 connect() - UDP Style Syntax
3635 *
3636 * An application may use the connect() call in the UDP model to initiate an
3637 * association without sending data.
3638 *
3639 * The syntax is:
3640 *
3641 * ret = connect(int sd, const struct sockaddr *nam, socklen_t len);
3642 *
3643 * sd: the socket descriptor to have a new association added to.
3644 *
3645 * nam: the address structure (either struct sockaddr_in or struct
3646 * sockaddr_in6 defined in RFC2553 [7]).
3647 *
3648 * len: the size of the address.
3649 */
3650SCTP_STATIC int sctp_connect(struct sock *sk, struct sockaddr *addr,
3651 int addr_len)
3652{
3653 int err = 0;
3654 struct sctp_af *af;
3655
3656 sctp_lock_sock(sk);
3657
3658 SCTP_DEBUG_PRINTK("%s - sk: %p, sockaddr: %p, addr_len: %d\n",
3659 __func__, sk, addr, addr_len);
3660
3661 /* Validate addr_len before calling common connect/connectx routine. */
3662 af = sctp_get_af_specific(addr->sa_family);
3663 if (!af || addr_len < af->sockaddr_len) {
3664 err = -EINVAL;
3665 } else {
3666 /* Pass correct addr len to common routine (so it knows there
3667 * is only one address being passed.
3668 */
3669 err = __sctp_connect(sk, addr, af->sockaddr_len, NULL);
3670 }
3671
3672 sctp_release_sock(sk);
3673 return err;
3674}
3675
3676/* FIXME: Write comments. */
3677SCTP_STATIC int sctp_disconnect(struct sock *sk, int flags)
3678{
3679 return -EOPNOTSUPP; /* STUB */
3680}
3681
3682/* 4.1.4 accept() - TCP Style Syntax
3683 *
3684 * Applications use accept() call to remove an established SCTP
3685 * association from the accept queue of the endpoint. A new socket
3686 * descriptor will be returned from accept() to represent the newly
3687 * formed association.
3688 */
3689SCTP_STATIC struct sock *sctp_accept(struct sock *sk, int flags, int *err)
3690{
3691 struct sctp_sock *sp;
3692 struct sctp_endpoint *ep;
3693 struct sock *newsk = NULL;
3694 struct sctp_association *asoc;
3695 long timeo;
3696 int error = 0;
3697
3698 sctp_lock_sock(sk);
3699
3700 sp = sctp_sk(sk);
3701 ep = sp->ep;
3702
3703 if (!sctp_style(sk, TCP)) {
3704 error = -EOPNOTSUPP;
3705 goto out;
3706 }
3707
3708 if (!sctp_sstate(sk, LISTENING)) {
3709 error = -EINVAL;
3710 goto out;
3711 }
3712
3713 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
3714
3715 error = sctp_wait_for_accept(sk, timeo);
3716 if (error)
3717 goto out;
3718
3719 /* We treat the list of associations on the endpoint as the accept
3720 * queue and pick the first association on the list.
3721 */
3722 asoc = list_entry(ep->asocs.next, struct sctp_association, asocs);
3723
3724 newsk = sp->pf->create_accept_sk(sk, asoc);
3725 if (!newsk) {
3726 error = -ENOMEM;
3727 goto out;
3728 }
3729
3730 /* Populate the fields of the newsk from the oldsk and migrate the
3731 * asoc to the newsk.
3732 */
3733 sctp_sock_migrate(sk, newsk, asoc, SCTP_SOCKET_TCP);
3734
3735out:
3736 sctp_release_sock(sk);
3737 *err = error;
3738 return newsk;
3739}
3740
3741/* The SCTP ioctl handler. */
3742SCTP_STATIC int sctp_ioctl(struct sock *sk, int cmd, unsigned long arg)
3743{
3744 int rc = -ENOTCONN;
3745
3746 sctp_lock_sock(sk);
3747
3748 /*
3749 * SEQPACKET-style sockets in LISTENING state are valid, for
3750 * SCTP, so only discard TCP-style sockets in LISTENING state.
3751 */
3752 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
3753 goto out;
3754
3755 switch (cmd) {
3756 case SIOCINQ: {
3757 struct sk_buff *skb;
3758 unsigned int amount = 0;
3759
3760 skb = skb_peek(&sk->sk_receive_queue);
3761 if (skb != NULL) {
3762 /*
3763 * We will only return the amount of this packet since
3764 * that is all that will be read.
3765 */
3766 amount = skb->len;
3767 }
3768 rc = put_user(amount, (int __user *)arg);
3769 break;
3770 }
3771 default:
3772 rc = -ENOIOCTLCMD;
3773 break;
3774 }
3775out:
3776 sctp_release_sock(sk);
3777 return rc;
3778}
3779
3780/* This is the function which gets called during socket creation to
3781 * initialized the SCTP-specific portion of the sock.
3782 * The sock structure should already be zero-filled memory.
3783 */
3784SCTP_STATIC int sctp_init_sock(struct sock *sk)
3785{
3786 struct sctp_endpoint *ep;
3787 struct sctp_sock *sp;
3788
3789 SCTP_DEBUG_PRINTK("sctp_init_sock(sk: %p)\n", sk);
3790
3791 sp = sctp_sk(sk);
3792
3793 /* Initialize the SCTP per socket area. */
3794 switch (sk->sk_type) {
3795 case SOCK_SEQPACKET:
3796 sp->type = SCTP_SOCKET_UDP;
3797 break;
3798 case SOCK_STREAM:
3799 sp->type = SCTP_SOCKET_TCP;
3800 break;
3801 default:
3802 return -ESOCKTNOSUPPORT;
3803 }
3804
3805 /* Initialize default send parameters. These parameters can be
3806 * modified with the SCTP_DEFAULT_SEND_PARAM socket option.
3807 */
3808 sp->default_stream = 0;
3809 sp->default_ppid = 0;
3810 sp->default_flags = 0;
3811 sp->default_context = 0;
3812 sp->default_timetolive = 0;
3813
3814 sp->default_rcv_context = 0;
3815 sp->max_burst = sctp_max_burst;
3816
3817 /* Initialize default setup parameters. These parameters
3818 * can be modified with the SCTP_INITMSG socket option or
3819 * overridden by the SCTP_INIT CMSG.
3820 */
3821 sp->initmsg.sinit_num_ostreams = sctp_max_outstreams;
3822 sp->initmsg.sinit_max_instreams = sctp_max_instreams;
3823 sp->initmsg.sinit_max_attempts = sctp_max_retrans_init;
3824 sp->initmsg.sinit_max_init_timeo = sctp_rto_max;
3825
3826 /* Initialize default RTO related parameters. These parameters can
3827 * be modified for with the SCTP_RTOINFO socket option.
3828 */
3829 sp->rtoinfo.srto_initial = sctp_rto_initial;
3830 sp->rtoinfo.srto_max = sctp_rto_max;
3831 sp->rtoinfo.srto_min = sctp_rto_min;
3832
3833 /* Initialize default association related parameters. These parameters
3834 * can be modified with the SCTP_ASSOCINFO socket option.
3835 */
3836 sp->assocparams.sasoc_asocmaxrxt = sctp_max_retrans_association;
3837 sp->assocparams.sasoc_number_peer_destinations = 0;
3838 sp->assocparams.sasoc_peer_rwnd = 0;
3839 sp->assocparams.sasoc_local_rwnd = 0;
3840 sp->assocparams.sasoc_cookie_life = sctp_valid_cookie_life;
3841
3842 /* Initialize default event subscriptions. By default, all the
3843 * options are off.
3844 */
3845 memset(&sp->subscribe, 0, sizeof(struct sctp_event_subscribe));
3846
3847 /* Default Peer Address Parameters. These defaults can
3848 * be modified via SCTP_PEER_ADDR_PARAMS
3849 */
3850 sp->hbinterval = sctp_hb_interval;
3851 sp->pathmaxrxt = sctp_max_retrans_path;
3852 sp->pathmtu = 0; // allow default discovery
3853 sp->sackdelay = sctp_sack_timeout;
3854 sp->sackfreq = 2;
3855 sp->param_flags = SPP_HB_ENABLE |
3856 SPP_PMTUD_ENABLE |
3857 SPP_SACKDELAY_ENABLE;
3858
3859 /* If enabled no SCTP message fragmentation will be performed.
3860 * Configure through SCTP_DISABLE_FRAGMENTS socket option.
3861 */
3862 sp->disable_fragments = 0;
3863
3864 /* Enable Nagle algorithm by default. */
3865 sp->nodelay = 0;
3866
3867 /* Enable by default. */
3868 sp->v4mapped = 1;
3869
3870 /* Auto-close idle associations after the configured
3871 * number of seconds. A value of 0 disables this
3872 * feature. Configure through the SCTP_AUTOCLOSE socket option,
3873 * for UDP-style sockets only.
3874 */
3875 sp->autoclose = 0;
3876
3877 /* User specified fragmentation limit. */
3878 sp->user_frag = 0;
3879
3880 sp->adaptation_ind = 0;
3881
3882 sp->pf = sctp_get_pf_specific(sk->sk_family);
3883
3884 /* Control variables for partial data delivery. */
3885 atomic_set(&sp->pd_mode, 0);
3886 skb_queue_head_init(&sp->pd_lobby);
3887 sp->frag_interleave = 0;
3888
3889 /* Create a per socket endpoint structure. Even if we
3890 * change the data structure relationships, this may still
3891 * be useful for storing pre-connect address information.
3892 */
3893 ep = sctp_endpoint_new(sk, GFP_KERNEL);
3894 if (!ep)
3895 return -ENOMEM;
3896
3897 sp->ep = ep;
3898 sp->hmac = NULL;
3899
3900 SCTP_DBG_OBJCNT_INC(sock);
3901
3902 local_bh_disable();
3903 percpu_counter_inc(&sctp_sockets_allocated);
3904 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
3905 if (sctp_default_auto_asconf) {
3906 list_add_tail(&sp->auto_asconf_list,
3907 &sctp_auto_asconf_splist);
3908 sp->do_auto_asconf = 1;
3909 } else
3910 sp->do_auto_asconf = 0;
3911 local_bh_enable();
3912
3913 return 0;
3914}
3915
3916/* Cleanup any SCTP per socket resources. */
3917SCTP_STATIC void sctp_destroy_sock(struct sock *sk)
3918{
3919 struct sctp_sock *sp;
3920
3921 SCTP_DEBUG_PRINTK("sctp_destroy_sock(sk: %p)\n", sk);
3922
3923 /* Release our hold on the endpoint. */
3924 sp = sctp_sk(sk);
3925 if (sp->do_auto_asconf) {
3926 sp->do_auto_asconf = 0;
3927 list_del(&sp->auto_asconf_list);
3928 }
3929 sctp_endpoint_free(sp->ep);
3930 local_bh_disable();
3931 percpu_counter_dec(&sctp_sockets_allocated);
3932 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
3933 local_bh_enable();
3934}
3935
3936/* API 4.1.7 shutdown() - TCP Style Syntax
3937 * int shutdown(int socket, int how);
3938 *
3939 * sd - the socket descriptor of the association to be closed.
3940 * how - Specifies the type of shutdown. The values are
3941 * as follows:
3942 * SHUT_RD
3943 * Disables further receive operations. No SCTP
3944 * protocol action is taken.
3945 * SHUT_WR
3946 * Disables further send operations, and initiates
3947 * the SCTP shutdown sequence.
3948 * SHUT_RDWR
3949 * Disables further send and receive operations
3950 * and initiates the SCTP shutdown sequence.
3951 */
3952SCTP_STATIC void sctp_shutdown(struct sock *sk, int how)
3953{
3954 struct sctp_endpoint *ep;
3955 struct sctp_association *asoc;
3956
3957 if (!sctp_style(sk, TCP))
3958 return;
3959
3960 if (how & SEND_SHUTDOWN) {
3961 ep = sctp_sk(sk)->ep;
3962 if (!list_empty(&ep->asocs)) {
3963 asoc = list_entry(ep->asocs.next,
3964 struct sctp_association, asocs);
3965 sctp_primitive_SHUTDOWN(asoc, NULL);
3966 }
3967 }
3968}
3969
3970/* 7.2.1 Association Status (SCTP_STATUS)
3971
3972 * Applications can retrieve current status information about an
3973 * association, including association state, peer receiver window size,
3974 * number of unacked data chunks, and number of data chunks pending
3975 * receipt. This information is read-only.
3976 */
3977static int sctp_getsockopt_sctp_status(struct sock *sk, int len,
3978 char __user *optval,
3979 int __user *optlen)
3980{
3981 struct sctp_status status;
3982 struct sctp_association *asoc = NULL;
3983 struct sctp_transport *transport;
3984 sctp_assoc_t associd;
3985 int retval = 0;
3986
3987 if (len < sizeof(status)) {
3988 retval = -EINVAL;
3989 goto out;
3990 }
3991
3992 len = sizeof(status);
3993 if (copy_from_user(&status, optval, len)) {
3994 retval = -EFAULT;
3995 goto out;
3996 }
3997
3998 associd = status.sstat_assoc_id;
3999 asoc = sctp_id2assoc(sk, associd);
4000 if (!asoc) {
4001 retval = -EINVAL;
4002 goto out;
4003 }
4004
4005 transport = asoc->peer.primary_path;
4006
4007 status.sstat_assoc_id = sctp_assoc2id(asoc);
4008 status.sstat_state = asoc->state;
4009 status.sstat_rwnd = asoc->peer.rwnd;
4010 status.sstat_unackdata = asoc->unack_data;
4011
4012 status.sstat_penddata = sctp_tsnmap_pending(&asoc->peer.tsn_map);
4013 status.sstat_instrms = asoc->c.sinit_max_instreams;
4014 status.sstat_outstrms = asoc->c.sinit_num_ostreams;
4015 status.sstat_fragmentation_point = asoc->frag_point;
4016 status.sstat_primary.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4017 memcpy(&status.sstat_primary.spinfo_address, &transport->ipaddr,
4018 transport->af_specific->sockaddr_len);
4019 /* Map ipv4 address into v4-mapped-on-v6 address. */
4020 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4021 (union sctp_addr *)&status.sstat_primary.spinfo_address);
4022 status.sstat_primary.spinfo_state = transport->state;
4023 status.sstat_primary.spinfo_cwnd = transport->cwnd;
4024 status.sstat_primary.spinfo_srtt = transport->srtt;
4025 status.sstat_primary.spinfo_rto = jiffies_to_msecs(transport->rto);
4026 status.sstat_primary.spinfo_mtu = transport->pathmtu;
4027
4028 if (status.sstat_primary.spinfo_state == SCTP_UNKNOWN)
4029 status.sstat_primary.spinfo_state = SCTP_ACTIVE;
4030
4031 if (put_user(len, optlen)) {
4032 retval = -EFAULT;
4033 goto out;
4034 }
4035
4036 SCTP_DEBUG_PRINTK("sctp_getsockopt_sctp_status(%d): %d %d %d\n",
4037 len, status.sstat_state, status.sstat_rwnd,
4038 status.sstat_assoc_id);
4039
4040 if (copy_to_user(optval, &status, len)) {
4041 retval = -EFAULT;
4042 goto out;
4043 }
4044
4045out:
4046 return retval;
4047}
4048
4049
4050/* 7.2.2 Peer Address Information (SCTP_GET_PEER_ADDR_INFO)
4051 *
4052 * Applications can retrieve information about a specific peer address
4053 * of an association, including its reachability state, congestion
4054 * window, and retransmission timer values. This information is
4055 * read-only.
4056 */
4057static int sctp_getsockopt_peer_addr_info(struct sock *sk, int len,
4058 char __user *optval,
4059 int __user *optlen)
4060{
4061 struct sctp_paddrinfo pinfo;
4062 struct sctp_transport *transport;
4063 int retval = 0;
4064
4065 if (len < sizeof(pinfo)) {
4066 retval = -EINVAL;
4067 goto out;
4068 }
4069
4070 len = sizeof(pinfo);
4071 if (copy_from_user(&pinfo, optval, len)) {
4072 retval = -EFAULT;
4073 goto out;
4074 }
4075
4076 transport = sctp_addr_id2transport(sk, &pinfo.spinfo_address,
4077 pinfo.spinfo_assoc_id);
4078 if (!transport)
4079 return -EINVAL;
4080
4081 pinfo.spinfo_assoc_id = sctp_assoc2id(transport->asoc);
4082 pinfo.spinfo_state = transport->state;
4083 pinfo.spinfo_cwnd = transport->cwnd;
4084 pinfo.spinfo_srtt = transport->srtt;
4085 pinfo.spinfo_rto = jiffies_to_msecs(transport->rto);
4086 pinfo.spinfo_mtu = transport->pathmtu;
4087
4088 if (pinfo.spinfo_state == SCTP_UNKNOWN)
4089 pinfo.spinfo_state = SCTP_ACTIVE;
4090
4091 if (put_user(len, optlen)) {
4092 retval = -EFAULT;
4093 goto out;
4094 }
4095
4096 if (copy_to_user(optval, &pinfo, len)) {
4097 retval = -EFAULT;
4098 goto out;
4099 }
4100
4101out:
4102 return retval;
4103}
4104
4105/* 7.1.12 Enable/Disable message fragmentation (SCTP_DISABLE_FRAGMENTS)
4106 *
4107 * This option is a on/off flag. If enabled no SCTP message
4108 * fragmentation will be performed. Instead if a message being sent
4109 * exceeds the current PMTU size, the message will NOT be sent and
4110 * instead a error will be indicated to the user.
4111 */
4112static int sctp_getsockopt_disable_fragments(struct sock *sk, int len,
4113 char __user *optval, int __user *optlen)
4114{
4115 int val;
4116
4117 if (len < sizeof(int))
4118 return -EINVAL;
4119
4120 len = sizeof(int);
4121 val = (sctp_sk(sk)->disable_fragments == 1);
4122 if (put_user(len, optlen))
4123 return -EFAULT;
4124 if (copy_to_user(optval, &val, len))
4125 return -EFAULT;
4126 return 0;
4127}
4128
4129/* 7.1.15 Set notification and ancillary events (SCTP_EVENTS)
4130 *
4131 * This socket option is used to specify various notifications and
4132 * ancillary data the user wishes to receive.
4133 */
4134static int sctp_getsockopt_events(struct sock *sk, int len, char __user *optval,
4135 int __user *optlen)
4136{
4137 if (len < sizeof(struct sctp_event_subscribe))
4138 return -EINVAL;
4139 len = sizeof(struct sctp_event_subscribe);
4140 if (put_user(len, optlen))
4141 return -EFAULT;
4142 if (copy_to_user(optval, &sctp_sk(sk)->subscribe, len))
4143 return -EFAULT;
4144 return 0;
4145}
4146
4147/* 7.1.8 Automatic Close of associations (SCTP_AUTOCLOSE)
4148 *
4149 * This socket option is applicable to the UDP-style socket only. When
4150 * set it will cause associations that are idle for more than the
4151 * specified number of seconds to automatically close. An association
4152 * being idle is defined an association that has NOT sent or received
4153 * user data. The special value of '0' indicates that no automatic
4154 * close of any associations should be performed. The option expects an
4155 * integer defining the number of seconds of idle time before an
4156 * association is closed.
4157 */
4158static int sctp_getsockopt_autoclose(struct sock *sk, int len, char __user *optval, int __user *optlen)
4159{
4160 /* Applicable to UDP-style socket only */
4161 if (sctp_style(sk, TCP))
4162 return -EOPNOTSUPP;
4163 if (len < sizeof(int))
4164 return -EINVAL;
4165 len = sizeof(int);
4166 if (put_user(len, optlen))
4167 return -EFAULT;
4168 if (copy_to_user(optval, &sctp_sk(sk)->autoclose, sizeof(int)))
4169 return -EFAULT;
4170 return 0;
4171}
4172
4173/* Helper routine to branch off an association to a new socket. */
4174SCTP_STATIC int sctp_do_peeloff(struct sctp_association *asoc,
4175 struct socket **sockp)
4176{
4177 struct sock *sk = asoc->base.sk;
4178 struct socket *sock;
4179 struct sctp_af *af;
4180 int err = 0;
4181
4182 /* An association cannot be branched off from an already peeled-off
4183 * socket, nor is this supported for tcp style sockets.
4184 */
4185 if (!sctp_style(sk, UDP))
4186 return -EINVAL;
4187
4188 /* Create a new socket. */
4189 err = sock_create(sk->sk_family, SOCK_SEQPACKET, IPPROTO_SCTP, &sock);
4190 if (err < 0)
4191 return err;
4192
4193 sctp_copy_sock(sock->sk, sk, asoc);
4194
4195 /* Make peeled-off sockets more like 1-1 accepted sockets.
4196 * Set the daddr and initialize id to something more random
4197 */
4198 af = sctp_get_af_specific(asoc->peer.primary_addr.sa.sa_family);
4199 af->to_sk_daddr(&asoc->peer.primary_addr, sk);
4200
4201 /* Populate the fields of the newsk from the oldsk and migrate the
4202 * asoc to the newsk.
4203 */
4204 sctp_sock_migrate(sk, sock->sk, asoc, SCTP_SOCKET_UDP_HIGH_BANDWIDTH);
4205
4206 *sockp = sock;
4207
4208 return err;
4209}
4210
4211static int sctp_getsockopt_peeloff(struct sock *sk, int len, char __user *optval, int __user *optlen)
4212{
4213 sctp_peeloff_arg_t peeloff;
4214 struct socket *newsock;
4215 int retval = 0;
4216 struct sctp_association *asoc;
4217
4218 if (len < sizeof(sctp_peeloff_arg_t))
4219 return -EINVAL;
4220 len = sizeof(sctp_peeloff_arg_t);
4221 if (copy_from_user(&peeloff, optval, len))
4222 return -EFAULT;
4223
4224 asoc = sctp_id2assoc(sk, peeloff.associd);
4225 if (!asoc) {
4226 retval = -EINVAL;
4227 goto out;
4228 }
4229
4230 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p\n", __func__, sk, asoc);
4231
4232 retval = sctp_do_peeloff(asoc, &newsock);
4233 if (retval < 0)
4234 goto out;
4235
4236 /* Map the socket to an unused fd that can be returned to the user. */
4237 retval = sock_map_fd(newsock, 0);
4238 if (retval < 0) {
4239 sock_release(newsock);
4240 goto out;
4241 }
4242
4243 SCTP_DEBUG_PRINTK("%s: sk: %p asoc: %p newsk: %p sd: %d\n",
4244 __func__, sk, asoc, newsock->sk, retval);
4245
4246 /* Return the fd mapped to the new socket. */
4247 peeloff.sd = retval;
4248 if (put_user(len, optlen))
4249 return -EFAULT;
4250 if (copy_to_user(optval, &peeloff, len))
4251 retval = -EFAULT;
4252
4253out:
4254 return retval;
4255}
4256
4257/* 7.1.13 Peer Address Parameters (SCTP_PEER_ADDR_PARAMS)
4258 *
4259 * Applications can enable or disable heartbeats for any peer address of
4260 * an association, modify an address's heartbeat interval, force a
4261 * heartbeat to be sent immediately, and adjust the address's maximum
4262 * number of retransmissions sent before an address is considered
4263 * unreachable. The following structure is used to access and modify an
4264 * address's parameters:
4265 *
4266 * struct sctp_paddrparams {
4267 * sctp_assoc_t spp_assoc_id;
4268 * struct sockaddr_storage spp_address;
4269 * uint32_t spp_hbinterval;
4270 * uint16_t spp_pathmaxrxt;
4271 * uint32_t spp_pathmtu;
4272 * uint32_t spp_sackdelay;
4273 * uint32_t spp_flags;
4274 * };
4275 *
4276 * spp_assoc_id - (one-to-many style socket) This is filled in the
4277 * application, and identifies the association for
4278 * this query.
4279 * spp_address - This specifies which address is of interest.
4280 * spp_hbinterval - This contains the value of the heartbeat interval,
4281 * in milliseconds. If a value of zero
4282 * is present in this field then no changes are to
4283 * be made to this parameter.
4284 * spp_pathmaxrxt - This contains the maximum number of
4285 * retransmissions before this address shall be
4286 * considered unreachable. If a value of zero
4287 * is present in this field then no changes are to
4288 * be made to this parameter.
4289 * spp_pathmtu - When Path MTU discovery is disabled the value
4290 * specified here will be the "fixed" path mtu.
4291 * Note that if the spp_address field is empty
4292 * then all associations on this address will
4293 * have this fixed path mtu set upon them.
4294 *
4295 * spp_sackdelay - When delayed sack is enabled, this value specifies
4296 * the number of milliseconds that sacks will be delayed
4297 * for. This value will apply to all addresses of an
4298 * association if the spp_address field is empty. Note
4299 * also, that if delayed sack is enabled and this
4300 * value is set to 0, no change is made to the last
4301 * recorded delayed sack timer value.
4302 *
4303 * spp_flags - These flags are used to control various features
4304 * on an association. The flag field may contain
4305 * zero or more of the following options.
4306 *
4307 * SPP_HB_ENABLE - Enable heartbeats on the
4308 * specified address. Note that if the address
4309 * field is empty all addresses for the association
4310 * have heartbeats enabled upon them.
4311 *
4312 * SPP_HB_DISABLE - Disable heartbeats on the
4313 * speicifed address. Note that if the address
4314 * field is empty all addresses for the association
4315 * will have their heartbeats disabled. Note also
4316 * that SPP_HB_ENABLE and SPP_HB_DISABLE are
4317 * mutually exclusive, only one of these two should
4318 * be specified. Enabling both fields will have
4319 * undetermined results.
4320 *
4321 * SPP_HB_DEMAND - Request a user initiated heartbeat
4322 * to be made immediately.
4323 *
4324 * SPP_PMTUD_ENABLE - This field will enable PMTU
4325 * discovery upon the specified address. Note that
4326 * if the address feild is empty then all addresses
4327 * on the association are effected.
4328 *
4329 * SPP_PMTUD_DISABLE - This field will disable PMTU
4330 * discovery upon the specified address. Note that
4331 * if the address feild is empty then all addresses
4332 * on the association are effected. Not also that
4333 * SPP_PMTUD_ENABLE and SPP_PMTUD_DISABLE are mutually
4334 * exclusive. Enabling both will have undetermined
4335 * results.
4336 *
4337 * SPP_SACKDELAY_ENABLE - Setting this flag turns
4338 * on delayed sack. The time specified in spp_sackdelay
4339 * is used to specify the sack delay for this address. Note
4340 * that if spp_address is empty then all addresses will
4341 * enable delayed sack and take on the sack delay
4342 * value specified in spp_sackdelay.
4343 * SPP_SACKDELAY_DISABLE - Setting this flag turns
4344 * off delayed sack. If the spp_address field is blank then
4345 * delayed sack is disabled for the entire association. Note
4346 * also that this field is mutually exclusive to
4347 * SPP_SACKDELAY_ENABLE, setting both will have undefined
4348 * results.
4349 */
4350static int sctp_getsockopt_peer_addr_params(struct sock *sk, int len,
4351 char __user *optval, int __user *optlen)
4352{
4353 struct sctp_paddrparams params;
4354 struct sctp_transport *trans = NULL;
4355 struct sctp_association *asoc = NULL;
4356 struct sctp_sock *sp = sctp_sk(sk);
4357
4358 if (len < sizeof(struct sctp_paddrparams))
4359 return -EINVAL;
4360 len = sizeof(struct sctp_paddrparams);
4361 if (copy_from_user(¶ms, optval, len))
4362 return -EFAULT;
4363
4364 /* If an address other than INADDR_ANY is specified, and
4365 * no transport is found, then the request is invalid.
4366 */
4367 if (!sctp_is_any(sk, ( union sctp_addr *)¶ms.spp_address)) {
4368 trans = sctp_addr_id2transport(sk, ¶ms.spp_address,
4369 params.spp_assoc_id);
4370 if (!trans) {
4371 SCTP_DEBUG_PRINTK("Failed no transport\n");
4372 return -EINVAL;
4373 }
4374 }
4375
4376 /* Get association, if assoc_id != 0 and the socket is a one
4377 * to many style socket, and an association was not found, then
4378 * the id was invalid.
4379 */
4380 asoc = sctp_id2assoc(sk, params.spp_assoc_id);
4381 if (!asoc && params.spp_assoc_id && sctp_style(sk, UDP)) {
4382 SCTP_DEBUG_PRINTK("Failed no association\n");
4383 return -EINVAL;
4384 }
4385
4386 if (trans) {
4387 /* Fetch transport values. */
4388 params.spp_hbinterval = jiffies_to_msecs(trans->hbinterval);
4389 params.spp_pathmtu = trans->pathmtu;
4390 params.spp_pathmaxrxt = trans->pathmaxrxt;
4391 params.spp_sackdelay = jiffies_to_msecs(trans->sackdelay);
4392
4393 /*draft-11 doesn't say what to return in spp_flags*/
4394 params.spp_flags = trans->param_flags;
4395 } else if (asoc) {
4396 /* Fetch association values. */
4397 params.spp_hbinterval = jiffies_to_msecs(asoc->hbinterval);
4398 params.spp_pathmtu = asoc->pathmtu;
4399 params.spp_pathmaxrxt = asoc->pathmaxrxt;
4400 params.spp_sackdelay = jiffies_to_msecs(asoc->sackdelay);
4401
4402 /*draft-11 doesn't say what to return in spp_flags*/
4403 params.spp_flags = asoc->param_flags;
4404 } else {
4405 /* Fetch socket values. */
4406 params.spp_hbinterval = sp->hbinterval;
4407 params.spp_pathmtu = sp->pathmtu;
4408 params.spp_sackdelay = sp->sackdelay;
4409 params.spp_pathmaxrxt = sp->pathmaxrxt;
4410
4411 /*draft-11 doesn't say what to return in spp_flags*/
4412 params.spp_flags = sp->param_flags;
4413 }
4414
4415 if (copy_to_user(optval, ¶ms, len))
4416 return -EFAULT;
4417
4418 if (put_user(len, optlen))
4419 return -EFAULT;
4420
4421 return 0;
4422}
4423
4424/*
4425 * 7.1.23. Get or set delayed ack timer (SCTP_DELAYED_SACK)
4426 *
4427 * This option will effect the way delayed acks are performed. This
4428 * option allows you to get or set the delayed ack time, in
4429 * milliseconds. It also allows changing the delayed ack frequency.
4430 * Changing the frequency to 1 disables the delayed sack algorithm. If
4431 * the assoc_id is 0, then this sets or gets the endpoints default
4432 * values. If the assoc_id field is non-zero, then the set or get
4433 * effects the specified association for the one to many model (the
4434 * assoc_id field is ignored by the one to one model). Note that if
4435 * sack_delay or sack_freq are 0 when setting this option, then the
4436 * current values will remain unchanged.
4437 *
4438 * struct sctp_sack_info {
4439 * sctp_assoc_t sack_assoc_id;
4440 * uint32_t sack_delay;
4441 * uint32_t sack_freq;
4442 * };
4443 *
4444 * sack_assoc_id - This parameter, indicates which association the user
4445 * is performing an action upon. Note that if this field's value is
4446 * zero then the endpoints default value is changed (effecting future
4447 * associations only).
4448 *
4449 * sack_delay - This parameter contains the number of milliseconds that
4450 * the user is requesting the delayed ACK timer be set to. Note that
4451 * this value is defined in the standard to be between 200 and 500
4452 * milliseconds.
4453 *
4454 * sack_freq - This parameter contains the number of packets that must
4455 * be received before a sack is sent without waiting for the delay
4456 * timer to expire. The default value for this is 2, setting this
4457 * value to 1 will disable the delayed sack algorithm.
4458 */
4459static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
4460 char __user *optval,
4461 int __user *optlen)
4462{
4463 struct sctp_sack_info params;
4464 struct sctp_association *asoc = NULL;
4465 struct sctp_sock *sp = sctp_sk(sk);
4466
4467 if (len >= sizeof(struct sctp_sack_info)) {
4468 len = sizeof(struct sctp_sack_info);
4469
4470 if (copy_from_user(¶ms, optval, len))
4471 return -EFAULT;
4472 } else if (len == sizeof(struct sctp_assoc_value)) {
4473 pr_warn("Use of struct sctp_assoc_value in delayed_ack socket option deprecated\n");
4474 pr_warn("Use struct sctp_sack_info instead\n");
4475 if (copy_from_user(¶ms, optval, len))
4476 return -EFAULT;
4477 } else
4478 return - EINVAL;
4479
4480 /* Get association, if sack_assoc_id != 0 and the socket is a one
4481 * to many style socket, and an association was not found, then
4482 * the id was invalid.
4483 */
4484 asoc = sctp_id2assoc(sk, params.sack_assoc_id);
4485 if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
4486 return -EINVAL;
4487
4488 if (asoc) {
4489 /* Fetch association values. */
4490 if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
4491 params.sack_delay = jiffies_to_msecs(
4492 asoc->sackdelay);
4493 params.sack_freq = asoc->sackfreq;
4494
4495 } else {
4496 params.sack_delay = 0;
4497 params.sack_freq = 1;
4498 }
4499 } else {
4500 /* Fetch socket values. */
4501 if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
4502 params.sack_delay = sp->sackdelay;
4503 params.sack_freq = sp->sackfreq;
4504 } else {
4505 params.sack_delay = 0;
4506 params.sack_freq = 1;
4507 }
4508 }
4509
4510 if (copy_to_user(optval, ¶ms, len))
4511 return -EFAULT;
4512
4513 if (put_user(len, optlen))
4514 return -EFAULT;
4515
4516 return 0;
4517}
4518
4519/* 7.1.3 Initialization Parameters (SCTP_INITMSG)
4520 *
4521 * Applications can specify protocol parameters for the default association
4522 * initialization. The option name argument to setsockopt() and getsockopt()
4523 * is SCTP_INITMSG.
4524 *
4525 * Setting initialization parameters is effective only on an unconnected
4526 * socket (for UDP-style sockets only future associations are effected
4527 * by the change). With TCP-style sockets, this option is inherited by
4528 * sockets derived from a listener socket.
4529 */
4530static int sctp_getsockopt_initmsg(struct sock *sk, int len, char __user *optval, int __user *optlen)
4531{
4532 if (len < sizeof(struct sctp_initmsg))
4533 return -EINVAL;
4534 len = sizeof(struct sctp_initmsg);
4535 if (put_user(len, optlen))
4536 return -EFAULT;
4537 if (copy_to_user(optval, &sctp_sk(sk)->initmsg, len))
4538 return -EFAULT;
4539 return 0;
4540}
4541
4542
4543static int sctp_getsockopt_peer_addrs(struct sock *sk, int len,
4544 char __user *optval, int __user *optlen)
4545{
4546 struct sctp_association *asoc;
4547 int cnt = 0;
4548 struct sctp_getaddrs getaddrs;
4549 struct sctp_transport *from;
4550 void __user *to;
4551 union sctp_addr temp;
4552 struct sctp_sock *sp = sctp_sk(sk);
4553 int addrlen;
4554 size_t space_left;
4555 int bytes_copied;
4556
4557 if (len < sizeof(struct sctp_getaddrs))
4558 return -EINVAL;
4559
4560 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4561 return -EFAULT;
4562
4563 /* For UDP-style sockets, id specifies the association to query. */
4564 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4565 if (!asoc)
4566 return -EINVAL;
4567
4568 to = optval + offsetof(struct sctp_getaddrs,addrs);
4569 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4570
4571 list_for_each_entry(from, &asoc->peer.transport_addr_list,
4572 transports) {
4573 memcpy(&temp, &from->ipaddr, sizeof(temp));
4574 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4575 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4576 if (space_left < addrlen)
4577 return -ENOMEM;
4578 if (copy_to_user(to, &temp, addrlen))
4579 return -EFAULT;
4580 to += addrlen;
4581 cnt++;
4582 space_left -= addrlen;
4583 }
4584
4585 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num))
4586 return -EFAULT;
4587 bytes_copied = ((char __user *)to) - optval;
4588 if (put_user(bytes_copied, optlen))
4589 return -EFAULT;
4590
4591 return 0;
4592}
4593
4594static int sctp_copy_laddrs(struct sock *sk, __u16 port, void *to,
4595 size_t space_left, int *bytes_copied)
4596{
4597 struct sctp_sockaddr_entry *addr;
4598 union sctp_addr temp;
4599 int cnt = 0;
4600 int addrlen;
4601
4602 rcu_read_lock();
4603 list_for_each_entry_rcu(addr, &sctp_local_addr_list, list) {
4604 if (!addr->valid)
4605 continue;
4606
4607 if ((PF_INET == sk->sk_family) &&
4608 (AF_INET6 == addr->a.sa.sa_family))
4609 continue;
4610 if ((PF_INET6 == sk->sk_family) &&
4611 inet_v6_ipv6only(sk) &&
4612 (AF_INET == addr->a.sa.sa_family))
4613 continue;
4614 memcpy(&temp, &addr->a, sizeof(temp));
4615 if (!temp.v4.sin_port)
4616 temp.v4.sin_port = htons(port);
4617
4618 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sctp_sk(sk),
4619 &temp);
4620 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4621 if (space_left < addrlen) {
4622 cnt = -ENOMEM;
4623 break;
4624 }
4625 memcpy(to, &temp, addrlen);
4626
4627 to += addrlen;
4628 cnt ++;
4629 space_left -= addrlen;
4630 *bytes_copied += addrlen;
4631 }
4632 rcu_read_unlock();
4633
4634 return cnt;
4635}
4636
4637
4638static int sctp_getsockopt_local_addrs(struct sock *sk, int len,
4639 char __user *optval, int __user *optlen)
4640{
4641 struct sctp_bind_addr *bp;
4642 struct sctp_association *asoc;
4643 int cnt = 0;
4644 struct sctp_getaddrs getaddrs;
4645 struct sctp_sockaddr_entry *addr;
4646 void __user *to;
4647 union sctp_addr temp;
4648 struct sctp_sock *sp = sctp_sk(sk);
4649 int addrlen;
4650 int err = 0;
4651 size_t space_left;
4652 int bytes_copied = 0;
4653 void *addrs;
4654 void *buf;
4655
4656 if (len < sizeof(struct sctp_getaddrs))
4657 return -EINVAL;
4658
4659 if (copy_from_user(&getaddrs, optval, sizeof(struct sctp_getaddrs)))
4660 return -EFAULT;
4661
4662 /*
4663 * For UDP-style sockets, id specifies the association to query.
4664 * If the id field is set to the value '0' then the locally bound
4665 * addresses are returned without regard to any particular
4666 * association.
4667 */
4668 if (0 == getaddrs.assoc_id) {
4669 bp = &sctp_sk(sk)->ep->base.bind_addr;
4670 } else {
4671 asoc = sctp_id2assoc(sk, getaddrs.assoc_id);
4672 if (!asoc)
4673 return -EINVAL;
4674 bp = &asoc->base.bind_addr;
4675 }
4676
4677 to = optval + offsetof(struct sctp_getaddrs,addrs);
4678 space_left = len - offsetof(struct sctp_getaddrs,addrs);
4679
4680 addrs = kmalloc(space_left, GFP_KERNEL);
4681 if (!addrs)
4682 return -ENOMEM;
4683
4684 /* If the endpoint is bound to 0.0.0.0 or ::0, get the valid
4685 * addresses from the global local address list.
4686 */
4687 if (sctp_list_single_entry(&bp->address_list)) {
4688 addr = list_entry(bp->address_list.next,
4689 struct sctp_sockaddr_entry, list);
4690 if (sctp_is_any(sk, &addr->a)) {
4691 cnt = sctp_copy_laddrs(sk, bp->port, addrs,
4692 space_left, &bytes_copied);
4693 if (cnt < 0) {
4694 err = cnt;
4695 goto out;
4696 }
4697 goto copy_getaddrs;
4698 }
4699 }
4700
4701 buf = addrs;
4702 /* Protection on the bound address list is not needed since
4703 * in the socket option context we hold a socket lock and
4704 * thus the bound address list can't change.
4705 */
4706 list_for_each_entry(addr, &bp->address_list, list) {
4707 memcpy(&temp, &addr->a, sizeof(temp));
4708 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp, &temp);
4709 addrlen = sctp_get_af_specific(temp.sa.sa_family)->sockaddr_len;
4710 if (space_left < addrlen) {
4711 err = -ENOMEM; /*fixme: right error?*/
4712 goto out;
4713 }
4714 memcpy(buf, &temp, addrlen);
4715 buf += addrlen;
4716 bytes_copied += addrlen;
4717 cnt ++;
4718 space_left -= addrlen;
4719 }
4720
4721copy_getaddrs:
4722 if (copy_to_user(to, addrs, bytes_copied)) {
4723 err = -EFAULT;
4724 goto out;
4725 }
4726 if (put_user(cnt, &((struct sctp_getaddrs __user *)optval)->addr_num)) {
4727 err = -EFAULT;
4728 goto out;
4729 }
4730 if (put_user(bytes_copied, optlen))
4731 err = -EFAULT;
4732out:
4733 kfree(addrs);
4734 return err;
4735}
4736
4737/* 7.1.10 Set Primary Address (SCTP_PRIMARY_ADDR)
4738 *
4739 * Requests that the local SCTP stack use the enclosed peer address as
4740 * the association primary. The enclosed address must be one of the
4741 * association peer's addresses.
4742 */
4743static int sctp_getsockopt_primary_addr(struct sock *sk, int len,
4744 char __user *optval, int __user *optlen)
4745{
4746 struct sctp_prim prim;
4747 struct sctp_association *asoc;
4748 struct sctp_sock *sp = sctp_sk(sk);
4749
4750 if (len < sizeof(struct sctp_prim))
4751 return -EINVAL;
4752
4753 len = sizeof(struct sctp_prim);
4754
4755 if (copy_from_user(&prim, optval, len))
4756 return -EFAULT;
4757
4758 asoc = sctp_id2assoc(sk, prim.ssp_assoc_id);
4759 if (!asoc)
4760 return -EINVAL;
4761
4762 if (!asoc->peer.primary_path)
4763 return -ENOTCONN;
4764
4765 memcpy(&prim.ssp_addr, &asoc->peer.primary_path->ipaddr,
4766 asoc->peer.primary_path->af_specific->sockaddr_len);
4767
4768 sctp_get_pf_specific(sk->sk_family)->addr_v4map(sp,
4769 (union sctp_addr *)&prim.ssp_addr);
4770
4771 if (put_user(len, optlen))
4772 return -EFAULT;
4773 if (copy_to_user(optval, &prim, len))
4774 return -EFAULT;
4775
4776 return 0;
4777}
4778
4779/*
4780 * 7.1.11 Set Adaptation Layer Indicator (SCTP_ADAPTATION_LAYER)
4781 *
4782 * Requests that the local endpoint set the specified Adaptation Layer
4783 * Indication parameter for all future INIT and INIT-ACK exchanges.
4784 */
4785static int sctp_getsockopt_adaptation_layer(struct sock *sk, int len,
4786 char __user *optval, int __user *optlen)
4787{
4788 struct sctp_setadaptation adaptation;
4789
4790 if (len < sizeof(struct sctp_setadaptation))
4791 return -EINVAL;
4792
4793 len = sizeof(struct sctp_setadaptation);
4794
4795 adaptation.ssb_adaptation_ind = sctp_sk(sk)->adaptation_ind;
4796
4797 if (put_user(len, optlen))
4798 return -EFAULT;
4799 if (copy_to_user(optval, &adaptation, len))
4800 return -EFAULT;
4801
4802 return 0;
4803}
4804
4805/*
4806 *
4807 * 7.1.14 Set default send parameters (SCTP_DEFAULT_SEND_PARAM)
4808 *
4809 * Applications that wish to use the sendto() system call may wish to
4810 * specify a default set of parameters that would normally be supplied
4811 * through the inclusion of ancillary data. This socket option allows
4812 * such an application to set the default sctp_sndrcvinfo structure.
4813
4814
4815 * The application that wishes to use this socket option simply passes
4816 * in to this call the sctp_sndrcvinfo structure defined in Section
4817 * 5.2.2) The input parameters accepted by this call include
4818 * sinfo_stream, sinfo_flags, sinfo_ppid, sinfo_context,
4819 * sinfo_timetolive. The user must provide the sinfo_assoc_id field in
4820 * to this call if the caller is using the UDP model.
4821 *
4822 * For getsockopt, it get the default sctp_sndrcvinfo structure.
4823 */
4824static int sctp_getsockopt_default_send_param(struct sock *sk,
4825 int len, char __user *optval,
4826 int __user *optlen)
4827{
4828 struct sctp_sndrcvinfo info;
4829 struct sctp_association *asoc;
4830 struct sctp_sock *sp = sctp_sk(sk);
4831
4832 if (len < sizeof(struct sctp_sndrcvinfo))
4833 return -EINVAL;
4834
4835 len = sizeof(struct sctp_sndrcvinfo);
4836
4837 if (copy_from_user(&info, optval, len))
4838 return -EFAULT;
4839
4840 asoc = sctp_id2assoc(sk, info.sinfo_assoc_id);
4841 if (!asoc && info.sinfo_assoc_id && sctp_style(sk, UDP))
4842 return -EINVAL;
4843
4844 if (asoc) {
4845 info.sinfo_stream = asoc->default_stream;
4846 info.sinfo_flags = asoc->default_flags;
4847 info.sinfo_ppid = asoc->default_ppid;
4848 info.sinfo_context = asoc->default_context;
4849 info.sinfo_timetolive = asoc->default_timetolive;
4850 } else {
4851 info.sinfo_stream = sp->default_stream;
4852 info.sinfo_flags = sp->default_flags;
4853 info.sinfo_ppid = sp->default_ppid;
4854 info.sinfo_context = sp->default_context;
4855 info.sinfo_timetolive = sp->default_timetolive;
4856 }
4857
4858 if (put_user(len, optlen))
4859 return -EFAULT;
4860 if (copy_to_user(optval, &info, len))
4861 return -EFAULT;
4862
4863 return 0;
4864}
4865
4866/*
4867 *
4868 * 7.1.5 SCTP_NODELAY
4869 *
4870 * Turn on/off any Nagle-like algorithm. This means that packets are
4871 * generally sent as soon as possible and no unnecessary delays are
4872 * introduced, at the cost of more packets in the network. Expects an
4873 * integer boolean flag.
4874 */
4875
4876static int sctp_getsockopt_nodelay(struct sock *sk, int len,
4877 char __user *optval, int __user *optlen)
4878{
4879 int val;
4880
4881 if (len < sizeof(int))
4882 return -EINVAL;
4883
4884 len = sizeof(int);
4885 val = (sctp_sk(sk)->nodelay == 1);
4886 if (put_user(len, optlen))
4887 return -EFAULT;
4888 if (copy_to_user(optval, &val, len))
4889 return -EFAULT;
4890 return 0;
4891}
4892
4893/*
4894 *
4895 * 7.1.1 SCTP_RTOINFO
4896 *
4897 * The protocol parameters used to initialize and bound retransmission
4898 * timeout (RTO) are tunable. sctp_rtoinfo structure is used to access
4899 * and modify these parameters.
4900 * All parameters are time values, in milliseconds. A value of 0, when
4901 * modifying the parameters, indicates that the current value should not
4902 * be changed.
4903 *
4904 */
4905static int sctp_getsockopt_rtoinfo(struct sock *sk, int len,
4906 char __user *optval,
4907 int __user *optlen) {
4908 struct sctp_rtoinfo rtoinfo;
4909 struct sctp_association *asoc;
4910
4911 if (len < sizeof (struct sctp_rtoinfo))
4912 return -EINVAL;
4913
4914 len = sizeof(struct sctp_rtoinfo);
4915
4916 if (copy_from_user(&rtoinfo, optval, len))
4917 return -EFAULT;
4918
4919 asoc = sctp_id2assoc(sk, rtoinfo.srto_assoc_id);
4920
4921 if (!asoc && rtoinfo.srto_assoc_id && sctp_style(sk, UDP))
4922 return -EINVAL;
4923
4924 /* Values corresponding to the specific association. */
4925 if (asoc) {
4926 rtoinfo.srto_initial = jiffies_to_msecs(asoc->rto_initial);
4927 rtoinfo.srto_max = jiffies_to_msecs(asoc->rto_max);
4928 rtoinfo.srto_min = jiffies_to_msecs(asoc->rto_min);
4929 } else {
4930 /* Values corresponding to the endpoint. */
4931 struct sctp_sock *sp = sctp_sk(sk);
4932
4933 rtoinfo.srto_initial = sp->rtoinfo.srto_initial;
4934 rtoinfo.srto_max = sp->rtoinfo.srto_max;
4935 rtoinfo.srto_min = sp->rtoinfo.srto_min;
4936 }
4937
4938 if (put_user(len, optlen))
4939 return -EFAULT;
4940
4941 if (copy_to_user(optval, &rtoinfo, len))
4942 return -EFAULT;
4943
4944 return 0;
4945}
4946
4947/*
4948 *
4949 * 7.1.2 SCTP_ASSOCINFO
4950 *
4951 * This option is used to tune the maximum retransmission attempts
4952 * of the association.
4953 * Returns an error if the new association retransmission value is
4954 * greater than the sum of the retransmission value of the peer.
4955 * See [SCTP] for more information.
4956 *
4957 */
4958static int sctp_getsockopt_associnfo(struct sock *sk, int len,
4959 char __user *optval,
4960 int __user *optlen)
4961{
4962
4963 struct sctp_assocparams assocparams;
4964 struct sctp_association *asoc;
4965 struct list_head *pos;
4966 int cnt = 0;
4967
4968 if (len < sizeof (struct sctp_assocparams))
4969 return -EINVAL;
4970
4971 len = sizeof(struct sctp_assocparams);
4972
4973 if (copy_from_user(&assocparams, optval, len))
4974 return -EFAULT;
4975
4976 asoc = sctp_id2assoc(sk, assocparams.sasoc_assoc_id);
4977
4978 if (!asoc && assocparams.sasoc_assoc_id && sctp_style(sk, UDP))
4979 return -EINVAL;
4980
4981 /* Values correspoinding to the specific association */
4982 if (asoc) {
4983 assocparams.sasoc_asocmaxrxt = asoc->max_retrans;
4984 assocparams.sasoc_peer_rwnd = asoc->peer.rwnd;
4985 assocparams.sasoc_local_rwnd = asoc->a_rwnd;
4986 assocparams.sasoc_cookie_life = (asoc->cookie_life.tv_sec
4987 * 1000) +
4988 (asoc->cookie_life.tv_usec
4989 / 1000);
4990
4991 list_for_each(pos, &asoc->peer.transport_addr_list) {
4992 cnt ++;
4993 }
4994
4995 assocparams.sasoc_number_peer_destinations = cnt;
4996 } else {
4997 /* Values corresponding to the endpoint */
4998 struct sctp_sock *sp = sctp_sk(sk);
4999
5000 assocparams.sasoc_asocmaxrxt = sp->assocparams.sasoc_asocmaxrxt;
5001 assocparams.sasoc_peer_rwnd = sp->assocparams.sasoc_peer_rwnd;
5002 assocparams.sasoc_local_rwnd = sp->assocparams.sasoc_local_rwnd;
5003 assocparams.sasoc_cookie_life =
5004 sp->assocparams.sasoc_cookie_life;
5005 assocparams.sasoc_number_peer_destinations =
5006 sp->assocparams.
5007 sasoc_number_peer_destinations;
5008 }
5009
5010 if (put_user(len, optlen))
5011 return -EFAULT;
5012
5013 if (copy_to_user(optval, &assocparams, len))
5014 return -EFAULT;
5015
5016 return 0;
5017}
5018
5019/*
5020 * 7.1.16 Set/clear IPv4 mapped addresses (SCTP_I_WANT_MAPPED_V4_ADDR)
5021 *
5022 * This socket option is a boolean flag which turns on or off mapped V4
5023 * addresses. If this option is turned on and the socket is type
5024 * PF_INET6, then IPv4 addresses will be mapped to V6 representation.
5025 * If this option is turned off, then no mapping will be done of V4
5026 * addresses and a user will receive both PF_INET6 and PF_INET type
5027 * addresses on the socket.
5028 */
5029static int sctp_getsockopt_mappedv4(struct sock *sk, int len,
5030 char __user *optval, int __user *optlen)
5031{
5032 int val;
5033 struct sctp_sock *sp = sctp_sk(sk);
5034
5035 if (len < sizeof(int))
5036 return -EINVAL;
5037
5038 len = sizeof(int);
5039 val = sp->v4mapped;
5040 if (put_user(len, optlen))
5041 return -EFAULT;
5042 if (copy_to_user(optval, &val, len))
5043 return -EFAULT;
5044
5045 return 0;
5046}
5047
5048/*
5049 * 7.1.29. Set or Get the default context (SCTP_CONTEXT)
5050 * (chapter and verse is quoted at sctp_setsockopt_context())
5051 */
5052static int sctp_getsockopt_context(struct sock *sk, int len,
5053 char __user *optval, int __user *optlen)
5054{
5055 struct sctp_assoc_value params;
5056 struct sctp_sock *sp;
5057 struct sctp_association *asoc;
5058
5059 if (len < sizeof(struct sctp_assoc_value))
5060 return -EINVAL;
5061
5062 len = sizeof(struct sctp_assoc_value);
5063
5064 if (copy_from_user(¶ms, optval, len))
5065 return -EFAULT;
5066
5067 sp = sctp_sk(sk);
5068
5069 if (params.assoc_id != 0) {
5070 asoc = sctp_id2assoc(sk, params.assoc_id);
5071 if (!asoc)
5072 return -EINVAL;
5073 params.assoc_value = asoc->default_rcv_context;
5074 } else {
5075 params.assoc_value = sp->default_rcv_context;
5076 }
5077
5078 if (put_user(len, optlen))
5079 return -EFAULT;
5080 if (copy_to_user(optval, ¶ms, len))
5081 return -EFAULT;
5082
5083 return 0;
5084}
5085
5086/*
5087 * 8.1.16. Get or Set the Maximum Fragmentation Size (SCTP_MAXSEG)
5088 * This option will get or set the maximum size to put in any outgoing
5089 * SCTP DATA chunk. If a message is larger than this size it will be
5090 * fragmented by SCTP into the specified size. Note that the underlying
5091 * SCTP implementation may fragment into smaller sized chunks when the
5092 * PMTU of the underlying association is smaller than the value set by
5093 * the user. The default value for this option is '0' which indicates
5094 * the user is NOT limiting fragmentation and only the PMTU will effect
5095 * SCTP's choice of DATA chunk size. Note also that values set larger
5096 * than the maximum size of an IP datagram will effectively let SCTP
5097 * control fragmentation (i.e. the same as setting this option to 0).
5098 *
5099 * The following structure is used to access and modify this parameter:
5100 *
5101 * struct sctp_assoc_value {
5102 * sctp_assoc_t assoc_id;
5103 * uint32_t assoc_value;
5104 * };
5105 *
5106 * assoc_id: This parameter is ignored for one-to-one style sockets.
5107 * For one-to-many style sockets this parameter indicates which
5108 * association the user is performing an action upon. Note that if
5109 * this field's value is zero then the endpoints default value is
5110 * changed (effecting future associations only).
5111 * assoc_value: This parameter specifies the maximum size in bytes.
5112 */
5113static int sctp_getsockopt_maxseg(struct sock *sk, int len,
5114 char __user *optval, int __user *optlen)
5115{
5116 struct sctp_assoc_value params;
5117 struct sctp_association *asoc;
5118
5119 if (len == sizeof(int)) {
5120 pr_warn("Use of int in maxseg socket option deprecated\n");
5121 pr_warn("Use struct sctp_assoc_value instead\n");
5122 params.assoc_id = 0;
5123 } else if (len >= sizeof(struct sctp_assoc_value)) {
5124 len = sizeof(struct sctp_assoc_value);
5125 if (copy_from_user(¶ms, optval, sizeof(params)))
5126 return -EFAULT;
5127 } else
5128 return -EINVAL;
5129
5130 asoc = sctp_id2assoc(sk, params.assoc_id);
5131 if (!asoc && params.assoc_id && sctp_style(sk, UDP))
5132 return -EINVAL;
5133
5134 if (asoc)
5135 params.assoc_value = asoc->frag_point;
5136 else
5137 params.assoc_value = sctp_sk(sk)->user_frag;
5138
5139 if (put_user(len, optlen))
5140 return -EFAULT;
5141 if (len == sizeof(int)) {
5142 if (copy_to_user(optval, ¶ms.assoc_value, len))
5143 return -EFAULT;
5144 } else {
5145 if (copy_to_user(optval, ¶ms, len))
5146 return -EFAULT;
5147 }
5148
5149 return 0;
5150}
5151
5152/*
5153 * 7.1.24. Get or set fragmented interleave (SCTP_FRAGMENT_INTERLEAVE)
5154 * (chapter and verse is quoted at sctp_setsockopt_fragment_interleave())
5155 */
5156static int sctp_getsockopt_fragment_interleave(struct sock *sk, int len,
5157 char __user *optval, int __user *optlen)
5158{
5159 int val;
5160
5161 if (len < sizeof(int))
5162 return -EINVAL;
5163
5164 len = sizeof(int);
5165
5166 val = sctp_sk(sk)->frag_interleave;
5167 if (put_user(len, optlen))
5168 return -EFAULT;
5169 if (copy_to_user(optval, &val, len))
5170 return -EFAULT;
5171
5172 return 0;
5173}
5174
5175/*
5176 * 7.1.25. Set or Get the sctp partial delivery point
5177 * (chapter and verse is quoted at sctp_setsockopt_partial_delivery_point())
5178 */
5179static int sctp_getsockopt_partial_delivery_point(struct sock *sk, int len,
5180 char __user *optval,
5181 int __user *optlen)
5182{
5183 u32 val;
5184
5185 if (len < sizeof(u32))
5186 return -EINVAL;
5187
5188 len = sizeof(u32);
5189
5190 val = sctp_sk(sk)->pd_point;
5191 if (put_user(len, optlen))
5192 return -EFAULT;
5193 if (copy_to_user(optval, &val, len))
5194 return -EFAULT;
5195
5196 return 0;
5197}
5198
5199/*
5200 * 7.1.28. Set or Get the maximum burst (SCTP_MAX_BURST)
5201 * (chapter and verse is quoted at sctp_setsockopt_maxburst())
5202 */
5203static int sctp_getsockopt_maxburst(struct sock *sk, int len,
5204 char __user *optval,
5205 int __user *optlen)
5206{
5207 struct sctp_assoc_value params;
5208 struct sctp_sock *sp;
5209 struct sctp_association *asoc;
5210
5211 if (len == sizeof(int)) {
5212 pr_warn("Use of int in max_burst socket option deprecated\n");
5213 pr_warn("Use struct sctp_assoc_value instead\n");
5214 params.assoc_id = 0;
5215 } else if (len >= sizeof(struct sctp_assoc_value)) {
5216 len = sizeof(struct sctp_assoc_value);
5217 if (copy_from_user(¶ms, optval, len))
5218 return -EFAULT;
5219 } else
5220 return -EINVAL;
5221
5222 sp = sctp_sk(sk);
5223
5224 if (params.assoc_id != 0) {
5225 asoc = sctp_id2assoc(sk, params.assoc_id);
5226 if (!asoc)
5227 return -EINVAL;
5228 params.assoc_value = asoc->max_burst;
5229 } else
5230 params.assoc_value = sp->max_burst;
5231
5232 if (len == sizeof(int)) {
5233 if (copy_to_user(optval, ¶ms.assoc_value, len))
5234 return -EFAULT;
5235 } else {
5236 if (copy_to_user(optval, ¶ms, len))
5237 return -EFAULT;
5238 }
5239
5240 return 0;
5241
5242}
5243
5244static int sctp_getsockopt_hmac_ident(struct sock *sk, int len,
5245 char __user *optval, int __user *optlen)
5246{
5247 struct sctp_hmacalgo __user *p = (void __user *)optval;
5248 struct sctp_hmac_algo_param *hmacs;
5249 __u16 data_len = 0;
5250 u32 num_idents;
5251
5252 if (!sctp_auth_enable)
5253 return -EACCES;
5254
5255 hmacs = sctp_sk(sk)->ep->auth_hmacs_list;
5256 data_len = ntohs(hmacs->param_hdr.length) - sizeof(sctp_paramhdr_t);
5257
5258 if (len < sizeof(struct sctp_hmacalgo) + data_len)
5259 return -EINVAL;
5260
5261 len = sizeof(struct sctp_hmacalgo) + data_len;
5262 num_idents = data_len / sizeof(u16);
5263
5264 if (put_user(len, optlen))
5265 return -EFAULT;
5266 if (put_user(num_idents, &p->shmac_num_idents))
5267 return -EFAULT;
5268 if (copy_to_user(p->shmac_idents, hmacs->hmac_ids, data_len))
5269 return -EFAULT;
5270 return 0;
5271}
5272
5273static int sctp_getsockopt_active_key(struct sock *sk, int len,
5274 char __user *optval, int __user *optlen)
5275{
5276 struct sctp_authkeyid val;
5277 struct sctp_association *asoc;
5278
5279 if (!sctp_auth_enable)
5280 return -EACCES;
5281
5282 if (len < sizeof(struct sctp_authkeyid))
5283 return -EINVAL;
5284 if (copy_from_user(&val, optval, sizeof(struct sctp_authkeyid)))
5285 return -EFAULT;
5286
5287 asoc = sctp_id2assoc(sk, val.scact_assoc_id);
5288 if (!asoc && val.scact_assoc_id && sctp_style(sk, UDP))
5289 return -EINVAL;
5290
5291 if (asoc)
5292 val.scact_keynumber = asoc->active_key_id;
5293 else
5294 val.scact_keynumber = sctp_sk(sk)->ep->active_key_id;
5295
5296 len = sizeof(struct sctp_authkeyid);
5297 if (put_user(len, optlen))
5298 return -EFAULT;
5299 if (copy_to_user(optval, &val, len))
5300 return -EFAULT;
5301
5302 return 0;
5303}
5304
5305static int sctp_getsockopt_peer_auth_chunks(struct sock *sk, int len,
5306 char __user *optval, int __user *optlen)
5307{
5308 struct sctp_authchunks __user *p = (void __user *)optval;
5309 struct sctp_authchunks val;
5310 struct sctp_association *asoc;
5311 struct sctp_chunks_param *ch;
5312 u32 num_chunks = 0;
5313 char __user *to;
5314
5315 if (!sctp_auth_enable)
5316 return -EACCES;
5317
5318 if (len < sizeof(struct sctp_authchunks))
5319 return -EINVAL;
5320
5321 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5322 return -EFAULT;
5323
5324 to = p->gauth_chunks;
5325 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5326 if (!asoc)
5327 return -EINVAL;
5328
5329 ch = asoc->peer.peer_chunks;
5330 if (!ch)
5331 goto num;
5332
5333 /* See if the user provided enough room for all the data */
5334 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5335 if (len < num_chunks)
5336 return -EINVAL;
5337
5338 if (copy_to_user(to, ch->chunks, num_chunks))
5339 return -EFAULT;
5340num:
5341 len = sizeof(struct sctp_authchunks) + num_chunks;
5342 if (put_user(len, optlen)) return -EFAULT;
5343 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5344 return -EFAULT;
5345 return 0;
5346}
5347
5348static int sctp_getsockopt_local_auth_chunks(struct sock *sk, int len,
5349 char __user *optval, int __user *optlen)
5350{
5351 struct sctp_authchunks __user *p = (void __user *)optval;
5352 struct sctp_authchunks val;
5353 struct sctp_association *asoc;
5354 struct sctp_chunks_param *ch;
5355 u32 num_chunks = 0;
5356 char __user *to;
5357
5358 if (!sctp_auth_enable)
5359 return -EACCES;
5360
5361 if (len < sizeof(struct sctp_authchunks))
5362 return -EINVAL;
5363
5364 if (copy_from_user(&val, optval, sizeof(struct sctp_authchunks)))
5365 return -EFAULT;
5366
5367 to = p->gauth_chunks;
5368 asoc = sctp_id2assoc(sk, val.gauth_assoc_id);
5369 if (!asoc && val.gauth_assoc_id && sctp_style(sk, UDP))
5370 return -EINVAL;
5371
5372 if (asoc)
5373 ch = (struct sctp_chunks_param*)asoc->c.auth_chunks;
5374 else
5375 ch = sctp_sk(sk)->ep->auth_chunk_list;
5376
5377 if (!ch)
5378 goto num;
5379
5380 num_chunks = ntohs(ch->param_hdr.length) - sizeof(sctp_paramhdr_t);
5381 if (len < sizeof(struct sctp_authchunks) + num_chunks)
5382 return -EINVAL;
5383
5384 if (copy_to_user(to, ch->chunks, num_chunks))
5385 return -EFAULT;
5386num:
5387 len = sizeof(struct sctp_authchunks) + num_chunks;
5388 if (put_user(len, optlen))
5389 return -EFAULT;
5390 if (put_user(num_chunks, &p->gauth_number_of_chunks))
5391 return -EFAULT;
5392
5393 return 0;
5394}
5395
5396/*
5397 * 8.2.5. Get the Current Number of Associations (SCTP_GET_ASSOC_NUMBER)
5398 * This option gets the current number of associations that are attached
5399 * to a one-to-many style socket. The option value is an uint32_t.
5400 */
5401static int sctp_getsockopt_assoc_number(struct sock *sk, int len,
5402 char __user *optval, int __user *optlen)
5403{
5404 struct sctp_sock *sp = sctp_sk(sk);
5405 struct sctp_association *asoc;
5406 u32 val = 0;
5407
5408 if (sctp_style(sk, TCP))
5409 return -EOPNOTSUPP;
5410
5411 if (len < sizeof(u32))
5412 return -EINVAL;
5413
5414 len = sizeof(u32);
5415
5416 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5417 val++;
5418 }
5419
5420 if (put_user(len, optlen))
5421 return -EFAULT;
5422 if (copy_to_user(optval, &val, len))
5423 return -EFAULT;
5424
5425 return 0;
5426}
5427
5428/*
5429 * 8.1.23 SCTP_AUTO_ASCONF
5430 * See the corresponding setsockopt entry as description
5431 */
5432static int sctp_getsockopt_auto_asconf(struct sock *sk, int len,
5433 char __user *optval, int __user *optlen)
5434{
5435 int val = 0;
5436
5437 if (len < sizeof(int))
5438 return -EINVAL;
5439
5440 len = sizeof(int);
5441 if (sctp_sk(sk)->do_auto_asconf && sctp_is_ep_boundall(sk))
5442 val = 1;
5443 if (put_user(len, optlen))
5444 return -EFAULT;
5445 if (copy_to_user(optval, &val, len))
5446 return -EFAULT;
5447 return 0;
5448}
5449
5450/*
5451 * 8.2.6. Get the Current Identifiers of Associations
5452 * (SCTP_GET_ASSOC_ID_LIST)
5453 *
5454 * This option gets the current list of SCTP association identifiers of
5455 * the SCTP associations handled by a one-to-many style socket.
5456 */
5457static int sctp_getsockopt_assoc_ids(struct sock *sk, int len,
5458 char __user *optval, int __user *optlen)
5459{
5460 struct sctp_sock *sp = sctp_sk(sk);
5461 struct sctp_association *asoc;
5462 struct sctp_assoc_ids *ids;
5463 u32 num = 0;
5464
5465 if (sctp_style(sk, TCP))
5466 return -EOPNOTSUPP;
5467
5468 if (len < sizeof(struct sctp_assoc_ids))
5469 return -EINVAL;
5470
5471 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5472 num++;
5473 }
5474
5475 if (len < sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num)
5476 return -EINVAL;
5477
5478 len = sizeof(struct sctp_assoc_ids) + sizeof(sctp_assoc_t) * num;
5479
5480 ids = kmalloc(len, GFP_KERNEL);
5481 if (unlikely(!ids))
5482 return -ENOMEM;
5483
5484 ids->gaids_number_of_ids = num;
5485 num = 0;
5486 list_for_each_entry(asoc, &(sp->ep->asocs), asocs) {
5487 ids->gaids_assoc_id[num++] = asoc->assoc_id;
5488 }
5489
5490 if (put_user(len, optlen) || copy_to_user(optval, ids, len)) {
5491 kfree(ids);
5492 return -EFAULT;
5493 }
5494
5495 kfree(ids);
5496 return 0;
5497}
5498
5499SCTP_STATIC int sctp_getsockopt(struct sock *sk, int level, int optname,
5500 char __user *optval, int __user *optlen)
5501{
5502 int retval = 0;
5503 int len;
5504
5505 SCTP_DEBUG_PRINTK("sctp_getsockopt(sk: %p... optname: %d)\n",
5506 sk, optname);
5507
5508 /* I can hardly begin to describe how wrong this is. This is
5509 * so broken as to be worse than useless. The API draft
5510 * REALLY is NOT helpful here... I am not convinced that the
5511 * semantics of getsockopt() with a level OTHER THAN SOL_SCTP
5512 * are at all well-founded.
5513 */
5514 if (level != SOL_SCTP) {
5515 struct sctp_af *af = sctp_sk(sk)->pf->af;
5516
5517 retval = af->getsockopt(sk, level, optname, optval, optlen);
5518 return retval;
5519 }
5520
5521 if (get_user(len, optlen))
5522 return -EFAULT;
5523
5524 sctp_lock_sock(sk);
5525
5526 switch (optname) {
5527 case SCTP_STATUS:
5528 retval = sctp_getsockopt_sctp_status(sk, len, optval, optlen);
5529 break;
5530 case SCTP_DISABLE_FRAGMENTS:
5531 retval = sctp_getsockopt_disable_fragments(sk, len, optval,
5532 optlen);
5533 break;
5534 case SCTP_EVENTS:
5535 retval = sctp_getsockopt_events(sk, len, optval, optlen);
5536 break;
5537 case SCTP_AUTOCLOSE:
5538 retval = sctp_getsockopt_autoclose(sk, len, optval, optlen);
5539 break;
5540 case SCTP_SOCKOPT_PEELOFF:
5541 retval = sctp_getsockopt_peeloff(sk, len, optval, optlen);
5542 break;
5543 case SCTP_PEER_ADDR_PARAMS:
5544 retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
5545 optlen);
5546 break;
5547 case SCTP_DELAYED_SACK:
5548 retval = sctp_getsockopt_delayed_ack(sk, len, optval,
5549 optlen);
5550 break;
5551 case SCTP_INITMSG:
5552 retval = sctp_getsockopt_initmsg(sk, len, optval, optlen);
5553 break;
5554 case SCTP_GET_PEER_ADDRS:
5555 retval = sctp_getsockopt_peer_addrs(sk, len, optval,
5556 optlen);
5557 break;
5558 case SCTP_GET_LOCAL_ADDRS:
5559 retval = sctp_getsockopt_local_addrs(sk, len, optval,
5560 optlen);
5561 break;
5562 case SCTP_SOCKOPT_CONNECTX3:
5563 retval = sctp_getsockopt_connectx3(sk, len, optval, optlen);
5564 break;
5565 case SCTP_DEFAULT_SEND_PARAM:
5566 retval = sctp_getsockopt_default_send_param(sk, len,
5567 optval, optlen);
5568 break;
5569 case SCTP_PRIMARY_ADDR:
5570 retval = sctp_getsockopt_primary_addr(sk, len, optval, optlen);
5571 break;
5572 case SCTP_NODELAY:
5573 retval = sctp_getsockopt_nodelay(sk, len, optval, optlen);
5574 break;
5575 case SCTP_RTOINFO:
5576 retval = sctp_getsockopt_rtoinfo(sk, len, optval, optlen);
5577 break;
5578 case SCTP_ASSOCINFO:
5579 retval = sctp_getsockopt_associnfo(sk, len, optval, optlen);
5580 break;
5581 case SCTP_I_WANT_MAPPED_V4_ADDR:
5582 retval = sctp_getsockopt_mappedv4(sk, len, optval, optlen);
5583 break;
5584 case SCTP_MAXSEG:
5585 retval = sctp_getsockopt_maxseg(sk, len, optval, optlen);
5586 break;
5587 case SCTP_GET_PEER_ADDR_INFO:
5588 retval = sctp_getsockopt_peer_addr_info(sk, len, optval,
5589 optlen);
5590 break;
5591 case SCTP_ADAPTATION_LAYER:
5592 retval = sctp_getsockopt_adaptation_layer(sk, len, optval,
5593 optlen);
5594 break;
5595 case SCTP_CONTEXT:
5596 retval = sctp_getsockopt_context(sk, len, optval, optlen);
5597 break;
5598 case SCTP_FRAGMENT_INTERLEAVE:
5599 retval = sctp_getsockopt_fragment_interleave(sk, len, optval,
5600 optlen);
5601 break;
5602 case SCTP_PARTIAL_DELIVERY_POINT:
5603 retval = sctp_getsockopt_partial_delivery_point(sk, len, optval,
5604 optlen);
5605 break;
5606 case SCTP_MAX_BURST:
5607 retval = sctp_getsockopt_maxburst(sk, len, optval, optlen);
5608 break;
5609 case SCTP_AUTH_KEY:
5610 case SCTP_AUTH_CHUNK:
5611 case SCTP_AUTH_DELETE_KEY:
5612 retval = -EOPNOTSUPP;
5613 break;
5614 case SCTP_HMAC_IDENT:
5615 retval = sctp_getsockopt_hmac_ident(sk, len, optval, optlen);
5616 break;
5617 case SCTP_AUTH_ACTIVE_KEY:
5618 retval = sctp_getsockopt_active_key(sk, len, optval, optlen);
5619 break;
5620 case SCTP_PEER_AUTH_CHUNKS:
5621 retval = sctp_getsockopt_peer_auth_chunks(sk, len, optval,
5622 optlen);
5623 break;
5624 case SCTP_LOCAL_AUTH_CHUNKS:
5625 retval = sctp_getsockopt_local_auth_chunks(sk, len, optval,
5626 optlen);
5627 break;
5628 case SCTP_GET_ASSOC_NUMBER:
5629 retval = sctp_getsockopt_assoc_number(sk, len, optval, optlen);
5630 break;
5631 case SCTP_GET_ASSOC_ID_LIST:
5632 retval = sctp_getsockopt_assoc_ids(sk, len, optval, optlen);
5633 break;
5634 case SCTP_AUTO_ASCONF:
5635 retval = sctp_getsockopt_auto_asconf(sk, len, optval, optlen);
5636 break;
5637 default:
5638 retval = -ENOPROTOOPT;
5639 break;
5640 }
5641
5642 sctp_release_sock(sk);
5643 return retval;
5644}
5645
5646static void sctp_hash(struct sock *sk)
5647{
5648 /* STUB */
5649}
5650
5651static void sctp_unhash(struct sock *sk)
5652{
5653 /* STUB */
5654}
5655
5656/* Check if port is acceptable. Possibly find first available port.
5657 *
5658 * The port hash table (contained in the 'global' SCTP protocol storage
5659 * returned by struct sctp_protocol *sctp_get_protocol()). The hash
5660 * table is an array of 4096 lists (sctp_bind_hashbucket). Each
5661 * list (the list number is the port number hashed out, so as you
5662 * would expect from a hash function, all the ports in a given list have
5663 * such a number that hashes out to the same list number; you were
5664 * expecting that, right?); so each list has a set of ports, with a
5665 * link to the socket (struct sock) that uses it, the port number and
5666 * a fastreuse flag (FIXME: NPI ipg).
5667 */
5668static struct sctp_bind_bucket *sctp_bucket_create(
5669 struct sctp_bind_hashbucket *head, unsigned short snum);
5670
5671static long sctp_get_port_local(struct sock *sk, union sctp_addr *addr)
5672{
5673 struct sctp_bind_hashbucket *head; /* hash list */
5674 struct sctp_bind_bucket *pp; /* hash list port iterator */
5675 struct hlist_node *node;
5676 unsigned short snum;
5677 int ret;
5678
5679 snum = ntohs(addr->v4.sin_port);
5680
5681 SCTP_DEBUG_PRINTK("sctp_get_port() begins, snum=%d\n", snum);
5682 sctp_local_bh_disable();
5683
5684 if (snum == 0) {
5685 /* Search for an available port. */
5686 int low, high, remaining, index;
5687 unsigned int rover;
5688
5689 inet_get_local_port_range(&low, &high);
5690 remaining = (high - low) + 1;
5691 rover = net_random() % remaining + low;
5692
5693 do {
5694 rover++;
5695 if ((rover < low) || (rover > high))
5696 rover = low;
5697 if (inet_is_reserved_local_port(rover))
5698 continue;
5699 index = sctp_phashfn(rover);
5700 head = &sctp_port_hashtable[index];
5701 sctp_spin_lock(&head->lock);
5702 sctp_for_each_hentry(pp, node, &head->chain)
5703 if (pp->port == rover)
5704 goto next;
5705 break;
5706 next:
5707 sctp_spin_unlock(&head->lock);
5708 } while (--remaining > 0);
5709
5710 /* Exhausted local port range during search? */
5711 ret = 1;
5712 if (remaining <= 0)
5713 goto fail;
5714
5715 /* OK, here is the one we will use. HEAD (the port
5716 * hash table list entry) is non-NULL and we hold it's
5717 * mutex.
5718 */
5719 snum = rover;
5720 } else {
5721 /* We are given an specific port number; we verify
5722 * that it is not being used. If it is used, we will
5723 * exahust the search in the hash list corresponding
5724 * to the port number (snum) - we detect that with the
5725 * port iterator, pp being NULL.
5726 */
5727 head = &sctp_port_hashtable[sctp_phashfn(snum)];
5728 sctp_spin_lock(&head->lock);
5729 sctp_for_each_hentry(pp, node, &head->chain) {
5730 if (pp->port == snum)
5731 goto pp_found;
5732 }
5733 }
5734 pp = NULL;
5735 goto pp_not_found;
5736pp_found:
5737 if (!hlist_empty(&pp->owner)) {
5738 /* We had a port hash table hit - there is an
5739 * available port (pp != NULL) and it is being
5740 * used by other socket (pp->owner not empty); that other
5741 * socket is going to be sk2.
5742 */
5743 int reuse = sk->sk_reuse;
5744 struct sock *sk2;
5745
5746 SCTP_DEBUG_PRINTK("sctp_get_port() found a possible match\n");
5747 if (pp->fastreuse && sk->sk_reuse &&
5748 sk->sk_state != SCTP_SS_LISTENING)
5749 goto success;
5750
5751 /* Run through the list of sockets bound to the port
5752 * (pp->port) [via the pointers bind_next and
5753 * bind_pprev in the struct sock *sk2 (pp->sk)]. On each one,
5754 * we get the endpoint they describe and run through
5755 * the endpoint's list of IP (v4 or v6) addresses,
5756 * comparing each of the addresses with the address of
5757 * the socket sk. If we find a match, then that means
5758 * that this port/socket (sk) combination are already
5759 * in an endpoint.
5760 */
5761 sk_for_each_bound(sk2, node, &pp->owner) {
5762 struct sctp_endpoint *ep2;
5763 ep2 = sctp_sk(sk2)->ep;
5764
5765 if (sk == sk2 ||
5766 (reuse && sk2->sk_reuse &&
5767 sk2->sk_state != SCTP_SS_LISTENING))
5768 continue;
5769
5770 if (sctp_bind_addr_conflict(&ep2->base.bind_addr, addr,
5771 sctp_sk(sk2), sctp_sk(sk))) {
5772 ret = (long)sk2;
5773 goto fail_unlock;
5774 }
5775 }
5776 SCTP_DEBUG_PRINTK("sctp_get_port(): Found a match\n");
5777 }
5778pp_not_found:
5779 /* If there was a hash table miss, create a new port. */
5780 ret = 1;
5781 if (!pp && !(pp = sctp_bucket_create(head, snum)))
5782 goto fail_unlock;
5783
5784 /* In either case (hit or miss), make sure fastreuse is 1 only
5785 * if sk->sk_reuse is too (that is, if the caller requested
5786 * SO_REUSEADDR on this socket -sk-).
5787 */
5788 if (hlist_empty(&pp->owner)) {
5789 if (sk->sk_reuse && sk->sk_state != SCTP_SS_LISTENING)
5790 pp->fastreuse = 1;
5791 else
5792 pp->fastreuse = 0;
5793 } else if (pp->fastreuse &&
5794 (!sk->sk_reuse || sk->sk_state == SCTP_SS_LISTENING))
5795 pp->fastreuse = 0;
5796
5797 /* We are set, so fill up all the data in the hash table
5798 * entry, tie the socket list information with the rest of the
5799 * sockets FIXME: Blurry, NPI (ipg).
5800 */
5801success:
5802 if (!sctp_sk(sk)->bind_hash) {
5803 inet_sk(sk)->inet_num = snum;
5804 sk_add_bind_node(sk, &pp->owner);
5805 sctp_sk(sk)->bind_hash = pp;
5806 }
5807 ret = 0;
5808
5809fail_unlock:
5810 sctp_spin_unlock(&head->lock);
5811
5812fail:
5813 sctp_local_bh_enable();
5814 return ret;
5815}
5816
5817/* Assign a 'snum' port to the socket. If snum == 0, an ephemeral
5818 * port is requested.
5819 */
5820static int sctp_get_port(struct sock *sk, unsigned short snum)
5821{
5822 long ret;
5823 union sctp_addr addr;
5824 struct sctp_af *af = sctp_sk(sk)->pf->af;
5825
5826 /* Set up a dummy address struct from the sk. */
5827 af->from_sk(&addr, sk);
5828 addr.v4.sin_port = htons(snum);
5829
5830 /* Note: sk->sk_num gets filled in if ephemeral port request. */
5831 ret = sctp_get_port_local(sk, &addr);
5832
5833 return ret ? 1 : 0;
5834}
5835
5836/*
5837 * Move a socket to LISTENING state.
5838 */
5839SCTP_STATIC int sctp_listen_start(struct sock *sk, int backlog)
5840{
5841 struct sctp_sock *sp = sctp_sk(sk);
5842 struct sctp_endpoint *ep = sp->ep;
5843 struct crypto_hash *tfm = NULL;
5844
5845 /* Allocate HMAC for generating cookie. */
5846 if (!sctp_sk(sk)->hmac && sctp_hmac_alg) {
5847 tfm = crypto_alloc_hash(sctp_hmac_alg, 0, CRYPTO_ALG_ASYNC);
5848 if (IS_ERR(tfm)) {
5849 if (net_ratelimit()) {
5850 pr_info("failed to load transform for %s: %ld\n",
5851 sctp_hmac_alg, PTR_ERR(tfm));
5852 }
5853 return -ENOSYS;
5854 }
5855 sctp_sk(sk)->hmac = tfm;
5856 }
5857
5858 /*
5859 * If a bind() or sctp_bindx() is not called prior to a listen()
5860 * call that allows new associations to be accepted, the system
5861 * picks an ephemeral port and will choose an address set equivalent
5862 * to binding with a wildcard address.
5863 *
5864 * This is not currently spelled out in the SCTP sockets
5865 * extensions draft, but follows the practice as seen in TCP
5866 * sockets.
5867 *
5868 */
5869 sk->sk_state = SCTP_SS_LISTENING;
5870 if (!ep->base.bind_addr.port) {
5871 if (sctp_autobind(sk))
5872 return -EAGAIN;
5873 } else {
5874 if (sctp_get_port(sk, inet_sk(sk)->inet_num)) {
5875 sk->sk_state = SCTP_SS_CLOSED;
5876 return -EADDRINUSE;
5877 }
5878 }
5879
5880 sk->sk_max_ack_backlog = backlog;
5881 sctp_hash_endpoint(ep);
5882 return 0;
5883}
5884
5885/*
5886 * 4.1.3 / 5.1.3 listen()
5887 *
5888 * By default, new associations are not accepted for UDP style sockets.
5889 * An application uses listen() to mark a socket as being able to
5890 * accept new associations.
5891 *
5892 * On TCP style sockets, applications use listen() to ready the SCTP
5893 * endpoint for accepting inbound associations.
5894 *
5895 * On both types of endpoints a backlog of '0' disables listening.
5896 *
5897 * Move a socket to LISTENING state.
5898 */
5899int sctp_inet_listen(struct socket *sock, int backlog)
5900{
5901 struct sock *sk = sock->sk;
5902 struct sctp_endpoint *ep = sctp_sk(sk)->ep;
5903 int err = -EINVAL;
5904
5905 if (unlikely(backlog < 0))
5906 return err;
5907
5908 sctp_lock_sock(sk);
5909
5910 /* Peeled-off sockets are not allowed to listen(). */
5911 if (sctp_style(sk, UDP_HIGH_BANDWIDTH))
5912 goto out;
5913
5914 if (sock->state != SS_UNCONNECTED)
5915 goto out;
5916
5917 /* If backlog is zero, disable listening. */
5918 if (!backlog) {
5919 if (sctp_sstate(sk, CLOSED))
5920 goto out;
5921
5922 err = 0;
5923 sctp_unhash_endpoint(ep);
5924 sk->sk_state = SCTP_SS_CLOSED;
5925 if (sk->sk_reuse)
5926 sctp_sk(sk)->bind_hash->fastreuse = 1;
5927 goto out;
5928 }
5929
5930 /* If we are already listening, just update the backlog */
5931 if (sctp_sstate(sk, LISTENING))
5932 sk->sk_max_ack_backlog = backlog;
5933 else {
5934 err = sctp_listen_start(sk, backlog);
5935 if (err)
5936 goto out;
5937 }
5938
5939 err = 0;
5940out:
5941 sctp_release_sock(sk);
5942 return err;
5943}
5944
5945/*
5946 * This function is done by modeling the current datagram_poll() and the
5947 * tcp_poll(). Note that, based on these implementations, we don't
5948 * lock the socket in this function, even though it seems that,
5949 * ideally, locking or some other mechanisms can be used to ensure
5950 * the integrity of the counters (sndbuf and wmem_alloc) used
5951 * in this place. We assume that we don't need locks either until proven
5952 * otherwise.
5953 *
5954 * Another thing to note is that we include the Async I/O support
5955 * here, again, by modeling the current TCP/UDP code. We don't have
5956 * a good way to test with it yet.
5957 */
5958unsigned int sctp_poll(struct file *file, struct socket *sock, poll_table *wait)
5959{
5960 struct sock *sk = sock->sk;
5961 struct sctp_sock *sp = sctp_sk(sk);
5962 unsigned int mask;
5963
5964 poll_wait(file, sk_sleep(sk), wait);
5965
5966 /* A TCP-style listening socket becomes readable when the accept queue
5967 * is not empty.
5968 */
5969 if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING))
5970 return (!list_empty(&sp->ep->asocs)) ?
5971 (POLLIN | POLLRDNORM) : 0;
5972
5973 mask = 0;
5974
5975 /* Is there any exceptional events? */
5976 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
5977 mask |= POLLERR;
5978 if (sk->sk_shutdown & RCV_SHUTDOWN)
5979 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
5980 if (sk->sk_shutdown == SHUTDOWN_MASK)
5981 mask |= POLLHUP;
5982
5983 /* Is it readable? Reconsider this code with TCP-style support. */
5984 if (!skb_queue_empty(&sk->sk_receive_queue))
5985 mask |= POLLIN | POLLRDNORM;
5986
5987 /* The association is either gone or not ready. */
5988 if (!sctp_style(sk, UDP) && sctp_sstate(sk, CLOSED))
5989 return mask;
5990
5991 /* Is it writable? */
5992 if (sctp_writeable(sk)) {
5993 mask |= POLLOUT | POLLWRNORM;
5994 } else {
5995 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
5996 /*
5997 * Since the socket is not locked, the buffer
5998 * might be made available after the writeable check and
5999 * before the bit is set. This could cause a lost I/O
6000 * signal. tcp_poll() has a race breaker for this race
6001 * condition. Based on their implementation, we put
6002 * in the following code to cover it as well.
6003 */
6004 if (sctp_writeable(sk))
6005 mask |= POLLOUT | POLLWRNORM;
6006 }
6007 return mask;
6008}
6009
6010/********************************************************************
6011 * 2nd Level Abstractions
6012 ********************************************************************/
6013
6014static struct sctp_bind_bucket *sctp_bucket_create(
6015 struct sctp_bind_hashbucket *head, unsigned short snum)
6016{
6017 struct sctp_bind_bucket *pp;
6018
6019 pp = kmem_cache_alloc(sctp_bucket_cachep, GFP_ATOMIC);
6020 if (pp) {
6021 SCTP_DBG_OBJCNT_INC(bind_bucket);
6022 pp->port = snum;
6023 pp->fastreuse = 0;
6024 INIT_HLIST_HEAD(&pp->owner);
6025 hlist_add_head(&pp->node, &head->chain);
6026 }
6027 return pp;
6028}
6029
6030/* Caller must hold hashbucket lock for this tb with local BH disabled */
6031static void sctp_bucket_destroy(struct sctp_bind_bucket *pp)
6032{
6033 if (pp && hlist_empty(&pp->owner)) {
6034 __hlist_del(&pp->node);
6035 kmem_cache_free(sctp_bucket_cachep, pp);
6036 SCTP_DBG_OBJCNT_DEC(bind_bucket);
6037 }
6038}
6039
6040/* Release this socket's reference to a local port. */
6041static inline void __sctp_put_port(struct sock *sk)
6042{
6043 struct sctp_bind_hashbucket *head =
6044 &sctp_port_hashtable[sctp_phashfn(inet_sk(sk)->inet_num)];
6045 struct sctp_bind_bucket *pp;
6046
6047 sctp_spin_lock(&head->lock);
6048 pp = sctp_sk(sk)->bind_hash;
6049 __sk_del_bind_node(sk);
6050 sctp_sk(sk)->bind_hash = NULL;
6051 inet_sk(sk)->inet_num = 0;
6052 sctp_bucket_destroy(pp);
6053 sctp_spin_unlock(&head->lock);
6054}
6055
6056void sctp_put_port(struct sock *sk)
6057{
6058 sctp_local_bh_disable();
6059 __sctp_put_port(sk);
6060 sctp_local_bh_enable();
6061}
6062
6063/*
6064 * The system picks an ephemeral port and choose an address set equivalent
6065 * to binding with a wildcard address.
6066 * One of those addresses will be the primary address for the association.
6067 * This automatically enables the multihoming capability of SCTP.
6068 */
6069static int sctp_autobind(struct sock *sk)
6070{
6071 union sctp_addr autoaddr;
6072 struct sctp_af *af;
6073 __be16 port;
6074
6075 /* Initialize a local sockaddr structure to INADDR_ANY. */
6076 af = sctp_sk(sk)->pf->af;
6077
6078 port = htons(inet_sk(sk)->inet_num);
6079 af->inaddr_any(&autoaddr, port);
6080
6081 return sctp_do_bind(sk, &autoaddr, af->sockaddr_len);
6082}
6083
6084/* Parse out IPPROTO_SCTP CMSG headers. Perform only minimal validation.
6085 *
6086 * From RFC 2292
6087 * 4.2 The cmsghdr Structure *
6088 *
6089 * When ancillary data is sent or received, any number of ancillary data
6090 * objects can be specified by the msg_control and msg_controllen members of
6091 * the msghdr structure, because each object is preceded by
6092 * a cmsghdr structure defining the object's length (the cmsg_len member).
6093 * Historically Berkeley-derived implementations have passed only one object
6094 * at a time, but this API allows multiple objects to be
6095 * passed in a single call to sendmsg() or recvmsg(). The following example
6096 * shows two ancillary data objects in a control buffer.
6097 *
6098 * |<--------------------------- msg_controllen -------------------------->|
6099 * | |
6100 *
6101 * |<----- ancillary data object ----->|<----- ancillary data object ----->|
6102 *
6103 * |<---------- CMSG_SPACE() --------->|<---------- CMSG_SPACE() --------->|
6104 * | | |
6105 *
6106 * |<---------- cmsg_len ---------->| |<--------- cmsg_len ----------->| |
6107 *
6108 * |<--------- CMSG_LEN() --------->| |<-------- CMSG_LEN() ---------->| |
6109 * | | | | |
6110 *
6111 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6112 * |cmsg_|cmsg_|cmsg_|XX| |XX|cmsg_|cmsg_|cmsg_|XX| |XX|
6113 *
6114 * |len |level|type |XX|cmsg_data[]|XX|len |level|type |XX|cmsg_data[]|XX|
6115 *
6116 * +-----+-----+-----+--+-----------+--+-----+-----+-----+--+-----------+--+
6117 * ^
6118 * |
6119 *
6120 * msg_control
6121 * points here
6122 */
6123SCTP_STATIC int sctp_msghdr_parse(const struct msghdr *msg,
6124 sctp_cmsgs_t *cmsgs)
6125{
6126 struct cmsghdr *cmsg;
6127 struct msghdr *my_msg = (struct msghdr *)msg;
6128
6129 for (cmsg = CMSG_FIRSTHDR(msg);
6130 cmsg != NULL;
6131 cmsg = CMSG_NXTHDR(my_msg, cmsg)) {
6132 if (!CMSG_OK(my_msg, cmsg))
6133 return -EINVAL;
6134
6135 /* Should we parse this header or ignore? */
6136 if (cmsg->cmsg_level != IPPROTO_SCTP)
6137 continue;
6138
6139 /* Strictly check lengths following example in SCM code. */
6140 switch (cmsg->cmsg_type) {
6141 case SCTP_INIT:
6142 /* SCTP Socket API Extension
6143 * 5.2.1 SCTP Initiation Structure (SCTP_INIT)
6144 *
6145 * This cmsghdr structure provides information for
6146 * initializing new SCTP associations with sendmsg().
6147 * The SCTP_INITMSG socket option uses this same data
6148 * structure. This structure is not used for
6149 * recvmsg().
6150 *
6151 * cmsg_level cmsg_type cmsg_data[]
6152 * ------------ ------------ ----------------------
6153 * IPPROTO_SCTP SCTP_INIT struct sctp_initmsg
6154 */
6155 if (cmsg->cmsg_len !=
6156 CMSG_LEN(sizeof(struct sctp_initmsg)))
6157 return -EINVAL;
6158 cmsgs->init = (struct sctp_initmsg *)CMSG_DATA(cmsg);
6159 break;
6160
6161 case SCTP_SNDRCV:
6162 /* SCTP Socket API Extension
6163 * 5.2.2 SCTP Header Information Structure(SCTP_SNDRCV)
6164 *
6165 * This cmsghdr structure specifies SCTP options for
6166 * sendmsg() and describes SCTP header information
6167 * about a received message through recvmsg().
6168 *
6169 * cmsg_level cmsg_type cmsg_data[]
6170 * ------------ ------------ ----------------------
6171 * IPPROTO_SCTP SCTP_SNDRCV struct sctp_sndrcvinfo
6172 */
6173 if (cmsg->cmsg_len !=
6174 CMSG_LEN(sizeof(struct sctp_sndrcvinfo)))
6175 return -EINVAL;
6176
6177 cmsgs->info =
6178 (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
6179
6180 /* Minimally, validate the sinfo_flags. */
6181 if (cmsgs->info->sinfo_flags &
6182 ~(SCTP_UNORDERED | SCTP_ADDR_OVER |
6183 SCTP_ABORT | SCTP_EOF))
6184 return -EINVAL;
6185 break;
6186
6187 default:
6188 return -EINVAL;
6189 }
6190 }
6191 return 0;
6192}
6193
6194/*
6195 * Wait for a packet..
6196 * Note: This function is the same function as in core/datagram.c
6197 * with a few modifications to make lksctp work.
6198 */
6199static int sctp_wait_for_packet(struct sock * sk, int *err, long *timeo_p)
6200{
6201 int error;
6202 DEFINE_WAIT(wait);
6203
6204 prepare_to_wait_exclusive(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6205
6206 /* Socket errors? */
6207 error = sock_error(sk);
6208 if (error)
6209 goto out;
6210
6211 if (!skb_queue_empty(&sk->sk_receive_queue))
6212 goto ready;
6213
6214 /* Socket shut down? */
6215 if (sk->sk_shutdown & RCV_SHUTDOWN)
6216 goto out;
6217
6218 /* Sequenced packets can come disconnected. If so we report the
6219 * problem.
6220 */
6221 error = -ENOTCONN;
6222
6223 /* Is there a good reason to think that we may receive some data? */
6224 if (list_empty(&sctp_sk(sk)->ep->asocs) && !sctp_sstate(sk, LISTENING))
6225 goto out;
6226
6227 /* Handle signals. */
6228 if (signal_pending(current))
6229 goto interrupted;
6230
6231 /* Let another process have a go. Since we are going to sleep
6232 * anyway. Note: This may cause odd behaviors if the message
6233 * does not fit in the user's buffer, but this seems to be the
6234 * only way to honor MSG_DONTWAIT realistically.
6235 */
6236 sctp_release_sock(sk);
6237 *timeo_p = schedule_timeout(*timeo_p);
6238 sctp_lock_sock(sk);
6239
6240ready:
6241 finish_wait(sk_sleep(sk), &wait);
6242 return 0;
6243
6244interrupted:
6245 error = sock_intr_errno(*timeo_p);
6246
6247out:
6248 finish_wait(sk_sleep(sk), &wait);
6249 *err = error;
6250 return error;
6251}
6252
6253/* Receive a datagram.
6254 * Note: This is pretty much the same routine as in core/datagram.c
6255 * with a few changes to make lksctp work.
6256 */
6257static struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags,
6258 int noblock, int *err)
6259{
6260 int error;
6261 struct sk_buff *skb;
6262 long timeo;
6263
6264 timeo = sock_rcvtimeo(sk, noblock);
6265
6266 SCTP_DEBUG_PRINTK("Timeout: timeo: %ld, MAX: %ld.\n",
6267 timeo, MAX_SCHEDULE_TIMEOUT);
6268
6269 do {
6270 /* Again only user level code calls this function,
6271 * so nothing interrupt level
6272 * will suddenly eat the receive_queue.
6273 *
6274 * Look at current nfs client by the way...
6275 * However, this function was correct in any case. 8)
6276 */
6277 if (flags & MSG_PEEK) {
6278 spin_lock_bh(&sk->sk_receive_queue.lock);
6279 skb = skb_peek(&sk->sk_receive_queue);
6280 if (skb)
6281 atomic_inc(&skb->users);
6282 spin_unlock_bh(&sk->sk_receive_queue.lock);
6283 } else {
6284 skb = skb_dequeue(&sk->sk_receive_queue);
6285 }
6286
6287 if (skb)
6288 return skb;
6289
6290 /* Caller is allowed not to check sk->sk_err before calling. */
6291 error = sock_error(sk);
6292 if (error)
6293 goto no_packet;
6294
6295 if (sk->sk_shutdown & RCV_SHUTDOWN)
6296 break;
6297
6298 /* User doesn't want to wait. */
6299 error = -EAGAIN;
6300 if (!timeo)
6301 goto no_packet;
6302 } while (sctp_wait_for_packet(sk, err, &timeo) == 0);
6303
6304 return NULL;
6305
6306no_packet:
6307 *err = error;
6308 return NULL;
6309}
6310
6311/* If sndbuf has changed, wake up per association sndbuf waiters. */
6312static void __sctp_write_space(struct sctp_association *asoc)
6313{
6314 struct sock *sk = asoc->base.sk;
6315 struct socket *sock = sk->sk_socket;
6316
6317 if ((sctp_wspace(asoc) > 0) && sock) {
6318 if (waitqueue_active(&asoc->wait))
6319 wake_up_interruptible(&asoc->wait);
6320
6321 if (sctp_writeable(sk)) {
6322 wait_queue_head_t *wq = sk_sleep(sk);
6323
6324 if (wq && waitqueue_active(wq))
6325 wake_up_interruptible(wq);
6326
6327 /* Note that we try to include the Async I/O support
6328 * here by modeling from the current TCP/UDP code.
6329 * We have not tested with it yet.
6330 */
6331 if (!(sk->sk_shutdown & SEND_SHUTDOWN))
6332 sock_wake_async(sock,
6333 SOCK_WAKE_SPACE, POLL_OUT);
6334 }
6335 }
6336}
6337
6338/* Do accounting for the sndbuf space.
6339 * Decrement the used sndbuf space of the corresponding association by the
6340 * data size which was just transmitted(freed).
6341 */
6342static void sctp_wfree(struct sk_buff *skb)
6343{
6344 struct sctp_association *asoc;
6345 struct sctp_chunk *chunk;
6346 struct sock *sk;
6347
6348 /* Get the saved chunk pointer. */
6349 chunk = *((struct sctp_chunk **)(skb->cb));
6350 asoc = chunk->asoc;
6351 sk = asoc->base.sk;
6352 asoc->sndbuf_used -= SCTP_DATA_SNDSIZE(chunk) +
6353 sizeof(struct sk_buff) +
6354 sizeof(struct sctp_chunk);
6355
6356 atomic_sub(sizeof(struct sctp_chunk), &sk->sk_wmem_alloc);
6357
6358 /*
6359 * This undoes what is done via sctp_set_owner_w and sk_mem_charge
6360 */
6361 sk->sk_wmem_queued -= skb->truesize;
6362 sk_mem_uncharge(sk, skb->truesize);
6363
6364 sock_wfree(skb);
6365 __sctp_write_space(asoc);
6366
6367 sctp_association_put(asoc);
6368}
6369
6370/* Do accounting for the receive space on the socket.
6371 * Accounting for the association is done in ulpevent.c
6372 * We set this as a destructor for the cloned data skbs so that
6373 * accounting is done at the correct time.
6374 */
6375void sctp_sock_rfree(struct sk_buff *skb)
6376{
6377 struct sock *sk = skb->sk;
6378 struct sctp_ulpevent *event = sctp_skb2event(skb);
6379
6380 atomic_sub(event->rmem_len, &sk->sk_rmem_alloc);
6381
6382 /*
6383 * Mimic the behavior of sock_rfree
6384 */
6385 sk_mem_uncharge(sk, event->rmem_len);
6386}
6387
6388
6389/* Helper function to wait for space in the sndbuf. */
6390static int sctp_wait_for_sndbuf(struct sctp_association *asoc, long *timeo_p,
6391 size_t msg_len)
6392{
6393 struct sock *sk = asoc->base.sk;
6394 int err = 0;
6395 long current_timeo = *timeo_p;
6396 DEFINE_WAIT(wait);
6397
6398 SCTP_DEBUG_PRINTK("wait_for_sndbuf: asoc=%p, timeo=%ld, msg_len=%zu\n",
6399 asoc, (long)(*timeo_p), msg_len);
6400
6401 /* Increment the association's refcnt. */
6402 sctp_association_hold(asoc);
6403
6404 /* Wait on the association specific sndbuf space. */
6405 for (;;) {
6406 prepare_to_wait_exclusive(&asoc->wait, &wait,
6407 TASK_INTERRUPTIBLE);
6408 if (!*timeo_p)
6409 goto do_nonblock;
6410 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6411 asoc->base.dead)
6412 goto do_error;
6413 if (signal_pending(current))
6414 goto do_interrupted;
6415 if (msg_len <= sctp_wspace(asoc))
6416 break;
6417
6418 /* Let another process have a go. Since we are going
6419 * to sleep anyway.
6420 */
6421 sctp_release_sock(sk);
6422 current_timeo = schedule_timeout(current_timeo);
6423 BUG_ON(sk != asoc->base.sk);
6424 sctp_lock_sock(sk);
6425
6426 *timeo_p = current_timeo;
6427 }
6428
6429out:
6430 finish_wait(&asoc->wait, &wait);
6431
6432 /* Release the association's refcnt. */
6433 sctp_association_put(asoc);
6434
6435 return err;
6436
6437do_error:
6438 err = -EPIPE;
6439 goto out;
6440
6441do_interrupted:
6442 err = sock_intr_errno(*timeo_p);
6443 goto out;
6444
6445do_nonblock:
6446 err = -EAGAIN;
6447 goto out;
6448}
6449
6450void sctp_data_ready(struct sock *sk, int len)
6451{
6452 struct socket_wq *wq;
6453
6454 rcu_read_lock();
6455 wq = rcu_dereference(sk->sk_wq);
6456 if (wq_has_sleeper(wq))
6457 wake_up_interruptible_sync_poll(&wq->wait, POLLIN |
6458 POLLRDNORM | POLLRDBAND);
6459 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
6460 rcu_read_unlock();
6461}
6462
6463/* If socket sndbuf has changed, wake up all per association waiters. */
6464void sctp_write_space(struct sock *sk)
6465{
6466 struct sctp_association *asoc;
6467
6468 /* Wake up the tasks in each wait queue. */
6469 list_for_each_entry(asoc, &((sctp_sk(sk))->ep->asocs), asocs) {
6470 __sctp_write_space(asoc);
6471 }
6472}
6473
6474/* Is there any sndbuf space available on the socket?
6475 *
6476 * Note that sk_wmem_alloc is the sum of the send buffers on all of the
6477 * associations on the same socket. For a UDP-style socket with
6478 * multiple associations, it is possible for it to be "unwriteable"
6479 * prematurely. I assume that this is acceptable because
6480 * a premature "unwriteable" is better than an accidental "writeable" which
6481 * would cause an unwanted block under certain circumstances. For the 1-1
6482 * UDP-style sockets or TCP-style sockets, this code should work.
6483 * - Daisy
6484 */
6485static int sctp_writeable(struct sock *sk)
6486{
6487 int amt = 0;
6488
6489 amt = sk->sk_sndbuf - sk_wmem_alloc_get(sk);
6490 if (amt < 0)
6491 amt = 0;
6492 return amt;
6493}
6494
6495/* Wait for an association to go into ESTABLISHED state. If timeout is 0,
6496 * returns immediately with EINPROGRESS.
6497 */
6498static int sctp_wait_for_connect(struct sctp_association *asoc, long *timeo_p)
6499{
6500 struct sock *sk = asoc->base.sk;
6501 int err = 0;
6502 long current_timeo = *timeo_p;
6503 DEFINE_WAIT(wait);
6504
6505 SCTP_DEBUG_PRINTK("%s: asoc=%p, timeo=%ld\n", __func__, asoc,
6506 (long)(*timeo_p));
6507
6508 /* Increment the association's refcnt. */
6509 sctp_association_hold(asoc);
6510
6511 for (;;) {
6512 prepare_to_wait_exclusive(&asoc->wait, &wait,
6513 TASK_INTERRUPTIBLE);
6514 if (!*timeo_p)
6515 goto do_nonblock;
6516 if (sk->sk_shutdown & RCV_SHUTDOWN)
6517 break;
6518 if (sk->sk_err || asoc->state >= SCTP_STATE_SHUTDOWN_PENDING ||
6519 asoc->base.dead)
6520 goto do_error;
6521 if (signal_pending(current))
6522 goto do_interrupted;
6523
6524 if (sctp_state(asoc, ESTABLISHED))
6525 break;
6526
6527 /* Let another process have a go. Since we are going
6528 * to sleep anyway.
6529 */
6530 sctp_release_sock(sk);
6531 current_timeo = schedule_timeout(current_timeo);
6532 sctp_lock_sock(sk);
6533
6534 *timeo_p = current_timeo;
6535 }
6536
6537out:
6538 finish_wait(&asoc->wait, &wait);
6539
6540 /* Release the association's refcnt. */
6541 sctp_association_put(asoc);
6542
6543 return err;
6544
6545do_error:
6546 if (asoc->init_err_counter + 1 > asoc->max_init_attempts)
6547 err = -ETIMEDOUT;
6548 else
6549 err = -ECONNREFUSED;
6550 goto out;
6551
6552do_interrupted:
6553 err = sock_intr_errno(*timeo_p);
6554 goto out;
6555
6556do_nonblock:
6557 err = -EINPROGRESS;
6558 goto out;
6559}
6560
6561static int sctp_wait_for_accept(struct sock *sk, long timeo)
6562{
6563 struct sctp_endpoint *ep;
6564 int err = 0;
6565 DEFINE_WAIT(wait);
6566
6567 ep = sctp_sk(sk)->ep;
6568
6569
6570 for (;;) {
6571 prepare_to_wait_exclusive(sk_sleep(sk), &wait,
6572 TASK_INTERRUPTIBLE);
6573
6574 if (list_empty(&ep->asocs)) {
6575 sctp_release_sock(sk);
6576 timeo = schedule_timeout(timeo);
6577 sctp_lock_sock(sk);
6578 }
6579
6580 err = -EINVAL;
6581 if (!sctp_sstate(sk, LISTENING))
6582 break;
6583
6584 err = 0;
6585 if (!list_empty(&ep->asocs))
6586 break;
6587
6588 err = sock_intr_errno(timeo);
6589 if (signal_pending(current))
6590 break;
6591
6592 err = -EAGAIN;
6593 if (!timeo)
6594 break;
6595 }
6596
6597 finish_wait(sk_sleep(sk), &wait);
6598
6599 return err;
6600}
6601
6602static void sctp_wait_for_close(struct sock *sk, long timeout)
6603{
6604 DEFINE_WAIT(wait);
6605
6606 do {
6607 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
6608 if (list_empty(&sctp_sk(sk)->ep->asocs))
6609 break;
6610 sctp_release_sock(sk);
6611 timeout = schedule_timeout(timeout);
6612 sctp_lock_sock(sk);
6613 } while (!signal_pending(current) && timeout);
6614
6615 finish_wait(sk_sleep(sk), &wait);
6616}
6617
6618static void sctp_skb_set_owner_r_frag(struct sk_buff *skb, struct sock *sk)
6619{
6620 struct sk_buff *frag;
6621
6622 if (!skb->data_len)
6623 goto done;
6624
6625 /* Don't forget the fragments. */
6626 skb_walk_frags(skb, frag)
6627 sctp_skb_set_owner_r_frag(frag, sk);
6628
6629done:
6630 sctp_skb_set_owner_r(skb, sk);
6631}
6632
6633void sctp_copy_sock(struct sock *newsk, struct sock *sk,
6634 struct sctp_association *asoc)
6635{
6636 struct inet_sock *inet = inet_sk(sk);
6637 struct inet_sock *newinet;
6638
6639 newsk->sk_type = sk->sk_type;
6640 newsk->sk_bound_dev_if = sk->sk_bound_dev_if;
6641 newsk->sk_flags = sk->sk_flags;
6642 newsk->sk_no_check = sk->sk_no_check;
6643 newsk->sk_reuse = sk->sk_reuse;
6644
6645 newsk->sk_shutdown = sk->sk_shutdown;
6646 newsk->sk_destruct = inet_sock_destruct;
6647 newsk->sk_family = sk->sk_family;
6648 newsk->sk_protocol = IPPROTO_SCTP;
6649 newsk->sk_backlog_rcv = sk->sk_prot->backlog_rcv;
6650 newsk->sk_sndbuf = sk->sk_sndbuf;
6651 newsk->sk_rcvbuf = sk->sk_rcvbuf;
6652 newsk->sk_lingertime = sk->sk_lingertime;
6653 newsk->sk_rcvtimeo = sk->sk_rcvtimeo;
6654 newsk->sk_sndtimeo = sk->sk_sndtimeo;
6655
6656 newinet = inet_sk(newsk);
6657
6658 /* Initialize sk's sport, dport, rcv_saddr and daddr for
6659 * getsockname() and getpeername()
6660 */
6661 newinet->inet_sport = inet->inet_sport;
6662 newinet->inet_saddr = inet->inet_saddr;
6663 newinet->inet_rcv_saddr = inet->inet_rcv_saddr;
6664 newinet->inet_dport = htons(asoc->peer.port);
6665 newinet->pmtudisc = inet->pmtudisc;
6666 newinet->inet_id = asoc->next_tsn ^ jiffies;
6667
6668 newinet->uc_ttl = inet->uc_ttl;
6669 newinet->mc_loop = 1;
6670 newinet->mc_ttl = 1;
6671 newinet->mc_index = 0;
6672 newinet->mc_list = NULL;
6673}
6674
6675/* Populate the fields of the newsk from the oldsk and migrate the assoc
6676 * and its messages to the newsk.
6677 */
6678static void sctp_sock_migrate(struct sock *oldsk, struct sock *newsk,
6679 struct sctp_association *assoc,
6680 sctp_socket_type_t type)
6681{
6682 struct sctp_sock *oldsp = sctp_sk(oldsk);
6683 struct sctp_sock *newsp = sctp_sk(newsk);
6684 struct sctp_bind_bucket *pp; /* hash list port iterator */
6685 struct sctp_endpoint *newep = newsp->ep;
6686 struct sk_buff *skb, *tmp;
6687 struct sctp_ulpevent *event;
6688 struct sctp_bind_hashbucket *head;
6689 struct list_head tmplist;
6690
6691 /* Migrate socket buffer sizes and all the socket level options to the
6692 * new socket.
6693 */
6694 newsk->sk_sndbuf = oldsk->sk_sndbuf;
6695 newsk->sk_rcvbuf = oldsk->sk_rcvbuf;
6696 /* Brute force copy old sctp opt. */
6697 if (oldsp->do_auto_asconf) {
6698 memcpy(&tmplist, &newsp->auto_asconf_list, sizeof(tmplist));
6699 inet_sk_copy_descendant(newsk, oldsk);
6700 memcpy(&newsp->auto_asconf_list, &tmplist, sizeof(tmplist));
6701 } else
6702 inet_sk_copy_descendant(newsk, oldsk);
6703
6704 /* Restore the ep value that was overwritten with the above structure
6705 * copy.
6706 */
6707 newsp->ep = newep;
6708 newsp->hmac = NULL;
6709
6710 /* Hook this new socket in to the bind_hash list. */
6711 head = &sctp_port_hashtable[sctp_phashfn(inet_sk(oldsk)->inet_num)];
6712 sctp_local_bh_disable();
6713 sctp_spin_lock(&head->lock);
6714 pp = sctp_sk(oldsk)->bind_hash;
6715 sk_add_bind_node(newsk, &pp->owner);
6716 sctp_sk(newsk)->bind_hash = pp;
6717 inet_sk(newsk)->inet_num = inet_sk(oldsk)->inet_num;
6718 sctp_spin_unlock(&head->lock);
6719 sctp_local_bh_enable();
6720
6721 /* Copy the bind_addr list from the original endpoint to the new
6722 * endpoint so that we can handle restarts properly
6723 */
6724 sctp_bind_addr_dup(&newsp->ep->base.bind_addr,
6725 &oldsp->ep->base.bind_addr, GFP_KERNEL);
6726
6727 /* Move any messages in the old socket's receive queue that are for the
6728 * peeled off association to the new socket's receive queue.
6729 */
6730 sctp_skb_for_each(skb, &oldsk->sk_receive_queue, tmp) {
6731 event = sctp_skb2event(skb);
6732 if (event->asoc == assoc) {
6733 __skb_unlink(skb, &oldsk->sk_receive_queue);
6734 __skb_queue_tail(&newsk->sk_receive_queue, skb);
6735 sctp_skb_set_owner_r_frag(skb, newsk);
6736 }
6737 }
6738
6739 /* Clean up any messages pending delivery due to partial
6740 * delivery. Three cases:
6741 * 1) No partial deliver; no work.
6742 * 2) Peeling off partial delivery; keep pd_lobby in new pd_lobby.
6743 * 3) Peeling off non-partial delivery; move pd_lobby to receive_queue.
6744 */
6745 skb_queue_head_init(&newsp->pd_lobby);
6746 atomic_set(&sctp_sk(newsk)->pd_mode, assoc->ulpq.pd_mode);
6747
6748 if (atomic_read(&sctp_sk(oldsk)->pd_mode)) {
6749 struct sk_buff_head *queue;
6750
6751 /* Decide which queue to move pd_lobby skbs to. */
6752 if (assoc->ulpq.pd_mode) {
6753 queue = &newsp->pd_lobby;
6754 } else
6755 queue = &newsk->sk_receive_queue;
6756
6757 /* Walk through the pd_lobby, looking for skbs that
6758 * need moved to the new socket.
6759 */
6760 sctp_skb_for_each(skb, &oldsp->pd_lobby, tmp) {
6761 event = sctp_skb2event(skb);
6762 if (event->asoc == assoc) {
6763 __skb_unlink(skb, &oldsp->pd_lobby);
6764 __skb_queue_tail(queue, skb);
6765 sctp_skb_set_owner_r_frag(skb, newsk);
6766 }
6767 }
6768
6769 /* Clear up any skbs waiting for the partial
6770 * delivery to finish.
6771 */
6772 if (assoc->ulpq.pd_mode)
6773 sctp_clear_pd(oldsk, NULL);
6774
6775 }
6776
6777 sctp_skb_for_each(skb, &assoc->ulpq.reasm, tmp)
6778 sctp_skb_set_owner_r_frag(skb, newsk);
6779
6780 sctp_skb_for_each(skb, &assoc->ulpq.lobby, tmp)
6781 sctp_skb_set_owner_r_frag(skb, newsk);
6782
6783 /* Set the type of socket to indicate that it is peeled off from the
6784 * original UDP-style socket or created with the accept() call on a
6785 * TCP-style socket..
6786 */
6787 newsp->type = type;
6788
6789 /* Mark the new socket "in-use" by the user so that any packets
6790 * that may arrive on the association after we've moved it are
6791 * queued to the backlog. This prevents a potential race between
6792 * backlog processing on the old socket and new-packet processing
6793 * on the new socket.
6794 *
6795 * The caller has just allocated newsk so we can guarantee that other
6796 * paths won't try to lock it and then oldsk.
6797 */
6798 lock_sock_nested(newsk, SINGLE_DEPTH_NESTING);
6799 sctp_assoc_migrate(assoc, newsk);
6800
6801 /* If the association on the newsk is already closed before accept()
6802 * is called, set RCV_SHUTDOWN flag.
6803 */
6804 if (sctp_state(assoc, CLOSED) && sctp_style(newsk, TCP))
6805 newsk->sk_shutdown |= RCV_SHUTDOWN;
6806
6807 newsk->sk_state = SCTP_SS_ESTABLISHED;
6808 sctp_release_sock(newsk);
6809}
6810
6811
6812/* This proto struct describes the ULP interface for SCTP. */
6813struct proto sctp_prot = {
6814 .name = "SCTP",
6815 .owner = THIS_MODULE,
6816 .close = sctp_close,
6817 .connect = sctp_connect,
6818 .disconnect = sctp_disconnect,
6819 .accept = sctp_accept,
6820 .ioctl = sctp_ioctl,
6821 .init = sctp_init_sock,
6822 .destroy = sctp_destroy_sock,
6823 .shutdown = sctp_shutdown,
6824 .setsockopt = sctp_setsockopt,
6825 .getsockopt = sctp_getsockopt,
6826 .sendmsg = sctp_sendmsg,
6827 .recvmsg = sctp_recvmsg,
6828 .bind = sctp_bind,
6829 .backlog_rcv = sctp_backlog_rcv,
6830 .hash = sctp_hash,
6831 .unhash = sctp_unhash,
6832 .get_port = sctp_get_port,
6833 .obj_size = sizeof(struct sctp_sock),
6834 .sysctl_mem = sysctl_sctp_mem,
6835 .sysctl_rmem = sysctl_sctp_rmem,
6836 .sysctl_wmem = sysctl_sctp_wmem,
6837 .memory_pressure = &sctp_memory_pressure,
6838 .enter_memory_pressure = sctp_enter_memory_pressure,
6839 .memory_allocated = &sctp_memory_allocated,
6840 .sockets_allocated = &sctp_sockets_allocated,
6841};
6842
6843#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
6844
6845struct proto sctpv6_prot = {
6846 .name = "SCTPv6",
6847 .owner = THIS_MODULE,
6848 .close = sctp_close,
6849 .connect = sctp_connect,
6850 .disconnect = sctp_disconnect,
6851 .accept = sctp_accept,
6852 .ioctl = sctp_ioctl,
6853 .init = sctp_init_sock,
6854 .destroy = sctp_destroy_sock,
6855 .shutdown = sctp_shutdown,
6856 .setsockopt = sctp_setsockopt,
6857 .getsockopt = sctp_getsockopt,
6858 .sendmsg = sctp_sendmsg,
6859 .recvmsg = sctp_recvmsg,
6860 .bind = sctp_bind,
6861 .backlog_rcv = sctp_backlog_rcv,
6862 .hash = sctp_hash,
6863 .unhash = sctp_unhash,
6864 .get_port = sctp_get_port,
6865 .obj_size = sizeof(struct sctp6_sock),
6866 .sysctl_mem = sysctl_sctp_mem,
6867 .sysctl_rmem = sysctl_sctp_rmem,
6868 .sysctl_wmem = sysctl_sctp_wmem,
6869 .memory_pressure = &sctp_memory_pressure,
6870 .enter_memory_pressure = sctp_enter_memory_pressure,
6871 .memory_allocated = &sctp_memory_allocated,
6872 .sockets_allocated = &sctp_sockets_allocated,
6873};
6874#endif /* defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) */